Explicitly link with libatomic when needed.
[official-gcc.git] / gcc / ChangeLog
blob7d95db8623fcc0d73389021c3c2f8d2657fd623b
1 2020-02-28  Michael Meissner  <meissner@linux.ibm.com>
3         PR target/93937
4         * config/rs6000/vsx.md (vsx_extract_<mode>_<VS_scalar>mode_var):
5         Delete insn.
7 2020-02-28  Martin Liska  <mliska@suse.cz>
9         PR other/93965
10         * configure.ac: Improve detection of ld_date by requiring
11         either two dashes or none.
12         * configure: Regenerate.
14 2020-02-28  Vladimir Makarov  <vmakarov@redhat.com>
16         PR rtl-optimization/93564
17         * ira-color.c (assign_hard_reg): Prefer smaller hard regno when we
18         do not honor reg alloc order.
20 2020-02-27  Joel Hutton  <Joel.Hutton@arm.com>
22         PR target/87612
23         * config/aarch64/aarch64.c (aarch64_override_options): Fix
24         misleading warning string.
26 2020-02-27  Martin Sebor  <msebor@redhat.com>
28         * doc/invoke.texi (-Wbuiltin-declaration-mismatch): Fix a typo.
30 2020-02-27  Michael Meissner  <meissner@linux.ibm.com>
32         PR target/93932
33         * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
34         Split the insn into two parts.  This insn only does variable
35         extract from a register.
36         (vsx_extract_<mode>_var_load, VSX_D iterator): New insn, do
37         variable extract from memory.
38         (vsx_extract_v4sf_var): Split the insn into two parts.  This insn
39         only does variable extract from a register.
40         (vsx_extract_v4sf_var_load): New insn, do variable extract from
41         memory.
42         (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Split the insn
43         into two parts.  This insn only does variable extract from a
44         register.
45         (vsx_extract_<mode>_var_load, VSX_EXTRACT_I iterator): New insn,
46         do variable extract from memory.
48 2020-02-27  Martin Jambor  <mjambor@suse.cz>
49             Feng Xue  <fxue@os.amperecomputing.com>
51         PR ipa/93707
52         * ipa-cp.c (same_node_or_its_all_contexts_clone_p): Replaced with
53         new function calls_same_node_or_its_all_contexts_clone_p.
54         (cgraph_edge_brings_value_p): Use it.
55         (cgraph_edge_brings_value_p): Likewise.
56         (self_recursive_pass_through_p): Return false if caller is a clone.
57         (self_recursive_agg_pass_through_p): Likewise.
59 2020-02-27  Jan Hubicka  <hubicka@ucw.cz>
61         PR middle-end/92152
62         * alias.c (ends_tbaa_access_path_p): Break out from ...
63         (component_uses_parent_alias_set_from): ... here.
64         * alias.h (ends_tbaa_access_path_p): Declare.
65         * tree-ssa-alias.c (access_path_may_continue_p): Break out from ...;
66         handle trailing arrays past end of tbaa access path.
67         (aliasing_component_refs_p): ... here; likewise.
68         (nonoverlapping_refs_since_match_p): Track TBAA segment of the access
69         path; disambiguate also past end of it.
70         (nonoverlapping_component_refs_p): Use only TBAA segment of the access
71         path.
73 2020-02-27  Mihail Ionescu  <mihail.ionescu@arm.com>
75         * (__ARM_NUM_LANES, __arm_lane, __arm_lane_q): Move to the
76         beginning of the file.
77         (vcreate_bf16, vcombine_bf16): New.
78         (vdup_n_bf16, vdupq_n_bf16): New.
79         (vdup_lane_bf16, vdup_laneq_bf16): New.
80         (vdupq_lane_bf16, vdupq_laneq_bf16): New.
81         (vduph_lane_bf16, vduph_laneq_bf16): New.
82         (vset_lane_bf16, vsetq_lane_bf16): New.
83         (vget_lane_bf16, vgetq_lane_bf16): New.
84         (vget_high_bf16, vget_low_bf16): New.
85         (vreinterpret_bf16_u8, vreinterpretq_bf16_u8): New.
86         (vreinterpret_bf16_u16, vreinterpretq_bf16_u16): New.
87         (vreinterpret_bf16_u32, vreinterpretq_bf16_u32): New.
88         (vreinterpret_bf16_u64, vreinterpretq_bf16_u64): New.
89         (vreinterpret_bf16_s8, vreinterpretq_bf16_s8): New.
90         (vreinterpret_bf16_s16, vreinterpretq_bf16_s16): New.
91         (vreinterpret_bf16_s32, vreinterpretq_bf16_s32): New.
92         (vreinterpret_bf16_s64, vreinterpretq_bf16_s64): New.
93         (vreinterpret_bf16_p8, vreinterpretq_bf16_p8): New.
94         (vreinterpret_bf16_p16, vreinterpretq_bf16_p16): New.
95         (vreinterpret_bf16_p64, vreinterpretq_bf16_p64): New.
96         (vreinterpret_bf16_f32, vreinterpretq_bf16_f32): New.
97         (vreinterpret_bf16_f64, vreinterpretq_bf16_f64): New.
98         (vreinterpretq_bf16_p128): New.
99         (vreinterpret_s8_bf16, vreinterpretq_s8_bf16): New.
100         (vreinterpret_s16_bf16, vreinterpretq_s16_bf16): New.
101         (vreinterpret_s32_bf16, vreinterpretq_s32_bf16): New.
102         (vreinterpret_s64_bf16, vreinterpretq_s64_bf16): New.
103         (vreinterpret_u8_bf16, vreinterpretq_u8_bf16): New.
104         (vreinterpret_u16_bf16, vreinterpretq_u16_bf16): New.
105         (vreinterpret_u32_bf16, vreinterpretq_u32_bf16): New.
106         (vreinterpret_u64_bf16, vreinterpretq_u64_bf16): New.
107         (vreinterpret_p8_bf16, vreinterpretq_p8_bf16): New.
108         (vreinterpret_p16_bf16, vreinterpretq_p16_bf16): New.
109         (vreinterpret_p64_bf16, vreinterpretq_p64_bf16): New.
110         (vreinterpret_f32_bf16, vreinterpretq_f32_bf16): New.
111         (vreinterpretq_p128_bf16): New.
112         * config/arm/arm_neon_builtins.def (VDX): Add V4BF.
113         (V_elem): Likewise.
114         (V_elem_l): Likewise.
115         (VD_LANE): Likewise.
116         (VQX) Add V8BF.
117         (V_DOUBLE): Likewise.
118         (VDQX): Add V4BF and V8BF.
119         (V_two_elem, V_three_elem, V_four_elem): Likewise.
120         (V_reg): Likewise.
121         (V_HALF): Likewise.
122         (V_double_vector_mode): Likewise.
123         (V_cmp_result): Likewise.
124         (V_uf_sclr): Likewise.
125         (V_sz_elem): Likewise.
126         (Is_d_reg): Likewise.
127         (V_mode_nunits): Likewise.
128         * config/arm/neon.md (neon_vdup_lane): Enable for BFloat16.
130 2020-02-27  Andrew Stubbs  <ams@codesourcery.com>
132         * config/gcn/gcn-valu.md (VEC_SUBDWORD_MODE): New mode iterator.
133         (<expander><mode>2<exec>): Change modes to VEC_ALL1REG_INT_MODE.
134         (<expander><mode>3<exec>): Likewise.
135         (<expander><mode>3): New.
136         (v<expander><mode>3): New.
137         (<expander><mode>3): New.
138         (<expander><mode>3<exec>): Rename to ...
139         (<expander>v64si3<exec>): ... this, and change modes to V64SI.
140         * config/gcn/gcn.md (mnemonic): Use '%B' for not.
142 2020-02-27  Alexandre Oliva <oliva@adacore.com>
144         * config/vx-common.h (NO_DOLLAR_IN_LABEL, NO_DOT_IN_LABEL): Leave
145         them alone on vx7.
147 2020-02-27  Richard Biener  <rguenther@suse.de>
149         PR tree-optimization/93508
150         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle _CHK like
151         non-_CHK variants.  Valueize their length arguments.
153 2020-02-27  Richard Biener  <rguenther@suse.de>
155         PR tree-optimization/93953
156         * tree-vect-slp.c (slp_copy_subtree): Avoid keeping a reference
157         to the hash-map entry.
159 2020-02-27  Andrew Stubbs  <ams@codesourcery.com>
161         * config/gcn/gcn.md (mov<mode>): Add transformations for BI subregs.
163 2020-02-27  Mark Williams  <mwilliams@fb.com>
165         * dwarf2out.c (file_name_acquire): Call remap_debug_filename.
166         * lto-opts.c (lto_write_options): Drop -fdebug-prefix-map,
167         -ffile-prefix-map and -fmacro-prefix-map.
168         * lto-streamer-out.c: Include file-prefix-map.h.
169         (lto_output_location): Remap the file part of locations.
171 2020-02-27  Jakub Jelinek  <jakub@redhat.com>
173         PR c/93949
174         * gimplify.c (gimplify_init_constructor): Don't promote readonly
175         DECL_REGISTER variables to TREE_STATIC.
177         PR tree-optimization/93582
178         PR tree-optimization/93945
179         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle memset with
180         non-zero INTEGER_CST second argument and ref->offset or ref->size
181         not a multiple of BITS_PER_UNIT.
183 2020-02-27  Jonathan Wakely  <jwakely@redhat.com>
185         * doc/install.texi (Binaries): Update description of BullFreeware.
187 2020-02-26  Sandra Loosemore  <sandra@codesourcery.com>
189         PR c++/90467
191         * doc/invoke.texi (Option Summary): Re-alphabetize warnings in
192         C++ Language Options, Warning Options, and Static Analyzer
193         Options lists.  Document negative form of options enabled by
194         default.  Move some things around to more accurately sort
195         warnings by category.
196         (C++ Dialect Options, Warning Options, Static Analyzer
197         Options): Document negative form of options when enabled by
198         default.  Move some things around to more accurately sort
199         warnings by category.  Add some missing index entries.
200         Light copy-editing.
202 2020-02-26  Carl Love  <cel@us.ibm.com>
204         PR target/91276
205         * doc/extend.texi (PowerPC AltiVec Built-in Functions available on
206         ISA 2.07): The builtin-function name __builtin_crypto_vpmsumb is only
207         for the vector unsigned short arguments.  It is also listed as the
208         name of the built-in for arguments vector unsigned short,
209         vector unsigned int and vector unsigned long long built-ins.  The
210         name of the builtins for these arguments should be:
211         __builtin_crypto_vpmsumh, __builtin_crypto_vpmsumw and
212         __builtin_crypto_vpmsumd respectively.
214 2020-02-26  Richard Biener  <rguenther@suse.de>
216         * tree-vect-slp.c (vect_print_slp_tree): Also dump ref count
217         and load permutation.
219 2020-02-26  Richard Sandiford  <richard.sandiford@arm.com>
221         PR middle-end/93843
222         * optabs-tree.c (supportable_convert_operation): Reject types with
223         scalar modes.
225 2020-02-26  David Malcolm  <dmalcolm@redhat.com>
227         * Makefile.in (ANALYZER_OBJS): Add analyzer/bar-chart.o.
229 2020-02-26  Jakub Jelinek  <jakub@redhat.com>
231         PR tree-optimization/93820
232         * gimple-ssa-store-merging.c (check_no_overlap): Change RHS_CODE
233         argument to ALL_INTEGER_CST_P boolean.
234         (imm_store_chain_info::try_coalesce_bswap): Adjust caller.
235         (imm_store_chain_info::coalesce_immediate_stores): Likewise.  Handle
236         adjacent INTEGER_CST store into merged_store->only_constants like
237         overlapping one.
239 2020-02-25  Jakub Jelinek  <jakub@redhat.com>
241         PR other/93912
242         * config/sh/sh.c (expand_cbranchdi4): Fix comment typo, probablity
243         -> probability.
244         * cfghooks.c (verify_flow_info): Likewise.
245         * predict.c (combine_predictions_for_bb): Likewise.
246         * bb-reorder.c (connect_better_edge_p): Likewise.  Fix comment typo,
247         sucessor -> successor.
248         (find_traces_1_round): Fix comment typo, destinarion -> destination.
249         * omp-expand.c (expand_oacc_for): Fix comment typo, sucessors ->
250         successors.
251         * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): Fix dump
252         message typo, sucessors -> successors.
254 2020-02-25  Martin Sebor  <msebor@redhat.com>
256         * doc/extend.texi (attribute access): Correct an example.
258 2020-02-25  Mihail Ionescu  <mihail.ionescu@arm.com>
260         * config/aarch64/aarch64-builtins.c (aarch64_scalar_builtin_types):
261         Add simd_bf.
262         (aarch64_init_simd_builtin_scalar_types): Register simd_bf.
263         (VAR15, VAR16): New.
264         * config/aarch64/iterators.md (VALLDIF): Enable for V4BF and V8BF.
265         (VD): Enable for V4BF.
266         (VDC): Likewise.
267         (VQ): Enable for V8BF.
268         (VQ2): Likewise.
269         (VQ_NO2E): Likewise.
270         (VDBL, Vdbl): Add V4BF.
271         (V_INT_EQUIV, v_int_equiv): Add V4BF and V8BF.
272         * config/aarch64/arm_neon.h (bfloat16x4x2_t): New typedef.
273         (bfloat16x8x2_t): Likewise.
274         (bfloat16x4x3_t): Likewise.
275         (bfloat16x8x3_t): Likewise.
276         (bfloat16x4x4_t): Likewise.
277         (bfloat16x8x4_t): Likewise.
278         (vcombine_bf16): New.
279         (vld1_bf16, vld1_bf16_x2): New.
280         (vld1_bf16_x3, vld1_bf16_x4): New.
281         (vld1q_bf16, vld1q_bf16_x2): New.
282         (vld1q_bf16_x3, vld1q_bf16_x4): New.
283         (vld1_lane_bf16): New.
284         (vld1q_lane_bf16): New.
285         (vld1_dup_bf16): New.
286         (vld1q_dup_bf16): New.
287         (vld2_bf16): New.
288         (vld2q_bf16): New.
289         (vld2_dup_bf16): New.
290         (vld2q_dup_bf16): New.
291         (vld3_bf16): New.
292         (vld3q_bf16): New.
293         (vld3_dup_bf16): New.
294         (vld3q_dup_bf16): New.
295         (vld4_bf16): New.
296         (vld4q_bf16): New.
297         (vld4_dup_bf16): New.
298         (vld4q_dup_bf16): New.
299         (vst1_bf16, vst1_bf16_x2): New.
300         (vst1_bf16_x3, vst1_bf16_x4): New.
301         (vst1q_bf16, vst1q_bf16_x2): New.
302         (vst1q_bf16_x3, vst1q_bf16_x4): New.
303         (vst1_lane_bf16): New.
304         (vst1q_lane_bf16): New.
305         (vst2_bf16): New.
306         (vst2q_bf16): New.
307         (vst3_bf16): New.
308         (vst3q_bf16): New.
309         (vst4_bf16): New.
310         (vst4q_bf16): New.
312 2020-02-25  Mihail Ionescu  <mihail.ionescu@arm.com>
314         * config/aarch64/iterators.md (VDQF_F16) Add V4BF and V8BF.
315         (VALL_F16): Likewise.
316         (VALLDI_F16): Likewise.
317         (Vtype): Likewise.
318         (Vetype): Likewise.
319         (vswap_width_name): Likewise.
320         (VSWAP_WIDTH): Likewise.
321         (Vel): Likewise.
322         (VEL): Likewise.
323         (q): Likewise.
324         * config/aarch64/arm_neon.h (vset_lane_bf16, vsetq_lane_bf16): New.
325         (vget_lane_bf16, vgetq_lane_bf16): New.
326         (vcreate_bf16): New.
327         (vdup_n_bf16, vdupq_n_bf16): New.
328         (vdup_lane_bf16, vdup_laneq_bf16): New.
329         (vdupq_lane_bf16, vdupq_laneq_bf16): New.
330         (vduph_lane_bf16, vduph_laneq_bf16): New.
331         (vreinterpret_bf16_u8, vreinterpretq_bf16_u8): New.
332         (vreinterpret_bf16_u16, vreinterpretq_bf16_u16): New.
333         (vreinterpret_bf16_u32, vreinterpretq_bf16_u32): New.
334         (vreinterpret_bf16_u64, vreinterpretq_bf16_u64): New.
335         (vreinterpret_bf16_s8, vreinterpretq_bf16_s8): New.
336         (vreinterpret_bf16_s16, vreinterpretq_bf16_s16): New.
337         (vreinterpret_bf16_s32, vreinterpretq_bf16_s32): New.
338         (vreinterpret_bf16_s64, vreinterpretq_bf16_s64): New.
339         (vreinterpret_bf16_p8, vreinterpretq_bf16_p8): New.
340         (vreinterpret_bf16_p16, vreinterpretq_bf16_p16): New.
341         (vreinterpret_bf16_p64, vreinterpretq_bf16_p64): New
342         (vreinterpret_bf16_f16, vreinterpretq_bf16_f16): New
343         (vreinterpret_bf16_f32, vreinterpretq_bf16_f32): New.
344         (vreinterpret_bf16_f64, vreinterpretq_bf16_f64): New.
345         (vreinterpretq_bf16_p128): New.
346         (vreinterpret_s8_bf16, vreinterpretq_s8_bf16): New.
347         (vreinterpret_s16_bf16, vreinterpretq_s16_bf16): New.
348         (vreinterpret_s32_bf16, vreinterpretq_s32_bf16): New.
349         (vreinterpret_s64_bf16, vreinterpretq_s64_bf16): New.
350         (vreinterpret_u8_bf16, vreinterpretq_u8_bf16): New.
351         (vreinterpret_u16_bf16, vreinterpretq_u16_bf16): New.
352         (vreinterpret_u32_bf16, vreinterpretq_u32_bf16): New.
353         (vreinterpret_u64_bf16, vreinterpretq_u64_bf16): New.
354         (vreinterpret_p8_bf16, vreinterpretq_p8_bf16): New.
355         (vreinterpret_p16_bf16, vreinterpretq_p16_bf16): New.
356         (vreinterpret_p64_bf16, vreinterpretq_p64_bf16): New.
357         (vreinterpret_f32_bf16, vreinterpretq_f32_bf16): New.
358         (vreinterpret_f64_bf16,vreinterpretq_f64_bf16): New.
359         (vreinterpret_f16_bf16,vreinterpretq_f16_bf16): New.
360         (vreinterpretq_p128_bf16): New.
362 2020-02-25  Dennis Zhang  <dennis.zhang@arm.com>
364         * config/arm/arm_neon.h (vbfdot_f32, vbfdotq_f32): New
365         (vbfdot_lane_f32, vbfdotq_laneq_f32): New.
366         (vbfdot_laneq_f32, vbfdotq_lane_f32): New.
367         * config/arm/arm_neon_builtins.def (vbfdot): New entry.
368         (vbfdot_lanev4bf, vbfdot_lanev8bf): Likewise.
369         * config/arm/iterators.md (VSF2BF): New attribute.
370         * config/arm/neon.md (neon_vbfdot<VCVTF:mode>): New entry.
371         (neon_vbfdot_lanev4bf<VCVTF:mode>): Likewise.
372         (neon_vbfdot_lanev8bf<VCVTF:mode>): Likewise.
374 2020-02-25  Christophe Lyon  <christophe.lyon@linaro.org>
376         * config/arm/arm.md (required_for_purecode): New attribute.
377         (enabled): Handle required_for_purecode.
378         * config/arm/thumb1.md (thumb1_movsi_insn): Add alternative to
379         work with -mpure-code.
381 2020-02-25  Jakub Jelinek  <jakub@redhat.com>
383         PR rtl-optimization/93908
384         * combine.c (find_split_point): For store into ZERO_EXTRACT, and src
385         with mask.
387 2019-02-25  Eric Botcazou  <ebotcazou@adacore.com>
389         * dwarf2out.c (dwarf2out_size_function): Run in early-DWARF mode.
391 2020-02-25  Roman Zhuykov  <zhroma@ispras.ru>
393         * doc/install.texi (--enable-checking): Adjust wording.
395 2020-02-25  Richard Biener  <rguenther@suse.de>
397         PR tree-optimization/93868
398         * tree-vect-slp.c (slp_copy_subtree): New function.
399         (vect_attempt_slp_rearrange_stmts): Copy the SLP tree before
400         re-arranging stmts in it.
402 2020-02-25  Jakub Jelinek  <jakub@redhat.com>
404         PR middle-end/93874
405         * passes.c (pass_manager::dump_passes): Create a cgraph node for the
406         dummy function and remove it at the end.
408         PR translation/93864
409         * config/lm32/lm32.c (lm32_setup_incoming_varargs): Fix comment typo
410         paramter -> parameter.
411         * config/aarch64/aarch64.c (aarch64_is_extend_from_extract): Likewise.
412         * ipa-prop.h (struct ipa_agg_replacement_value): Likewise.
414 2020-02-24  Roman Zhuykov  <zhroma@ispras.ru>
416         * doc/install.texi (--enable-checking): Properly document current
417         behavior.
418         (--enable-stage1-checking): Minor clarification about bootstrap.
420 2020-02-24  David Malcolm  <dmalcolm@redhat.com>
422         PR analyzer/93032
423         * doc/invoke.texi (-Wnanalyzer-tainted-array-index): Note that
424         -fanalyzer-checker=taint is also required.
425         (-fanalyzer-checker=): Note that providing this option enables the
426         given checker, and doing so may be required for checkers that are
427         disabled by default.
429 2020-02-24  David Malcolm  <dmalcolm@redhat.com>
431         * doc/invoke.texi (-fanalyzer-verbosity=): "2" only shows
432         significant control flow events; add a "3" which shows all
433         control flow events; the old "3" becomes "4".
435 2020-02-24  Jakub Jelinek  <jakub@redhat.com>
437         PR tree-optimization/93582
438         * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Consider
439         pd.offset and pd.size to be counted in bits rather than bytes, add
440         support for maxsizei that is not a multiple of BITS_PER_UNIT and
441         handle bitfield stores and loads.
442         (vn_reference_lookup_3): Don't call ranges_known_overlap_p with
443         uncomparable quantities - bytes vs. bits.  Allow push_partial_def
444         on offsets/sizes that aren't multiple of BITS_PER_UNIT and adjust
445         pd.offset/pd.size to be counted in bits rather than bytes.
446         Formatting fix.  Rename shadowed len variable to buflen.
448 2020-02-24  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
449             Kugan Vivekandarajah  <kugan.vivekanandarajah@linaro.org>
451         PR driver/47785
452         * gcc.c (putenv_COLLECT_AS_OPTIONS): New function.
453         (driver::main): Call putenv_COLLECT_AS_OPTIONS.
454         * opts-common.c (parse_options_from_collect_gcc_options): New function.
455         (prepend_xassembler_to_collect_as_options): Likewise.
456         * opts.h (parse_options_from_collect_gcc_options): Declare prototype.
457         (prepend_xassembler_to_collect_as_options): Likewise.
458         * lto-opts.c (lto_write_options): Stream assembler options
459         in COLLECT_AS_OPTIONS.
460         * lto-wrapper.c (xassembler_options_error): New static variable.
461         (get_options_from_collect_gcc_options): Move parsing options code to
462         parse_options_from_collect_gcc_options and call it.
463         (merge_and_complain): Validate -Xassembler options.
464         (append_compiler_options): Handle OPT_Xassembler.
465         (run_gcc): Append command line -Xassembler options to
466         collect_gcc_options.
467         * doc/invoke.texi: Add documentation about using Xassembler
468         options with LTO.
470 2020-02-24  Kito Cheng  <kito.cheng@sifive.com>
472         * config/riscv/riscv.c (riscv_emit_float_compare): Change the code gen
473         for LTGT.
474         (riscv_rtx_costs): Update cost model for LTGT.
476 2020-02-23  Vladimir Makarov  <vmakarov@redhat.com>
478         PR rtl-optimization/93564
479         * ira-color.c (struct update_cost_queue_elem): New member start.
480         (queue_update_cost, get_next_update_cost): Add new arg start.
481         (allocnos_conflict_p): New function.
482         (update_costs_from_allocno): Add new arg conflict_cost_update_p.
483         Add checking conflicts with allocnos_conflict_p.
484         (update_costs_from_prefs, restore_costs_from_copies): Adjust
485         update_costs_from_allocno calls.
486         (update_conflict_hard_regno_costs): Add checking conflicts with
487         allocnos_conflict_p.  Adjust calls of queue_update_cost and
488         get_next_update_cost.
489         (assign_hard_reg): Adjust calls of queue_update_cost.  Add
490         debugging print.
491         (bucket_allocno_compare_func): Restore previous version.
493 2020-02-21  John David Anglin  <danglin@gcc.gnu.org>
495         * gcc/config/pa/pa.c (pa_function_value): Fix check for word and
496         double-word size when handling aggregate return values.
497         * gcc/config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Fix to indicate
498         that homogeneous SFmode and DFmode aggregates are passed and returned
499         in general registers.
501 2020-02-21  Jakub Jelinek  <jakub@redhat.com>
503         PR translation/93759
504         * opts.c (print_filtered_help): Translate help before appending
505         messages to it rather than after that.
507 2020-02-19  Richard Sandiford  <richard.sandiford@arm.com>
509         PR rtl-optimization/PR92989
510         * lra-lives.c (process_bb_lives): Restore the original order
511         of the bb liveness update.  Call make_hard_regno_dead for each
512         register clobbered at the start of an EH receiver.
514 2020-02-18  Feng Xue  <fxue@os.amperecomputing.com>
516         PR ipa/93763
517         * ipa-cp.c (self_recursively_generated_p): Mark self-dependent value as
518         self-recursively generated.
520 2020-02-21  Iain Sandoe  <iain@sandoe.co.uk>
522         PR target/93860
523         * config/darwin-c.c (pop_field_alignment): Adjust quoting of
524         error string.
526 2020-02-21  Mihail Ionescu  <mihail.ionescu@arm.com>
528         * doc/sourcebuild.texi (arm_v8_1m_mve_ok):
529         Document new target supports option.
531 2020-02-21  Dennis Zhang  <dennis.zhang@arm.com>
533         * config/arm/arm_neon.h (vmmlaq_s32, vmmlaq_u32, vusmmlaq_s32): New.
534         * config/arm/arm_neon_builtins.def (smmla, ummla, usmmla): New.
535         * config/arm/iterators.md (MATMUL): New iterator.
536         (sup): Add UNSPEC_MATMUL_S, UNSPEC_MATMUL_U, and UNSPEC_MATMUL_US.
537         (mmla_sfx): New attribute.
538         * config/arm/neon.md (neon_<sup>mmlav16qi): New.
539         * config/arm/unspecs.md (UNSPEC_MATMUL_S, UNSPEC_MATMUL_U): New.
540         (UNSPEC_MATMUL_US): New.
542 2020-02-21  Mihail-Calin Ionescu  <mihail.ionescu@arm.com>
544         * config/arm/arm.md: Prevent scalar shifts from being used when big
545         endian is enabled.
547 2020-02-21  Jan Hubicka  <hubicka@ucw.cz>
548             Richard Biener  <rguenther@suse.de>
550         PR tree-optimization/93586
551         * tree-ssa-alias.c (nonoverlapping_array_refs_p): Finish array walk
552         after mismatched array refs; do not sure type size information to
553         recover from unmatched referneces with !flag_strict_aliasing_p.
555 2020-02-21  Andrew Stubbs  <ams@codesourcery.com>
557         * config/gcn/gcn-valu.md (gather_load<mode>): Rename to ...
558         (gather_load<mode>v64si): ... this and set operand 2 to V64SI.
559         (scatter_store<mode>): Rename to ...
560         (scatter_store<mode>v64si): ... this and set operand 1 to V64SI.
561         (scatter<mode>_exec): Delete. Move contents ...
562         (mask_scatter_store<mode>): ... here, and rename that to ...
563         (mask_gather_load<mode>v64si): ... this. Set operand 2 to V64SI.
564         Remove mode conversion.
565         (mask_gather_load<mode>): Rename to ...
566         (mask_scatter_store<mode>v64si): ... this. Set operand 1 to V64SI.
567         Remove mode conversion.
568         * config/gcn/gcn.c (gcn_expand_scaled_offsets): Remove mode conversion.
570 2020-02-21  Martin Jambor  <mjambor@suse.cz>
572         PR tree-optimization/93845
573         * tree-sra.c (verify_sra_access_forest): Only test access size of
574         scalar types.
576 2020-02-21  Andrew Stubbs  <ams@codesourcery.com>
578         * config/gcn/gcn.c (gcn_hard_regno_mode_ok): Align VGPR pairs.
579         * config/gcn/gcn-valu.md (addv64di3): Remove early-clobber.
580         (addv64di3_exec): Likewise.
581         (subv64di3): Likewise.
582         (subv64di3_exec): Likewise.
583         (addv64di3_zext): Likewise.
584         (addv64di3_zext_exec): Likewise.
585         (addv64di3_zext_dup): Likewise.
586         (addv64di3_zext_dup_exec): Likewise.
587         (addv64di3_zext_dup2): Likewise.
588         (addv64di3_zext_dup2_exec): Likewise.
589         (addv64di3_sext_dup2): Likewise.
590         (addv64di3_sext_dup2_exec): Likewise.
591         (<expander>v64di3): Likewise.
592         (<expander>v64di3_exec): Likewise.
593         (*<reduc_op>_dpp_shr_v64di): Likewise.
594         (*plus_carry_dpp_shr_v64di): Likewise.
595         * config/gcn/gcn.md (adddi3): Likewise.
596         (addptrdi3): Likewise.
597         (<expander>di3): Likewise.
599 2020-02-21  Andrew Stubbs  <ams@codesourcery.com>
601         * config/gcn/gcn-valu.md (vec_seriesv64di): Use gen_vec_duplicatev64di.
603 2020-02-21  Richard Sandiford  <richard.sandiford@arm.com>
605         * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Add SVE
606         support.  Use aarch64_emit_mult instead of emitting multiplication
607         instructions directly.
608         * config/aarch64/aarch64-sve.md (sqrt<mode>2, rsqrt<mode>2)
609         (@aarch64_rsqrte<mode>, @aarch64_rsqrts<mode>): New expanders.
611 2020-02-21  Richard Sandiford  <richard.sandiford@arm.com>
613         * config/aarch64/aarch64.c (aarch64_emit_mult): New function.
614         (aarch64_emit_approx_div): Add SVE support.  Use aarch64_emit_mult
615         instead of emitting multiplication instructions directly.
616         * config/aarch64/iterators.md (SVE_COND_FP_BINARY_OPTAB): New iterator.
617         * config/aarch64/aarch64-sve.md (div<mode>3, @aarch64_frecpe<mode>)
618         (@aarch64_frecps<mode>): New expanders.
620 2020-02-21  Richard Sandiford  <richard.sandiford@arm.com>
622         * config/aarch64/aarch64-protos.h (AARCH64_APPROX_MODE): Operate
623         on and produce uint64_ts rather than ints.
624         (AARCH64_APPROX_NONE, AARCH64_APPROX_ALL): Change to uint64_ts.
625         (cpu_approx_modes): Change the fields from unsigned int to uint64_t.
627 2020-02-21  Richard Sandiford  <richard.sandiford@arm.com>
629         * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Don't create
630         an unused xmsk register when handling approximate rsqrt.
632 2020-02-21  Richard Sandiford  <richard.sandiford@arm.com>
634         * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Fix inverted
635         flag_finite_math_only condition.
637 2020-02-20  UroÅ¡ Bizjak  <ubizjak@gmail.com>
639         PR target/93828
640         * config/i386/mmx.md (*vec_extractv2sf_1): Match source operand
641         to destination operand for shufps alternative.
642         (*vec_extractv2si_1): Ditto.
644 2020-02-20  Peter Bergner  <bergner@linux.ibm.com>
646         PR target/93658
647         * config/rs6000/rs6000.c (rs6000_legitimate_address_p): Handle VSX
648         vector modes.
