rs6000: Correct logic to disable NO_SUM_IN_TOC and NO_FP_IN_TOC [PR94065]
[official-gcc.git] / gcc / ChangeLog
blob843c49ee9f0aefd445808356d2eaf2266b380bf8
1 2020-03-06  David Edelsohn  <dje.gcc@gmail.com>
3         PR target/94065
4         * config/rs6000/aix61.h (TARGET_NO_SUM_IN_TOC): Set to 1 for
5         cmodel=large.
6         (TARGET_NO_FP_IN_TOC): Same.
7         * config/rs6000/aix71.h: Same.
8         * config/rs6000/aix72.h: Same.
10 2020-03-06  Andrew Pinski  <apinski@marvell.com>
11             Jeff Law  <law@redhat.com>
13         PR rtl-optimization/93996
14         * haifa-sched.c (remove_notes): Be more careful when adding
15         REG_SAVE_NOTE.
17 2020-03-06  Delia Burduv  <delia.burduv@arm.com>
19         * config/arm/arm_neon.h (vld2_bf16): New.
20         (vld2q_bf16): New.
21         (vld3_bf16): New.
22         (vld3q_bf16): New.
23         (vld4_bf16): New.
24         (vld4q_bf16): New.
25         (vld2_dup_bf16): New.
26         (vld2q_dup_bf16): New.
27         (vld3_dup_bf16): New.
28         (vld3q_dup_bf16): New.
29         (vld4_dup_bf16): New.
30         (vld4q_dup_bf16): New.
31         * config/arm/arm_neon_builtins.def
32         (vld2): Changed to VAR13 and added v4bf, v8bf
33         (vld2_dup): Changed to VAR8 and added v4bf, v8bf
34         (vld3): Changed to VAR13 and added v4bf, v8bf
35         (vld3_dup): Changed to VAR8 and added v4bf, v8bf
36         (vld4): Changed to VAR13 and added v4bf, v8bf
37         (vld4_dup): Changed to VAR8 and added v4bf, v8bf
38         * config/arm/iterators.md (VDXBF2): New iterator.
39         *config/arm/neon.md (neon_vld2): Use new iterators.
40         (neon_vld2_dup<mode): Use new iterators.
41         (neon_vld3<mode>): Likewise.
42         (neon_vld3qa<mode>): Likewise.
43         (neon_vld3qb<mode>): Likewise.
44         (neon_vld3_dup<mode>): Likewise.
45         (neon_vld4<mode>): Likewise.
46         (neon_vld4qa<mode>): Likewise.
47         (neon_vld4qb<mode>): Likewise.
48         (neon_vld4_dup<mode>): Likewise.
49         (neon_vld2_dupv8bf): New.
50         (neon_vld3_dupv8bf): Likewise.
51         (neon_vld4_dupv8bf): Likewise.
53 2020-03-06  Delia Burduv  <delia.burduv@arm.com>
55         * config/arm/arm_neon.h (bfloat16x4x2_t): New typedef.
56         (bfloat16x8x2_t): New typedef.
57         (bfloat16x4x3_t): New typedef.
58         (bfloat16x8x3_t): New typedef.
59         (bfloat16x4x4_t): New typedef.
60         (bfloat16x8x4_t): New typedef.
61         (vst2_bf16): New.
62         (vst2q_bf16): New.
63         (vst3_bf16): New.
64         (vst3q_bf16): New.
65         (vst4_bf16): New.
66         (vst4q_bf16): New.
67         * config/arm/arm-builtins.c (v2bf_UP): Define.
68         (VAR13): New.
69         (arm_init_simd_builtin_types): Init Bfloat16x2_t eltype.
70         * config/arm/arm-modes.def (V2BF): New mode.
71         * config/arm/arm-simd-builtin-types.def
72         (Bfloat16x2_t): New entry.
73         * config/arm/arm_neon_builtins.def
74         (vst2): Changed to VAR13 and added v4bf, v8bf
75         (vst3): Changed to VAR13 and added v4bf, v8bf
76         (vst4): Changed to VAR13 and added v4bf, v8bf
77         * config/arm/iterators.md (VDXBF): New iterator.
78         (VQ2BF): New iterator.
79         *config/arm/neon.md (neon_vst2<mode>): Used new iterators.
80         (neon_vst2<mode>): Used new iterators.