650 2020-02-20  Martin Liska  <mliska@suse.cz>
652         PR translation/93831
653         * config/darwin.c (darwin_override_options): Change 64b to 64-bit mode.
655 2020-02-20  Martin Liska  <mliska@suse.cz>
657         PR translation/93830
658         * common/config/avr/avr-common.c: Remote trailing "|".
660 2020-02-19  Bernd Edlinger  <bernd.edlinger@hotmail.de>
662         * collect2.c (maybe_run_lto_and_relink): Fix typo in
663         comment.
665 2020-02-19  Richard Sandiford  <richard.sandiford@arm.com>
667         PR tree-optimization/93767
668         * tree-vect-data-refs.c (vect_compile_time_alias): Remove the
669         access-size bias from the offset calculations for negative strides.
671 2020-02-19  Bernd Edlinger  <bernd.edlinger@hotmail.de>
673         * collect2.c (c_file, o_file): Make const again.
674         (ldout,lderrout, dump_ld_file): Remove.
675         (tool_cleanup): Avoid calling not signal-safe functions.
676         (maybe_run_lto_and_relink): Avoid possible signal handler
677         access to unintialzed memory (lto_o_files).
678         (main): Avoid leaking temp files in $TMPDIR.
679         Initialize c_file/o_file with concat, which avoids exposing
680         uninitialized memory to signal handler, which calls unlink(!).
681         Avoid calling maybe_unlink when the main function returns,
682         since the atexit handler is already doing this.
683         * collect2.h (dump_ld_file, ldout, lderrout): Remove.
685 2020-02-19  Martin Jambor  <mjambor@suse.cz>
687         PR tree-optimization/93776
688         * tree-sra.c (create_access): Do not create zero size accesses.
689         (get_access_for_expr): Do not search for zero sized accesses.
691 2020-02-19  Martin Jambor  <mjambor@suse.cz>
693         PR tree-optimization/93667
694         * tree-sra.c (scalarizable_type_p): Return false if record fields
695         do not follow wach other.
697 2020-01-21  Kito Cheng  <kito.cheng@sifive.com>
699         * config/riscv/riscv.c (riscv_output_move) Using fmv.x.w/fmv.w.x
700         rather than fmv.x.s/fmv.s.x.
702 2020-02-18  James Greenhalgh  <james.greenhalgh@arm.com>
704         * config/aarch64/aarch64-simd-builtins.def
705         (intrinsic_vec_smult_lo_): New.
706         (intrinsic_vec_umult_lo_): Likewise.
707         (vec_widen_smult_hi_): Likewise.
708         (vec_widen_umult_hi_): Likewise.
709         * config/aarch64/aarch64-simd.md
710         (aarch64_intrinsic_vec_<su>mult_lo_<mode>): New.
711         * config/aarch64/arm_neon.h (vmull_high_s8): Use intrinsics.
712         (vmull_high_s16): Likewise.
713         (vmull_high_s32): Likewise.
714         (vmull_high_u8): Likewise.
715         (vmull_high_u16): Likewise.
716         (vmull_high_u32): Likewise.
717         (vmull_s8): Likewise.
718         (vmull_s16): Likewise.
719         (vmull_s32): Likewise.
720         (vmull_u8): Likewise.
721         (vmull_u16): Likewise.
722         (vmull_u32): Likewise.
724 2020-02-18  Martin Liska  <mliska@suse.cz>
726         * value-prof.c (stream_out_histogram_value): Restore LTO PGO
727         bootstrap by missing removal of invalid sanity check.
729 2020-02-18  Martin Liska  <mliska@suse.cz>
731         PR ipa/92518
732         * ipa-icf-gimple.c (func_checker::compare_gimple_assign):
733         Always compare LHS of gimple_assign.
735 2020-02-18  Martin Liska  <mliska@suse.cz>
737         PR ipa/93583
738         * cgraph.c (cgraph_node::verify_node): Verify MALLOC attribute
739         and return type of functions.
740         * ipa-param-manipulation.c (ipa_param_adjustments::adjust_decl):
741         Drop MALLOC attribute for void functions.
742         * ipa-pure-const.c (funct_state_summary_t::duplicate): Drop
743         malloc_state for a new VOID clone.
745 2020-02-18  Martin Liska  <mliska@suse.cz>
747         PR ipa/92924
748         * common.opt: Add -fprofile-reproducibility.
749         * doc/invoke.texi: Document it.
750         * value-prof.c (dump_histogram_value):
751         Document and support behavior for counters[0]
752         being a negative value.
753         (get_nth_most_common_value): Handle negative
754         counters[0] in respect to flag_profile_reproducible.
756 2020-02-18  Jakub Jelinek  <jakub@redhat.com>
758         PR ipa/93797
759         * cgraph.c (verify_speculative_call): Use speculative_id instead of
760         speculative_uid in messages.  Remove trailing whitespace from error
761         message.  Use num_speculative_call_targets instead of
762         num_speculative_targets in a message.
763         (cgraph_node::verify_node): Use call_stmt instead of cal_stmt in
764         edge messages and stmt instead of cal_stmt in reference message.
766         PR tree-optimization/93780
767         * tree-ssa.c (non_rewritable_lvalue_p): Check valid_vector_subparts_p
768         before calling build_vector_type.
769         (execute_update_addresses_taken): Likewise.
771         PR driver/93796
772         * params.opt (-param=ipa-max-switch-predicate-bounds=): Fix help
773         typo, functoin -> function.
774         * tree.c (free_lang_data_in_decl): Fix comment typo,
775         functoin -> function.
776         * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
778 2020-02-17  David Malcolm  <dmalcolm@redhat.com>
780         * diagnostic.c (print_any_cwe): Don't call get_cwe_url if URLs
781         won't be printed.
782         (print_option_information): Don't call get_option_url if URLs
783         won't be printed.
785 2020-02-17  Alexandre Oliva  <oliva@adacore.com>
787         * tree-emutls.c (new_emutls_decl, emutls_common_1): Complete
788         handling of register_common-less targets.
790 2020-02-17  Martin Liska  <mliska@suse.cz>
792         PR ipa/93760
793         * ipa-devirt.c (odr_types_equivalent_p): Fix grammar.
795 2020-02-17  Martin Liska  <mliska@suse.cz>
797         PR translation/93755
798         * config/rs6000/rs6000.c (rs6000_option_override_internal):
799         Fix double quotes.
801 2020-02-17  Martin Liska  <mliska@suse.cz>
803         PR other/93756
804         * config/rx/elf.opt: Fix typo.
806 2020-02-17  Richard Biener  <rguenther@suse.de>
808         PR c/86134
809         * opts-global.c (print_ignored_options): Use inform and
810         amend message.
812 2020-02-17  Jiufu Guo  <guojiufu@linux.ibm.com>
814         PR target/93047
815         * config/rs6000/rs6000.md (untyped_call): Add emit_clobber.
817 2020-02-16  UroÅ¡ Bizjak  <ubizjak@gmail.com>
819         PR target/93743
820         * config/i386/i386.md (atan2xf3): Swap operands 1 and 2.
821         (atan2<mode>3): Update operand order in the call to gen_atan2xf3.
823 2020-02-15  Jason Merrill  <jason@redhat.com>
825         * doc/invoke.texi (C Dialect Options): Add -std=c++20.
827 2020-02-15  Jakub Jelinek  <jakub@redhat.com>
829         PR tree-optimization/93744
830         * match.pd (((m1 >/</>=/<= m2) * d -> (m1 >/</>=/<= m2) ? d : 0,
831         A - ((A - B) & -(C cmp D)) -> (C cmp D) ? B : A,
832         A + ((B - A) & -(C cmp D)) -> (C cmp D) ? B : A): For GENERIC, make
833         sure @2 in the first and @1 in the other patterns has no side-effects.
835 2020-02-15  David Malcolm  <dmalcolm@redhat.com>
836             Bernd Edlinger  <bernd.edlinger@hotmail.de>
838         PR 87488
839         PR other/93168
840         * config.in (DIAGNOSTICS_URLS_DEFAULT): New define.
841         * configure.ac (--with-diagnostics-urls): New configuration
842         option, based on --with-diagnostics-color.
843         (DIAGNOSTICS_URLS_DEFAULT): New define.
844         * config.h: Regenerate.
845         * configure: Regenerate.
846         * diagnostic.c (diagnostic_urls_init): Handle -1 for
847         DIAGNOSTICS_URLS_DEFAULT from configure-time
848         --with-diagnostics-urls=auto-if-env by querying for a GCC_URLS
849         and TERM_URLS environment variable.
850         * diagnostic-url.h (diagnostic_url_format): New enum type.
851         (diagnostic_urls_enabled_p): rename to...
852         (determine_url_format): ... this, and change return type.
853         * diagnostic-color.c (parse_env_vars_for_urls): New helper function.
854         (auto_enable_urls): Disable URLs on xfce4-terminal, gnome-terminal,
855         the linux console, and mingw.
856         (diagnostic_urls_enabled_p): rename to...
857         (determine_url_format): ... this, and adjust.
858         * pretty-print.h (pretty_printer::show_urls): rename to...
859         (pretty_printer::url_format): ... this, and change to enum.
860         * pretty-print.c (pretty_printer::pretty_printer,
861         pp_begin_url, pp_end_url, test_urls): Adjust.
862         * doc/install.texi (--with-diagnostics-urls): Document the new
863         configuration option.
864         (--with-diagnostics-color): Document the existing interaction
865         with GCC_COLORS better.
866         * doc/invoke.texi (-fdiagnostics-urls): Add GCC_URLS and TERM_URLS
867         vindex reference.  Update description of defaults based on the above.
868         (-fdiagnostics-color): Update description of how -fdiagnostics-color
869         interacts with GCC_COLORS.
871 2020-02-14  Eric Botcazou  <ebotcazou@adacore.com>
873         PR target/93704
874         * config/sparc/sparc.c (eligible_for_call_delay): Test HAVE_GNU_LD in
875         conjunction with TARGET_GNU_TLS in early return.
877 2020-02-14  Alexander Monakov  <amonakov@ispras.ru>
879         * rtlanal.c (rtx_cost): Handle a SET up front. Avoid division if
880         the mode is not wider than UNITS_PER_WORD.
882 2020-02-14  Martin Jambor  <mjambor@suse.cz>
884         PR tree-optimization/93516
885         * tree-sra.c (propagate_subaccesses_from_rhs): Do not create
886         access of the same type as the parent.
887         (propagate_subaccesses_from_lhs): Likewise.
889 2020-02-14 Hongtao Liu  <hongtao.liu@intel.com>
891         PR target/93724
892         * config/i386/avx512vbmi2intrin.h
893         (_mm512_shrdi_epi16, _mm512_mask_shrdi_epi16,
894         _mm512_maskz_shrdi_epi16, _mm512_shrdi_epi32,
895         _mm512_mask_shrdi_epi32, _mm512_maskz_shrdi_epi32,
896         _m512_shrdi_epi64, _m512_mask_shrdi_epi64,
897         _m512_maskz_shrdi_epi64, _mm512_shldi_epi16,
898         _mm512_mask_shldi_epi16, _mm512_maskz_shldi_epi16,
899         _mm512_shldi_epi32, _mm512_mask_shldi_epi32,
900         _mm512_maskz_shldi_epi32, _mm512_shldi_epi64,
901         _mm512_mask_shldi_epi64, _mm512_maskz_shldi_epi64): Fix typo
902         of lacking a closing parenthesis.
903         * config/i386/avx512vbmi2vlintrin.h
904         (_mm256_shrdi_epi16, _mm256_mask_shrdi_epi16,
905         _mm256_maskz_shrdi_epi16, _mm256_shrdi_epi32,
906         _mm256_mask_shrdi_epi32, _mm256_maskz_shrdi_epi32,
907         _m256_shrdi_epi64, _m256_mask_shrdi_epi64,
908         _m256_maskz_shrdi_epi64, _mm256_shldi_epi16,
909         _mm256_mask_shldi_epi16, _mm256_maskz_shldi_epi16,
910         _mm256_shldi_epi32, _mm256_mask_shldi_epi32,
911         _mm256_maskz_shldi_epi32, _mm256_shldi_epi64,
912         _mm256_mask_shldi_epi64, _mm256_maskz_shldi_epi64,
913         _mm_shrdi_epi16, _mm_mask_shrdi_epi16,
914         _mm_maskz_shrdi_epi16, _mm_shrdi_epi32,
915         _mm_mask_shrdi_epi32, _mm_maskz_shrdi_epi32,
916         _mm_shrdi_epi64, _mm_mask_shrdi_epi64,
917         _m_maskz_shrdi_epi64, _mm_shldi_epi16,
918         _mm_mask_shldi_epi16, _mm_maskz_shldi_epi16,
919         _mm_shldi_epi32, _mm_mask_shldi_epi32,
920         _mm_maskz_shldi_epi32, _mm_shldi_epi64,
921         _mm_mask_shldi_epi64, _mm_maskz_shldi_epi64): Ditto.
923 2020-02-13  H.J. Lu  <hongjiu.lu@intel.com>
925         PR target/93656
926         * config/i386/i386.c (ix86_trampoline_init): Skip ENDBR32 at
927         the target function entry.
929 2020-02-13  Claudiu Zissulescu  <claziss@synopsys.com>
931         * common/config/arc/arc-common.c (arc_option_optimization_table):
932         Disable if-conversion step when optimized for size.
934 2020-02-13  Claudiu Zissulescu  <claziss@synopsys.com>
936         * config/arc/arc.c (arc_conditional_register_usage): R0-R3 and
937         R12-R15 are always in ARCOMPACT16_REGS register class.
938         * config/arc/arc.opt (mq-class): Deprecate.
939         * config/arc/constraint.md ("q"): Remove dependency on mq-class
940         option.
941         * doc/invoke.texi (mq-class): Update text.
942         * common/config/arc/arc-common.c (arc_option_optimization_table):
943         Update list.
945 2020-02-13  Claudiu Zissulescu  <claziss@synopsys.com>
947         * config/arc/arc.c (arc_insn_cost): New function.
948         (TARGET_INSN_COST): Define.
949         * config/arc/arc.md (cost): New attribute.
950         (add_n): Use arc_nonmemory_operand.
951         (ashlsi3_insn): Likewise, also update constraints.
952         (ashrsi3_insn): Likewise.
953         (rotrsi3): Likewise.
954         (add_shift): Likewise.
955         * config/arc/predicates.md (arc_nonmemory_operand): New predicate.
957 2020-02-13  Claudiu Zissulescu  <claziss@synopsys.com>
959         * config/arc/arc.md (mulsidi_600): Correctly select mlo/mhi
960         registers.
961         (umulsidi_600): Likewise.
963 2020-02-13  Jakub Jelinek  <jakub@redhat.com>
965         PR target/93696
966         * config/i386/avx512bitalgintrin.h (_mm512_mask_popcnt_epi8,
967         _mm512_mask_popcnt_epi16, _mm256_mask_popcnt_epi8,
968         _mm256_mask_popcnt_epi16, _mm_mask_popcnt_epi8,
969         _mm_mask_popcnt_epi16): Rename __B argument to __A and __A to __W,
970         pass __A to the builtin followed by __W instead of __A followed by
971         __B.
972         * config/i386/avx512vpopcntdqintrin.h (_mm512_mask_popcnt_epi32,
973         _mm512_mask_popcnt_epi64): Likewise.
974         * config/i386/avx512vpopcntdqvlintrin.h (_mm_mask_popcnt_epi32,
975         _mm256_mask_popcnt_epi32, _mm_mask_popcnt_epi64,
976         _mm256_mask_popcnt_epi64): Likewise.
978         PR tree-optimization/93582
979         * fold-const.h (shift_bytes_in_array_left,
980         shift_bytes_in_array_right): Declare.
981         * fold-const.c (shift_bytes_in_array_left,
982         shift_bytes_in_array_right): New function, moved from
983         gimple-ssa-store-merging.c, no longer static.
984         * gimple-ssa-store-merging.c (shift_bytes_in_array): Move
985         to gimple-ssa-store-merging.c and rename to shift_bytes_in_array_left.
986         (shift_bytes_in_array_right): Move to gimple-ssa-store-merging.c.
987         (encode_tree_to_bitpos): Use shift_bytes_in_array_left instead of
988         shift_bytes_in_array.
989         (verify_shift_bytes_in_array): Rename to ...
990         (verify_shift_bytes_in_array_left): ... this.  Use
991         shift_bytes_in_array_left instead of shift_bytes_in_array.
992         (store_merging_c_tests): Call verify_shift_bytes_in_array_left
993         instead of verify_shift_bytes_in_array.
994         * tree-ssa-sccvn.c (vn_reference_lookup_3): For native_encode_expr
995         / native_interpret_expr where the store covers all needed bits,
996         punt on PDP-endian, otherwise allow all involved offsets and sizes
997         not to be byte-aligned.
999         PR target/93673
1000         * config/i386/sse.md (k<code><mode>): Drop mode from last operand and
1001         use const_0_to_255_operand predicate instead of immediate_operand.
1002         (avx512dq_fpclass<mode><mask_scalar_merge_name>,
1003         avx512dq_vmfpclass<mode><mask_scalar_merge_name>,
1004         vgf2p8affineinvqb_<mode><mask_name>,
1005         vgf2p8affineqb_<mode><mask_name>): Drop mode from
1006         const_0_to_255_operand predicated operands.
1008 2020-02-12  Jeff Law  <law@redhat.com>
1010         * config/h8300/h8300.md (comparison shortening peepholes): Use
1011         a mode iterator to merge the HImode and SImode peepholes.
1013 2020-02-12  Jakub Jelinek  <jakub@redhat.com>
1015         PR middle-end/93663
1016         * real.c (is_even): Make static.  Function comment fix.
1017         (is_halfway_below): Make static, don't assert R is not inf/nan,
1018         instead return false for those.  Small formatting fixes.
1020 2020-02-12  Martin Sebor  <msebor@redhat.com>
1022         PR middle-end/93646
1023         * tree-ssa-strlen.c (handle_builtin_stxncpy): Rename...
1024         (handle_builtin_stxncpy_strncat): ...to this.  Change first argument.
1025         Issue only -Wstringop-overflow strncat, never -Wstringop-truncation.
1026         (strlen_check_and_optimize_call): Adjust callee name.
1028 2020-02-12  Jeff Law  <law@redhat.com>
1030         * config/h8300/h8300.md (comparison shortening peepholes): Drop
1031         (and (xor)) variant.  Combine other two into single peephole.
1033 2020-02-12  Wilco Dijkstra  <wdijkstr@arm.com>
1035         PR rtl-optimization/93565
1036         * config/aarch64/aarch64.c (aarch64_rtx_costs): Add CTZ costs.
1038 2020-02-12  Wilco Dijkstra  <wdijkstr@arm.com>
1040         * config/aarch64/aarch64-simd.md
1041         (aarch64_zero_extend<GPI:mode>_reduc_plus_<VDQV_E:mode>): New pattern.
1042         * config/aarch64/aarch64.md (popcount<mode>2): Use it instead of
1043         generating separate ADDV and zero_extend patterns.
1044         * config/aarch64/iterators.md (VDQV_E): New iterator.
1046 2020-02-12  Jeff Law  <law@redhat.com>
1048         * config/h8300/h8300.md (cpymemsi, movmd): Remove dead patterns,
1049         expanders, splits, etc.
1050         (movmd_internal_<mode>, movmd splitter, movstr, movsd): Likewise.
1051         (stpcpy_internal_<mode>, stpcpy splitter): Likewise.
1052         (peepholes to convert QI/HI mode pushes to SI mode pushes): Likewise.
1053         * config/h8300/h8300.c (h8300_swap_into_er6): Remove unused function.
1054         (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise
1055         * config/h8300/h8300-protos.h (h8300_swap_into_er6): Remove unused
1056         function prototype.
1057         (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise.
1059 2020-02-12  Jakub Jelinek  <jakub@redhat.com>
1061         PR target/93670
1062         * config/i386/sse.md (VI48F_256_DQ): New mode iterator.
1063         (avx512vl_vextractf128<mode>): Use it instead of VI48F_256.  Remove
1064         TARGET_AVX512DQ from condition.
1065         (vec_extract_lo_<mode><mask_name>): Use <mask_avx512dq_condition>
1066         instead of <mask_mode512bit_condition> in condition.  If
1067         TARGET_AVX512DQ is false, emit vextract*64x4 instead of
1068         vextract*32x8.
1069         (vec_extract_lo_<mode><mask_name>): Drop <mask_avx512dq_condition>
1070         from condition.
1072 2020-02-12  Kewen Lin  <linkw@gcc.gnu.org>
1074         PR target/91052
1075         * ira.c (combine_and_move_insns): Skip multiple_sets def_insn.
1077 2020-02-12  Segher Boessenkool  <segher@kernel.crashing.org>
1079         * config/rs6000/rs6000.c (rs6000_debug_print_mode): Don't use sizeof
1080         where strlen is more legible.
1081         (rs6000_builtin_vectorized_libmass): Ditto.
1082         (rs6000_print_options_internal): Ditto.
1084 2020-02-11  Martin Sebor  <msebor@redhat.com>
1086         PR tree-optimization/93683
1087         * tree-ssa-alias.c (stmt_kills_ref_p): Avoid using LHS when not set.
1089 2020-02-11  Michael Meissner  <meissner@linux.ibm.com>
1091         * config/rs6000/predicates.md (cint34_operand): Rename the
1092         -mprefixed-addr option to be -mprefixed.
1093         * config/rs6000/rs6000-cpus.def (ISA_FUTURE_MASKS_SERVER): Rename
1094         the -mprefixed-addr option to be -mprefixed.
1095         (OTHER_FUTURE_MASKS): Likewise.
1096         (POWERPC_MASKS): Likewise.
1097         * config/rs6000/rs6000.c (rs6000_option_override_internal): Rename
1098         the -mprefixed-addr option to be -mprefixed.  Change error
1099         messages to refer to -mprefixed.
1100         (num_insns_constant_gpr): Rename the -mprefixed-addr option to be
1101         -mprefixed.
1102         (rs6000_legitimate_offset_address_p): Likewise.
1103         (rs6000_mode_dependent_address): Likewise.
1104         (rs6000_opt_masks): Change the spelling of "-mprefixed-addr" to be
1105         "-mprefixed" for target attributes and pragmas.
1106         (address_to_insn_form): Rename the -mprefixed-addr option to be
1107         -mprefixed.
1108         (rs6000_adjust_insn_length): Likewise.
1109         * config/rs6000/rs6000.h (FINAL_PRESCAN_INSN): Rename the
1110         -mprefixed-addr option to be -mprefixed.
1111         (ASM_OUTPUT_OPCODE): Likewise.
1112         * config/rs6000/rs6000.md (prefixed insn attribute): Rename the
1113         -mprefixed-addr option to be -mprefixed.
1114         * config/rs6000/rs6000.opt (-mprefixed): Rename the
1115         -mprefixed-addr option to be prefixed.  Change the option from
1116         being undocumented to being documented.
1117         * doc/invoke.texi (RS/6000 and PowerPC Options): Document the
1118         -mprefixed option.  Update the -mpcrel documentation to mention
1119         -mprefixed.
1121 2020-02-11  Hans-Peter Nilsson  <hp@axis.com>
1123         * ira-conflicts.c (print_hard_reg_set): Correct output for sets
1124         including FIRST_PSEUDO_REGISTER - 1.
1125         * ira-color.c (print_hard_reg_set): Ditto.
1127 2020-02-11  Stam Markianos-Wright  <stam.markianos-wright@arm.com>
1129         * config/arm/arm-builtins.c (enum arm_type_qualifiers): 
1130         (USTERNOP_QUALIFIERS): New define.
1131         (USMAC_LANE_QUADTUP_QUALIFIERS): New define.
1132         (SUMAC_LANE_QUADTUP_QUALIFIERS): New define.
1133         (arm_expand_builtin_args): Add case ARG_BUILTIN_LANE_QUADTUP_INDEX.
1134         (arm_expand_builtin_1): Add qualifier_lane_quadtup_index.
1135         * config/arm/arm_neon.h (vusdot_s32): New.
1136         (vusdot_lane_s32): New.
1137         (vusdotq_lane_s32): New.
1138         (vsudot_lane_s32): New.
1139         (vsudotq_lane_s32): New.
1140         * config/arm/arm_neon_builtins.def (usdot, usdot_lane,sudot_lane): New.
1141         * config/arm/iterators.md (DOTPROD_I8MM): New.
1142         (sup, opsuffix): Add <us/su>.
1143         * config/arm/neon.md (neon_usdot, <us/su>dot_lane: New.
1144         * config/arm/unspecs.md (UNSPEC_DOT_US, UNSPEC_DOT_SU): New.
1146 2020-02-11  Richard Biener  <rguenther@suse.de>
1148         PR tree-optimization/93661
1149         PR tree-optimization/93662
1150         * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly guard
1151         tree_to_poly_int64.
1152         * tree-sra.c (get_access_for_expr): Likewise.
1154 2020-02-10  Jakub Jelinek  <jakub@redhat.com>
1156         PR target/93637
1157         * config/i386/sse.md (VI_256_AVX2): New mode iterator.
1158         (vcond_mask_<mode><sseintvecmodelower>): Use it instead of VI_256.
1159         Change condition from TARGET_AVX2 to TARGET_AVX.
1161 2020-02-10  Iain Sandoe  <iain@sandoe.co.uk>
1163         PR other/93641
1164         * config/darwin-c.c (darwin_cfstring_ref_p): Fix up last
1165         argument of strncmp.
1167 2020-02-10  Hans-Peter Nilsson  <hp@axis.com>
1169         Try to generate zero-based comparisons.
1170         * config/cris/cris.c (cris_reduce_compare): New function.
1171         * config/cris/cris-protos.h  (cris_reduce_compare): Add prototype.
1172         * config/cris/cris.md ("cbranch<mode>4", "cbranchdi4", "cstoredi4")
1173         (cstore<mode>4"): Apply cris_reduce_compare in expanders.
1175 2020-02-10  Richard Earnshaw  <rearnsha@arm.com>
1177         PR target/91913
1178         * config/arm/arm.md (movsi_compare0): Allow SP as a source register
1179         in Thumb state and also as a destination in Arm state.  Add T16
1180         variants.
1182 2020-02-10  Hans-Peter Nilsson  <hp@axis.com>
1184         * md.texi (Define Subst): Match closing paren in example.
1186 2020-02-10  Jakub Jelinek  <jakub@redhat.com>
1188         PR target/58218
1189         PR other/93641
1190         * config/i386/i386.c (x86_64_elf_section_type_flags): Fix up last
1191         arguments of strncmp.
1193 2020-02-10  Feng Xue  <fxue@os.amperecomputing.com>
1195         PR ipa/93203
1196         * ipa-cp.c (ipcp_lattice::add_value): Add source with same call edge
1197         but different source value.
1198         (adjust_callers_for_value_intersection): New function.
1199         (gather_edges_for_value): Adjust order of callers to let a
1200         non-self-recursive caller be the first element.
1201         (self_recursive_pass_through_p): Add a new parameter "simple", and
1202         check generalized self-recursive pass-through jump function.
1203         (self_recursive_agg_pass_through_p): Likewise.
1204         (find_more_scalar_values_for_callers_subset): Compute value from
1205         pass-through jump function for self-recursive.
1206         (intersect_with_plats): Cleanup previous implementation code for value
1207         itersection with self-recursive call edge.
1208         (intersect_with_agg_replacements): Likewise.
1209         (intersect_aggregates_with_edge): Deduce value from pass-through jump
1210         function for self-recursive call edge.  Cleanup previous implementation
1211         code for value intersection with self-recursive call edge.
1212         (decide_whether_version_node): Remove dead callers and adjust order
1213         to let a non-self-recursive caller be the first element.
1215 2020-02-09  UroÅ¡ Bizjak  <ubizjak@gmail.com>
1217         * recog.c: Move pass_split_before_sched2 code in front of
1218         pass_split_before_regstack.
1219         (pass_data_split_before_sched2): Rename pass to split3 from split4.
1220         (pass_data_split_before_regstack): Rename pass to split4 from split3.
1221         (rest_of_handle_split_before_sched2): Remove.
1222         (pass_split_before_sched2::execute): Unconditionally call
1223         split_all_insns.
1224         (enable_split_before_sched2): New function.
1225         (pass_split_before_sched2::gate): Use enable_split_before_sched2.
1226         (pass_split_before_regstack::gate): Ditto.
1227         * config/nds32/nds32.c (nds32_split_double_word_load_store_p):
1228         Update name check for renamed split4 pass.
1229         * config/sh/sh.c (register_sh_passes): Update pass insertion
1230         point for renamed split4 pass.
1232 2020-02-09  Jakub Jelinek  <jakub@redhat.com>
1234         * gimplify.c (gimplify_adjust_omp_clauses_1): Promote
1235         DECL_IN_CONSTANT_POOL variables into "omp declare target" to avoid
1236         copying them around between host and target.
1238 2020-02-08  Andrew Pinski  <apinski@marvell.com>
1240         PR target/91927
1241         * config/aarch64/aarch64-simd.md (movmisalign<mode>): Check
1242         STRICT_ALIGNMENT also.
1244 2020-02-08  Jim Wilson  <jimw@sifive.com>
1246         PR target/93532
1247         * config/riscv/riscv.h (HARD_REGNO_CALLER_SAVE_MODE): Define.
1249 2020-02-08  UroÅ¡ Bizjak  <ubizjak@gmail.com>
1250             Jakub Jelinek  <jakub@redhat.com>
1252         PR target/65782
1253         * config/i386/i386.h (CALL_USED_REGISTERS): Make
1254         xmm16-xmm31 call-used even in 64-bit ms-abi.
1256 2020-02-07  Dennis Zhang  <dennis.zhang@arm.com>
1258         * config/aarch64/aarch64-simd-builtins.def (simd_smmla): New entry.
1259         (simd_ummla, simd_usmmla): Likewise.
1260         * config/aarch64/aarch64-simd.md (aarch64_simd_<sur>mmlav16qi): New.
1261         * config/aarch64/arm_neon.h (vmmlaq_s32, vmmlaq_u32): New.
1262         (vusmmlaq_s32): New.
1264 2020-02-07  Richard Biener  <rguenther@suse.de>
1266         PR middle-end/93519
1267         * tree-inline.c (fold_marked_statements): Do a PRE walk,
1268         skipping unreachable regions.
1269         (optimize_inline_calls): Skip folding stmts when we didn't
1270         inline.
1272 2020-02-07  H.J. Lu  <hongjiu.lu@intel.com>
1274         PR target/85667
1275         * config/i386/i386.c (function_arg_ms_64): Add a type argument.
1276         Don't return aggregates with only SFmode and DFmode in SSE
1277         register.
1278         (ix86_function_arg): Pass arg.type to function_arg_ms_64.
1280 2020-02-07  Jakub Jelinek  <jakub@redhat.com>
1282         PR target/93122
1283         * config/rs6000/rs6000-logue.c
1284         (rs6000_emit_probe_stack_range_stack_clash): Always use gen_add3_insn,
1285         if it fails, move rs into end_addr and retry.  Add
1286         REG_FRAME_RELATED_EXPR note whenever it returns more than one insn or
1287         the insn pattern doesn't describe well what exactly happens to
1288         dwarf2cfi.c.
1290         PR target/93594
1291         * config/i386/predicates.md (avx_identity_operand): Remove.
1292         * config/i386/sse.md (*avx_vec_concat<mode>_1): Remove.
1293         (avx_<castmode><avxsizesuffix>_<castmode>,
1294         avx512f_<castmode><avxsizesuffix>_256<castmode>): Change patterns to
1295         a VEC_CONCAT of the operand and UNSPEC_CAST.
1296         (avx512f_<castmode><avxsizesuffix>_<castmode>): Change pattern to
1297         a VEC_CONCAT of VEC_CONCAT of the operand and UNSPEC_CAST with
1298         UNSPEC_CAST.
1300         PR target/93611
1301         * config/i386/i386.c (ix86_lea_outperforms): Make sure to clear
1302         recog_data.insn if distance_non_agu_define changed it.
1304 2020-02-06  Michael Meissner  <meissner@linux.ibm.com>
1306         PR target/93569
1307         * config/rs6000/rs6000.c (reg_to_non_prefixed): Before ISA 3.0
1308         we only had X-FORM (reg+reg) addressing for vectors.  Also before
1309         ISA 3.0, we only had X-FORM addressing for scalars in the
1310         traditional Altivec registers.
1312 2020-02-06  <zhongyunde@huawei.com>
1313             Vladimir Makarov  <vmakarov@redhat.com>
1315         PR rtl-optimization/93561
1316         * lra-assigns.c (spill_for): Check that tested hard regno is not out of
1317         hard register range.
1319 2020-02-06  Richard Sandiford  <richard.sandiford@arm.com>
1321         * config/aarch64/aarch64.md (aarch64_movk<mode>): Add a type
1322         attribute.
1324 2020-02-06  Segher Boessenkool  <segher@kernel.crashing.org>
1326         * config/rs6000/rs6000.c (rs6000_emit_set_long_const): Handle the case
1327         where the low and the high 32 bits are equal to each other specially,
1328         with an rldimi instruction.
1330 2020-02-06  Mihail Ionescu  <mihail.ionescu@arm.com>
1332         * config/arm/arm-cpus.in: Set profile M for armv8.1-m.main.
1334 2020-02-06  Mihail Ionescu  <mihail.ionescu@arm.com>
1336         * config/arm/arm-tables.opt: Regenerate.
1338 2020-02-06  Richard Sandiford  <richard.sandiford@arm.com>
1340         PR target/87763
1341         * config/aarch64/aarch64-protos.h (aarch64_movk_shift): Declare.
1342         * config/aarch64/aarch64.c (aarch64_movk_shift): New function.
1343         * config/aarch64/aarch64.md (aarch64_movk<mode>): New pattern.
1345 2020-02-06  Richard Sandiford  <richard.sandiford@arm.com>
1347         PR rtl-optimization/87763
1348         * config/aarch64/aarch64.md (*ashiftsi_extvdi_bfiz): New pattern.
1350 2020-02-06  Delia Burduv  <delia.burduv@arm.com>
1352         * config/aarch64/aarch64-simd-builtins.def
1353         (bfmlaq): New built-in function.
1354         (bfmlalb): New built-in function.
1355         (bfmlalt): New built-in function.
1356         (bfmlalb_lane): New built-in function.
1357         (bfmlalt_lane): New built-in function.
1358         * config/aarch64/aarch64-simd.md
1359         (aarch64_bfmmlaqv4sf): New pattern.
1360         (aarch64_bfmlal<bt>v4sf): New pattern.
1361         (aarch64_bfmlal<bt>_lane<q>v4sf): New pattern.
1362         * config/aarch64/arm_neon.h (vbfmmlaq_f32): New intrinsic.
1363         (vbfmlalbq_f32): New intrinsic.
1364         (vbfmlaltq_f32): New intrinsic.
1365         (vbfmlalbq_lane_f32): New intrinsic.
1366         (vbfmlaltq_lane_f32): New intrinsic.
1367         (vbfmlalbq_laneq_f32): New intrinsic.
1368         (vbfmlaltq_laneq_f32): New intrinsic.
1369         * config/aarch64/iterators.md (BF_MLA): New int iterator.
1370         (bt): New int attribute.
1372 2020-02-06  UroÅ¡ Bizjak  <ubizjak@gmail.com>
1374         * config/i386/i386.md (*pushtf): Emit "#" instead of
1375         calling gcc_unreachable in insn output.
1376         (*pushxf): Ditto.
1377         (*pushdf): Ditto.
1378         (*pushsf_rex64): Ditto for alternatives other than 1.
1379         (*pushsf): Ditto for alternatives other than 1.
1381 2020-02-06  Martin Liska  <mliska@suse.cz>
1383         PR gcov-profile/91971
1384         PR gcov-profile/93466
1385         * coverage.c (coverage_init): Revert mangling of
1386         path into filename.  It can lead to huge filename length.
1387         Creation of subfolders seem more natural.
1389 2020-02-06  Stam Markianos-Wright  <stam.markianos-wright@arm.com>
1391         PR target/93300
1392         * config/arm/arm.c (arm_block_arith_comp_libfuncs_for_mode): New.
1393         (arm_init_libfuncs): Add BFmode support to block spurious BF libfuncs.
1394         Use arm_block_arith_comp_libfuncs_for_mode for HFmode.
1396 2020-02-06  Jakub Jelinek  <jakub@redhat.com>
1398         PR target/93594
1399         * config/i386/predicates.md (avx_identity_operand): New predicate.
1400         * config/i386/sse.md (*avx_vec_concat<mode>_1): New
1401         define_insn_and_split.
1403         PR libgomp/93515
1404         * omp-low.c (use_pointer_for_field): For nested constructs, also
1405         look for map clauses on target construct.
1406         (scan_omp_1_stmt) <case GIMPLE_OMP_TARGET>: Bump temporarily
1407         taskreg_nesting_level.
1409         PR libgomp/93515
1410         * gimplify.c (gimplify_scan_omp_clauses) <do_notice>: If adding
1411         shared clause, call omp_notice_variable on outer context if any.
1413 2020-02-05  Jason Merrill  <jason@redhat.com>
1415         PR c++/92003
1416         * symtab.c (symtab_node::nonzero_address): A DECL_COMDAT decl has
1417         non-zero address even if weak and not yet defined.
1419 2020-02-05  Martin Sebor  <msebor@redhat.com>
1421         PR tree-optimization/92765
1422         * gimple-fold.c (get_range_strlen_tree): Handle MEM_REF and PARM_DECL.
1423         * tree-ssa-strlen.c (compute_string_length): Remove.
1424         (determine_min_objsize): Remove.
1425         (get_len_or_size): Add an argument.  Call get_range_strlen_dynamic.
1426         Avoid using type size as the upper bound on string length.
1427         (handle_builtin_string_cmp): Add an argument.  Adjust.
1428         (strlen_check_and_optimize_call): Pass additional argument to
1429         handle_builtin_string_cmp.
1431 2020-02-05  UroÅ¡ Bizjak  <ubizjak@gmail.com>
1433         * config/i386/i386.md (*pushdi2_rex64 peephole2): Remove.
1434         (*pushdi2_rex64 peephole2): Unconditionally split after
1435         epilogue_completed.
1436         (*ashl<mode>3_doubleword): Ditto.
1437         (*<shift_insn><mode>3_doubleword): Ditto.
1439 2020-02-05  Michael Meissner  <meissner@linux.ibm.com>
1441         PR target/93568
1442         * config/rs6000/rs6000.c (get_vector_offset): Fix
1444 2020-02-05  Andrew Stubbs  <ams@codesourcery.com>
1446         * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Use / not space.
1448 2020-02-05  David Malcolm  <dmalcolm@redhat.com>
1450         * doc/analyzer.texi
1451         (Special Functions for Debugging the Analyzer): Update description
1452         of __analyzer_dump_exploded_nodes.
1454 2020-02-05  Jakub Jelinek  <jakub@redhat.com>
1456         PR target/92190
1457         * config/i386/i386-features.c (ix86_add_reg_usage_to_vzeroupper): Only
1458         include sets and not clobbers in the vzeroupper pattern.
1459         * config/i386/sse.md (*avx_vzeroupper): Require in insn condition that
1460         the parallel has 17 (64-bit) or 9 (32-bit) elts.
1461         (*avx_vzeroupper_1): New define_insn_and_split.
1463         PR target/92190
1464         * recog.c (pass_split_after_reload::gate): For STACK_REGS targets,
1465         don't run when !optimize.
1466         (pass_split_before_regstack::gate): For STACK_REGS targets, run even
1467         when !optimize.
1469 2020-02-05  Richard Biener  <rguenther@suse.de>
1471         PR middle-end/90648
1472         * genmatch.c (dt_node::gen_kids_1): Emit number of argument
1473         checks before matching calls.
1475 2020-02-05  Jakub Jelinek  <jakub@redhat.com>
1477         * tree-ssa-alias.c (aliasing_matching_component_refs_p): Fix up
1478         function comment typo.
1480         PR middle-end/93555
1481         * omp-simd-clone.c (expand_simd_clones): If simd_clone_mangle or
1482         simd_clone_create failed when i == 0, adjust clone->nargs by
1483         clone->inbranch.
1485 2020-02-05  Martin Liska  <mliska@suse.cz>
1487         PR c++/92717
1488         * doc/invoke.texi: Document that one should
1489         not combine ASLR and -fpch.
1491 2020-02-04  Richard Biener  <rguenther@suse.de>
1493         PR tree-optimization/93538
1494         * match.pd (addr EQ/NE ptr): Amend to handle &ptr->x EQ/NE ptr.
1496 2020-02-04  Richard Biener  <rguenther@suse.de>
1498         PR tree-optimization/91123
1499         * tree-ssa-sccvn.c (vn_walk_cb_data::finish): New method.
1500         (vn_walk_cb_data::last_vuse): New member.
1501         (vn_walk_cb_data::saved_operands): Likewsie.
1502         (vn_walk_cb_data::~vn_walk_cb_data): Release saved_operands.
1503         (vn_walk_cb_data::push_partial_def): Use finish.
1504         (vn_reference_lookup_2): Update last_vuse and use finish if
1505         we've saved operands.
1506         (vn_reference_lookup_3): Use finish and update calls to
1507         push_partial_defs everywhere.  When translating through
1508         memcpy or aggregate copies save off operands and alias-set.
1509         (eliminate_dom_walker::eliminate_stmt): Restore VN_WALKREWRITE
1510         operation for redundant store removal.
1512 2020-02-04  Richard Biener  <rguenther@suse.de>
1514         PR tree-optimization/92819
1515         * tree-ssa-forwprop.c (simplify_vector_constructor): Avoid
1516         generating more stmts than before.
1518 2020-02-04  Martin Liska  <mliska@suse.cz>
1520         * config/arm/arm.c (arm_gen_far_branch): Move the function
1521         outside of selftests.
1523 2020-02-03  Michael Meissner  <meissner@linux.ibm.com>
1525         * config/rs6000/rs6000.c (adjust_vec_address_pcrel): New helper
1526         function to adjust PC-relative vector addresses.
1527         (rs6000_adjust_vec_address): Call adjust_vec_address_pcrel to
1528         handle vectors with PC-relative addresses.
1530 2020-02-03  Michael Meissner  <meissner@linux.ibm.com>
1532         * config/rs6000/rs6000.c (reg_to_non_prefixed): Add forward
1533         reference.
1534         (hard_reg_and_mode_to_addr_mask): Delete.
1535         (rs6000_adjust_vec_address): If the original vector address
1536         was REG+REG or REG+OFFSET and the element is not zero, do the add
1537         of the elements in the original address before adding the offset
1538         for the vector element.  Use address_to_insn_form to validate the
1539         address using the register being loaded, rather than guessing
1540         whether the address is a DS-FORM or DQ-FORM address.
1542 2020-02-03  Michael Meissner  <meissner@linux.ibm.com>
1544         * config/rs6000/rs6000.c (get_vector_offset): New helper function
1545         to calculate the offset in memory from the start of a vector of a
1546         particular element.  Add code to keep the element number in
1547         bounds if the element number is variable.
1548         (rs6000_adjust_vec_address): Move calculation of offset of the
1549         vector element to get_vector_offset.
1550         (rs6000_split_vec_extract_var): Do not do the initial AND of
1551         element here, move the code to get_vector_offset.
1553 2020-02-03  Michael Meissner  <meissner@linux.ibm.com>
1555         * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add some
1556         gcc_asserts.
1558 2020-02-03  Segher Boessenkool  <segher@kernel.crashing.org>
1560         * config/rs6000/constraints.md: Improve documentation.
1562 2020-02-03  Richard Earnshaw  <rearnsha@arm.com>
1564         PR target/93548
1565         * config/arm/t-arm: ($(srcdir)/config/arm/arm-tune.md)
1566         ($(srcdir)/config/arm/arm-tables.opt): Use move-if-change.
1568 2020-02-03  Andrew Stubbs  <ams@codesourcery.com>
1570         * config.gcc: Remove "carrizo" support.
1571         * config/gcn/gcn-opts.h (processor_type): Likewise.
1572         * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Likewise.
1573         * config/gcn/gcn.opt (gpu_type): Likewise.
1574         * config/gcn/t-omp-device: Likewise.
1576 2020-02-03  Stam Markianos-Wright  <stam.markianos-wright@arm.com>
1578         PR target/91816
1579         * config/arm/arm-protos.h: New function arm_gen_far_branch prototype.
1580         * config/arm/arm.c (arm_gen_far_branch): New function
1581         arm_gen_far_branch.
1582         * config/arm/arm.md: Update b<cond> for Thumb2 range checks.
1584 2020-02-03  Julian Brown  <julian@codesourcery.com>
1585             Tobias Burnus  <tobias@codesourcery.com>
1587         * doc/invoke.texi: Update mention of OpenACC version to 2.6.
1589 2020-02-03  Jakub Jelinek  <jakub@redhat.com>
1591         PR target/93533
1592         * config/s390/s390.md (popcounthi2_z196): Fix up expander to emit
1593         valid RTL to sum up the lowest and second lowest bytes of the popcnt
1594         result.
1596 2020-02-02  Vladimir Makarov  <vmakarov@redhat.com>
1598         PR rtl-optimization/91333
1599         * ira-color.c (struct allocno_color_data): Add member
1600         hard_reg_prefs.
1601         (init_allocno_threads): Set the member up.
1602         (bucket_allocno_compare_func): Add compare hard reg
1603         prefs.
1605 2020-01-31  Sandra Loosemore  <sandra@codesourcery.com>
1607         nios2: Support for GOT-relative DW_EH_PE_datarel encoding.
1609         * configure.ac [nios2-*-*]: Check HAVE_AS_NIOS2_GOTOFF_RELOCATION.
1610         * config.in: Regenerated.
1611         * configure: Regenerated.
1612         * config/nios2/nios2.h (ASM_PREFERRED_EH_DATA_FORMAT): Fix handling
1613         for PIC when HAVE_AS_NIOS2_GOTOFF_RELOCATION.
1614         (ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX): New.
1616 2020-02-01  Andrew Burgess  <andrew.burgess@embecosm.com>
1618         * configure: Regenerate.
1620 2020-01-31  Vladimir Makarov  <vmakarov@redhat.com>
1622         PR rtl-optimization/91333
1623         * ira-color.c (bucket_allocno_compare_func): Move conflict hard
1624         reg preferences comparison up.
1626 2020-01-31  Richard Sandiford  <richard.sandiford@arm.com>
1628         * config/aarch64/aarch64.h (TARGET_SVE_BF16): New macro.
1629         * config/aarch64/aarch64-sve-builtins-sve2.h (svcvtnt): Move to
1630         aarch64-sve-builtins-base.h.
1631         * config/aarch64/aarch64-sve-builtins-sve2.cc (svcvtnt): Move to
1632         aarch64-sve-builtins-base.cc.
1633         * config/aarch64/aarch64-sve-builtins-base.h (svbfdot, svbfdot_lane)
1634         (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1635         (svcvtnt): Declare.
1636         * config/aarch64/aarch64-sve-builtins-base.cc (svbfdot, svbfdot_lane)
1637         (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1638         (svcvtnt): New functions.
1639         * config/aarch64/aarch64-sve-builtins-base.def (svbfdot, svbfdot_lane)
1640         (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1641         (svcvtnt): New functions.
1642         (svcvt): Add a form that converts f32 to bf16.
1643         * config/aarch64/aarch64-sve-builtins-shapes.h (ternary_bfloat)
1644         (ternary_bfloat_lane, ternary_bfloat_lanex2, ternary_bfloat_opt_n):
1645         Declare.
1646         * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_element_type):
1647         Treat B as bfloat16_t.
1648         (ternary_bfloat_lane_base): New class.
1649         (ternary_bfloat_def): Likewise.
1650         (ternary_bfloat): New shape.
1651         (ternary_bfloat_lane_def): New class.
1652         (ternary_bfloat_lane): New shape.
1653         (ternary_bfloat_lanex2_def): New class.
1654         (ternary_bfloat_lanex2): New shape.
1655         (ternary_bfloat_opt_n_def): New class.
1656         (ternary_bfloat_opt_n): New shape.
1657         * config/aarch64/aarch64-sve-builtins.cc (TYPES_cvt_bfloat): New macro.
1658         * config/aarch64/aarch64-sve.md (@aarch64_sve_<sve_fp_op>vnx4sf)
1659         (@aarch64_sve_<sve_fp_op>_lanevnx4sf): New patterns.
1660         (@aarch64_sve_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>)
1661         (@cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
1662         (*cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
1663         (@aarch64_sve_cvtnt<VNx8BF_ONLY:mode>): Likewise.
1664         * config/aarch64/aarch64-sve2.md (@aarch64_sve2_cvtnt<mode>): Key
1665         the pattern off the narrow mode instead of the wider one.
1666         * config/aarch64/iterators.md (VNx8BF_ONLY): New mode iterator.
1667         (UNSPEC_BFMLALB, UNSPEC_BFMLALT, UNSPEC_BFMMLA): New unspecs.
1668         (sve_fp_op): Handle them.
1669         (SVE_BFLOAT_TERNARY_LONG): New int itertor.
1670         (SVE_BFLOAT_TERNARY_LONG_LANE): Likewise.
1672 2020-01-31  Richard Sandiford  <richard.sandiford@arm.com>
1674         * config/aarch64/arm_sve.h: Include arm_bf16.h.
1675         * config/aarch64/aarch64-modes.def (BF): Move definition before
1676         VECTOR_MODES.  Remove separate VECTOR_MODES for V4BF and V8BF.
1677         (SVE_MODES): Handle BF modes.
1678         * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
1679         BF modes.
1680         (aarch64_full_sve_mode): Likewise.
1681         * config/aarch64/iterators.md (SVE_STRUCT): Add VNx16BF, VNx24BF
1682         and VNx32BF.
1683         (SVE_FULL, SVE_FULL_HSD, SVE_ALL): Add VNx8BF.
1684         (Vetype, Vesize, Vctype, VEL, Vel, VEL_INT, V128, v128, vwcore)
1685         (V_INT_EQUIV, v_int_equiv, V_FP_EQUIV, v_fp_equiv, vector_count)
1686         (insn_length, VSINGLE, vsingle, VPRED, vpred, VDOUBLE): Handle the
1687         new SVE BF modes.
1688         * config/aarch64/aarch64-sve-builtins.h (TYPE_bfloat): New
1689         type_class_index.
1690         * config/aarch64/aarch64-sve-builtins.cc (TYPES_all_arith): New macro.
1691         (TYPES_all_data): Add bf16.
1692         (TYPES_reinterpret1, TYPES_reinterpret): Likewise.
1693         (register_tuple_type): Increase buffer size.
1694         * config/aarch64/aarch64-sve-builtins.def (svbfloat16_t): New type.
1695         (bf16): New type suffix.
1696         * config/aarch64/aarch64-sve-builtins-base.def (svabd, svadd, svaddv)
1697         (svcmpeq, svcmpge, svcmpgt, svcmple, svcmplt, svcmpne, svmad, svmax)
1698         (svmaxv, svmin, svminv, svmla, svmls, svmsb, svmul, svsub, svsubr):
1699         Change type from all_data to all_arith.
1700         * config/aarch64/aarch64-sve-builtins-sve2.def (svaddp, svmaxp)
1701         (svminp): Likewise.
1703 2020-01-31  Dennis Zhang  <dennis.zhang@arm.com>
1704             Matthew Malcomson  <matthew.malcomson@arm.com>
1705             Richard Sandiford  <richard.sandiford@arm.com>
1707         * doc/invoke.texi (f32mm): Document new AArch64 -march= extension.
1708         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Define
1709         __ARM_FEATURE_SVE_MATMUL_INT8, __ARM_FEATURE_SVE_MATMUL_FP32 and
1710         __ARM_FEATURE_SVE_MATMUL_FP64 as appropriate.  Don't define
1711         __ARM_FEATURE_MATMUL_FP64.
1712         * config/aarch64/aarch64-option-extensions.def (fp, simd, fp16)
1713         (sve): Add AARCH64_FL_F32MM to the list of extensions that should
1714         be disabled at the same time.
1715         (f32mm): New extension.
1716         * config/aarch64/aarch64.h (AARCH64_FL_F32MM): New macro.
1717         (AARCH64_FL_F64MM): Bump to the next bit up.
1718         (AARCH64_ISA_F32MM, TARGET_SVE_I8MM, TARGET_F32MM, TARGET_SVE_F32MM)
1719         (TARGET_SVE_F64MM): New macros.
1720         * config/aarch64/iterators.md (SVE_MATMULF): New mode iterator.
1721         (UNSPEC_FMMLA, UNSPEC_SMATMUL, UNSPEC_UMATMUL, UNSPEC_USMATMUL)
1722         (UNSPEC_TRN1Q, UNSPEC_TRN2Q, UNSPEC_UZP1Q, UNSPEC_UZP2Q, UNSPEC_ZIP1Q)
1723         (UNSPEC_ZIP2Q): New unspeccs.
1724         (DOTPROD_US_ONLY, PERMUTEQ, MATMUL, FMMLA): New int iterators.
1725         (optab, sur, perm_insn): Handle the new unspecs.
1726         (sve_fp_op): Handle UNSPEC_FMMLA.  Resort.
1727         * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use
1728         TARGET_SVE_F64MM instead of separate tests.
1729         (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod<vsi2qi>): New pattern.
1730         (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod_lane<vsi2qi>): Likewise.
1731         (@aarch64_sve_add_<MATMUL:optab><vsi2qi>): Likewise.
1732         (@aarch64_sve_<FMMLA:sve_fp_op><mode>): Likewise.
1733         (@aarch64_sve_<PERMUTEQ:optab><mode>): Likewise.
1734         * config/aarch64/aarch64-sve-builtins.cc (TYPES_s_float): New macro.
1735         (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): Use it.
1736         (TYPES_s_signed): New macro.
1737         (TYPES_s_integer): Use it.
1738         (TYPES_d_float): New macro.
1739         (TYPES_d_data): Use it.
1740         * config/aarch64/aarch64-sve-builtins-shapes.h (mmla): Declare.
1741         (ternary_intq_uintq_lane, ternary_intq_uintq_opt_n, ternary_uintq_intq)
1742         (ternary_uintq_intq_lane, ternary_uintq_intq_opt_n): Likewise.
1743         * config/aarch64/aarch64-sve-builtins-shapes.cc (mmla_def): New class.
1744         (svmmla): New shape.
1745         (ternary_resize2_opt_n_base): Add TYPE_CLASS2 and TYPE_CLASS3
1746         template parameters.
1747         (ternary_resize2_lane_base): Likewise.
1748         (ternary_resize2_base): New class.
1749         (ternary_qq_lane_base): Likewise.
1750         (ternary_intq_uintq_lane_def): Likewise.
1751         (ternary_intq_uintq_lane): New shape.
1752         (ternary_intq_uintq_opt_n_def): New class
1753         (ternary_intq_uintq_opt_n): New shape.
1754         (ternary_qq_lane_def): Inherit from ternary_qq_lane_base.
1755         (ternary_uintq_intq_def): New class.
1756         (ternary_uintq_intq): New shape.
1757         (ternary_uintq_intq_lane_def): New class.
1758         (ternary_uintq_intq_lane): New shape.
1759         (ternary_uintq_intq_opt_n_def): New class.
1760         (ternary_uintq_intq_opt_n): New shape.
1761         * config/aarch64/aarch64-sve-builtins-base.h (svmmla, svsudot)
1762         (svsudot_lane, svtrn1q, svtrn2q, svusdot, svusdot_lane, svusmmla)
1763         (svuzp1q, svuzp2q, svzip1q, svzip2q): Declare.
1764         * config/aarch64/aarch64-sve-builtins-base.cc (svdot_lane_impl):
1765         Generalize to...
1766         (svdotprod_lane_impl): ...this new class.
1767         (svmmla_impl, svusdot_impl): New classes.
1768         (svdot_lane): Update to use svdotprod_lane_impl.
1769         (svmmla, svsudot, svsudot_lane, svtrn1q, svtrn2q, svusdot)
1770         (svusdot_lane, svusmmla, svuzp1q, svuzp2q, svzip1q, svzip2q): New
1771         functions.
1772         * config/aarch64/aarch64-sve-builtins-base.def (svmmla): New base
1773         function, with no types defined.
1774         (svmmla, svusmmla, svsudot, svsudot_lane, svusdot, svusdot_lane): New
1775         AARCH64_FL_I8MM functions.
1776         (svmmla): New AARCH64_FL_F32MM function.
1777         (svld1ro): Depend only on AARCH64_FL_F64MM, not on AARCH64_FL_V8_6.
1778         (svmmla, svtrn1q, svtrn2q, svuz1q, svuz2q, svzip1q, svzip2q): New
1779         AARCH64_FL_F64MM function.
1780         (REQUIRED_EXTENSIONS):
1782 2020-01-31  Andrew Stubbs  <ams@codesourcery.com>
1784         * config/gcn/gcn-valu.md (addv64di3_exec): Allow one '0' in each
1785         alternative only.
1787 2020-01-31  UroÅ¡ Bizjak  <ubizjak@gmail.com>
1789         * config/i386/i386.md (*movoi_internal_avx): Do not check for
1790         TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.  Remove MODE_V8SF handling.
1791         (*movti_internal): Do not check for
1792         TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
1793         (*movtf_internal): Move check for TARGET_SSE2 and size optimization
1794         just after check for TARGET_AVX.
1795         (*movdf_internal): Ditto.
1796         * config/i386/mmx.md (*mov<mode>_internal): Do not check for
1797         TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
1798         * config/i386/sse.md (mov<mode>_internal): Only check
1799         TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL with V2DFmode.  Move check
1800         for TARGET_SSE2 and size optimization just after check for TARGET_AVX.
1801         (<sse>_andnot<mode>3<mask_name>): Move check for
1802         TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL after check for TARGET_AVX.
1803         (<code><mode>3<mask_name>): Ditto.
1804         (*andnot<mode>3): Ditto.
1805         (*andnottf3): Ditto.
1806         (*<code><mode>3): Ditto.
1807         (*<code>tf3): Ditto.
1808         (*andnot<VI:mode>3): Remove
1809         TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL handling.
1810         (<mask_codefor><code><VI48_AVX_AVX512F:mode>3<mask_name>): Ditto.
1811         (*<code><VI12_AVX_AVX512F:mode>3): Ditto.
1812         (sse4_1_blendv<ssemodesuffix>): Ditto.
1813         * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL):
1814         Explain that tune applies to 128bit instructions only.
1816 2020-01-31  Kwok Cheung Yeung  <kcy@codesourcery.com>
1818         * config/gcn/mkoffload.c (process_asm): Add sgpr_count and vgpr_count
1819         to definition of hsa_kernel_description.  Parse assembly to find SGPR
1820         and VGPR count of kernel and store in hsa_kernel_description.
1822 2020-01-31  Tamar Christina  <tamar.christina@arm.com>
1824         PR rtl-optimization/91838
1825         * simplify-rtx.c (simplify_binary_operation_1): Update LSHIFTRT case
1826         to truncate if allowed or reject combination.
1828 2020-01-31  Andrew Stubbs  <ams@codesourcery.com>
1830         * tree-ssa-loop-ivopts.c (get_iv): Use sizetype for zero-step.
1831         (find_inv_vars_cb): Likewise.
1833 2020-01-31  David Malcolm  <dmalcolm@redhat.com>
1835         * calls.c (special_function_p): Split out the check for DECL_NAME
1836         being non-NULL and fndecl being extern at file scope into a
1837         new maybe_special_function_p and call it.  Drop check for fndecl
1838         being non-NULL that was after a usage of DECL_NAME (fndecl).
1839         * tree.h (maybe_special_function_p): New inline function.
1841 2020-01-30  Andrew Stubbs  <ams@codesourcery.com>
1843         * config/gcn/gcn-valu.md (gather<mode>_exec): Move contents ...
1844         (mask_gather_load<mode>): ... here, and zero-initialize the
1845         destination.
1846         (maskload<mode>di): Zero-initialize the destination.
1847         * config/gcn/gcn.c:
1849 2020-01-30  David Malcolm  <dmalcolm@redhat.com>
1851         PR analyzer/93356
1852         * doc/analyzer.texi (Limitations): Note that constraints on
1853         floating-point values are currently ignored.
1855 2020-01-30  Jakub Jelinek  <jakub@redhat.com>
1857         PR lto/93384
1858         * symtab.c (symtab_node::noninterposable_alias): If localalias
1859         already exists, but is not usable, append numbers after it until
1860         a unique name is found.  Formatting fix.
1862         PR middle-end/93505
1863         * combine.c (simplify_comparison) <case ROTATE>: Punt on out of range
1864         rotate counts.
1866 2020-01-30  Andrew Stubbs  <ams@codesourcery.com>
1868         * config/gcn/gcn.c (print_operand): Handle LTGT.
1869         * config/gcn/predicates.md (gcn_fp_compare_operator): Allow ltgt.
1871 2020-01-30  Richard Biener  <rguenther@suse.de>
1873         * tree-pretty-print.c (dump_generic_node): Wrap VECTOR_CST
1874         and CONSTRUCTOR in _Literal (type) with TDF_GIMPLE.
1876 2020-01-30  John David Anglin  <danglin@gcc.gnu.org>
1878         * config/pa/pa.c (pa_elf_select_rtx_section): Place function pointers
1879         without a DECL in .data.rel.ro.local.
1881 2020-01-30  Jakub Jelinek  <jakub@redhat.com>
1883         PR target/93494
1884         * config/arm/arm.md (uaddvdi4): Actually emit what gen_uaddvsi4
1885         returned.
1887         PR target/91824
1888         * config/i386/sse.md
1889         (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext): Renamed to ...
1890         (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext): ... this.  Use
1891         any_extend code iterator instead of always zero_extend.
1892         (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_lt): Renamed to ...
1893         (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_lt): ... this.
1894         Use any_extend code iterator instead of always zero_extend.
1895         (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_shift): Renamed to ...
1896         (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_shift): ... this.
1897         Use any_extend code iterator instead of always zero_extend.
1898         (*sse2_pmovmskb_ext): New define_insn.
1899         (*sse2_pmovmskb_ext_lt): New define_insn_and_split.
1901         PR target/91824
1902         * config/i386/i386.md (*popcountsi2_zext): New define_insn_and_split.
1903         (*popcountsi2_zext_falsedep): New define_insn.
1905 2020-01-30  Dragan Mladjenovic  <dmladjenovic@wavecomp.com>
1907         * config.in: Regenerated.
1908         * configure: Regenerated.
1910 2020-01-29  Tobias Burnus  <tobias@codesourcery.com>
1912         PR bootstrap/93409
1913         * config/gcn/gcn-hsa.h (ASM_SPEC): Add -mattr=-code-object-v3 as
1914         LLVM's assembler changed the default in version 9.
1916 2020-01-24  Jeff Law  <law@redhat.com>
1918         PR tree-optimization/89689
1919         * builtins.def (BUILT_IN_OBJECT_SIZE): Make it const rather than pure.
1921 2020-01-29  Richard Sandiford  <richard.sandiford@arm.com>
1923         Revert:
1925         2020-01-28  Richard Sandiford  <richard.sandiford@arm.com>
1927         PR rtl-optimization/87763
1928         * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
1929         simplification to handle subregs as well as bare regs.
1930         * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
1932 2020-01-29  Joel Hutton  <Joel.Hutton@arm.com>
1934         PR target/93221
1935         * ira.c (ira): Revert use of simplified LRA algorithm.