81         (neon_vst3<mode>): Used new iterators.
82         (neon_vst3<mode>): Used new iterators.
83         (neon_vst3qa<mode>): Used new iterators.
84         (neon_vst3qb<mode>): Used new iterators.
85         (neon_vst4<mode>): Used new iterators.
86         (neon_vst4<mode>): Used new iterators.
87         (neon_vst4qa<mode>): Used new iterators.
88         (neon_vst4qb<mode>): Used new iterators.
90 2020-03-06  Delia Burduv  <delia.burduv@arm.com>
92         * config/aarch64/aarch64-simd-builtins.def
93         (bfcvtn): New built-in function.
94         (bfcvtn_q): New built-in function.
95         (bfcvtn2): New built-in function.
96         (bfcvt): New built-in function.
97         * config/aarch64/aarch64-simd.md
98         (aarch64_bfcvtn<q><mode>): New pattern.
99         (aarch64_bfcvtn2v8bf): New pattern.
100         (aarch64_bfcvtbf): New pattern.
101         * config/aarch64/arm_bf16.h (float32_t): New typedef.
102         (vcvth_bf16_f32): New intrinsic.
103         * config/aarch64/arm_bf16.h (vcvt_bf16_f32): New intrinsic.
104         (vcvtq_low_bf16_f32): New intrinsic.
105         (vcvtq_high_bf16_f32): New intrinsic.
106         * config/aarch64/iterators.md (V4SF_TO_BF): New mode iterator.
107         (UNSPEC_BFCVTN): New UNSPEC.
108         (UNSPEC_BFCVTN2): New UNSPEC.
109         (UNSPEC_BFCVT): New UNSPEC.
110         * config/arm/types.md (bf_cvt): New type.
112 2020-03-06  Andreas Krebbel  <krebbel@linux.ibm.com>
114         * config/s390/s390.md ("tabort"): Get rid of two consecutive
115         blanks in format string.
117 2020-03-05  H.J. Lu  <hongjiu.lu@intel.com>
119         PR target/89229
120         PR target/89346
121         * config/i386/i386-protos.h (ix86_output_ssemov): New prototype.
122         * config/i386/i386.c (ix86_get_ssemov): New function.
123         (ix86_output_ssemov): Likewise.
124         * config/i386/sse.md (VMOVE:mov<mode>_internal): Call
125         ix86_output_ssemov for TYPE_SSEMOV.  Remove TARGET_AVX512VL
126         check.
127         (*movxi_internal_avx512f): Call ix86_output_ssemov for TYPE_SSEMOV.
128         (*movoi_internal_avx): Call ix86_output_ssemov for TYPE_SSEMOV.
129         Remove ext_sse_reg_operand and TARGET_AVX512VL check.
130         (*movti_internal): Likewise.
131         (*movtf_internal): Call ix86_output_ssemov for TYPE_SSEMOV.
133 2020-03-05  Jeff Law  <law@redhat.com>
135         PR tree-optimization/91890
136         * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Remove LOC argument.
137         Use gimple_or_expr_nonartificial_location.
138         (check_bounds_overlap): Drop LOC argument to maybe_diag_access_bounds.
139         Use gimple_or_expr_nonartificial_location.
140         * gimple.c (gimple_or_expr_nonartificial_location): New function.
141         * gimple.h (gimple_or_expr_nonartificial_location): Declare it.
142         * tree-ssa-strlen.c (maybe_warn_overflow): Use
143         gimple_or_expr_nonartificial_location.
144         (maybe_diag_stxncpy_trunc, handle_builtin_stxncpy_strncat): Likewise.
145         (maybe_warn_pointless_strcmp): Likewise.
147 2020-03-05  Jakub Jelinek  <jakub@redhat.com>
149         PR target/94046
150         * config/i386/avx2intrin.h (_mm_mask_i32gather_ps): Fix first cast of
151         SRC and MASK arguments to __m128 from __m128d.
152         (_mm256_mask_i32gather_ps): Fix first cast of MASK argument to __m256
153         from __m256d.
154         (_mm_mask_i64gather_ps): Fix first cast of MASK argument to __m128
155         from __m128d.
156         * config/i386/xopintrin.h (_mm_permute2_pd): Fix first cast of C
157         argument to __m128i from __m128d.