1937 2020-01-29  Martin Jambor  <mjambor@suse.cz>
1939         PR tree-optimization/92706
1940         * tree-sra.c (struct access): Fields first_link, last_link,
1941         next_queued and grp_queued renamed to first_rhs_link, last_rhs_link,
1942         next_rhs_queued and grp_rhs_queued respectively, new fields
1943         first_lhs_link, last_lhs_link, next_lhs_queued and grp_lhs_queued.
1944         (struct assign_link): Field next renamed to next_rhs, new field
1945         next_lhs.  Updated comment.
1946         (work_queue_head): Renamed to rhs_work_queue_head.
1947         (lhs_work_queue_head): New variable.
1948         (add_link_to_lhs): New function.
1949         (relink_to_new_repr): Also relink LHS lists.
1950         (add_access_to_work_queue): Renamed to add_access_to_rhs_work_queue.
1951         (add_access_to_lhs_work_queue): New function.
1952         (pop_access_from_work_queue): Renamed to
1953         pop_access_from_rhs_work_queue.
1954         (pop_access_from_lhs_work_queue): New function.
1955         (build_accesses_from_assign): Also add links to LHS lists and to LHS
1956         work_queue.
1957         (child_would_conflict_in_lacc): Renamed to
1958         child_would_conflict_in_acc.  Adjusted parameter names.
1959         (create_artificial_child_access): New parameter set_grp_read, use it.
1960         (subtree_mark_written_and_enqueue): Renamed to
1961         subtree_mark_written_and_rhs_enqueue.
1962         (propagate_subaccesses_across_link): Renamed to
1963         propagate_subaccesses_from_rhs.
1964         (propagate_subaccesses_from_lhs): New function.
1965         (propagate_all_subaccesses): Also propagate subaccesses from LHSs to
1966         RHSs.
1968 2020-01-29  Martin Jambor  <mjambor@suse.cz>
1970         PR tree-optimization/92706
1971         * tree-sra.c (struct access): Adjust comment of
1972         grp_total_scalarization.
1973         (find_access_in_subtree): Look for single children spanning an entire
1974         access.
1975         (scalarizable_type_p): Allow register accesses, adjust callers.
1976         (completely_scalarize): Remove function.
1977         (scalarize_elem): Likewise.
1978         (create_total_scalarization_access): Likewise.
1979         (sort_and_splice_var_accesses): Do not track total scalarization
1980         flags.
1981         (analyze_access_subtree): New parameter totally, adjust to new meaning
1982         of grp_total_scalarization.
1983         (analyze_access_trees): Pass new parameter to analyze_access_subtree.
1984         (can_totally_scalarize_forest_p): New function.
1985         (create_total_scalarization_access): Likewise.
1986         (create_total_access_and_reshape): Likewise.
1987         (total_should_skip_creating_access): Likewise.
1988         (totally_scalarize_subtree): Likewise.
1989         (analyze_all_variable_accesses): Perform total scalarization after
1990         subaccess propagation using the new functions above.
1991         (initialize_constant_pool_replacements): Output initializers by
1992         traversing the access tree.
1994 2020-01-29  Martin Jambor  <mjambor@suse.cz>
1996         * tree-sra.c (verify_sra_access_forest): New function.
1997         (verify_all_sra_access_forests): Likewise.
1998         (create_artificial_child_access): Set parent.
1999         (analyze_all_variable_accesses): Call the verifier.
2001 2020-01-28  Jan Hubicka  <hubicka@ucw.cz>
2003         * cgraph.c (cgraph_edge::resolve_speculation): Only lookup direct edge
2004         if called on indirect edge.
2005         (cgraph_edge::redirect_call_stmt_to_callee): Lookup indirect edge of
2006         speculative call if needed.
2008 2020-01-29  Richard Biener  <rguenther@suse.de>
2010         PR tree-optimization/93428
2011         * tree-vect-slp.c (vect_build_slp_tree_2): Compute the load
2012         permutation when the load node is created.
2013         (vect_analyze_slp_instance): Re-use it here.
2015 2020-01-28  Jan Hubicka  <hubicka@ucw.cz>
2017         * ipa-prop.c (update_indirect_edges_after_inlining): Fix warning.
2019 2020-01-28  Vladimir Makarov  <vmakarov@redhat.com>
2021         PR rtl-optimization/93272
2022         * ira-lives.c (process_out_of_region_eh_regs): New function.
2023         (process_bb_node_lives): Call it.
2025 2020-01-28  Jan Hubicka  <hubicka@ucw.cz>
2027         * coverage.c (read_counts_file): Make error message lowercase.
2029 2020-01-28  Jan Hubicka  <hubicka@ucw.cz>
2031         * profile-count.c (profile_quality_display_names): Fix ordering.
2033 2020-01-28  Jan Hubicka  <hubicka@ucw.cz>
2035         PR lto/93318    
2036         * cgraph.c (cgraph_add_edge_to_call_site_hash): Update call site
2037         hash only when edge is first within the sequence.
2038         (cgraph_edge::set_call_stmt): Update handling of speculative calls.
2039         (symbol_table::create_edge): Do not set target_prob.
2040         (cgraph_edge::remove_caller): Watch for speculative calls when updating
2041         the call site hash.
2042         (cgraph_edge::make_speculative): Drop target_prob parameter.
2043         (cgraph_edge::speculative_call_info): Remove.
2044         (cgraph_edge::first_speculative_call_target): New member function.
2045         (update_call_stmt_hash_for_removing_direct_edge): New function.
2046         (cgraph_edge::resolve_speculation): Rewrite to new API.
2047         (cgraph_edge::speculative_call_for_target): New member function.
2048         (cgraph_edge::make_direct): Rewrite to new API; fix handling of
2049         multiple speculation targets.
2050         (cgraph_edge::redirect_call_stmt_to_callee): Likewise; fix updating
2051         of profile.
2052         (verify_speculative_call): Verify that targets form an interval.
2053         * cgraph.h (cgraph_edge::speculative_call_info): Remove.
2054         (cgraph_edge::first_speculative_call_target): New member function.
2055         (cgraph_edge::next_speculative_call_target): New member function.
2056         (cgraph_edge::speculative_call_target_ref): New member function.
2057         (cgraph_edge;:speculative_call_indirect_edge): New member funtion.
2058         (cgraph_edge): Remove target_prob.
2059         * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
2060         Fix handling of speculative calls.
2061         * ipa-devirt.c (ipa_devirt): Fix handling of speculative cals.
2062         * ipa-fnsummary.c (analyze_function_body): Likewise.
2063         * ipa-inline.c (speculation_useful_p): Use new speculative call API.
2064         * ipa-profile.c (dump_histogram): Fix formating.
2065         (ipa_profile_generate_summary): Watch for overflows.
2066         (ipa_profile): Do not require probablity to be 1/2; update to new API.
2067         * ipa-prop.c (ipa_make_edge_direct_to_target): Update to new API.
2068         (update_indirect_edges_after_inlining): Update to new API.
2069         * ipa-utils.c (ipa_merge_profiles): Rewrite merging of speculative call
2070         profiles.
2071         * profile-count.h: (profile_probability::adjusted): New.
2072         * tree-inline.c (copy_bb): Update to new speculative call API; fix
2073         updating of profile.
2074         * value-prof.c (gimple_ic_transform): Rename to ...
2075         (dump_ic_profile): ... this one; update dumping.
2076         (stream_in_histogram_value): Fix formating.
2077         (gimple_value_profile_transformations): Update.
2079 2020-01-28  H.J. Lu  <hongjiu.lu@intel.com>
2081         PR target/91461
2082         * config/i386/i386.md (*movoi_internal_avx): Remove
2083         TARGET_SSE_TYPELESS_STORES check.
2084         (*movti_internal): Prefer TARGET_AVX over
2085         TARGET_SSE_TYPELESS_STORES.
2086         (*movtf_internal): Likewise.
2087         * config/i386/sse.md (mov<mode>_internal): Prefer TARGET_AVX over
2088         TARGET_SSE_TYPELESS_STORES.  Remove "<MODE_SIZE> == 16" check
2089         from TARGET_SSE_TYPELESS_STORES.
2091 2020-01-28  David Malcolm  <dmalcolm@redhat.com>
2093         * diagnostic-core.h (warning_at): Rename overload to...
2094         (warning_meta): ...this.
2095         (emit_diagnostic_valist): Delete decl of overload taking
2096         diagnostic_metadata.
2097         * diagnostic.c (emit_diagnostic_valist): Likewise for defn.
2098         (warning_at): Rename overload taking diagnostic_metadata to...
2099         (warning_meta): ...this.
2101 2020-01-28  Richard Biener  <rguenther@suse.de>
2103         PR tree-optimization/93439
2104         * tree-parloops.c (create_loop_fn): Move clique bookkeeping...
2105         * tree-cfg.c (move_sese_region_to_fn): ... here.
2106         (verify_types_in_gimple_reference): Verify used cliques are
2107         tracked.
2109 2020-01-28  H.J. Lu  <hongjiu.lu@intel.com>
2111         PR target/91399
2112         * config/i386/i386-options.c (set_ix86_tune_features): Add an
2113         argument of a pointer to struct gcc_options and pass it to
2114         parse_mtune_ctrl_str.
2115         (ix86_function_specific_restore): Pass opts to
2116         set_ix86_tune_features.
2117         (ix86_option_override_internal): Likewise.
2118         (parse_mtune_ctrl_str): Add an argument of a pointer to struct
2119         gcc_options and use it for x_ix86_tune_ctrl_string.
2121 2020-01-28  Richard Sandiford  <richard.sandiford@arm.com>
2123         PR rtl-optimization/87763
2124         * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
2125         simplification to handle subregs as well as bare regs.
2126         * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
2128 2020-01-28  Richard Sandiford  <richard.sandiford@arm.com>
2130         * tree-vect-loop.c (vectorizable_reduction): Fail gracefully
2131         for reduction chains that (now) include a call.
2133 2020-01-28  Richard Sandiford  <richard.sandiford@arm.com>
2135         PR tree-optimization/92822
2136         * tree-ssa-forwprop.c (simplify_vector_constructor): When filling
2137         out the don't-care elements of a vector whose significant elements
2138         are duplicates, make the don't-care elements duplicates too.
2140 2020-01-28  Richard Sandiford  <richard.sandiford@arm.com>
2142         PR tree-optimization/93434
2143         * tree-predcom.c (split_data_refs_to_components): Record which
2144         components have had aliasing loads removed.  Prevent store-store
2145         commoning for all such components.
2147 2020-01-28  Jakub Jelinek  <jakub@redhat.com>
2149         PR target/93418
2150         * config/i386/i386.c (ix86_fold_builtin) <do_shift>: If mask is not
2151         -1 or is_vshift is true, use new_vector with number of elts npatterns
2152         rather than new_unary_operation.
2154         PR tree-optimization/93454
2155         * gimple-fold.c (fold_array_ctor_reference): Perform
2156         elt_size.to_uhwi () just once, instead of calling it in every
2157         iteration.  Punt if that value is above size of the temporary
2158         buffer.  Decrease third native_encode_expr argument when
2159         bufoff + elt_sz is above size of buf.
2161 2020-01-27  Joseph Myers  <joseph@codesourcery.com>
2163         * config/mips/mips.c (mips_declare_object_name)
2164         [USE_GNU_UNIQUE_OBJECT]: Support use of gnu_unique_object.
2166 2020-01-27  Martin Liska  <mliska@suse.cz>
2168         PR gcov-profile/93403
2169         * tree-profile.c (gimple_init_gcov_profiler): Generate
2170         both __gcov_indirect_call_profiler_v4 and
2171         __gcov_indirect_call_profiler_v4_atomic.
2173 2020-01-27  Richard Sandiford  <richard.sandiford@arm.com>
2175         PR target/92822
2176         * config/aarch64/aarch64-simd.md (aarch64_get_half<mode>): New
2177         expander.
2178         (@aarch64_split_simd_mov<mode>): Use it.
2179         (aarch64_simd_mov_from_<mode>low): Add a GPR alternative.
2180         Leave the vec_extract patterns to handle 2-element vectors.
2181         (aarch64_simd_mov_from_<mode>high): Likewise.
2182         (vec_extract<VQMOV_NO2E:mode><Vhalf>): New expander.
2183         (vec_extractv2dfv1df): Likewise.
2185 2020-01-27  Richard Sandiford  <richard.sandiford@arm.com>
2187         * config/aarch64/aarch64.c (aarch64_if_then_else_costs): Match
2188         jump conditions for *compare_condjump<GPI:mode>.
2190 2020-01-27  David Malcolm  <dmalcolm@redhat.com>
2192         PR analyzer/93276
2193         * digraph.cc (test_edge::test_edge): Specify template for base
2194         class initializer.
2196 2020-01-27  Claudiu Zissulescu  <claziss@synopsys.com>
2198         * config/arc/arc.c (arc_rtx_costs): Update mul64 cost.
2200 2020-01-27  Claudiu Zissulescu  <claziss@synopsys.com>
2202         * config/arc/arc-protos.h (gen_mlo): Remove.
2203         (gen_mhi): Likewise.
2204         * config/arc/arc.c (AUX_MULHI): Define.
2205         (arc_must_save_reister): Special handling for r58/59.
2206         (arc_compute_frame_size): Consider mlo/mhi registers.
2207         (arc_save_callee_saves): Emit fp/sp move only when emit_move
2208         paramter is true.
2209         (arc_conditional_register_usage): Remove TARGET_BIG_ENDIAN from
2210         mlo/mhi name selection.
2211         (arc_restore_callee_saves): Don't early restore blink when ISR.
2212         (arc_expand_prologue): Add mlo/mhi saving.
2213         (arc_expand_epilogue): Add mlo/mhi restoring.
2214         (gen_mlo): Remove.
2215         (gen_mhi): Remove.
2216         * config/arc/arc.h (DBX_REGISTER_NUMBER): Correct register
2217         numbering when MUL64 option is used.
2218         (DWARF2_FRAME_REG_OUT): Define.
2219         * config/arc/arc.md (arc600_stall): New pattern.
2220         (VUNSPEC_ARC_ARC600_STALL): Define.
2221         (mulsi64): Use correct mlo/mhi registers.
2222         (mulsi_600): Clean it up.
2223         * config/arc/predicates.md (mlo_operand): Remove any dependency on
2224         TARGET_BIG_ENDIAN.
2225         (mhi_operand): Likewise.
2227 2020-01-27  Claudiu Zissulescu  <claziss@synopsys.com>
2228             Petro Karashchenko  <petro.karashchenko@ring.com>
2230         * config/arc/arc.c (arc_is_uncached_mem_p): Check struct
2231         attributes if needed.
2232         (prepare_move_operands): Generate special unspec instruction for
2233         direct access.
2234         (arc_isuncached_mem_p): Propagate uncached attribute to each
2235         structure member.
2236         * config/arc/arc.md (VUNSPEC_ARC_LDDI): Define.
2237         (VUNSPEC_ARC_STDI): Likewise.
2238         (ALLI): New mode iterator.
2239         (mALLI): New mode attribute.
2240         (lddi): New instruction pattern.
2241         (stdi): Likewise.
2242         (stdidi_split): Split instruction for architectures which are not
2243         supporting ll64 option.
2244         (lddidi_split): Likewise.
2246 2020-01-27  Richard Sandiford  <richard.sandiford@arm.com>
2248         PR rtl-optimization/92989
2249         * lra-lives.c (process_bb_lives): Update the live-in set before
2250         processing additional clobbers.
2252 2020-01-27  Richard Sandiford  <richard.sandiford@arm.com>
2254         PR rtl-optimization/93170
2255         * cselib.c (cselib_invalidate_regno_val): New function, split out
2256         from...
2257         (cselib_invalidate_regno): ...here.
2258         (cselib_invalidated_by_call_p): New function.
2259         (cselib_process_insn): Iterate over all the hard-register entries in
2260         REG_VALUES and invalidate any that cross call-clobbered registers.
2262 2020-01-27  Richard Sandiford  <richard.sandiford@arm.com>
2264         * dojump.c (split_comparison): Use HONOR_NANS rather than
2265         HONOR_SNANS when splitting LTGT.
2267 2020-01-27  Martin Liska  <mliska@suse.cz>
2269         PR driver/91220
2270         * opts.c (print_filtered_help): Exclude language-specific
2271         options from --help=common unless enabled in all FEs.
2273 2020-01-27  Martin Liska  <mliska@suse.cz>
2275         * opts.c (print_help): Exclude params from
2276         all except --help=param.
2278 2020-01-27  Martin Liska  <mliska@suse.cz>
2280         PR target/93274
2281         * config/i386/i386-features.c (make_resolver_func):
2282         Align the code with ppc64 target implementation.
2283         Do not generate a unique name for resolver function.
2285 2020-01-27  Richard Biener  <rguenther@suse.de>
2287         PR tree-optimization/93397
2288         * tree-vect-slp.c (vect_analyze_slp_instance): Delay
2289         converted reduction chain SLP graph adjustment.
2291 2020-01-26  Marek Polacek  <polacek@redhat.com>
2293         PR sanitizer/93436
2294         * sanopt.c (sanitize_rewrite_addressable_params): Avoid crash on
2295         null DECL_NAME.
2297 2020-01-26  Jason Merrill  <jason@redhat.com>
2299         PR c++/92601
2300         * tree.c (verify_type_variant): Only verify TYPE_NEEDS_CONSTRUCTING
2301         of complete types.
2303 2020-01-26  Darius Galis  <darius.galis@cyberthorstudios.com>
2305         * config/rx/rx.md (setmemsi): Added rx_allow_string_insns constraint
2306         (rx_setmem): Likewise.
2308 2020-01-26  Jakub Jelinek  <jakub@redhat.com>
2310         PR target/93412
2311         * config/i386/i386.md (*addv<dwi>4_doubleword, *subv<dwi>4_doubleword):
2312         Use nonimmediate_operand instead of x86_64_hilo_general_operand and
2313         drop <di> from constraint of last operand.
2315         PR target/93430
2316         * config/i386/sse.md (*avx_vperm_broadcast_<mode>): Disallow for
2317         TARGET_AVX2 and V4DFmode not in the split condition, but in the
2318         pattern condition, though allow { 0, 0, 0, 0 } broadcast always.
2320 2020-01-25  Feng Xue  <fxue@os.amperecomputing.com>
2322         PR ipa/93166
2323         * ipa-cp.c (get_info_about_necessary_edges): Remove value
2324         check assertion.
2326 2020-01-24  Jeff Law  <law@redhat.com>
2328         PR tree-optimization/92788
2329         * tree-ssa-threadedge.c (thread_across_edge): Check EDGE_COMPLEX
2330         not EDGE_ABNORMAL.
2332 2020-01-24  Jakub Jelinek  <jakub@redhat.com>
2334         PR target/93395
2335         * config/i386/sse.md (*avx_vperm_broadcast_v4sf,
2336         *avx_vperm_broadcast_<mode>,
2337         <sse2_avx_avx512f>_vpermil<mode><mask_name>,
2338         *<sse2_avx_avx512f>_vpermilp<mode><mask_name>):
2339         Move before avx2_perm<mode>/avx512f_perm<mode>.
2341         PR target/93376
2342         * simplify-rtx.c (simplify_const_unary_operation,
2343         simplify_const_binary_operation): Punt for mode precision above
2344         MAX_BITSIZE_MODE_ANY_INT.
2346 2020-01-24  Andrew Pinski  <apinski@marvell.com>
2348         * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): Change
2349         alu.shift_reg to 0.
2351 2020-01-24  Jeff Law  <law@redhat.com>
2353         PR target/13721
2354         * config/h8300/h8300.c (h8300_print_operand): Only call byte_reg
2355         for REGs.  Call output_operand_lossage to get more reasonable
2356         diagnostics.
2358 2020-01-24  Andrew Stubbs  <ams@codesourcery.com>
2360         * config/gcn/gcn-valu.md (vec_cmp<mode>di): Use
2361         gcn_fp_compare_operator.
2362         (vec_cmpu<mode>di): Use gcn_compare_operator.
2363         (vec_cmp<u>v64qidi): Use gcn_compare_operator.
2364         (vec_cmp<mode>di_exec): Use gcn_fp_compare_operator.
2365         (vec_cmpu<mode>di_exec): Use gcn_compare_operator.
2366         (vec_cmp<u>v64qidi_exec): Use gcn_compare_operator.
2367         (vec_cmp<mode>di_dup): Use gcn_fp_compare_operator.
2368         (vec_cmp<mode>di_dup_exec): Use gcn_fp_compare_operator.
2369         (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): Use
2370         gcn_fp_compare_operator.
2371         (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): Use
2372         gcn_fp_compare_operator.
2373         (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): Use
2374         gcn_fp_compare_operator.
2375         (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): Use
2376         gcn_fp_compare_operator.
2378 2020-01-24  Maciej W. Rozycki  <macro@wdc.com>
2380         * doc/install.texi (Cross-Compiler-Specific Options): Document
2381         `--with-toolexeclibdir' option.
2383 2020-01-24  Hans-Peter Nilsson  <hp@axis.com>
2385         * target.def (flags_regnum): Also mention effect on delay slot filling.
2386         * doc/tm.texi: Regenerate.
2388 2020-01-23  Jeff Law  <law@redhat.com>
2390         PR translation/90162
2391         * config/h8300/h8300.c (h8300_option_override): Fix diagnostic text.
2393 2020-01-23  Mikael Tillenius  <mti-1@tillenius.com>
2395         PR target/92269
2396         * config/h8300/h8300.h (FUNCTION_PROFILER): Fix emission of
2397         profiling label
2399 2020-01-23  Jakub Jelinek  <jakub@redhat.com>
2401         PR rtl-optimization/93402
2402         * postreload.c (reload_combine_recognize_pattern): Don't try to adjust
2403         USE insns.
2405 2020-01-23  Dragan Mladjenovic  <dmladjenovic@wavecomp.com>
2407         * config.in: Regenerated.
2408         * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to 1
2409         for TARGET_LIBC_GNUSTACK.
2410         * configure: Regenerated.
2411         * configure.ac: Define TARGET_LIBC_GNUSTACK if glibc version is
2412         found to be 2.31 or greater.
2414 2020-01-23  Dragan Mladjenovic  <dmladjenovic@wavecomp.com>
2416         * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to
2417         TARGET_SOFT_FLOAT.
2418         * config/mips/mips.c (TARGET_ASM_FILE_END): Define to ...
2419         (mips_asm_file_end): New function. Delegate to
2420         file_end_indicate_exec_stack if NEED_INDICATE_EXEC_STACK is true.
2421         * config/mips/mips.h (NEED_INDICATE_EXEC_STACK): Define to 0.
2423 2020-01-23  Jakub Jelinek  <jakub@redhat.com>
2425         PR target/93376
2426         * config/i386/i386-modes.def (POImode): New mode.
2427         (MAX_BITSIZE_MODE_ANY_INT): Change from 128 to 160.
2428         * config/i386/i386.md (DPWI): New mode attribute.
2429         (addv<mode>4, subv<mode>4): Use <DPWI> instead of <DWI>.
2430         (QWI): Rename to...
2431         (QPWI): ... this.  Use POI instead of OI for TImode.
2432         (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1,
2433         *subv<dwi>4_doubleword, *subv<dwi>4_doubleword_1): Use <QPWI>
2434         instead of <QWI>.
2436 2020-01-23  Richard Sandiford  <richard.sandiford@arm.com>
2438         PR target/93341
2439         * config/aarch64/aarch64.md (UNSPEC_SPECULATION_TRACKER_REV): New
2440         unspec.
2441         (speculation_tracker_rev): New pattern.
2442         * config/aarch64/aarch64-speculation.cc (aarch64_do_track_speculation):
2443         Use speculation_tracker_rev to track the inverse condition.
2445 2020-01-23  Richard Biener  <rguenther@suse.de>
2447         PR tree-optimization/93381
2448         * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Take
2449         alias-set of the def as argument and record the first one.
2450         (vn_walk_cb_data::first_set): New member.
2451         (vn_reference_lookup_3): Pass the alias-set of the current def
2452         to push_partial_def.  Fix alias-set used in the aggregate copy
2453         case.
2454         (vn_reference_lookup): Consistently set *last_vuse_ptr.
2455         * real.c (clear_significand_below): Fix out-of-bound access.
2457 2020-01-23  Jakub Jelinek  <jakub@redhat.com>
2459         PR target/93346
2460         * config/i386/i386.md (*bmi2_bzhi_<mode>3_2, *bmi2_bzhi_<mode>3_3):
2461         New define_insn patterns.
2463 2020-01-23  Richard Sandiford  <richard.sandiford@arm.com>
2465         * doc/sourcebuild.texi (check-function-bodies): Add an
2466         optional target/xfail selector.
2468 2020-01-23  Richard Sandiford  <richard.sandiford@arm.com>
2470         PR rtl-optimization/93124
2471         * auto-inc-dec.c (merge_in_block): Don't add auto inc/decs to
2472         bare USE and CLOBBER insns.
2474 2020-01-22  Andrew Pinski  <apinski@marvell.com>
2476         * config/arc/arc.c (output_short_suffix): Check insn for nullness.
2478 2020-01-22  David Malcolm  <dmalcolm@redhat.com>
2480         PR analyzer/93307
2481         * gdbinit.in (break-on-saved-diagnostic): Update for move of
2482         diagnostic_manager into "ana" namespace.
2483         * selftest-run-tests.c (selftest::run_tests): Update for move of
2484         selftest::run_analyzer_selftests to
2485         ana::selftest::run_analyzer_selftests.
2487 2020-01-22  Richard Sandiford  <richard.sandiford@arm.com>
2489         * cfgexpand.c (union_stack_vars): Update the size.
2491 2020-01-22  Richard Biener  <rguenther@suse.de>
2493         PR tree-optimization/93381
2494         * tree-ssa-structalias.c (find_func_aliases): Assume offsetting
2495         throughout, handle all conversions the same.
2497 2020-01-22  Jakub Jelinek  <jakub@redhat.com>
2499         PR target/93335
2500         * config/aarch64/aarch64.c (aarch64_expand_subvti): Only use
2501         gen_subdi3_compare1_imm if low_in2 satisfies aarch64_plus_immediate
2502         predicate, not whenever it is CONST_INT.  Otherwise, force_reg it.
2503         Call force_reg on high_in2 unconditionally.
2505 2020-01-22  Martin Liska  <mliska@suse.cz>
2507         PR tree-optimization/92924
2508         * profile.c (compute_value_histograms): Divide
2509         all counter values.
2511 2020-01-22  Jakub Jelinek  <jakub@redhat.com>
2513         PR target/91298
2514         * output.h (assemble_name_resolve): Declare.
2515         * varasm.c (assemble_name_resolve): New function.
2516         (assemble_name): Use it.
2517         * config/i386/i386.h (ASM_OUTPUT_SYMBOL_REF): Define.
2519 2020-01-22  Joseph Myers  <joseph@codesourcery.com>
2521         * doc/sourcebuild.texi (Texinfo Manuals, Front End): Refer to
2522         update_web_docs_git instead of update_web_docs_svn.
2524 2020-01-21  Andrew Pinski  <apinski@marvell.com>
2526         PR target/9311
2527         * config/aarch64/aarch64.md (tlsgd_small_<mode>): Have operand 0
2528         as PTR mode. Have operand 1 as being modeless, it can be P mode.
2529         (*tlsgd_small_<mode>): Likewise.
2530         * config/aarch64/aarch64.c (aarch64_load_symref_appropriately)
2531         <case SYMBOL_SMALL_TLSGD>: Call gen_tlsgd_small_* with a ptr_mode
2532         register.  Convert that register back to dest using convert_mode.
2534 2020-01-21  Jim Wilson  <jimw@sifive.com>
2536         * config/riscv/riscv-sr.c (riscv_sr_match_prologue): Use INTVAL
2537         instead of XINT.
2539 2020-01-21  H.J. Lu  <hongjiu.lu@intel.com>
2540             Uros Bizjak    <ubizjak@gmail.com>
2542         PR target/93319
2543         * config/i386/i386.c (ix86_tls_module_base): Replace Pmode
2544         with ptr_mode.
2545         (legitimize_tls_address): Do GNU2 TLS address computation in
2546         ptr_mode and zero-extend result to Pmode.
2547         *  config/i386/i386.md (@tls_dynamic_gnu2_64_<mode>): Replace
2548         :P with :PTR and Pmode with ptr_mode.
2549         (*tls_dynamic_gnu2_lea_64_<mode>): Likewise.
2550         (*tls_dynamic_gnu2_call_64_<mode>): Likewise.
2551         (*tls_dynamic_gnu2_combine_64_<mode>): Likewise.
2553 2020-01-21  Jakub Jelinek  <jakub@redhat.com>
2555         PR target/93333
2556         * config/riscv/riscv.c (riscv_rtx_costs) <case ZERO_EXTRACT>: Verify
2557         the last two operands are CONST_INT_P before using them as such.
2559 2020-01-21  Richard Sandiford  <richard.sandiford@arm.com>
2561         * config/aarch64/aarch64-sve-builtins.def: Use get_typenode_from_name
2562         to get the integer element types.
2564 2020-01-21  Richard Sandiford  <richard.sandiford@arm.com>
2566         * config/aarch64/aarch64-sve-builtins.h
2567         (function_expander::convert_to_pmode): Declare.
2568         * config/aarch64/aarch64-sve-builtins.cc
2569         (function_expander::convert_to_pmode): New function.
2570         (function_expander::get_contiguous_base): Use it.
2571         (function_expander::prepare_gather_address_operands): Likewise.
2572         * config/aarch64/aarch64-sve-builtins-sve2.cc
2573         (svwhilerw_svwhilewr_impl::expand): Likewise.
2575 2020-01-21  Szabolcs Nagy  <szabolcs.nagy@arm.com>
2577         PR target/92424
2578         * config/aarch64/aarch64.c (aarch64_declare_function_name): Set
2579         cfun->machine->label_is_assembled.
2580         (aarch64_print_patchable_function_entry): New.
2581         (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY): Define.
2582         * config/aarch64/aarch64.h (struct machine_function): New field,
2583         label_is_assembled.
2585 2020-01-21  David Malcolm  <dmalcolm@redhat.com>
2587         PR ipa/93315
2588         * ipa-profile.c (ipa_profile): Delete call_sums and set it to
2589         NULL on exit.
2591 2020-01-18  Jan Hubicka  <hubicka@ucw.cz>
2593         PR lto/93318    
2594         * cgraph.c (cgraph_edge::resolve_speculation,
2595         cgraph_edge::redirect_call_stmt_to_callee): Fix update of
2596         call_stmt_site_hash.
2598 2020-01-21  Martin Liska  <mliska@suse.cz>
2600         * config/rs6000/rs6000.c (common_mode_defined): Remove
2601         unused variable.
2603 2020-01-21  Richard Biener  <rguenther@suse.de>
2605         PR tree-optimization/92328
2606         * tree-ssa-sccvn.c (vn_reference_lookup_3): Preserve
2607         type when value-numbering same-sized store by inserting a
2608         VIEW_CONVERT_EXPR.
2609         (eliminate_dom_walker::eliminate_stmt): When eliminating
2610         a redundant store handle bit-reinterpretation of the same value.
2612 2020-01-21  Andrew Pinski  <apinski@marvel.com>
2614         PR tree-opt/93321
2615         * tree-into-ssa.c (prepare_block_for_update_1): Split out
2616         from ...
2617         (prepare_block_for_update): This.  Use a worklist instead of
2618         recursing.
2620 2020-01-21  Mihail-Calin Ionescu  <mihail.ionescu@arm.com>
2622         * gcc/config/arm/arm.c (clear_operation_p):
2623         Initialise last_regno, skip first iteration
2624         based on the first_set value and use ints instead
2625         of the unnecessary HOST_WIDE_INTs.
2627 2020-01-21  Jakub Jelinek  <jakub@redhat.com>
2629         PR target/93073
2630         * config/rs6000/rs6000.c (rs6000_emit_cmove): If using fsel, punt for
2631         compare_mode other than SFmode or DFmode.
2633 2020-01-21  Kito Cheng  <kito.cheng@sifive.com>
2635         PR target/93304
2636         * config/riscv/riscv-protos.h (riscv_hard_regno_rename_ok): New.
2637         * config/riscv/riscv.c (riscv_hard_regno_rename_ok): New.
2638         * config/riscv/riscv.h (HARD_REGNO_RENAME_OK): Defined.
2640 2020-01-20  Wilco Dijkstra  <wdijkstr@arm.com>
2642         * config/aarch64/aarch64.c (neoversen1_tunings): Set jump_align to 4.
2644 2020-01-20  Andrew Pinski  <apinski@marvell.com>
2646         PR middle-end/93242
2647         * targhooks.c (default_print_patchable_function_entry): Use
2648         output_asm_insn to emit the nop instruction.
2650 2020-01-20  Fangrui Song  <maskray@google.com>
2652         PR middle-end/93194
2653         * targhooks.c (default_print_patchable_function_entry): Align to
2654         POINTER_SIZE.
2656 2020-01-20  H.J. Lu  <hongjiu.lu@intel.com>
2658         PR target/93319
2659         * config/i386/i386.c (legitimize_tls_address): Pass Pmode to
2660         gen_tls_dynamic_gnu2_64.  Compute GNU2 TLS address in ptr_mode.
2661         * config/i386/i386.md (tls_dynamic_gnu2_64): Renamed to ...
2662         (@tls_dynamic_gnu2_64_<mode>): This.  Replace DI with P.
2663         (*tls_dynamic_gnu2_lea_64): Renamed to ...
2664         (*tls_dynamic_gnu2_lea_64_<mode>): This.  Replace DI with P.
2665         Remove the {q} suffix from lea.
2666         (*tls_dynamic_gnu2_call_64): Renamed to ...
2667         (*tls_dynamic_gnu2_call_64_<mode>): This.  Replace DI with P.
2668         (*tls_dynamic_gnu2_combine_64): Renamed to ...
2669         (*tls_dynamic_gnu2_combine_64_<mode>): This.  Replace DI with P.
2670         Pass Pmode to gen_tls_dynamic_gnu2_64.
2672 2020-01-20  Wilco Dijkstra  <wdijkstr@arm.com>
2674         * config/aarch64/aarch64.h (SLOW_BYTE_ACCESS): Set to 1.
2676 2020-01-20  Richard Sandiford  <richard.sandiford@arm.com>
2678         * config/aarch64/aarch64-sve-builtins-base.cc
2679         (svld1ro_impl::memory_vector_mode): Remove parameter name.
2681 2020-01-20  Richard Biener  <rguenther@suse.de>
2683         PR debug/92763
2684         * dwarf2out.c (prune_unused_types): Unconditionally mark
2685         called function DIEs.
2687 2020-01-20  Martin Liska  <mliska@suse.cz>
2689         PR tree-optimization/93199
2690         * tree-eh.c (struct leh_state): Add
2691         new field outer_non_cleanup.
2692         (cleanup_is_dead_in): Pass leh_state instead
2693         of eh_region.  Add a checking that state->outer_non_cleanup
2694         points to outer non-clean up region.
2695         (lower_try_finally): Record outer_non_cleanup
2696         for this_state.
2697         (lower_catch): Likewise.
2698         (lower_eh_filter): Likewise.
2699         (lower_eh_must_not_throw): Likewise.
2700         (lower_cleanup): Likewise.
2702 2020-01-20  Richard Biener  <rguenther@suse.de>
2704         PR tree-optimization/93094
2705         * tree-vectorizer.h (vect_loop_versioning): Adjust.
2706         (vect_transform_loop): Likewise.
2707         * tree-vectorizer.c (try_vectorize_loop_1): Pass down
2708         loop_vectorized_call to vect_transform_loop.
2709         * tree-vect-loop.c (vect_transform_loop): Pass down
2710         loop_vectorized_call to vect_loop_versioning.
2711         * tree-vect-loop-manip.c (vect_loop_versioning): Use
2712         the earlier discovered loop_vectorized_call.
2714 2020-01-19  Eric S. Raymond <esr@thyrsus.com>
2716         * doc/contribute.texi: Update for SVN -> Git transition.
2717         * doc/install.texi: Likewise.
2719 2020-01-18  Jan Hubicka  <hubicka@ucw.cz>
2721         PR lto/93318
2722         * cgraph.c (cgraph_edge::make_speculative): Increase number of
2723         speculative targets.
2724         (verify_speculative_call): New function
2725         (cgraph_node::verify_node): Use it.
2726         * ipa-profile.c (ipa_profile): Fix formating; do not set number of
2727         speculations.
2729 2020-01-18  Jan Hubicka  <hubicka@ucw.cz>
2731         PR lto/93318
2732         * cgraph.c (cgraph_edge::resolve_speculation): Fix foramting.
2733         (cgraph_edge::make_direct): Remove all indirect targets.
2734         (cgraph_edge::redirect_call_stmt_to_callee): Use make_direct..
2735         (cgraph_node::verify_node): Verify that only one call_stmt or
2736         lto_stmt_uid is set.
2737         * cgraphclones.c (cgraph_edge::clone): Set only one call_stmt or
2738         lto_stmt_uid.
2739         * lto-cgraph.c (lto_output_edge): Simplify streaming of stmt.
2740         (lto_output_ref): Simplify streaming of stmt.
2741         * lto-streamer-in.c (fixup_call_stmt_edges_1): Clear lto_stmt_uid.
2743 2020-01-18  Tamar Christina  <tamar.christina@arm.com>
2745         * config/aarch64/aarch64-sve-builtins-base.cc (memory_vector_mode):
2746         Mark parameter unused.
2748 2020-01-18  Hans-Peter Nilsson  <hp@axis.com>
2750         * config.gcc <obsolete targets>: Add crisv32-*-* and cris-*-linux*
2752 2019-01-18  Gerald Pfeifer  <gerald@pfeifer.com>
2754         * varpool.c (ctor_useable_for_folding_p): Fix grammar.
2756 2020-01-18  Iain Sandoe  <iain@sandoe.co.uk>
2758         * Makefile.in: Add coroutine-passes.o.
2759         * builtin-types.def (BT_CONST_SIZE): New.
2760         (BT_FN_BOOL_PTR): New.
2761         (BT_FN_PTR_PTR_CONST_SIZE_BOOL): New.
2762         * builtins.def (DEF_COROUTINE_BUILTIN): New.
2763         * coroutine-builtins.def: New file.
2764         * coroutine-passes.cc: New file.
2765         * function.h (struct GTY function): Add a bit to indicate that the
2766         function is a coroutine component.
2767         * internal-fn.c (expand_CO_FRAME): New.
2768         (expand_CO_YIELD): New.
2769         (expand_CO_SUSPN): New.
2770         (expand_CO_ACTOR): New.
2771         * internal-fn.def (CO_ACTOR): New.
2772         (CO_YIELD): New.
2773         (CO_SUSPN): New.
2774         (CO_FRAME): New.
2775         * passes.def: Add pass_coroutine_lower_builtins,
2776         pass_coroutine_early_expand_ifns.
2777         * tree-pass.h (make_pass_coroutine_lower_builtins): New.
2778         (make_pass_coroutine_early_expand_ifns): New.
2779         * doc/invoke.texi: Document the fcoroutines command line
2780         switch.
2782 2020-01-18  Jakub Jelinek  <jakub@redhat.com>
2784         * config/arm/vfp.md (*clear_vfp_multiple): Remove unused variable.
2786         PR target/93312
2787         * config/arm/arm.c (clear_operation_p): Don't use REGNO until
2788         after checking the argument is a REG.  Don't use REGNO (reg)
2789         again to set last_regno, reuse regno variable instead.
2791 2020-01-17  David Malcolm  <dmalcolm@redhat.com>
2793         * doc/analyzer.texi (Limitations): Add note about NaN.
2795 2020-01-17  Mihail-Calin Ionescu  <mihail.ionescu@arm.com>
2796             Sudakshina Das  <sudi.das@arm.com>
2798         * config/arm/arm.md (ashldi3): Generate thumb2_lsll for both reg
2799         and valid immediate.
2800         (ashrdi3): Generate thumb2_asrl for both reg and valid immediate.
2801         (lshrdi3): Generate thumb2_lsrl for valid immediates.
2802         * config/arm/constraints.md (Pg): New.
2803         * config/arm/predicates.md (long_shift_imm): New.
2804         (arm_reg_or_long_shift_imm): Likewise.
2805         * config/arm/thumb2.md (thumb2_asrl): New immediate alternative.
2806         (thumb2_lsll): Likewise.
2807         (thumb2_lsrl): New.
2809 2020-01-17  Mihail-Calin Ionescu  <mihail.ionescu@arm.com>
2810             Sudakshina Das  <sudi.das@arm.com>
2812         * config/arm/arm.md (ashldi3): Generate thumb2_lsll for TARGET_HAVE_MVE.
2813         (ashrdi3): Generate thumb2_asrl for TARGET_HAVE_MVE.
2814         * config/arm/arm.c (arm_hard_regno_mode_ok): Allocate even odd
2815         register pairs for doubleword quantities for ARMv8.1M-Mainline.
2816         * config/arm/thumb2.md (thumb2_asrl): New.
2817         (thumb2_lsll): Likewise.
2819 2020-01-17  Jakub Jelinek  <jakub@redhat.com>
2821         * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear): Remove
2822         unused variable.
2824 2020-01-17  Alexander Monakov  <amonakov@ispras.ru>
2826         * gdbinit.in (help-gcc-hooks): New command.
2827         (pp, pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, ptc, pdn, ptn, pdd, prc,
2828         pi, pbm, pel, trt): Take $arg0 instead of $ if supplied. Update
2829         documentation.
2831 2020-01-17  Matthew Malcomson  <matthew.malcomson@arm.com>
2833         * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use the
2834         correct target macro.
2836 2020-01-17  Matthew Malcomson  <matthew.malcomson@arm.com>
2838         * config/aarch64/aarch64-protos.h
2839         (aarch64_sve_ld1ro_operand_p): New.
2840         * config/aarch64/aarch64-sve-builtins-base.cc
2841         (class load_replicate): New.
2842         (class svld1ro_impl): New.
2843         (class svld1rq_impl): Change to inherit from load_replicate.
2844         (svld1ro): New sve intrinsic function base.
2845         * config/aarch64/aarch64-sve-builtins-base.def (svld1ro):
2846         New DEF_SVE_FUNCTION.
2847         * config/aarch64/aarch64-sve-builtins-base.h
2848         (svld1ro): New decl.
2849         * config/aarch64/aarch64-sve-builtins.cc
2850         (function_expander::add_mem_operand): Modify assert to allow
2851         OImode.
2852         * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): New
2853         pattern.
2854         * config/aarch64/aarch64.c
2855         (aarch64_sve_ld1rq_operand_p): Implement in terms of ...
2856         (aarch64_sve_ld1rq_ld1ro_operand_p): This.
2857         (aarch64_sve_ld1ro_operand_p): New.
2858         * config/aarch64/aarch64.md (UNSPEC_LD1RO): New unspec.
2859         * config/aarch64/constraints.md (UOb,UOh,UOw,UOd): New.
2860         * config/aarch64/predicates.md
2861         (aarch64_sve_ld1ro_operand_{b,h,w,d}): New.
2863 2020-01-17  Matthew Malcomson  <matthew.malcomson@arm.com>
2865         * config/aarch64/aarch64-c.c (_ARM_FEATURE_MATMUL_FLOAT64):
2866         Introduce this ACLE specified predefined macro.
2867         * config/aarch64/aarch64-option-extensions.def (f64mm): New.
2868         (fp): Disabling this disables f64mm.
2869         (simd): Disabling this disables f64mm.
2870         (fp16): Disabling this disables f64mm.
2871         (sve): Disabling this disables f64mm.
2872         * config/aarch64/aarch64.h (AARCH64_FL_F64MM): New.
2873         (AARCH64_ISA_F64MM): New.
2874         (TARGET_F64MM): New.
2875         * doc/invoke.texi (f64mm): Document new option.
2877 2020-01-17  Wilco Dijkstra  <wdijkstr@arm.com>
2879         * config/aarch64/aarch64.c (generic_tunings): Add branch fusion.
2880         (neoversen1_tunings): Likewise.
2882 2020-01-17  Wilco Dijkstra  <wdijkstr@arm.com>
2884         PR target/92692
2885         * config/aarch64/aarch64.c (aarch64_split_compare_and_swap)
2886         Add assert to ensure prolog has been emitted.
2887         (aarch64_split_atomic_op): Likewise.
2888         * config/aarch64/atomics.md (aarch64_compare_and_swap<mode>)
2889         Use epilogue_completed rather than reload_completed.
2890         (aarch64_atomic_exchange<mode>): Likewise.
2891         (aarch64_atomic_<atomic_optab><mode>): Likewise.
2892         (atomic_nand<mode>): Likewise.
2893         (aarch64_atomic_fetch_<atomic_optab><mode>): Likewise.
2894         (atomic_fetch_nand<mode>): Likewise.
2895         (aarch64_atomic_<atomic_optab>_fetch<mode>): Likewise.
2896         (atomic_nand_fetch<mode>): Likewise.
2898 2020-01-17  Richard Sandiford  <richard.sandiford@arm.com>
2900         PR target/93133
2901         * config/aarch64/aarch64.h (REVERSIBLE_CC_MODE): Return false
2902         for FP modes.
2903         (REVERSE_CONDITION): Delete.
2904         * config/aarch64/iterators.md (CC_ONLY): New mode iterator.
2905         (CCFP_CCFPE): Likewise.
2906         (e): New mode attribute.
2907         * config/aarch64/aarch64.md (ccmp<GPI:mode>): Rename to...
2908         (@ccmp<CC_ONLY:mode><GPI:mode>): ...this, using CC_ONLY instead of CC.
2909         (fccmp<GPF:mode>, fccmpe<GPF:mode>): Merge into...
2910         (@ccmp<CCFP_CCFPE:mode><GPF:mode>): ...this combined pattern.
2911         (@ccmp<CC_ONLY:mode><GPI:mode>_rev): New pattern.
2912         (@ccmp<CCFP_CCFPE:mode><GPF:mode>_rev): Likewise.
2913         * config/aarch64/aarch64.c (aarch64_gen_compare_reg): Update
2914         name of generator from gen_ccmpdi to gen_ccmpccdi.
2915         (aarch64_gen_ccmp_next): Use code_for_ccmp.  If we want to reverse
2916         the previous comparison but aren't able to, use the new ccmp_rev
2917         patterns instead.
2919 2020-01-17  Richard Sandiford  <richard.sandiford@arm.com>
2921         * gimplify.c (gimplify_return_expr): Use poly_int_tree_p rather
2922         than testing directly for INTEGER_CST.
2923         (gimplify_target_expr, gimplify_omp_depend): Likewise.
2925 2020-01-17  Jakub Jelinek  <jakub@redhat.com>
2927         PR tree-optimization/93292
2928         * tree-vect-stmts.c (vectorizable_comparison): Punt also if
2929         get_vectype_for_scalar_type returns NULL.
2931 2020-01-16  Jan Hubicka  <hubicka@ucw.cz>
2933         * params.opt (-param=max-predicted-iterations): Increase range from 0.
2934         * predict.c (estimate_loops): Add 1 to param_max_predicted_iterations.
2936 2020-01-16  Jan Hubicka  <hubicka@ucw.cz>
2938         * ipa-fnsummary.c (estimate_calls_size_and_time): Fix formating of
2939         dump.
2940         * params.opt: (max-predicted-iterations): Set bounds.
2941         * predict.c (real_almost_one, real_br_prob_base,
2942         real_inv_br_prob_base, real_one_half, real_bb_freq_max): Remove.
2943         (propagate_freq): Add max_cyclic_prob parameter; cap cyclic
2944         probabilities; do not truncate to reg_br_prob_bases.
2945         (estimate_loops_at_level): Pass max_cyclic_prob.
2946         (estimate_loops): Compute max_cyclic_prob.
2947         (estimate_bb_frequencies): Do not initialize real_*; update calculation
2948         of back edge prob.
2949         * profile-count.c (profile_probability::to_sreal): New.
2950         * profile-count.h (class sreal): Move up in file.
2951         (profile_probability::to_sreal): Declare.
2953 2020-01-16  Stam Markianos-Wright  <stam.markianos-wright@arm.com>
2955         * config/arm/arm.c
2956         (arm_invalid_conversion): New function for target hook.
2957         (arm_invalid_unary_op): New function for target hook.
2958         (arm_invalid_binary_op): New function for target hook.
2960 2020-01-16  Stam Markianos-Wright  <stam.markianos-wright@arm.com>
2962         * config.gcc: Add arm_bf16.h.
2963         * config/arm/arm-builtins.c (arm_mangle_builtin_type): Fix comment.
2964         (arm_simd_builtin_std_type): Add BFmode.
2965         (arm_init_simd_builtin_types): Define element types for vector types.
2966         (arm_init_bf16_types): New function.
2967         (arm_init_builtins): Add arm_init_bf16_types function call.
2968         * config/arm/arm-modes.def: Add BFmode and V4BF, V8BF vector modes.
2969         * config/arm/arm-simd-builtin-types.def: Add V4BF, V8BF.
2970         * config/arm/arm.c (aapcs_vfp_sub_candidate):  Add BFmode.
2971         (arm_hard_regno_mode_ok): Add BFmode and tidy up statements.
2972         (arm_vector_mode_supported_p): Add V4BF, V8BF.
2973         (arm_mangle_type):  Add __bf16.
2974         * config/arm/arm.h: Add V4BF, V8BF to VALID_NEON_DREG_MODE, 
2975         VALID_NEON_QREG_MODE respectively. Add export arm_bf16_type_node,
2976         arm_bf16_ptr_type_node.
2977         * config/arm/arm.md: Add BFmode to movhf expand, mov pattern and
2978         define_split between ARM registers.
2979         * config/arm/arm_bf16.h: New file.
2980         * config/arm/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
2981         * config/arm/iterators.md: (ANY64_BF, VDXMOV, VHFBF, HFBF, fporbf): New.
2982         (VQXMOV): Add V8BF.
2983         * config/arm/neon.md: Add BF vector types to movhf NEON move patterns.
2984         * config/arm/vfp.md: Add BFmode to movhf patterns.
2986 2020-01-16  Mihail Ionescu  <mihail.ionescu@arm.com>
2987             Andre Vieira  <andre.simoesdiasvieira@arm.com>
2989         * config/arm/arm-cpus.in (mve, mve_float): New features.
2990         (dsp, mve, mve.fp): New options.
2991         * config/arm/arm.h (TARGET_HAVE_MVE, TARGET_HAVE_MVE_FLOAT): Define.
2992         * config/arm/t-rmprofile: Map v8.1-M multilibs to v8-M.
2993         * doc/invoke.texi: Document the armv8.1-m mve and dps options.
2995 2020-01-16  Mihail-Calin Ionescu  <mihail.ionescu@arm.com>
2996             Thomas Preud'homme  <thomas.preudhomme@arm.com>
2998         * config/arm/arm-cpus.in (ARMv8_1m_main): Redefine as an extension to
2999         Armv8-M Mainline.
3000         * config/arm/arm.c (arm_options_perform_arch_sanity_checks): Remove
3001         error for using -mcmse when targeting Armv8.1-M Mainline.
3003 2020-01-16  Mihail-Calin Ionescu  <mihail.ionescu@arm.com>
3004             Thomas Preud'homme  <thomas.preudhomme@arm.com>
3006         * config/arm/arm.md (nonsecure_call_internal): Do not force memory
3007         address in r4 when targeting Armv8.1-M Mainline.
3008         (nonsecure_call_value_internal): Likewise.
3009         * config/arm/thumb2.md (nonsecure_call_reg_thumb2): Make memory address
3010         a register match_operand again.  Emit BLXNS when targeting
3011         Armv8.1-M Mainline.
3012         (nonsecure_call_value_reg_thumb2): Likewise.
3014 2020-01-16  Mihail-Calin Ionescu  <mihail.ionescu@arm.com>
3015             Thomas Preud'homme  <thomas.preudhomme@arm.com>
3017         * config/arm/arm.c (arm_add_cfa_adjust_cfa_note): Declare early.
3018         (cmse_nonsecure_call_inline_register_clear): Define new lazy_fpclear
3019         variable as true when floating-point ABI is not hard.  Replace
3020         check against TARGET_HARD_FLOAT_ABI by checks against lazy_fpclear.
3021         Generate VLSTM and VLLDM instruction respectively before and
3022         after a function call to cmse_nonsecure_call function.
3023         * config/arm/unspecs.md (VUNSPEC_VLSTM): Define unspec.
3024         (VUNSPEC_VLLDM): Likewise.
3025         * config/arm/vfp.md (lazy_store_multiple_insn): New define_insn.
3026         (lazy_load_multiple_insn): Likewise.
3028 2020-01-16  Mihail-Calin Ionescu  <mihail.ionescu@arm.com>
3029             Thomas Preud'homme  <thomas.preudhomme@arm.com>
3031         * config/arm/arm.c (vfp_emit_fstmd): Declare early.
3032         (arm_emit_vfp_multi_reg_pop): Likewise.
3033         (cmse_nonsecure_call_inline_register_clear): Abstract number of VFP
3034         registers to clear in max_fp_regno.  Emit VPUSH and VPOP to save and
3035         restore callee-saved VFP registers.
3037 2020-01-16  Mihail-Calin Ionescu  <mihail.ionescu@arm.com>
3038             Thomas Preud'homme  <thomas.preudhomme@arm.com>
3040         * config/arm/arm.c (arm_emit_multi_reg_pop): Declare early.
3041         (cmse_nonsecure_call_clear_caller_saved): Rename into ...
3042         (cmse_nonsecure_call_inline_register_clear): This.  Save and clear
3043         callee-saved GPRs as well as clear ip register before doing a nonsecure
3044         call then restore callee-saved GPRs after it when targeting
3045         Armv8.1-M Mainline.
3046         (arm_reorg): Adapt to function rename.
3048 2020-01-16  Mihail-Calin Ionescu  <mihail.ionescu@arm.com>
3049             Thomas Preud'homme  <thomas.preudhomme@arm.com>
3051         * config/arm/arm-protos.h (clear_operation_p): Adapt prototype.
3052         * config/arm/arm.c (clear_operation_p): Extend to be able to check a
3053         clear_vfp_multiple pattern based on a new vfp parameter.
3054         (cmse_clear_registers): Generate VSCCLRM to clear VFP registers when
3055         targeting Armv8.1-M Mainline.
3056         (cmse_nonsecure_entry_clear_before_return): Clear VFP registers
3057         unconditionally when targeting Armv8.1-M Mainline architecture.  Check
3058         whether VFP registers are available before looking call_used_regs for a
3059         VFP register.
3060         * config/arm/predicates.md (clear_multiple_operation): Adapt to change
3061         of prototype of clear_operation_p.
3062         (clear_vfp_multiple_operation): New predicate.
3063         * config/arm/unspecs.md (VUNSPEC_VSCCLRM_VPR): New volatile unspec.
3064         * config/arm/vfp.md (clear_vfp_multiple): New define_insn.
3066 2020-01-16  Mihail-Calin Ionescu  <mihail.ionescu@arm.com>
3067             Thomas Preud'homme  <thomas.preudhomme@arm.com>
3069         * config/arm/arm-protos.h (clear_operation_p): Declare.
3070         * config/arm/arm.c (clear_operation_p): New function.
3071         (cmse_clear_registers): Generate clear_multiple instruction pattern if
3072         targeting Armv8.1-M Mainline or successor.
3073         (output_return_instruction): Only output APSR register clearing if
3074         Armv8.1-M Mainline instructions not available.
3075         (thumb_exit): Likewise.
3076         * config/arm/predicates.md (clear_multiple_operation): New predicate.
3077         * config/arm/thumb2.md (clear_apsr): New define_insn.
3078         (clear_multiple): Likewise.
3079         * config/arm/unspecs.md (VUNSPEC_CLRM_APSR): New volatile unspec.
3081 2020-01-16  Mihail-Calin Ionescu  <mihail.ionescu@arm.com>
3082             Thomas Preud'homme  <thomas.preudhomme@arm.com>
3084         * config/arm/arm.c (fp_sysreg_names): Declare and define.
3085         (use_return_insn): Also return false for Armv8.1-M Mainline.
3086         (output_return_instruction): Skip FPSCR clearing if Armv8.1-M
3087         Mainline instructions are available.
3088         (arm_compute_frame_layout): Allocate space in frame for FPCXTNS
3089         when targeting Armv8.1-M Mainline Security Extensions.
3090         (arm_expand_prologue): Save FPCXTNS if this is an Armv8.1-M
3091         Mainline entry function.
3092         (cmse_nonsecure_entry_clear_before_return): Clear IP and r4 if
3093         targeting Armv8.1-M Mainline or successor.
3094         (arm_expand_epilogue): Fix indentation of caller-saved register
3095         clearing.  Restore FPCXTNS if this is an Armv8.1-M Mainline
3096         entry function.
3097         * config/arm/arm.h (TARGET_HAVE_FP_CMSE): New macro.
3098         (FP_SYSREGS): Likewise.
3099         (enum vfp_sysregs_encoding): Define enum.
3100         (fp_sysreg_names): Declare.
3101         * config/arm/unspecs.md (VUNSPEC_VSTR_VLDR): New volatile unspec.
3102         * config/arm/vfp.md (push_fpsysreg_insn): New define_insn.
3103         (pop_fpsysreg_insn): Likewise.
3105 2020-01-16  Mihail-Calin Ionescu  <mihail.ionescu@arm.com>
3106             Thomas Preud'homme  <thomas.preudhomme@arm.com>
3108         * config/arm/arm-cpus.in (armv8_1m_main): New feature.
3109         (ARMv4, ARMv4t, ARMv5t, ARMv5te, ARMv5tej, ARMv6, ARMv6j, ARMv6k,
3110         ARMv6z, ARMv6kz, ARMv6zk, ARMv6t2, ARMv6m, ARMv7, ARMv7a, ARMv7ve,
3111         ARMv7r, ARMv7m, ARMv7em, ARMv8a, ARMv8_1a, ARMv8_2a, ARMv8_3a,
3112         ARMv8_4a, ARMv8_5a, ARMv8m_base, ARMv8m_main, ARMv8r): Reindent.
3113         (ARMv8_1m_main): New feature group.
3114         (armv8.1-m.main): New architecture.
3115         * config/arm/arm-tables.opt: Regenerate.
3116         * config/arm/arm.c (arm_arch8_1m_main): Define and default initialize.
3117         (arm_option_reconfigure_globals): Initialize arm_arch8_1m_main.
3118         (arm_options_perform_arch_sanity_checks): Error out when targeting
3119         Armv8.1-M Mainline Security Extensions.
3120         * config/arm/arm.h (arm_arch8_1m_main): Declare.
3122 2020-01-16  Stam Markianos-Wright  <stam.markianos-wright@arm.com>
3124         * config/aarch64/aarch64-simd-builtins.def (aarch64_bfdot,
3125         aarch64_bfdot_lane, aarch64_bfdot_laneq): New.
3126         * config/aarch64/aarch64-simd.md (aarch64_bfdot, aarch64_bfdot_lane,
3127         aarch64_bfdot_laneq): New.
3128         * config/aarch64/arm_bf16.h (vbfdot_f32, vbfdotq_f32,
3129         vbfdot_lane_f32, vbfdotq_lane_f32, vbfdot_laneq_f32,
3130         vbfdotq_laneq_f32): New.
3131         * config/aarch64/iterators.md (UNSPEC_BFDOT, Vbfdottype,
3132         VBFMLA_W, VBF): New.
3133         (isquadop): Add V4BF, V8BF.
3135 2020-01-16  Stam Markianos-Wright  <stam.markianos-wright@arm.com>
3137         * config/aarch64/aarch64-builtins.c: (enum aarch64_type_qualifiers):
3138         New qualifier_lane_quadtup_index, TYPES_TERNOP_SSUS,
3139         TYPES_QUADOPSSUS_LANE_QUADTUP, TYPES_QUADOPSSSU_LANE_QUADTUP.
3140         (aarch64_simd_expand_args): Add case SIMD_ARG_LANE_QUADTUP_INDEX.
3141         (aarch64_simd_expand_builtin): Add qualifier_lane_quadtup_index.
3142         * config/aarch64/aarch64-simd-builtins.def (usdot, usdot_lane,
3143         usdot_laneq, sudot_lane,sudot_laneq): New.
3144         * config/aarch64/aarch64-simd.md (aarch64_usdot): New.
3145         (aarch64_<sur>dot_lane): New.
3146         * config/aarch64/arm_neon.h (vusdot_s32): New.
3147         (vusdotq_s32): New.
3148         (vusdot_lane_s32): New.
3149         (vsudot_lane_s32): New.
3150         * config/aarch64/iterators.md (DOTPROD_I8MM): New iterator.
3151         (UNSPEC_USDOT, UNSPEC_SUDOT): New unspecs.
3153 2020-01-16  Martin Liska  <mliska@suse.cz>
3155         * value-prof.c (dump_histogram_value): Fix
3156         obvious spacing issue.
3158 2020-01-16  Andrew Pinski  <apinski@marvell.com>
3160         * tree-ssa-sccvn.c(vn_reference_lookup_3): Check lhs for
3161         !storage_order_barrier_p.
3163 2020-01-16  Andrew Pinski  <apinski@marvell.com>
3165         * sched-int.h (_dep): Add unused bit-field field for the padding.
3166         * sched-deps.c (init_dep_1): Init unused field.
3168 2020-01-16  Andrew Pinski  <apinski@marvell.com>
3170         * optabs.h (create_expand_operand): Initialize target field also.
3172 2020-01-16  Andre Vieira  <andre.simoesdiasvieira@arm.com>
3174         PR tree-optimization/92429
3175         * tree-ssa-loop-niter.h (simplify_replace_tree): Add parameter.
3176         * tree-ssa-loop-niter.c (simplify_replace_tree): Add parameter to
3177         control folding.
3178         * tree-vect-loop.c (update_epilogue_vinfo): Do not fold when replacing
3179         tree.
3181 2020-01-16  Richard Sandiford  <richard.sandiford@arm.com>
3183         * config/aarch64/aarch64.c (aarch64_split_sve_subreg_move): Apply
3184         aarch64_sve_int_mode to each mode.
3186 2020-01-15  David Malcolm  <dmalcolm@redhat.com>
3188         * doc/analyzer.texi (Overview): Add note about
3189         -fdump-ipa-analyzer.
3191 2020-01-15  Wilco Dijkstra  <wdijkstr@arm.com>
3193         PR tree-optimization/93231
3194         * tree-ssa-forwprop.c (optimize_count_trailing_zeroes): Check
3195         input_type is unsigned.  Use tree_to_shwi for shift constant.
3196         Check CST_STRING element size is CHAR_TYPE_SIZE bits.
3197         (simplify_count_trailing_zeroes): Add test to handle known non-zero
3198         inputs more efficiently.
3200 2020-01-15  UroÅ¡ Bizjak  <ubizjak@gmail.com>
3202         * config/i386/i386.md (*movsf_internal): Do not require
3203         SSE2 ISA for alternatives 14 and 15.
3205 2020-01-15  Richard Biener  <rguenther@suse.de>
3207         PR middle-end/93273
3208         * tree-eh.c (sink_clobbers): If we already visited the destination
3209         block do not defer insertion.
3210         (pass_lower_eh_dispatch::execute): Maintain BB_VISITED for
3211         the purpose of defered insertion.
3213 2020-01-15  Jakub Jelinek  <jakub@redhat.com>
3215         * BASE-VER: Bump to 10.0.1.
3217 2020-01-15  Richard Sandiford  <richard.sandiford@arm.com>
3219         PR tree-optimization/93247
3220         * tree-vect-loop.c (update_epilogue_loop_vinfo): Check the access
3221         type of the stmt that we're going to vectorize.
3223 2020-01-15  Richard Sandiford  <richard.sandiford@arm.com>
3225         * tree-vect-slp.c (vectorize_slp_instance_root_stmt): Use a
3226         VIEW_CONVERT_EXPR if the vectorized constructor has a diffeent
3227         type from the lhs.
3229 2020-01-15  Martin Liska  <mliska@suse.cz>
3231         * ipa-profile.c (ipa_profile_read_edge_summary): Do not allow
3232         2 calls of streamer_read_hwi in a function call.
3234 2020-01-15  Richard Biener  <rguenther@suse.de>
3236         * alias.c (record_alias_subset): Avoid redundant work when
3237         subset is already recorded.
3239 2020-01-14  David Malcolm  <dmalcolm@redhat.com>
3241         * doc/invoke.texi (-fdiagnostics-show-cwe): Add note that some of
3242         the analyzer options provide CWE identifiers.
3244 2020-01-14  David Malcolm  <dmalcolm@redhat.com>
3246         * tree-diagnostic-path.cc (path_summary::event_range::print):
3247         When testing for UNKNOWN_LOCATION, look through ad-hoc wrappers
3248         using get_pure_location.
3250 2020-01-15  Jakub Jelinek  <jakub@redhat.com>
3252         PR tree-optimization/93262
3253         * tree-ssa-dse.c (maybe_trim_memstar_call): For *_chk builtins,
3254         perform head trimming only if the last argument is constant,
3255         either all ones, or larger or equal to head trim, in the latter
3256         case decrease the last argument by head_trim.
3258         PR tree-optimization/93249
3259         * tree-ssa-dse.c: Include builtins.h and gimple-fold.h.
3260         (maybe_trim_memstar_call): Move head_trim and tail_trim vars to
3261         function body scope, reindent.  For BUILTIN_IN_STRNCPY*, don't
3262         perform head trim unless we can prove there are no '\0' chars
3263         from the source among the first head_trim chars.
3265 2020-01-14  David Malcolm  <dmalcolm@redhat.com>
3267         * Makefile.in (ANALYZER_OBJS): Add analyzer/function-set.o.
3269 2020-01-15  Jakub Jelinek  <jakub@redhat.com>
3271         PR target/93009
3272         * config/i386/sse.md
3273         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1,
3274         *<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1,
3275         *<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1,
3276         *<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Use
3277         just a single alternative instead of two, make operands 1 and 2
3278         commutative.
3280 2020-01-14  Jan Hubicka  <hubicka@ucw.cz>
3282         PR lto/91576
3283         * ipa-devirt.c (odr_types_equivalent_p): Compare TREE_ADDRESSABLE and
3284         TYPE_MODE.
3286 2020-01-14  David Malcolm  <dmalcolm@redhat.com>
3288         * Makefile.in (lang_opt_files): Add analyzer.opt.
3289         (ANALYZER_OBJS): New.
3290         (OBJS): Add digraph.o, graphviz.o, ordered-hash-map-tests.o,
3291         tristate.o and ANALYZER_OBJS.
3292         (TEXI_GCCINT_FILES): Add analyzer.texi.
3293         * common.opt (-fanalyzer): New driver option.
3294         * config.in: Regenerate.
3295         * configure: Regenerate.
3296         * configure.ac (--disable-analyzer, ENABLE_ANALYZER): New option.
3297         (gccdepdir): Also create depdir for "analyzer" subdir.
3298         * digraph.cc: New file.
3299         * digraph.h: New file.
3300         * doc/analyzer.texi: New file.
3301         * doc/gccint.texi ("Static Analyzer") New menu item.
3302         (analyzer.texi): Include it.
3303         * doc/invoke.texi ("Static Analyzer Options"): New list and new section.
3304         ("Warning Options"): Add static analysis warnings to the list.
3305         (-Wno-analyzer-double-fclose): New option.
3306         (-Wno-analyzer-double-free): New option.
3307         (-Wno-analyzer-exposure-through-output-file): New option.
3308         (-Wno-analyzer-file-leak): New option.
3309         (-Wno-analyzer-free-of-non-heap): New option.
3310         (-Wno-analyzer-malloc-leak): New option.
3311         (-Wno-analyzer-possible-null-argument): New option.
3312         (-Wno-analyzer-possible-null-dereference): New option.
3313         (-Wno-analyzer-null-argument): New option.
3314         (-Wno-analyzer-null-dereference): New option.
3315         (-Wno-analyzer-stale-setjmp-buffer): New option.
3316         (-Wno-analyzer-tainted-array-index): New option.
3317         (-Wno-analyzer-use-after-free): New option.
3318         (-Wno-analyzer-use-of-pointer-in-stale-stack-frame): New option.
3319         (-Wno-analyzer-use-of-uninitialized-value): New option.
3320         (-Wanalyzer-too-complex): New option.
3321         (-fanalyzer-call-summaries): New warning.
3322         (-fanalyzer-checker=): New warning.
3323         (-fanalyzer-fine-grained): New warning.
3324         (-fno-analyzer-state-merge): New warning.
3325         (-fno-analyzer-state-purge): New warning.
3326         (-fanalyzer-transitivity): New warning.
3327         (-fanalyzer-verbose-edges): New warning.
3328         (-fanalyzer-verbose-state-changes): New warning.
3329         (-fanalyzer-verbosity=): New warning.
3330         (-fdump-analyzer): New warning.
3331         (-fdump-analyzer-callgraph): New warning.
3332         (-fdump-analyzer-exploded-graph): New warning.
3333         (-fdump-analyzer-exploded-nodes): New warning.
3334         (-fdump-analyzer-exploded-nodes-2): New warning.
3335         (-fdump-analyzer-exploded-nodes-3): New warning.
3336         (-fdump-analyzer-supergraph): New warning.
3337         * doc/sourcebuild.texi (dg-require-dot): New.
3338         (dg-check-dot): New.
3339         * gdbinit.in (break-on-saved-diagnostic): New command.
3340         * graphviz.cc: New file.
3341         * graphviz.h: New file.
3342         * ordered-hash-map-tests.cc: New file.
3343         * ordered-hash-map.h: New file.
3344         * passes.def (pass_analyzer): Add before
3345         pass_ipa_whole_program_visibility.
3346         * selftest-run-tests.c (selftest::run_tests): Call
3347         selftest::ordered_hash_map_tests_cc_tests.
3348         * selftest.h (selftest::ordered_hash_map_tests_cc_tests): New
3349         decl.
3350         * shortest-paths.h: New file.
3351         * timevar.def (TV_ANALYZER): New timevar.
3352         (TV_ANALYZER_SUPERGRAPH): Likewise.
3353         (TV_ANALYZER_STATE_PURGE): Likewise.
3354         (TV_ANALYZER_PLAN): Likewise.
3355         (TV_ANALYZER_SCC): Likewise.
3356         (TV_ANALYZER_WORKLIST): Likewise.
3357         (TV_ANALYZER_DUMP): Likewise.
3358         (TV_ANALYZER_DIAGNOSTICS): Likewise.
3359         (TV_ANALYZER_SHORTEST_PATHS): Likewise.
3360         * tree-pass.h (make_pass_analyzer): New decl.
3361         * tristate.cc: New file.
3362         * tristate.h: New file.
3364 2020-01-14  UroÅ¡ Bizjak  <ubizjak@gmail.com>
3366         PR target/93254
3367         * config/i386/i386.md (*movsf_internal): Require SSE2 ISA for
3368         alternatives 9 and 10.
3370 2020-01-14  David Malcolm  <dmalcolm@redhat.com>
3372         * attribs.c (excl_hash_traits::empty_zero_p): New static constant.
3373         * gcov.c (function_start_pair_hash::empty_zero_p): Likewise.
3374         * graphite.c (struct sese_scev_hash::empty_zero_p): Likewise.
3375         * hash-map-tests.c (selftest::test_nonzero_empty_key): New selftest.
3376         (selftest::hash_map_tests_c_tests): Call it.
3377         * hash-map-traits.h (simple_hashmap_traits::empty_zero_p):
3378         New static constant, using the value of = H::empty_zero_p.
3379         (unbounded_hashmap_traits::empty_zero_p): Likewise, using the value
3380         from default_hash_traits <Value>.
3381         * hash-map.h (hash_map::empty_zero_p): Likewise, using the value
3382         from Traits.
3383         * hash-set-tests.c (value_hash_traits::empty_zero_p): Likewise.
3384         * hash-table.h (hash_table::alloc_entries): Guard the loop of
3385         calls to mark_empty with !Descriptor::empty_zero_p.
3386         (hash_table::empty_slow): Conditionalize the memset call with a
3387         check that Descriptor::empty_zero_p; otherwise, loop through the
3388         entries calling mark_empty on them.
3389         * hash-traits.h (int_hash::empty_zero_p): New static constant.
3390         (pointer_hash::empty_zero_p): Likewise.
3391         (pair_hash::empty_zero_p): Likewise.
3392         * ipa-devirt.c (default_hash_traits <type_pair>::empty_zero_p):
3393         Likewise.
3394         * ipa-prop.c (ipa_bit_ggc_hash_traits::empty_zero_p): Likewise.
3395         (ipa_vr_ggc_hash_traits::empty_zero_p): Likewise.
3396         * profile.c (location_triplet_hash::empty_zero_p): Likewise.
3397         * sanopt.c (sanopt_tree_triplet_hash::empty_zero_p): Likewise.
3398         (sanopt_tree_couple_hash::empty_zero_p): Likewise.
3399         * tree-hasher.h (int_tree_hasher::empty_zero_p): Likewise.
3400         * tree-ssa-sccvn.c (vn_ssa_aux_hasher::empty_zero_p): Likewise.
3401         * tree-vect-slp.c (bst_traits::empty_zero_p): Likewise.
3402         * tree-vectorizer.h
3403         (default_hash_traits<scalar_cond_masked_key>::empty_zero_p):
3404         Likewise.
3406 2020-01-14  Kewen Lin  <linkw@gcc.gnu.org>
3408         * cfgloopanal.c (average_num_loop_insns): Free bbs when early return,
3409         fix typo on return value.
3411 2020-01-14  Xiong Hu Luo  <luoxhu@linux.ibm.com>
3413         PR ipa/69678
3414         * cgraph.c (symbol_table::create_edge): Init speculative_id and
3415         target_prob.
3416         (cgraph_edge::make_speculative): Add param for setting speculative_id
3417         and target_prob.
3418         (cgraph_edge::speculative_call_info): Update comments and find reference
3419         by speculative_id for multiple indirect targets.
3420         (cgraph_edge::resolve_speculation): Decrease the speculations
3421         for indirect edge, drop it's speculative if not direct target
3422         left. Update comments.
3423         (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
3424         (cgraph_node::dump): Print num_speculative_call_targets.
3425         (cgraph_node::verify_node): Don't report error if speculative
3426         edge not include statement.
3427         (cgraph_edge::num_speculative_call_targets_p): New function.
3428         * cgraph.h (int common_target_id): Remove.
3429         (int common_target_probability): Remove.
3430         (num_speculative_call_targets): New variable.
3431         (make_speculative): Add param for setting speculative_id.
3432         (cgraph_edge::num_speculative_call_targets_p): New declare.
3433         (target_prob): New variable.
3434         (speculative_id): New variable.
3435         * ipa-fnsummary.c (analyze_function_body): Create and duplicate
3436           call summaries for multiple speculative call targets.
3437         * cgraphclones.c (cgraph_node::create_clone): Clone speculative_id.
3438         * ipa-profile.c (struct speculative_call_target): New struct.
3439         (class speculative_call_summary): New class.
3440         (class speculative_call_summaries): New class.
3441         (call_sums): New variable.
3442         (ipa_profile_generate_summary): Generate indirect multiple targets summaries.
3443         (ipa_profile_write_edge_summary): New function.
3444         (ipa_profile_write_summary): Stream out indirect multiple targets summaries.
3445         (ipa_profile_dump_all_summaries): New function.
3446         (ipa_profile_read_edge_summary): New function.
3447         (ipa_profile_read_summary_section): New function.
3448         (ipa_profile_read_summary): Stream in indirect multiple targets summaries.
3449         (ipa_profile): Generate num_speculative_call_targets from
3450         profile summaries.
3451         * ipa-ref.h (speculative_id): New variable.
3452         * ipa-utils.c (ipa_merge_profiles): Update with target_prob.
3453         * lto-cgraph.c (lto_output_edge): Remove indirect common_target_id and
3454         common_target_probability.   Stream out speculative_id and
3455         num_speculative_call_targets.
3456         (input_edge): Likewise.
3457         * predict.c (dump_prediction): Remove edges count assert to be
3458         precise.
3459         * symtab.c (symtab_node::create_reference): Init speculative_id.
3460         (symtab_node::clone_references): Clone speculative_id.
3461         (symtab_node::clone_referring): Clone speculative_id.
3462         (symtab_node::clone_reference): Clone speculative_id.
3463         (symtab_node::clear_stmts_in_references): Clear speculative_id.
3464         * tree-inline.c (copy_bb): Duplicate all the speculative edges
3465         if indirect call contains multiple speculative targets.
3466         * value-prof.h  (check_ic_target): Remove.
3467         * value-prof.c  (gimple_value_profile_transformations):
3468         Use void function gimple_ic_transform.
3469         * value-prof.c  (gimple_ic_transform): Handle topn case.
3470         Fix comment typos.  Change it to a void function.
3472 2020-01-13  Andrew Pinski  <apinski@marvell.com>
3474         * config/aarch64/aarch64-cores.def (octeontx2): New define.
3475         (octeontx2t98): New define.
3476         (octeontx2t96): New define.
3477         (octeontx2t93): New define.
3478         (octeontx2f95): New define.
3479         (octeontx2f95n): New define.
3480         (octeontx2f95mm): New define.
3481         * config/aarch64/aarch64-tune.md: Regenerate.
3482         * doc/invoke.texi (-mcpu=): Document the new cpu types.
3484 2020-01-13  Jason Merrill  <jason@redhat.com>
3486         PR c++/33799 - destroy return value if local cleanup throws.
3487         * gimplify.c (gimplify_return_expr): Handle COMPOUND_EXPR.
3489 2020-01-13  Martin Liska  <mliska@suse.cz>
3491         * ipa-cp.c (get_max_overall_size): Use newly
3492         renamed param param_ipa_cp_unit_growth.
3493         * params.opt: Remove legacy param name.
3495 2020-01-13  Martin Sebor  <msebor@redhat.com>
3497         PR tree-optimization/93213
3498         * tree-ssa-strlen.c (handle_store): Only allow single-byte nul-over-nul
3499         stores to be eliminated.
3501 2020-01-13  Martin Liska  <mliska@suse.cz>
3503         * opts.c (print_help): Do not print CL_PARAM
3504         and CL_WARNING for CL_OPTIMIZATION.
3506 2020-01-13  Jonathan Wakely  <jwakely@redhat.com>
3508         PR driver/92757
3509         * doc/invoke.texi (Warning Options): Add caveat about some warnings
3510         depending on optimization settings.
3512 2020-01-13  Jakub Jelinek  <jakub@redhat.com>
3514         PR tree-optimization/90838
3515         * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
3516         SCALAR_INT_TYPE_MODE directly in CTZ_DEFINED_VALUE_AT_ZERO macro
3517         argument rather than to initialize temporary for targets that
3518         don't use the mode argument at all.  Initialize ctzval to avoid
3519         warning at -O0.
3521 2020-01-10  Thomas Schwinge  <thomas@codesourcery.com>
3523         * tree.h (OMP_CLAUSE_USE_DEVICE_PTR_IF_PRESENT): New definition.
3524         * tree-core.h: Document it.
3525         * gimplify.c (gimplify_omp_workshare): Set it.
3526         * omp-low.c (lower_omp_target): Use it.
3527         * tree-pretty-print.c (dump_omp_clause): Print it.
3529         * omp-low.c (lower_omp_target) <OMP_CLAUSE_USE_DEVICE_PTR etc.>:
3530         Assert that for OpenACC we always have 'GOMP_MAP_USE_DEVICE_PTR'.
3532 2020-01-10  David Malcolm  <dmalcolm@redhat.com>
3534         * Makefile.in (OBJS): Add tree-diagnostic-path.o.
3535         * common.opt (fdiagnostics-path-format=): New option.
3536         (diagnostic_path_format): New enum.
3537         (fdiagnostics-show-path-depths): New option.
3538         * coretypes.h (diagnostic_event_id_t): New forward decl.
3539         * diagnostic-color.c (color_dict): Add "path".
3540         * diagnostic-event-id.h: New file.
3541         * diagnostic-format-json.cc (json_from_expanded_location): Make
3542         non-static.
3543         (json_end_diagnostic): Call context->make_json_for_path if it
3544         exists and the diagnostic has a path.
3545         (diagnostic_output_format_init): Clear context->print_path.
3546         * diagnostic-path.h: New file.
3547         * diagnostic-show-locus.c (colorizer::set_range): Special-case
3548         when printing a run of events in a diagnostic_path so that they
3549         all get the same color.
3550         (layout::m_diagnostic_path_p): New field.
3551         (layout::layout): Initialize it.
3552         (layout::print_any_labels): Don't colorize the label text for an
3553         event in a diagnostic_path.
3554         (gcc_rich_location::add_location_if_nearby): Add
3555         "restrict_to_current_line_spans" and "label" params.  Pass the
3556         former to layout.maybe_add_location_range; pass the latter
3557         when calling add_range.
3558         * diagnostic.c: Include "diagnostic-path.h".
3559         (diagnostic_initialize): Initialize context->path_format and
3560         context->show_path_depths.
3561         (diagnostic_show_any_path): New function.
3562         (diagnostic_path::interprocedural_p): New function.
3563         (diagnostic_report_diagnostic): Call diagnostic_show_any_path.
3564         (simple_diagnostic_path::num_events): New function.
3565         (simple_diagnostic_path::get_event): New function.
3566         (simple_diagnostic_path::add_event): New function.
3567         (simple_diagnostic_event::simple_diagnostic_event): New ctor.
3568         (simple_diagnostic_event::~simple_diagnostic_event): New dtor.
3569         (debug): New overload taking a diagnostic_path *.
3570         * diagnostic.def (DK_DIAGNOSTIC_PATH): New.
3571         * diagnostic.h (enum diagnostic_path_format): New enum.
3572         (json::value): New forward decl.
3573         (diagnostic_context::path_format): New field.
3574         (diagnostic_context::show_path_depths): New field.
3575         (diagnostic_context::print_path): New callback field.
3576         (diagnostic_context::make_json_for_path): New callback field.
3577         (diagnostic_show_any_path): New decl.
3578         (json_from_expanded_location): New decl.
3579         * doc/invoke.texi (-fdiagnostics-path-format=): New option.
3580         (-fdiagnostics-show-path-depths): New option.
3581         (-fdiagnostics-color): Add "path" to description of default
3582         GCC_COLORS; describe it.
3583         (-fdiagnostics-format=json): Document how diagnostic paths are
3584         represented in the JSON output format.
3585         * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
3586         Add optional params "restrict_to_current_line_spans" and "label".
3587         * opts.c (common_handle_option): Handle
3588         OPT_fdiagnostics_path_format_ and
3589         OPT_fdiagnostics_show_path_depths.
3590         * pretty-print.c: Include "diagnostic-event-id.h".
3591         (pp_format): Implement "%@" format code for printing
3592         diagnostic_event_id_t *.
3593         (selftest::test_pp_format): Add tests for "%@".
3594         * selftest-run-tests.c (selftest::run_tests): Call
3595         selftest::tree_diagnostic_path_cc_tests.
3596         * selftest.h (selftest::tree_diagnostic_path_cc_tests): New decl.
3597         * toplev.c (general_init): Initialize global_dc->path_format and
3598         global_dc->show_path_depths.
3599         * tree-diagnostic-path.cc: New file.
3600         * tree-diagnostic.c (maybe_unwind_expanded_macro_loc): Make
3601         non-static.  Drop "diagnostic" param in favor of storing the
3602         original value of "where" and re-using it.
3603         (virt_loc_aware_diagnostic_finalizer): Update for dropped param of
3604         maybe_unwind_expanded_macro_loc.
3605         (tree_diagnostics_defaults): Initialize context->print_path and
3606         context->make_json_for_path.
3607         * tree-diagnostic.h (default_tree_diagnostic_path_printer): New
3608         decl.
3609         (default_tree_make_json_for_path): New decl.
3610         (maybe_unwind_expanded_macro_loc): New decl.
3612 2020-01-10  Jakub Jelinek  <jakub@redhat.com>
3614         PR tree-optimization/93210
3615         * fold-const.h (native_encode_initializer,
3616         can_native_interpret_type_p): Declare.
3617         * fold-const.c (native_encode_string): Fix up handling with off != -1,
3618         simplify.
3619         (native_encode_initializer): New function, moved from dwarf2out.c.
3620         Adjust to native_encode_expr compatible arguments, including dry-run
3621         and partial extraction modes.  Don't handle STRING_CST.
3622         (can_native_interpret_type_p): No longer static.
3623         * gimple-fold.c (fold_ctor_reference): For native_encode_expr, verify
3624         offset / BITS_PER_UNIT fits into int and don't call it if
3625         can_native_interpret_type_p fails.  If suboff is NULL and for
3626         CONSTRUCTOR fold_{,non}array_ctor_reference returns NULL, retry with
3627         native_encode_initializer.
3628         (fold_const_aggregate_ref_1): Formatting fix.
3629         * dwarf2out.c (native_encode_initializer): Moved to fold-const.c.
3630         (tree_add_const_value_attribute): Adjust caller.
3632         PR tree-optimization/90838
3633         * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
3634         SCALAR_INT_TYPE_MODE instead of TYPE_MODE as operand of
3635         CTZ_DEFINED_VALUE_AT_ZERO.
3637 2020-01-10  Vladimir Makarov  <vmakarov@redhat.com>
3639         PR inline-asm/93027
3640         * lra-constraints.c (match_reload): Permit input operands have the
3641         same mode as output while other input operands have a different
3642         mode.
3644 2020-01-10  Wilco Dijkstra  <wdijkstr@arm.com>
3646         PR tree-optimization/90838
3647         * tree-ssa-forwprop.c (check_ctz_array): Add new function.
3648         (check_ctz_string): Likewise.
3649         (optimize_count_trailing_zeroes): Likewise.
3650         (simplify_count_trailing_zeroes): Likewise.
3651         (pass_forwprop::execute): Try ctz simplification.
3652         * match.pd: Add matching for ctz idioms.
3654 2020-01-10  Stam Markianos-Wright  <stam.markianos-wright@arm.com>
3656         * config/aarch64/aarch64.c (aarch64_invalid_conversion): New function
3657         for target hook.
3658         (aarch64_invalid_unary_op): New function for target hook.
3659         (aarch64_invalid_binary_op): New function for target hook.
3661 2020-01-10  Stam Markianos-Wright  <stam.markianos-wright@arm.com>
3663         * config.gcc: Add arm_bf16.h.
3664         * config/aarch64/aarch64-builtins.c
3665         (aarch64_simd_builtin_std_type): Add BFmode.
3666         (aarch64_init_simd_builtin_types): Define element types for vector
3667         types.
3668         (aarch64_init_bf16_types): New function.
3669         (aarch64_general_init_builtins): Add arm_init_bf16_types function call.
3670         * config/aarch64/aarch64-modes.def: Add BFmode and V4BF, V8BF vector
3671         modes.
3672         * config/aarch64/aarch64-simd-builtin-types.def: Add BF SIMD types.
3673         * config/aarch64/aarch64-simd.md: Add BF vector types to NEON move
3674         patterns.
3675         * config/aarch64/aarch64.h (AARCH64_VALID_SIMD_DREG_MODE): Add V4BF.
3676         (AARCH64_VALID_SIMD_QREG_MODE): Add V8BF.
3677         * config/aarch64/aarch64.c
3678         (aarch64_classify_vector_mode): Add support for BF types.
3679         (aarch64_gimplify_va_arg_expr): Add support for BF types.
3680         (aarch64_vq_mode): Add support for BF types.
3681         (aarch64_simd_container_mode): Add support for BF types.
3682         (aarch64_mangle_type): Add support for BF scalar type.
3683         * config/aarch64/aarch64.md: Add BFmode to movhf pattern.
3684         * config/aarch64/arm_bf16.h: New file.
3685         * config/aarch64/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
3686         * config/aarch64/iterators.md: Add BF types to mode attributes.
3687         (HFBF, GPF_TF_F16_MOV, VDMOV, VQMOV, VQMOV_NO2Em VALL_F16MOV): New.
3689 2020-01-10  Jason Merrill  <jason@redhat.com>
3691         PR c++/93173 - incorrect tree sharing.
3692         * gimplify.c (copy_if_shared): No longer static.
3693         * gimplify.h: Declare it.
3695 2020-01-10  Richard Sandiford  <richard.sandiford@arm.com>
3697         * doc/invoke.texi (-msve-vector-bits=): Document that
3698         -msve-vector-bits=128 now generates VL-specific code for
3699         little-endian targets.
3700         * config/aarch64/aarch64-sve-builtins.cc (register_builtin_types): Use
3701         build_vector_type_for_mode to construct the data vector types.
3702         * config/aarch64/aarch64.c (aarch64_convert_sve_vector_bits): Generate
3703         VL-specific code for -msve-vector-bits=128 on little-endian targets.
3704         (aarch64_simd_container_mode): Always prefer Advanced SIMD modes
3705         for 128-bit vectors.
3707 2020-01-10  Richard Sandiford  <richard.sandiford@arm.com>
3709         * config/aarch64/aarch64.c (aarch64_evpc_sel): Fix gen_vcond_mask
3710         invocation.
3712 2020-01-10  Richard Sandiford  <richard.sandiford@arm.com>
3714         * config/aarch64/aarch64-builtins.c
3715         (aarch64_builtin_vectorized_function): Check for specific vector modes,
3716         rather than checking the number of elements and the element mode.
3718 2020-01-10  Richard Sandiford  <richard.sandiford@arm.com>
3720         * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
3721         get_related_vectype_for_scalar_type rather than build_vector_type
3722         to create the index type for a conditional reduction.
3724 2020-01-10  Richard Sandiford  <richard.sandiford@arm.com>
3726         * tree-vect-loop.c (update_epilogue_loop_vinfo): Update DR_REF
3727         for any type of gather or scatter, including strided accesses.
3729 2020-01-10  Andre Vieira  <andre.simoesdiasvieira@arm.com>
3731         * tree-vectorizer.h (get_dr_vinfo_offset): Add missing function
3732          comment.
3734 2020-01-10  Andre Vieira  <andre.simoesdiasvieira@arm.com>
3736         * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref): Use
3737         get_dr_vinfo_offset
3738         * tree-vect-loop.c (update_epilogue_loop_vinfo):  Remove orig_drs_init
3739         parameter and its use to reset DR_OFFSET's.
3740         (vect_transform_loop): Remove orig_drs_init argument.
3741         * tree-vect-loop-manip.c (vect_update_init_of_dr): Update the offset
3742         member of dr_vec_info rather than the offset of the associated
3743         data_reference's innermost_loop_behavior.
3744         (vect_update_init_of_dr): Pass dr_vec_info instead of data_reference.
3745         (vect_do_peeling): Remove orig_drs_init parameter and its construction.
3746         * tree-vect-stmts.c (check_scan_store): Replace use of DR_OFFSET with
3747         get_dr_vinfo_offset.
3748         (vectorizable_store): Likewise.
3749         (vectorizable_load): Likewise.
3751 2020-01-10  Richard Biener  <rguenther@suse.de>
3753         * gimple-ssa-store-merging
3754         (pass_store_merging::terminate_all_aliasing_chains): Cache alias info.
3756 2020-01-10  Martin Liska  <mliska@suse.cz>
3758         PR ipa/93217
3759         * ipa-inline-analysis.c (offline_size): Make proper parenthesis
3760         encapsulation that was there before r280040.
3762 2020-01-10  Richard Biener  <rguenther@suse.de>
3764         PR middle-end/93199
3765         * tree-eh.c (sink_clobbers): Move clobbers to out-of-IL
3766         sequences to avoid walking them again for secondary opportunities.
3767         (pass_lower_eh_dispatch::execute): Instead actually insert
3768         them here.
3770 2020-01-10  Richard Biener  <rguenther@suse.de>
3772         PR middle-end/93199
3773         * tree-eh.c (redirect_eh_edge_1): Avoid some work if possible.
3774         (cleanup_all_empty_eh): Walk landing pads in reverse order to
3775         avoid quadraticness.
3777 2020-01-10  Martin Jambor  <mjambor@suse.cz>
3779         * params.opt (param_ipa_sra_max_replacements): Mark as Optimization.
3780         * ipa-sra.c (pull_accesses_from_callee): New parameter caller, use it
3781         to get param_ipa_sra_max_replacements.
3782         (param_splitting_across_edge): Pass the caller to
3783         pull_accesses_from_callee.
3785 2020-01-10  Martin Jambor  <mjambor@suse.cz>
3787         * params.opt (param_ipcp_unit_growth): Mark as Optimization.
3788         * ipa-cp.c (max_new_size): Removed.
3789         (orig_overall_size): New variable.
3790         (get_max_overall_size): New function.
3791         (estimate_local_effects): Use it.  Adjust dump.
3792         (decide_about_value): Likewise.
3793         (ipcp_propagate_stage): Do not calculate max_new_size, just store
3794         orig_overall_size.  Adjust dump.
3795         (ipa_cp_c_finalize): Clear orig_overall_size instead of max_new_size.
3797 2020-01-10  Martin Jambor  <mjambor@suse.cz>
3799         * params.opt (param_ipa_max_agg_items): Mark as Optimization
3800         * ipa-cp.c (merge_agg_lats_step): New parameter max_agg_items, use
3801         instead of param_ipa_max_agg_items.
3802         (merge_aggregate_lattices): Extract param_ipa_max_agg_items from
3803         optimization info for the callee.
3805 2020-01-09  Kwok Cheung Yeung  <kcy@codesourcery.com>
3807         * lto-streamer-in.c (input_function): Remove streamed-in inline debug
3808         markers if debug_inline_points is false.
3810 2020-01-09  Richard Sandiford  <richard.sandiford@arm.com>
3812         * config.gcc (aarch64*-*-*): Add aarch64-sve-builtins-sve2.o to
3813         extra_objs.
3814         * config/aarch64/t-aarch64 (aarch64-sve-builtins.o): Depend on
3815         aarch64-sve-builtins-base.def, aarch64-sve-builtins-sve2.def and
3816         aarch64-sve-builtins-sve2.h.
3817         (aarch64-sve-builtins-sve2.o): New rule.
3818         * config/aarch64/aarch64.h (AARCH64_ISA_SVE2_AES): New macro.
3819         (AARCH64_ISA_SVE2_BITPERM, AARCH64_ISA_SVE2_SHA3): Likewise.
3820         (AARCH64_ISA_SVE2_SM4, TARGET_SVE2_AES, TARGET_SVE2_BITPERM): Likewise.
3821         (TARGET_SVE2_SHA, TARGET_SVE2_SM4): Likewise.
3822         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
3823         TARGET_SVE2_AES, TARGET_SVE2_BITPERM, TARGET_SVE2_SHA3 and
3824         TARGET_SVE2_SM4.
3825         * config/aarch64/aarch64-sve.md: Update comments with SVE2
3826         instructions that are handled here.
3827         (@cond_asrd<mode>): Generalize to...
3828         (@cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>): ...this.
3829         (*cond_asrd<mode>_2): Generalize to...
3830         (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_2): ...this.
3831         (*cond_asrd<mode>_z): Generalize to...
3832         (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_z): ...this.
3833         * config/aarch64/aarch64.md (UNSPEC_LDNT1_GATHER): New unspec.
3834         (UNSPEC_STNT1_SCATTER, UNSPEC_WHILEGE, UNSPEC_WHILEGT): Likewise.
3835         (UNSPEC_WHILEHI, UNSPEC_WHILEHS): Likewise.
3836         * config/aarch64/aarch64-sve2.md (@aarch64_gather_ldnt<mode>): New
3837         pattern.
3838         (@aarch64_gather_ldnt_<ANY_EXTEND:optab><SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
3839         (@aarch64_scatter_stnt<mode>): Likewise.
3840         (@aarch64_scatter_stnt_<SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
3841         (@aarch64_mul_lane_<mode>): Likewise.
3842         (@aarch64_sve_suqadd<mode>_const): Likewise.
3843         (*<sur>h<addsub><mode>): Generalize to...
3844         (@aarch64_pred_<SVE2_COND_INT_BINARY_REV:sve_int_op><mode>): ...this
3845         new pattern.
3846         (@cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>): New expander.
3847         (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_2): New pattern.
3848         (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_3): Likewise.
3849         (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_any): Likewise.
3850         (*cond_<SVE2_COND_INT_BINARY_NOREV:sve_int_op><mode>_z): Likewise.
3851         (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op><mode>):: Likewise.
3852         (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op>_lane_<mode>): Likewise.
3853         (@aarch64_pred_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): Likewise.
3854         (@cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): New expander.
3855         (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_2): New pattern.
3856         (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_3): Likewise.
3857         (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_any): Likewise.
3858         (@aarch64_sve_<SVE2_INT_TERNARY:sve_int_op><mode>): Likewise.
3859         (@aarch64_sve_<SVE2_INT_TERNARY_LANE:sve_int_op>_lane_<mode>)
3860         (@aarch64_sve_add_mul_lane_<mode>): Likewise.
3861         (@aarch64_sve_sub_mul_lane_<mode>): Likewise.
3862         (@aarch64_sve2_xar<mode>): Likewise.
3863         (@aarch64_sve2_bcax<mode>): Likewise.
3864         (*aarch64_sve2_eor3<mode>): Rename to...
3865         (@aarch64_sve2_eor3<mode>): ...this.
3866         (@aarch64_sve2_bsl<mode>): New expander.
3867         (@aarch64_sve2_nbsl<mode>): Likewise.
3868         (@aarch64_sve2_bsl1n<mode>): Likewise.
3869         (@aarch64_sve2_bsl2n<mode>): Likewise.
3870         (@aarch64_sve_add_<SHIFTRT:sve_int_op><mode>): Likewise.
3871         (*aarch64_sve2_sra<mode>): Add MOVPRFX support.
3872         (@aarch64_sve_add_<VRSHR_N:sve_int_op><mode>): New pattern.
3873         (@aarch64_sve_<SVE2_INT_SHIFT_INSERT:sve_int_op><mode>): Likewise.
3874         (@aarch64_sve2_<USMAX:su>aba<mode>): New expander.
3875         (*aarch64_sve2_<USMAX:su>aba<mode>): New pattern.
3876         (@aarch64_sve_<SVE2_INT_BINARY_WIDE:sve_int_op><mode>): Likewise.
3877         (<su>mull<bt><Vwide>): Generalize to...
3878         (@aarch64_sve_<SVE2_INT_BINARY_LONG:sve_int_op><mode>): ...this new
3879         pattern.
3880         (@aarch64_sve_<SVE2_INT_BINARY_LONG_lANE:sve_int_op>_lane_<mode>)
3881         (@aarch64_sve_<SVE2_INT_SHIFT_IMM_LONG:sve_int_op><mode>)
3882         (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG:sve_int_op><mode>)
3883         (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3884         (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG:sve_int_op><mode>)
3885         (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3886         (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG:sve_int_op><mode>)
3887         (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3888         (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG:sve_int_op><mode>)
3889         (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3890         (@aarch64_sve_<SVE2_FP_TERNARY_LONG:sve_fp_op><mode>): New patterns.
3891         (@aarch64_<SVE2_FP_TERNARY_LONG_LANE:sve_fp_op>_lane_<mode>)
3892         (@aarch64_sve_<SVE2_INT_UNARY_NARROWB:sve_int_op><mode>): Likewise.
3893         (@aarch64_sve_<SVE2_INT_UNARY_NARROWT:sve_int_op><mode>): Likewise.
3894         (@aarch64_sve_<SVE2_INT_BINARY_NARROWB:sve_int_op><mode>): Likewise.
3895         (@aarch64_sve_<SVE2_INT_BINARY_NARROWT:sve_int_op><mode>): Likewise.
3896         (<SHRNB:r>shrnb<mode>): Generalize to...
3897         (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWB:sve_int_op><mode>): ...this
3898         new pattern.
3899         (<SHRNT:r>shrnt<mode>): Generalize to...
3900         (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWT:sve_int_op><mode>): ...this
3901         new pattern.
3902         (@aarch64_pred_<SVE2_INT_BINARY_PAIR:sve_int_op><mode>): New pattern.
3903         (@aarch64_pred_<SVE2_FP_BINARY_PAIR:sve_fp_op><mode>): Likewise.
3904         (@cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>): New expander.
3905         (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_2): New pattern.
3906         (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_z): Likewise.
3907         (@aarch64_sve_<SVE2_INT_CADD:optab><mode>): Likewise.
3908         (@aarch64_sve_<SVE2_INT_CMLA:optab><mode>): Likewise.
3909         (@aarch64_<SVE2_INT_CMLA:optab>_lane_<mode>): Likewise.
3910         (@aarch64_sve_<SVE2_INT_CDOT:optab><mode>): Likewise.
3911         (@aarch64_<SVE2_INT_CDOT:optab>_lane_<mode>): Likewise.
3912         (@aarch64_pred_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): Likewise.
3913         (@cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New expander.
3914         (*cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New pattern.
3915         (@aarch64_sve2_cvtnt<mode>): Likewise.
3916         (@aarch64_pred_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): Likewise.
3917         (@cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): New expander.
3918         (*cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>_any): New pattern.
3919         (@aarch64_sve2_cvtxnt<mode>): Likewise.
3920         (@aarch64_pred_<SVE2_U32_UNARY:sve_int_op><mode>): Likewise.
3921         (@cond_<SVE2_U32_UNARY:sve_int_op><mode>): New expander.
3922         (*cond_<SVE2_U32_UNARY:sve_int_op><mode>): New pattern.
3923         (@aarch64_pred_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): Likewise.
3924         (@cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New expander.
3925         (*cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New pattern.
3926         (@aarch64_sve2_pmul<mode>): Likewise.
3927         (@aarch64_sve_<SVE2_PMULL:optab><mode>): Likewise.
3928         (@aarch64_sve_<SVE2_PMULL_PAIR:optab><mode>): Likewise.
3929         (@aarch64_sve2_tbl2<mode>): Likewise.
3930         (@aarch64_sve2_tbx<mode>): Likewise.
3931         (@aarch64_sve_<SVE2_INT_BITPERM:sve_int_op><mode>): Likewise.
3932         (@aarch64_sve2_histcnt<mode>): Likewise.
3933         (@aarch64_sve2_histseg<mode>): Likewise.
3934         (@aarch64_pred_<SVE2_MATCH:sve_int_op><mode>): Likewise.
3935         (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_cc): Likewise.
3936         (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_ptest): Likewise.
3937         (aarch64_sve2_aes<CRYPTO_AES:aes_op>): Likewise.
3938         (aarch64_sve2_aes<CRYPTO_AESMC:aesmc_op>): Likewise.
3939         (*aarch64_sve2_aese_fused, *aarch64_sve2_aesd_fused): Likewise.
3940         (aarch64_sve2_rax1, aarch64_sve2_sm4e, aarch64_sve2_sm4ekey): Likewise.
3941         (<su>mulh<r>s<mode>3): Update after above pattern name changes.
3942         * config/aarch64/iterators.md (VNx16QI_ONLY, VNx4SF_ONLY)
3943         (SVE_STRUCT2, SVE_FULL_BHI, SVE_FULL_HSI, SVE_FULL_HDI)
3944         (SVE2_PMULL_PAIR_I): New mode iterators.
3945         (UNSPEC_ADCLB, UNSPEC_ADCLT, UNSPEC_ADDHNB, UNSPEC_ADDHNT, UNSPEC_BDEP)
3946         (UNSPEC_BEXT, UNSPEC_BGRP, UNSPEC_CADD90, UNSPEC_CADD270, UNSPEC_CDOT)
3947         (UNSPEC_CDOT90, UNSPEC_CDOT180, UNSPEC_CDOT270, UNSPEC_CMLA)
3948         (UNSPEC_CMLA90, UNSPEC_CMLA180, UNSPEC_CMLA270, UNSPEC_COND_FCVTLT)
3949         (UNSPEC_COND_FCVTNT, UNSPEC_COND_FCVTX, UNSPEC_COND_FCVTXNT)
3950         (UNSPEC_COND_FLOGB, UNSPEC_EORBT, UNSPEC_EORTB, UNSPEC_FADDP)
3951         (UNSPEC_FMAXP, UNSPEC_FMAXNMP, UNSPEC_FMLALB, UNSPEC_FMLALT)
3952         (UNSPEC_FMLSLB, UNSPEC_FMLSLT, UNSPEC_FMINP, UNSPEC_FMINNMP)
3953         (UNSPEC_HISTCNT, UNSPEC_HISTSEG, UNSPEC_MATCH, UNSPEC_NMATCH)
3954         (UNSPEC_PMULLB, UNSPEC_PMULLB_PAIR, UNSPEC_PMULLT, UNSPEC_PMULLT_PAIR)
3955         (UNSPEC_RADDHNB, UNSPEC_RADDHNT, UNSPEC_RSUBHNB, UNSPEC_RSUBHNT)
3956         (UNSPEC_SLI, UNSPEC_SRI, UNSPEC_SABDLB, UNSPEC_SABDLT, UNSPEC_SADDLB)
3957         (UNSPEC_SADDLBT, UNSPEC_SADDLT, UNSPEC_SADDWB, UNSPEC_SADDWT)
3958         (UNSPEC_SBCLB, UNSPEC_SBCLT, UNSPEC_SMAXP, UNSPEC_SMINP)
3959         (UNSPEC_SQCADD90, UNSPEC_SQCADD270, UNSPEC_SQDMULLB, UNSPEC_SQDMULLBT)
3960         (UNSPEC_SQDMULLT, UNSPEC_SQRDCMLAH, UNSPEC_SQRDCMLAH90)
3961         (UNSPEC_SQRDCMLAH180, UNSPEC_SQRDCMLAH270, UNSPEC_SQRSHRNB)
3962         (UNSPEC_SQRSHRNT, UNSPEC_SQRSHRUNB, UNSPEC_SQRSHRUNT, UNSPEC_SQSHRNB)
3963         (UNSPEC_SQSHRNT, UNSPEC_SQSHRUNB, UNSPEC_SQSHRUNT, UNSPEC_SQXTNB)
3964         (UNSPEC_SQXTNT, UNSPEC_SQXTUNB, UNSPEC_SQXTUNT, UNSPEC_SSHLLB)
3965         (UNSPEC_SSHLLT, UNSPEC_SSUBLB, UNSPEC_SSUBLBT, UNSPEC_SSUBLT)
3966         (UNSPEC_SSUBLTB, UNSPEC_SSUBWB, UNSPEC_SSUBWT, UNSPEC_SUBHNB)
3967         (UNSPEC_SUBHNT, UNSPEC_TBL2, UNSPEC_UABDLB, UNSPEC_UABDLT)
3968         (UNSPEC_UADDLB, UNSPEC_UADDLT, UNSPEC_UADDWB, UNSPEC_UADDWT)
3969         (UNSPEC_UMAXP, UNSPEC_UMINP, UNSPEC_UQRSHRNB, UNSPEC_UQRSHRNT)
3970         (UNSPEC_UQSHRNB, UNSPEC_UQSHRNT, UNSPEC_UQXTNB, UNSPEC_UQXTNT)
3971         (UNSPEC_USHLLB, UNSPEC_USHLLT, UNSPEC_USUBLB, UNSPEC_USUBLT)
3972         (UNSPEC_USUBWB, UNSPEC_USUBWT): New unspecs.
3973         (UNSPEC_SMULLB, UNSPEC_SMULLT, UNSPEC_UMULLB, UNSPEC_UMULLT)
3974         (UNSPEC_SMULHS, UNSPEC_SMULHRS, UNSPEC_UMULHS, UNSPEC_UMULHRS)
3975         (UNSPEC_RSHRNB, UNSPEC_RSHRNT, UNSPEC_SHRNB, UNSPEC_SHRNT): Move
3976         further down file.
3977         (VNARROW, Ventype): New mode attributes.
3978         (Vewtype): Handle VNx2DI.  Fix typo in comment.
3979         (VDOUBLE): New mode attribute.
3980         (sve_lane_con): Handle VNx8HI.
3981         (SVE_INT_UNARY): Include ss_abs and ss_neg for TARGET_SVE2.
3982         (SVE_INT_BINARY): Likewise ss_plus, us_plus, ss_minus and us_minus.
3983         (sve_int_op, sve_int_op_rev): Handle the above codes.
3984         (sve_pred_int_rhs2_operand): Likewise.
3985         (MULLBT, SHRNB, SHRNT): Delete.
3986         (SVE_INT_SHIFT_IMM): New int iterator.
3987         (SVE_WHILE): Add UNSPEC_WHILEGE, UNSPEC_WHILEGT, UNSPEC_WHILEHI
3988         and UNSPEC_WHILEHS for TARGET_SVE2.
3989         (SVE2_U32_UNARY, SVE2_INT_UNARY_NARROWB, SVE2_INT_UNARY_NARROWT)
3990         (SVE2_INT_BINARY, SVE2_INT_BINARY_LANE, SVE2_INT_BINARY_LONG)
3991         (SVE2_INT_BINARY_LONG_LANE, SVE2_INT_BINARY_NARROWB)
3992         (SVE2_INT_BINARY_NARROWT, SVE2_INT_BINARY_PAIR, SVE2_FP_BINARY_PAIR)
3993         (SVE2_INT_BINARY_PAIR_LONG, SVE2_INT_BINARY_WIDE): New int iterators.
3994         (SVE2_INT_SHIFT_IMM_LONG, SVE2_INT_SHIFT_IMM_NARROWB): Likewise.
3995         (SVE2_INT_SHIFT_IMM_NARROWT, SVE2_INT_SHIFT_INSERT, SVE2_INT_CADD)
3996         (SVE2_INT_BITPERM, SVE2_INT_TERNARY, SVE2_INT_TERNARY_LANE): Likewise.
3997         (SVE2_FP_TERNARY_LONG, SVE2_FP_TERNARY_LONG_LANE, SVE2_INT_CMLA)
3998         (SVE2_INT_CDOT, SVE2_INT_ADD_BINARY_LONG, SVE2_INT_QADD_BINARY_LONG)
3999         (SVE2_INT_SUB_BINARY_LONG, SVE2_INT_QSUB_BINARY_LONG): Likewise.
4000         (SVE2_INT_ADD_BINARY_LONG_LANE, SVE2_INT_QADD_BINARY_LONG_LANE)
4001         (SVE2_INT_SUB_BINARY_LONG_LANE, SVE2_INT_QSUB_BINARY_LONG_LANE)
4002         (SVE2_COND_INT_UNARY_FP, SVE2_COND_FP_UNARY_LONG): Likewise.
4003         (SVE2_COND_FP_UNARY_NARROWB, SVE2_COND_INT_BINARY): Likewise.
4004         (SVE2_COND_INT_BINARY_NOREV, SVE2_COND_INT_BINARY_REV): Likewise.
4005         (SVE2_COND_INT_SHIFT, SVE2_MATCH, SVE2_PMULL): Likewise.
4006         (optab): Handle the new unspecs.
4007         (su, r): Remove entries for UNSPEC_SHRNB, UNSPEC_SHRNT, UNSPEC_RSHRNB
4008         and UNSPEC_RSHRNT.
4009         (lr): Handle the new unspecs.
4010         (bt): Delete.
4011         (cmp_op, while_optab_cmp, sve_int_op): Handle the new unspecs.
4012         (sve_int_op_rev, sve_int_add_op, sve_int_qadd_op, sve_int_sub_op)
4013         (sve_int_qsub_op): New int attributes.
4014         (sve_fp_op, rot): Handle the new unspecs.
4015         * config/aarch64/aarch64-sve-builtins.h
4016         (function_resolver::require_matching_pointer_type): Declare.
4017         (function_resolver::resolve_unary): Add an optional boolean argument.
4018         (function_resolver::finish_opt_n_resolution): Add an optional
4019         type_suffix_index argument.
4020         (gimple_folder::redirect_call): Declare.
4021         (gimple_expander::prepare_gather_address_operands): Add an optional
4022         bool parameter.
4023         * config/aarch64/aarch64-sve-builtins.cc: Include
4024         aarch64-sve-builtins-sve2.h.
4025         (TYPES_b_unsigned, TYPES_b_integer, TYPES_bh_integer): New macros.
4026         (TYPES_bs_unsigned, TYPES_hs_signed, TYPES_hs_integer): Likewise.
4027         (TYPES_hd_unsigned, TYPES_hsd_signed): Likewise.
4028         (TYPES_hsd_integer): Use TYPES_hsd_signed.
4029         (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): New macros.
4030         (TYPES_s_unsigned): Likewise.
4031         (TYPES_s_integer): Use TYPES_s_unsigned.
4032         (TYPES_sd_signed, TYPES_sd_unsigned): New macros.
4033         (TYPES_sd_integer): Use them.
4034         (TYPES_d_unsigned): New macro.
4035         (TYPES_d_integer): Use it.
4036         (TYPES_d_data, TYPES_cvt_long, TYPES_cvt_narrow_s): New macros.
4037         (TYPES_cvt_narrow): Likewise.
4038         (DEF_SVE_TYPES_ARRAY): Include the new types macros above.
4039         (preds_mx): New variable.
4040         (function_builder::add_overloaded_function): Allow the new feature
4041         set to be more restrictive than the original one.
4042         (function_resolver::infer_pointer_type): Remove qualifiers from
4043         the pointer type before printing it.
4044         (function_resolver::require_matching_pointer_type): New function.
4045         (function_resolver::resolve_sv_displacement): Handle functions
4046         that don't support 32-bit vector indices or svint32_t vector offsets.
4047         (function_resolver::finish_opt_n_resolution): Take the inferred type
4048         as a separate argument.
4049         (function_resolver::resolve_unary): Optionally treat all forms in
4050         the same way as normal merging functions.
4051         (gimple_folder::redirect_call): New function.
4052         (function_expander::prepare_gather_address_operands): Add an argument
4053         that says whether scaled forms are available.  If they aren't,
4054         handle scaling of vector indices and don't add the extension and
4055         scaling operands.
4056         (function_expander::map_to_unspecs): If aarch64_sve isn't available,
4057         fall back to using cond_* instead.
4058         * config/aarch64/aarch64-sve-builtins-functions.h (rtx_code_function):
4059         Split out the member variables into...
4060         (rtx_code_function_base): ...this new base class.
4061         (rtx_code_function_rotated): Inherit rtx_code_function_base.
4062         (unspec_based_function): Split out the member variables into...
4063         (unspec_based_function_base): ...this new base class.
4064         (unspec_based_function_rotated): Inherit unspec_based_function_base.
4065         (unspec_based_function_exact_insn): New class.
4066         (unspec_based_add_function, unspec_based_add_lane_function)
4067         (unspec_based_lane_function, unspec_based_pred_function)
4068         (unspec_based_qadd_function, unspec_based_qadd_lane_function)
4069         (unspec_based_qsub_function, unspec_based_qsub_lane_function)
4070         (unspec_based_sub_function, unspec_based_sub_lane_function): New
4071         typedefs.
4072         (unspec_based_fused_function): New class.
4073         (unspec_based_mla_function, unspec_based_mls_function): New typedefs.
4074         (unspec_based_fused_lane_function): New class.
4075         (unspec_based_mla_lane_function, unspec_based_mls_lane_function): New
4076         typedefs.
4077         (CODE_FOR_MODE1): New macro.
4078         (fixed_insn_function): New class.
4079         (while_comparison): Likewise.
4080         * config/aarch64/aarch64-sve-builtins-shapes.h (binary_long_lane)
4081         (binary_long_opt_n, binary_narrowb_opt_n, binary_narrowt_opt_n)
4082         (binary_to_uint, binary_wide, binary_wide_opt_n, compare, compare_ptr)
4083         (load_ext_gather_index_restricted, load_ext_gather_offset_restricted)
4084         (load_gather_sv_restricted, shift_left_imm_long): Declare.
4085         (shift_left_imm_to_uint, shift_right_imm_narrowb): Likewise.
4086         (shift_right_imm_narrowt, shift_right_imm_narrowb_to_uint): Likewise.
4087         (shift_right_imm_narrowt_to_uint, store_scatter_index_restricted)
4088         (store_scatter_offset_restricted, tbl_tuple, ternary_long_lane)
4089         (ternary_long_opt_n, ternary_qq_lane_rotate, ternary_qq_rotate)
4090         (ternary_shift_left_imm, ternary_shift_right_imm, ternary_uint)
4091         (unary_convert_narrowt, unary_long, unary_narrowb, unary_narrowt)
4092         (unary_narrowb_to_uint, unary_narrowt_to_uint, unary_to_int): Likewise.
4093         * config/aarch64/aarch64-sve-builtins-shapes.cc (apply_predication):
4094         Also add an initial argument for unary_convert_narrowt, regardless
4095         of the predication type.
4096         (build_32_64): Allow loads and stores to specify MODE_none.
4097         (build_sv_index64, build_sv_uint_offset): New functions.
4098         (long_type_suffix): New function.
4099         (binary_imm_narrowb_base, binary_imm_narrowt_base): New classes.
4100         (binary_imm_long_base, load_gather_sv_base): Likewise.
4101         (shift_right_imm_narrow_wrapper, ternary_shift_imm_base): Likewise.
4102         (ternary_resize2_opt_n_base, ternary_resize2_lane_base): Likewise.
4103         (unary_narrowb_base, unary_narrowt_base): Likewise.
4104         (binary_long_lane_def, binary_long_lane): New shape.
4105         (binary_long_opt_n_def, binary_long_opt_n): Likewise.
4106         (binary_narrowb_opt_n_def, binary_narrowb_opt_n): Likewise.
4107         (binary_narrowt_opt_n_def, binary_narrowt_opt_n): Likewise.
4108         (binary_to_uint_def, binary_to_uint): Likewise.
4109         (binary_wide_def, binary_wide): Likewise.
4110         (binary_wide_opt_n_def, binary_wide_opt_n): Likewise.
4111         (compare_def, compare): Likewise.
4112         (compare_ptr_def, compare_ptr): Likewise.
4113         (load_ext_gather_index_restricted_def,
4114         load_ext_gather_index_restricted): Likewise.
4115         (load_ext_gather_offset_restricted_def,
4116         load_ext_gather_offset_restricted): Likewise.
4117         (load_gather_sv_def): Inherit from load_gather_sv_base.
4118         (load_gather_sv_restricted_def, load_gather_sv_restricted): New shape.
4119         (shift_left_imm_def, shift_left_imm): Likewise.
4120         (shift_left_imm_long_def, shift_left_imm_long): Likewise.
4121         (shift_left_imm_to_uint_def, shift_left_imm_to_uint): Likewise.
4122         (store_scatter_index_restricted_def,
4123         store_scatter_index_restricted): Likewise.
4124         (store_scatter_offset_restricted_def,
4125         store_scatter_offset_restricted): Likewise.
4126         (tbl_tuple_def, tbl_tuple): Likewise.
4127         (ternary_long_lane_def, ternary_long_lane): Likewise.
4128         (ternary_long_opt_n_def, ternary_long_opt_n): Likewise.
4129         (ternary_qq_lane_def): Inherit from ternary_resize2_lane_base.
4130         (ternary_qq_lane_rotate_def, ternary_qq_lane_rotate): New shape
4131         (ternary_qq_opt_n_def): Inherit from ternary_resize2_opt_n_base.
4132         (ternary_qq_rotate_def, ternary_qq_rotate): New shape.
4133         (ternary_shift_left_imm_def, ternary_shift_left_imm): Likewise.
4134         (ternary_shift_right_imm_def, ternary_shift_right_imm): Likewise.
4135         (ternary_uint_def, ternary_uint): Likewise.
4136         (unary_convert): Fix typo in comment.
4137         (unary_convert_narrowt_def, unary_convert_narrowt): New shape.
4138         (unary_long_def, unary_long): Likewise.
4139         (unary_narrowb_def, unary_narrowb): Likewise.
4140         (unary_narrowt_def, unary_narrowt): Likewise.
4141         (unary_narrowb_to_uint_def, unary_narrowb_to_uint): Likewise.
4142         (unary_narrowt_to_uint_def, unary_narrowt_to_uint): Likewise.
4143         (unary_to_int_def, unary_to_int): Likewise.
4144         * config/aarch64/aarch64-sve-builtins-base.cc (unspec_cmla)
4145         (unspec_fcmla, unspec_cond_fcmla, expand_mla_mls_lane): New functions.
4146         (svasrd_impl): Delete.
4147         (svcadd_impl::expand): Handle integer operations too.
4148         (svcmla_impl::expand, svcmla_lane::expand): Likewise, using the
4149         new functions to derive the unspec numbers.
4150         (svmla_svmls_lane_impl): Replace with...
4151         (svmla_lane_impl, svmls_lane_impl): ...these new classes.  Handle
4152         integer operations too.
4153         (svwhile_impl): Rename to...
4154         (svwhilelx_impl): ...this and inherit from while_comparison.
4155         (svasrd): Use unspec_based_function.
4156         (svmla_lane): Use svmla_lane_impl.
4157         (svmls_lane): Use svmls_lane_impl.
4158         (svrecpe, svrsqrte): Handle unsigned integer operations too.
4159         (svwhilele, svwhilelt): Use svwhilelx_impl.
4160         * config/aarch64/aarch64-sve-builtins-sve2.h: New file.
4161         * config/aarch64/aarch64-sve-builtins-sve2.cc: Likewise.
4162         * config/aarch64/aarch64-sve-builtins-sve2.def: Likewise.
4163         * config/aarch64/aarch64-sve-builtins.def: Include
4164         aarch64-sve-builtins-sve2.def.
4166 2020-01-09  Richard Sandiford  <richard.sandiford@arm.com>
4168         * config/aarch64/aarch64-protos.h (aarch64_sve_arith_immediate_p)
4169         (aarch64_sve_sqadd_sqsub_immediate_p): Add a machine_mode argument.
4170         * config/aarch64/aarch64.c (aarch64_sve_arith_immediate_p)
4171         (aarch64_sve_sqadd_sqsub_immediate_p): Likewise.  Handle scalar
4172         immediates as well as vector ones.
4173         * config/aarch64/predicates.md (aarch64_sve_arith_immediate)
4174         (aarch64_sve_sub_arith_immediate, aarch64_sve_qadd_immediate)
4175         (aarch64_sve_qsub_immediate): Update calls accordingly.
4177 2020-01-09  Richard Sandiford  <richard.sandiford@arm.com>
4179         * config/aarch64/aarch64-sve2.md: Add banner comments.
4180         (<su>mulh<r>s<mode>3): Move further up file.
4181         (<su>mull<bt><Vwide>, <r>shrnb<mode>, <r>shrnt<mode>)
4182         (*aarch64_sve2_sra<mode>): Move further down file.
4183         * config/aarch64/t-aarch64 (s-check-sve-md): Check aarch64-sve2.md too.
4185 2020-01-09  Richard Sandiford  <richard.sandiford@arm.com>
4187         * config/aarch64/iterators.md (SVE_WHILE): Add UNSPEC_WHILERW
4188         and UNSPEC_WHILEWR.
4189         (while_optab_cmp): Handle them.
4190         * config/aarch64/aarch64-sve.md
4191         (*while_<while_optab_cmp><GPI:mode><PRED_ALL:mode>_ptest): Make public
4192         and add a "@" marker.
4193         * config/aarch64/aarch64-sve2.md (check_<raw_war>_ptrs<mode>): Use it
4194         instead of gen_aarch64_sve2_while_ptest.
4195         (@aarch64_sve2_while<cmp_op><GPI:mode><PRED_ALL:mode>_ptest): Delete.
4197 2020-01-09  Richard Sandiford  <richard.sandiford@arm.com>
4199         * config/aarch64/aarch64.md (UNSPEC_WHILE_LE): Rename to...
4200         (UNSPEC_WHILELE): ...this.
4201         (UNSPEC_WHILE_LO): Rename to...
4202         (UNSPEC_WHILELO): ...this.
4203         (UNSPEC_WHILE_LS): Rename to...
4204         (UNSPEC_WHILELS): ...this.
4205         (UNSPEC_WHILE_LT): Rename to...
4206         (UNSPEC_WHILELT): ...this.
4207         * config/aarch64/iterators.md (SVE_WHILE): Update accordingly.
4208         (cmp_op, while_optab_cmp): Likewise.
4209         * config/aarch64/aarch64.c (aarch64_sve_move_pred_via_while): Likewise.
4210         * config/aarch64/aarch64-sve-builtins-base.cc (svwhilele): Likewise.
4211         (svwhilelt): Likewise.
4213 2020-01-09  Richard Sandiford  <richard.sandiford@arm.com>
4215         * config/aarch64/aarch64-sve-builtins-shapes.h (unary_count): Delete.
4216         (unary_to_uint): Define.
4217         * config/aarch64/aarch64-sve-builtins-shapes.cc (unary_count_def)
4218         (unary_count): Rename to...
4219         (unary_to_uint_def, unary_to_uint): ...this.
4220         * config/aarch64/aarch64-sve-builtins-base.def: Update accordingly.
4222 2020-01-09  Richard Sandiford  <richard.sandiford@arm.com>
4224         * config/aarch64/aarch64-sve-builtins-functions.h
4225         (code_for_mode_function): New class.
4226         (CODE_FOR_MODE0, QUIET_CODE_FOR_MODE0): New macros.
4227         * config/aarch64/aarch64-sve-builtins-base.cc (svcompact_impl)
4228         (svext_impl, svmul_lane_impl, svsplice_impl, svtmad_impl): Delete.
4229         (svcompact, svext, svsplice): Use QUIET_CODE_FOR_MODE0.
4230         (svmul_lane, svtmad): Use CODE_FOR_MODE0.
4232 2020-01-09  Richard Sandiford  <richard.sandiford@arm.com>
4234         * config/aarch64/iterators.md (addsub): New code attribute.
4235         * config/aarch64/aarch64-simd.md (aarch64_<su_optab><optab><mode>):
4236         Re-express as...
4237         (aarch64_<su_optab>q<addsub><mode>): ...this, making the same change
4238         in the asm string and attributes.  Fix indentation.
4239         * config/aarch64/aarch64-sve.md (@aarch64_<su_optab><optab><mode>):
4240         Re-express as...
4241         (@aarch64_sve_<optab><mode>): ...this.
4242         * config/aarch64/aarch64-sve-builtins.h
4243         (function_expander::expand_signed_unpred_op): Delete.
4244         * config/aarch64/aarch64-sve-builtins.cc
4245         (function_expander::expand_signed_unpred_op): Likewise.
4246         (function_expander::map_to_rtx_codes): If the optab isn't defined,
4247         try using code_for_aarch64_sve instead.
4248         * config/aarch64/aarch64-sve-builtins-base.cc (svqadd_impl): Delete.
4249         (svqsub_impl): Likewise.
4250         (svqadd, svqsub): Use rtx_code_function instead.
4252 2020-01-09  Richard Sandiford  <richard.sandiford@arm.com>
4254         * config/aarch64/iterators.md (SRHSUB, URHSUB): Delete.
4255         (HADDSUB, sur, addsub): Remove them.
4257 2020-01-09  Richard Sandiford  <richard.sandiford@arm.com>
4259         * tree-nrv.c (pass_return_slot::execute): Handle all internal
4260         functions the same way, rather than singling out those that
4261         aren't mapped directly to optabs.
4263 2020-01-09  Richard Sandiford  <richard.sandiford@arm.com>
4265         * target.def (compatible_vector_types_p): New target hook.
4266         * hooks.h (hook_bool_const_tree_const_tree_true): Declare.
4267         * hooks.c (hook_bool_const_tree_const_tree_true): New function.
4268         * doc/tm.texi.in (TARGET_COMPATIBLE_VECTOR_TYPES_P): New hook.
4269         * doc/tm.texi: Regenerate.
4270         * gimple-expr.c: Include target.h.
4271         (useless_type_conversion_p): Use targetm.compatible_vector_types_p.
4272         * config/aarch64/aarch64.c (aarch64_compatible_vector_types_p): New
4273         function.
4274         (TARGET_COMPATIBLE_VECTOR_TYPES_P): Define.
4275         * config/aarch64/aarch64-sve-builtins.cc (gimple_folder::convert_pred):
4276         Use the original predicate if it already has a suitable type.
4278 2020-01-09  Martin Jambor  <mjambor@suse.cz>
4280         * cgraph.h (cgraph_edge): Make remove, set_call_stmt, make_direct,
4281         resolve_speculation and redirect_call_stmt_to_callee static.  Change
4282         return type of set_call_stmt to cgraph_edge *.
4283         * auto-profile.c (afdo_indirect_call): Adjust call to
4284         redirect_call_stmt_to_callee.
4285         * cgraph.c (cgraph_edge::set_call_stmt): Make return cgraph-edge *,
4286         make the this pointer explicit, adjust self-recursive calls and the
4287         call top make_direct.  Return the resulting edge.
4288         (cgraph_edge::remove): Make this pointer explicit.
4289         (cgraph_edge::resolve_speculation): Likewise, adjust call to remove.
4290         (cgraph_edge::make_direct): Likewise, adjust call to
4291         resolve_speculation.
4292         (cgraph_edge::redirect_call_stmt_to_callee): Likewise, also adjust
4293         call to set_call_stmt.
4294         (cgraph_update_edges_for_call_stmt_node): Update call to
4295         set_call_stmt and remove.
4296         * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
4297         Renamed edge to master_edge.  Adjusted calls to set_call_stmt.
4298         (cgraph_node::create_edge_including_clones): Moved "first" definition
4299         of edge to the block where it was used.  Adjusted calls to
4300         set_call_stmt.
4301         (cgraph_node::remove_symbol_and_inline_clones): Adjust call to
4302         cgraph_edge::remove.
4303         * cgraphunit.c (walk_polymorphic_call_targets): Adjusted calls to
4304         make_direct and redirect_call_stmt_to_callee.
4305         * ipa-fnsummary.c (redirect_to_unreachable): Adjust calls to
4306         resolve_speculation and make_direct.
4307         * ipa-inline-transform.c (inline_transform): Adjust call to
4308         redirect_call_stmt_to_callee.
4309         (check_speculations_1):: Adjust call to resolve_speculation.
4310         * ipa-inline.c (resolve_noninline_speculation): Adjust call to
4311         resolve-speculation.
4312         (inline_small_functions): Adjust call to resolve_speculation.
4313         (ipa_inline): Likewise.
4314         * ipa-prop.c (ipa_make_edge_direct_to_target): Adjust call to
4315         make_direct.
4316         * ipa-visibility.c (function_and_variable_visibility): Make iteration
4317         safe with regards to edge removal, adjust calls to
4318         redirect_call_stmt_to_callee.
4319         * ipa.c (walk_polymorphic_call_targets): Adjust calls to make_direct
4320         and redirect_call_stmt_to_callee.
4321         * multiple_target.c (create_dispatcher_calls): Adjust call to
4322         redirect_call_stmt_to_callee
4323         (redirect_to_specific_clone): Likewise.
4324         * tree-cfgcleanup.c (delete_unreachable_blocks_update_callgraph):
4325         Adjust calls to cgraph_edge::remove.
4326         * tree-inline.c (copy_bb): Adjust call to set_call_stmt.
4327         (redirect_all_calls): Adjust call to redirect_call_stmt_to_callee.
4328         (expand_call_inline): Adjust call to cgraph_edge::remove.
4330 2020-01-09  Martin Liska  <mliska@suse.cz>
4332         * params.opt: Set Optimization for
4333         param_max_speculative_devirt_maydefs.
4335 2020-01-09  Martin Sebor  <msebor@redhat.com>
4337         PR middle-end/93200
4338         PR fortran/92956
4339         * builtins.c (compute_objsize): Avoid handling MEM_REFs of vector type.
4341 2020-01-09  Martin Liska  <mliska@suse.cz>
4343         * auto-profile.c (auto_profile): Use opt_for_fn
4344         for a parameter.
4345         * ipa-cp.c (ipcp_lattice::add_value): Likewise.
4346         (propagate_vals_across_arith_jfunc): Likewise.
4347         (hint_time_bonus): Likewise.
4348         (incorporate_penalties): Likewise.
4349         (good_cloning_opportunity_p): Likewise.
4350         (perform_estimation_of_a_value): Likewise.
4351         (estimate_local_effects): Likewise.
4352         (ipcp_propagate_stage): Likewise.
4353         * ipa-fnsummary.c (decompose_param_expr): Likewise.
4354         (set_switch_stmt_execution_predicate): Likewise.
4355         (analyze_function_body): Likewise.
4356         * ipa-inline-analysis.c (offline_size): Likewise.
4357         * ipa-inline.c (early_inliner): Likewise.
4358         * ipa-prop.c (ipa_analyze_node): Likewise.
4359         (ipcp_transform_function): Likewise.
4360         * ipa-sra.c (process_scan_results): Likewise.
4361         (ipa_sra_summarize_function): Likewise.
4362         * params.opt: Rename ipcp-unit-growth to
4363         ipa-cp-unit-growth.  Add Optimization for various
4364         IPA-related parameters.
4366 2020-01-09  Richard Biener  <rguenther@suse.de>
4368         PR middle-end/93054
4369         * gimplify.c (gimplify_expr): Deal with NOP definitions.
4371 2020-01-09  Richard Biener  <rguenther@suse.de>
4373         PR tree-optimization/93040
4374         * gimple-ssa-store-merging.c (find_bswap_or_nop): Raise search limit.
4376 2020-01-09  Georg-Johann Lay  <avr@gjlay.de>
4378         * common/config/avr/avr-common.c (avr_option_optimization_table)
4379         [OPT_LEVELS_1_PLUS]: Set -fsplit-wide-types-early.
4381 2020-01-09  Martin Liska  <mliska@suse.cz>
4383         * cgraphclones.c (symbol_table::materialize_all_clones):
4384         Use cgraph_node::dump_name.
4386 2020-01-09  Jakub Jelinek  <jakub@redhat.com>
4388         PR inline-asm/93202
4389         * config/riscv/riscv.c (riscv_print_operand_reloc): Use
4390         output_operand_lossage instead of gcc_unreachable.
4391         * doc/md.texi (riscv f constraint): Fix typo.
4393         PR target/93141
4394         * config/i386/i386.md (subv<mode>4): Use SWIDWI iterator instead of
4395         SWI.  Use <general_hilo_operand> instead of <general_operand>.  Use
4396         CONST_SCALAR_INT_P instead of CONST_INT_P.
4397         (*subv<mode>4_1): Rename to ...
4398         (subv<mode>4_1): ... this.
4399         (*subv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
4400         define_insn_and_split patterns.
4401         (*subv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
4402         patterns.
4404 2020-01-08  David Malcolm  <dmalcolm@redhat.com>
4406         * vec.c (class selftest::count_dtor): New class.
4407         (selftest::test_auto_delete_vec): New test.
4408         (selftest::vec_c_tests): Call it.
4409         * vec.h (class auto_delete_vec): New class template.
4410         (auto_delete_vec<T>::~auto_delete_vec): New dtor.
4412 2020-01-08  David Malcolm  <dmalcolm@redhat.com>
4414         * sbitmap.h (auto_sbitmap): Add operator const_sbitmap.
4416 2020-01-08  Jim Wilson  <jimw@sifive.com>
4418         * config/riscv/riscv.c (riscv_legitimize_tls_address): Ifdef out
4419         use of TLS_MODEL_LOCAL_EXEC when not pic.
4421 2020-01-08  David Malcolm  <dmalcolm@redhat.com>
4423         * hash-map-tests.c (selftest::test_map_of_strings_to_int): Fix
4424         memory leak.
4426 2020-01-08  Jakub Jelinek  <jakub@redhat.com>
4428         PR target/93187
4429         * config/i386/i386.md (*stack_protect_set_2_<mode> peephole2,
4430         *stack_protect_set_3 peephole2): Also check that the second
4431         insns source is general_operand.
4433         PR target/93174
4434         * config/i386/i386.md (addcarry<mode>_0): Use nonimmediate_operand
4435         predicate for output operand instead of register_operand.
4436         (addcarry<mode>, addcarry<mode>_1): Likewise.  Add alternative with
4437         memory destination and non-memory operands[2].
4439 2020-01-08  Martin Liska  <mliska@suse.cz>
4441         * cgraph.c (cgraph_node::dump): Use ::dump_name or
4442         ::dump_asm_name instead of (::name or ::asm_name).
4443         * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
4444         * cgraphunit.c (walk_polymorphic_call_targets): Likewise.
4445         (analyze_functions): Likewise.
4446         (expand_all_functions): Likewise.
4447         * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
4448         (propagate_bits_across_jump_function): Likewise.
4449         (dump_profile_updates): Likewise.
4450         (ipcp_store_bits_results): Likewise.
4451         (ipcp_store_vr_results): Likewise.
4452         * ipa-devirt.c (dump_targets): Likewise.
4453         * ipa-fnsummary.c (analyze_function_body): Likewise.
4454         * ipa-hsa.c (check_warn_node_versionable): Likewise.
4455         (process_hsa_functions): Likewise.
4456         * ipa-icf.c (sem_item_optimizer::merge_classes): Likewise.
4457         (set_alias_uids): Likewise.
4458         * ipa-inline-transform.c (save_inline_function_body): Likewise.
4459         * ipa-inline.c (recursive_inlining): Likewise.
4460         (inline_to_all_callers_1): Likewise.
4461         (ipa_inline): Likewise.
4462         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
4463         (ipa_propagate_frequency): Likewise.
4464         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
4465         (remove_described_reference): Likewise.
4466         * ipa-pure-const.c (worse_state): Likewise.
4467         (check_retval_uses): Likewise.
4468         (analyze_function): Likewise.
4469         (propagate_pure_const): Likewise.
4470         (propagate_nothrow): Likewise.
4471         (dump_malloc_lattice): Likewise.
4472         (propagate_malloc): Likewise.
4473         (pass_local_pure_const::execute): Likewise.
4474         * ipa-visibility.c (optimize_weakref): Likewise.
4475         (function_and_variable_visibility): Likewise.
4476         * ipa.c (symbol_table::remove_unreachable_nodes): Likewise.
4477         (ipa_discover_variable_flags): Likewise.
4478         * lto-streamer-out.c (output_function): Likewise.
4479         (output_constructor): Likewise.
4480         * tree-inline.c (copy_bb): Likewise.
4481         * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
4482         * varpool.c (symbol_table::remove_unreferenced_decls): Likewise.
4484 2020-01-08  Richard Biener  <rguenther@suse.de>
4486         PR middle-end/93199
4487         * tree-eh.c (sink_clobbers): Update virtual operands for
4488         the first and last stmt only.  Add a dry-run capability.
4489         (pass_lower_eh_dispatch::execute): Perform clobber sinking
4490         after CFG manipulations and in RPO order to catch all
4491         secondary opportunities reliably.
4493 2020-01-08  Georg-Johann Lay  <avr@gjlay.de>
4495         PR target/93182
4496         * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
4498 2019-01-08  Richard Biener  <rguenther@suse.de>
4500         PR middle-end/93199
4501         * gimple-fold.c (rewrite_to_defined_overflow): Mark stmt modified.
4502         * tree-ssa-loop-im.c (move_computations_worker): Properly adjust
4503         virtual operand, also updating SSA use.
4504         * gimple-loop-interchange.cc (loop_cand::undo_simple_reduction):
4505         Update stmt after resetting virtual operand.
4506         (tree_loop_interchange::move_code_to_inner_loop): Likewise.
4507         * gimple-iterator.c (gsi_remove): When not removing the stmt
4508         permanently do not delink immediate uses or mark the stmt modified.
4510 2020-01-08  Martin Liska  <mliska@suse.cz>
4512         * ipa-fnsummary.c (dump_ipa_call_summary): Use symtab_node::dump_name.
4513         (ipa_call_context::estimate_size_and_time): Likewise.
4514         (inline_analyze_function): Likewise.
4516 2020-01-08  Martin Liska  <mliska@suse.cz>
4518         * cgraph.c (cgraph_node::dump): Use systematically
4519         dump_asm_name.
4521 2020-01-08  Georg-Johann Lay  <avr@gjlay.de>
4523         Add -nodevicespecs option for avr.
4525         PR target/93182
4526         * config/avr/avr.opt (-nodevicespecs): New driver option.
4527         * config/avr/driver-avr.c (avr_devicespecs_file): Only issue
4528         "-specs=device-specs/..." if that option is not set.
4529         * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
4531 2020-01-08  Georg-Johann Lay  <avr@gjlay.de>
4533         Implement 64-bit double functions for avr.
4535         PR target/92055
4536         * config.gcc (tm_defines) [target=avr]: Support --with-libf7,
4537         --with-double-comparison.
4538         * doc/install.texi: Document them.
4539         * config/avr/avr-c.c (avr_cpu_cpp_builtins)
4540         <WITH_LIBF7_LIBGCC, WITH_LIBF7_MATH, WITH_LIBF7_MATH_SYMBOLS>
4541         <WITH_DOUBLE_COMPARISON>: New built-in defines.
4542         * doc/invoke.texi (AVR Built-in Macros): Document them.
4543         * config/avr/avr-protos.h (avr_float_lib_compare_returns_bool): New.
4544         * config/avr/avr.c (avr_float_lib_compare_returns_bool): New function.
4545         * config/avr/avr.h (FLOAT_LIB_COMPARE_RETURNS_BOOL): New macro.
4547 2020-01-08  Richard Earnshaw  <rearnsha@arm.com>
4549         PR target/93188
4550         * config/arm/t-multilib (MULTILIB_MATCHES): Add rules to match
4551         armv7-a{+mp,+sec,+mp+sec} to appropriate armv7 multilib variants
4552         when only building rm-profile multilibs.
4554 2020-01-08  Feng Xue  <fxue@os.amperecomputing.com>
4556         PR ipa/93084
4557         * ipa-cp.c (self_recursively_generated_p): Find matched aggregate
4558         lattice for a value to check.
4559         (propagate_vals_across_arith_jfunc): Add an assertion to ensure
4560         finite propagation in self-recursive scc.
4562 2020-01-08  Luo Xiong Hu  <luoxhu@linux.ibm.com>
4564         * ipa-inline.c (caller_growth_limits): Restore the AND.
4566 2020-01-07  Andrew Stubbs  <ams@codesourcery.com>
4568         * config/gcn/gcn-valu.md (VEC_1REG_INT_ALT): Delete iterator.
4569         (VEC_ALLREG_ALT): New iterator.
4570         (VEC_ALLREG_INT_MODE): New iterator.
4571         (VCMP_MODE): New iterator.
4572         (VCMP_MODE_INT): New iterator.
4573         (vec_cmpu<mode>di): Use VCMP_MODE_INT.
4574         (vec_cmp<u>v64qidi): New define_expand.
4575         (vec_cmp<mode>di_exec): Use VCMP_MODE.
4576         (vec_cmpu<mode>di_exec): New define_expand.
4577         (vec_cmp<u>v64qidi_exec): New define_expand.
4578         (vec_cmp<mode>di_dup): Use VCMP_MODE.
4579         (vec_cmp<mode>di_dup_exec): Use VCMP_MODE.
4580         (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>): Rename ...
4581         (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): ... to this.
4582         (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>_exec): Rename ...
4583         (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): ... to this.
4584         (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>): Rename ...
4585         (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): ... to this.
4586         (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>_exec): Rename ...
4587         (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): ... to
4588         this.
4589         * config/gcn/gcn.c (print_operand): Fix 8 and 16 bit suffixes.
4590         * config/gcn/gcn.md (expander): Add sign_extend and zero_extend.
4592 2020-01-07  Andrew Stubbs  <ams@codesourcery.com>
4594         * config/gcn/constraints.md (DA): Update description and match.
4595         (DB): Likewise.
4596         (Db): New constraint.
4597         * config/gcn/gcn-protos.h (gcn_inline_constant64_p): Add second
4598         parameter.
4599         * config/gcn/gcn.c (gcn_inline_constant64_p): Add 'mixed' parameter.
4600         Implement 'Db' mixed immediate type.
4601         * config/gcn/gcn-valu.md (addcv64si3<exec_vcc>): Rework constraints.
4602         (addcv64si3_dup<exec_vcc>): Delete.
4603         (subcv64si3<exec_vcc>): Rework constraints.
4604         (addv64di3): Rework constraints.
4605         (addv64di3_exec): Rework constraints.
4606         (subv64di3): Rework constraints.
4607         (addv64di3_dup): Delete.
4608         (addv64di3_dup_exec): Delete.
4609         (addv64di3_zext): Rework constraints.
4610         (addv64di3_zext_exec): Rework constraints.
4611         (addv64di3_zext_dup): Rework constraints.
4612         (addv64di3_zext_dup_exec): Rework constraints.
4613         (addv64di3_zext_dup2): Rework constraints.
4614         (addv64di3_zext_dup2_exec): Rework constraints.
4615         (addv64di3_sext_dup2): Rework constraints.
4616         (addv64di3_sext_dup2_exec): Rework constraints.
4618 2020-01-07  Andre Vieira  <andre.simoesdiasvieira@arm.com>
4620         * doc/sourcebuild.texi (arm_little_endian, arm_nothumb): Documented
4621         existing target checks.
4623 2020-01-07  Richard Biener  <rguenther@suse.de>
4625         * doc/install.texi: Bump minimal supported MPC version.
4627 2020-01-07  Richard Sandiford  <richard.sandiford@arm.com>
4629         * langhooks-def.h (lhd_simulate_enum_decl): Declare.
4630         (LANG_HOOKS_SIMULATE_ENUM_DECL): Use it.
4631         * langhooks.c: Include stor-layout.h.
4632         (lhd_simulate_enum_decl): New function.
4633         * config/aarch64/aarch64-sve-builtins.cc (init_builtins): Call
4634         handle_arm_sve_h for the LTO frontend.
4635         (register_vector_type): Cope with null returns from pushdecl.
4637 2020-01-07  Richard Sandiford  <richard.sandiford@arm.com>
4639         * config/aarch64/aarch64-protos.h (aarch64_sve::svbool_type_p)
4640         (aarch64_sve::nvectors_if_data_type): Replace with...
4641         (aarch64_sve::builtin_type_p): ...this.
4642         * config/aarch64/aarch64-sve-builtins.cc: Include attribs.h.
4643         (find_vector_type): Delete.
4644         (add_sve_type_attribute): New function.
4645         (lookup_sve_type_attribute): Likewise.
4646         (register_builtin_types): Add an "SVE type" attribute to each type.
4647         (register_tuple_type): Likewise.
4648         (svbool_type_p, nvectors_if_data_type): Delete.
4649         (mangle_builtin_type): Use lookup_sve_type_attribute.
4650         (builtin_type_p): Likewise.  Add an overload that returns the
4651         number of constituent vector and predicate registers.
4652         * config/aarch64/aarch64.c (aarch64_sve_argument_p): Delete.
4653         (aarch64_returns_value_in_sve_regs_p): Use aarch64_sve::builtin_type_p
4654         instead of aarch64_sve_argument_p.
4655         (aarch64_takes_arguments_in_sve_regs_p): Likewise.
4656         (aarch64_pass_by_reference): Likewise.
4657         (aarch64_function_value_1): Likewise.
4658         (aarch64_return_in_memory): Likewise.
4659         (aarch64_layout_arg): Likewise.
4661 2020-01-07  Jakub Jelinek  <jakub@redhat.com>
4663         PR tree-optimization/93156
4664         * tree-ssa-ccp.c (bit_value_binop): For x * x note that the second
4665         least significant bit is always clear.
4667         PR tree-optimization/93118
4668         * match.pd ((x >> c) << c -> x & (-1<<c)): Add nop_convert?.  Add new
4669         simplifier with two intermediate conversions.
4671 2020-01-07  Martin Liska  <mliska@suse.cz>
4673         * params.opt: Add Optimization for various parameters.
4675 2020-01-07  Martin Liska  <mliska@suse.cz>
4677         PR ipa/83411
4678         * doc/extend.texi: Explain cloning for target_clone
4679         attribute.
4681 2020-01-07  Martin Liska  <mliska@suse.cz>
4683         PR tree-optimization/92860
4684         * common.opt: Make in Optimization option
4685         as it is affected by -O0, which is an Optimization
4686         option.
4687         * tree-inline.c (tree_inlinable_function_p):
4688         Use opt_for_fn for warn_inline.
4689         (expand_call_inline): Likewise.
4691 2020-01-07  Martin Liska  <mliska@suse.cz>
4693         PR tree-optimization/92860
4694         * common.opt: Make flag_ree as optimization
4695         attribute. 
4697 2020-01-07  Martin Liska  <mliska@suse.cz>
4699         PR optimization/92860
4700         * params.opt: Mark param_min_crossjump_insns with Optimization
4701         keyword.
4703 2020-01-07  Luo Xiong Hu  <luoxhu@linux.ibm.com>
4705         * ipa-inline-analysis.c (estimate_growth): Fix typo.
4706         * ipa-inline.c (caller_growth_limits): Use OR instead of AND.
4708 2020-01-06  Michael Meissner  <meissner@linux.ibm.com>
4710         * config/rs6000/rs6000.c (hard_reg_and_mode_to_addr_mask): New
4711         helper function to return the valid addressing formats for a given
4712         hard register and mode.
4713         (rs6000_adjust_vec_address): Call hard_reg_and_mode_to_addr_mask.
4715         * config/rs6000/constraints.md (Q constraint): Update
4716         documentation.
4717         * doc/md.texi (RS/6000 constraints): Update 'Q' cosntraint
4718         documentation.
4720         * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
4721         Use 'Q' for doing vector extract from memory.
4722         (vsx_extract_v4sf_var): Use 'Q' for doing vector extract from
4723         memory.
4724         (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Use 'Q' for
4725         doing vector extract from memory.
4726         (vsx_extract_<mode>_<VS_scalar>mode_var): Use 'Q' for doing vector
4727         extract from memory.
4729         * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add support
4730         for the offset being 34-bits when -mcpu=future is used.
4732 2020-01-06  John David Anglin  <danglin@gcc.gnu.org>
4734         * config/pa/pa.md: Revert change to use ordered_comparison_operator
4735         instead of cmpib_comparison_operator in cmpib patterns.
4736         * config/pa/predicates.md (cmpib_comparison_operator): Revert removal
4737         of cmpib_comparison_operator.  Revise comment.
4739 2020-01-06  Richard Sandiford  <richard.sandiford@arm.com>
4741         * tree-vect-slp.c (vect_build_slp_tree_1): Require all shifts
4742         in an IFN_DIV_POW2 node to be equal.
4744 2020-01-06  Richard Sandiford  <richard.sandiford@arm.com>
4746         * tree-vect-stmts.c (vect_check_load_store_mask): Rename to...
4747         (vect_check_scalar_mask): ...this.
4748         (vectorizable_store, vectorizable_load): Update call accordingly.
4749         (vectorizable_call): Use vect_check_scalar_mask to check the mask
4750         argument in calls to conditional internal functions.
4752 2020-01-06  Andrew Stubbs  <ams@codesourcery.com>
4754         * config/gcn/gcn-valu.md (subv64di3): Use separate alternatives for
4755         '0' matching inputs.
4756         (subv64di3_exec): Likewise.
4758 2020-01-06  Bryan Stenson  <bryan@siliconvortex.com>
4760         * config/mips/mips.c (vr4130_align_insns): Fix typo.
4761         * doc/md.texi (movstr): Likewise.
4763 2020-01-06  Andrew Stubbs  <ams@codesourcery.com>
4765         * config/gcn/gcn-valu.md (vec_extract<mode><scalar_mode>): Add early
4766         clobber.
4768 2020-01-06  Richard Sandiford  <richard.sandiford@arm.com>
4770         * config/aarch64/t-aarch64 ($(srcdir)/config/aarch64/aarch64-tune.md):
4771         Depend on...
4772         (s-aarch64-tune-md): ...this new stamp file.  Pipe the new contents
4773         to a temporary file and use move-if-change to update the real
4774         file where necessary.
4776 2020-01-06  Richard Sandiford  <richard.sandiford@arm.com>
4778         * config/aarch64/aarch64-sve.md (@aarch64_sel_dup<mode>): Use Upl
4779         rather than Upa for CPY /M.
4781 2020-01-06  Andrew Stubbs  <ams@codesourcery.com>
4783         * config/gcn/gcn.c (gcn_inline_constant_p): Allow 64 as an inline
4784         immediate.
4786 2020-01-06  Martin Liska  <mliska@suse.cz>
4788     PR tree-optimization/92860
4789     * params.opt: Mark param_max_combine_insns with Optimization
4790     keyword. 
4792 2020-01-05  Jakub Jelinek  <jakub@redhat.com>
4794         PR target/93141
4795         * config/i386/i386.md (SWIDWI): New mode iterator.
4796         (DWI, dwi): Add TImode variants.
4797         (addv<mode>4): Use SWIDWI iterator instead of SWI.  Use
4798         <general_hilo_operand> instead of <general_operand>.  Use
4799         CONST_SCALAR_INT_P instead of CONST_INT_P.
4800         (*addv<mode>4_1): Rename to ...
4801         (addv<mode>4_1): ... this.
4802         (QWI): New mode attribute.
4803         (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
4804         define_insn_and_split patterns.
4805         (*addv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
4806         patterns.
4807         (uaddv<mode>4): Use SWIDWI iterator instead of SWI.  Use
4808         <general_hilo_operand> instead of <general_operand>.
4809         (*addcarry<mode>_1): New define_insn.
4810         (*add<dwi>3_doubleword_cc_overflow_1): New define_insn_and_split.
4812 2020-01-03  Konstantin Kharlamov  <Hi-Angel@yandex.ru>
4814         * gdbinit.in (pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, pdd, pbs, pbm):
4815         Use "call" instead of "set".
4817 2020-01-03  Martin Jambor  <mjambor@suse.cz>
4819         PR ipa/92917
4820         * ipa-cp.c (print_all_lattices): Skip functions without info.
4822 2020-01-03  Jakub Jelinek  <jakub@redhat.com>
4824         PR target/93089
4825         * config/i386/i386-options.c (ix86_simd_clone_adjust): If
4826         TARGET_PREFER_AVX128, use prefer-vector-width=256 for 'c' and 'd'
4827         simd clones.  If TARGET_PREFER_AVX256, use prefer-vector-width=512
4828         for 'e' simd clones.
4830         PR target/93089
4831         * config/i386/i386.opt (x_prefer_vector_width_type): Remove TargetSave
4832         entry.
4833         (mprefer-vector-width=): Add Save.
4834         * config/i386/i386-options.c (ix86_target_string): Add PVW argument, print
4835         -mprefer-vector-width= if non-zero.  Fix up -mfpmath= comment.
4836         (ix86_debug_options, ix86_function_specific_print): Adjust
4837         ix86_target_string callers.
4838         (ix86_valid_target_attribute_inner_p): Handle prefer-vector-width=.
4839         (ix86_valid_target_attribute_tree): Likewise.
4840         * config/i386/i386-options.h (ix86_target_string): Add PVW argument.
4841         * config/i386/i386-expand.c (ix86_expand_builtin): Adjust
4842         ix86_target_string caller.
4844         PR target/93110
4845         * config/i386/i386.md (abs<mode>2): Use expand_simple_binop instead of
4846         emitting ASHIFTRT, XOR and MINUS by hand.  Use gen_int_mode with QImode
4847         instead of gen_int_shift_amount + convert_modes.
4849         PR rtl-optimization/93088
4850         * loop-iv.c (find_single_def_src): Punt after looking through
4851         128 reg copies for regs with single definitions.  Move definitions
4852         to first uses.
4854 2020-01-02  Dennis Zhang  <dennis.zhang@arm.com>
4856         * config/arm/arm-c.c (arm_cpu_builtins): Define
4857         __ARM_FEATURE_MATMUL_INT8, __ARM_FEATURE_BF16_VECTOR_ARITHMETIC,
4858         __ARM_FEATURE_BF16_SCALAR_ARITHMETIC, and
4859         __ARM_BF16_FORMAT_ALTERNATIVE when enabled.
4860         * config/arm/arm-cpus.in (armv8_6, i8mm, bf16): New features.
4861         * config/arm/arm-tables.opt: Regenerated.
4862         * config/arm/arm.c (arm_option_reconfigure_globals): Initialize
4863         arm_arch_i8mm and arm_arch_bf16 when enabled.
4864         * config/arm/arm.h (TARGET_I8MM): New macro.
4865         (TARGET_BF16_FP, TARGET_BF16_SIMD): Likewise.
4866         * config/arm/t-aprofile: Add matching rules for -march=armv8.6-a.
4867         * config/arm/t-arm-elf (all_v8_archs): Add armv8.6-a.
4868         * config/arm/t-multilib: Add matching rules for -march=armv8.6-a.
4869         (v8_6_a_simd_variants): New.
4870         (v8_*_a_simd_variants): Add i8mm and bf16.
4871         * doc/invoke.texi (armv8.6-a, i8mm, bf16): Document new options.
4873 2020-01-02  Jakub Jelinek  <jakub@redhat.com>
4875         PR ipa/93087
4876         * predict.c (compute_function_frequency): Don't call
4877         warn_function_cold on functions that already have cold attribute.
4879 2020-01-01  John David Anglin  <danglin@gcc.gnu.org>
4881         PR target/67834
4882         * config/pa/pa.c (pa_elf_select_rtx_section): New.  Put references to
4883         COMDAT group function labels in .data.rel.ro.local section.
4884         * config/pa/pa32-linux.h (TARGET_ASM_SELECT_RTX_SECTION): Define.
4886         PR target/93111
4887         * config/pa/pa.md (scc): Use ordered_comparison_operator instead of
4888         comparison_operator in B and S integer comparisons.  Likewise, use
4889         ordered_comparison_operator instead of cmpib_comparison_operator in
4890         cmpib patterns.
4891         * config/pa/predicates.md (cmpib_comparison_operator): Remove.
4893 2020-01-01  Jakub Jelinek  <jakub@redhat.com>
4895         Update copyright years.
4897         * gcc.c (process_command): Update copyright notice dates.
4898         * gcov-dump.c (print_version): Ditto.
4899         * gcov.c (print_version): Ditto.
4900         * gcov-tool.c (print_version): Ditto.
4901         * gengtype.c (create_file): Ditto.
4902         * doc/cpp.texi: Bump @copying's copyright year.
4903         * doc/cppinternals.texi: Ditto.
4904         * doc/gcc.texi: Ditto.
4905         * doc/gccint.texi: Ditto.
4906         * doc/gcov.texi: Ditto.
4907         * doc/install.texi: Ditto.
4908         * doc/invoke.texi: Ditto.
4910 2020-01-01  Jan Hubicka  <hubicka@ucw.cz>
4912         * ipa.c (walk_polymorphic_call_targets): Fix updating of overall
4913         summary.
4915 2020-01-01  Jakub Jelinek  <jakub@redhat.com>
4917         PR tree-optimization/93098
4918         * match.pd (popcount): For shift amounts, use integer_onep
4919         or wi::to_widest () == cst instead of tree_to_uhwi () == cst
4920         tests.  Make sure that precision is power of two larger than or equal
4921         to 16.  Ensure shift is never negative.  Use HOST_WIDE_INT_UC macro
4922         instead of ULL suffixed constants.  Formatting fixes.
4924 Copyright (C) 2020 Free Software Foundation, Inc.
4926 Copying and distribution of this file, with or without modification,
4927 are permitted in any medium without royalty provided the copyright
4928 notice and this notice are preserved.