158         (_mm256_permute2_pd): Fix first cast of C argument to __m256i from
159         __m256d.
160         (_mm_permute2_ps): Fix first cast of C argument to __m128i from __m128.
161         (_mm256_permute2_ps): Fix first cast of C argument to __m256i from
162         __m256.
164 2020-03-05  Delia Burduv  <delia.burduv@arm.com>
166         * config/arm/arm_neon.h (vbfmmlaq_f32): New.
167         (vbfmlalbq_f32): New.
168         (vbfmlaltq_f32): New.
169         (vbfmlalbq_lane_f32): New.
170         (vbfmlaltq_lane_f32): New.
171         (vbfmlalbq_laneq_f32): New.
172         (vbfmlaltq_laneq_f32): New.
173         * config/arm/arm_neon_builtins.def (vmmla): New.
174         (vfmab): New.
175         (vfmat): New.
176         (vfmab_lane): New.
177         (vfmat_lane): New.
178         (vfmab_laneq): New.
179         (vfmat_laneq): New.
180         * config/arm/iterators.md (BF_MA): New int iterator.
181         (bt): New int attribute.
182         (VQXBF): Copy of VQX with V8BF.
183         * config/arm/neon.md (neon_vmmlav8bf): New insn.
184         (neon_vfma<bt>v8bf): New insn.
185         (neon_vfma<bt>_lanev8bf): New insn.
186         (neon_vfma<bt>_laneqv8bf): New expand.
187         (neon_vget_high<mode>): Changed iterator to VQXBF.
188         * config/arm/unspecs.md (UNSPEC_BFMMLA): New UNSPEC.
189         (UNSPEC_BFMAB): New UNSPEC.
190         (UNSPEC_BFMAT): New UNSPEC.
192 2020-03-05  Jakub Jelinek  <jakub@redhat.com>
194         PR middle-end/93399
195         * tree-pretty-print.h (pretty_print_string): Declare.
196         * tree-pretty-print.c (pretty_print_string): Remove forward
197         declaration, no longer static.  Change nbytes parameter type
198         from unsigned to size_t.
199         * print-rtl.c (print_value) <case CONST_STRING>: Use
200         pretty_print_string and for shrink way too long strings.
202 2020-03-05  Richard Biener  <rguenther@suse.de>
203             Jakub Jelinek  <jakub@redhat.com>
205         PR tree-optimization/93582
206         * tree-ssa-sccvn.c (vn_reference_lookup_3): Treat POINTER_PLUS_EXPR
207         last operand as signed when looking for memset offset.  Formatting
208         fix.
210 2020-03-04  Andrew Pinski  <apinski@marvell.com>
212         PR bootstrap/93962
213         * value-prof.c (dump_histogram_value): Use std::abs.
215 2020-03-04  Martin Sebor  <msebor@redhat.com>
217         PR tree-optimization/93986
218         * tree-ssa-strlen.c (maybe_warn_overflow): Convert all wide_int
219         operands to the same precision widest_int to avoid ICEs.
221 2020-03-04  Bill Schmidt  <wschmidt@linux.ibm.com>
223         PR target/87560
224         * rs6000-cpus.def (OTHER_ALTIVEC_MASKS): New #define.
225         * rs6000.c (rs6000_disable_incompatible_switches): Add table entry
226         for OPTION_MASK_ALTIVEC.
228 2020-03-04  Andreas Krebbel  <krebbel@linux.ibm.com>
230         * config.gcc: Include the glibc-stdint.h header for zTPF.
232 2020-03-04  Andreas Krebbel  <krebbel@linux.ibm.com>
234         * config/s390/s390.c (s390_secondary_memory_needed): Disallow
235         direct FPR-GPR copies.
236         (s390_register_info_gprtofpr): Disallow GPR content to be saved in
237         FPRs.
239 2020-03-04  Andreas Krebbel  <krebbel@linux.ibm.com>
241         * config/s390/s390.c (s390_emit_prologue): Specify the 2 new
242         operands to the prologue_tpf expander.
243         (s390_emit_epilogue): Likewise.
244         (s390_option_override_internal): Do error checking and setup for
245         the new options.
246         * config/s390/tpf.h (TPF_TRACE_PROLOGUE_CHECK)
247         (TPF_TRACE_EPILOGUE_CHECK, TPF_TRACE_PROLOGUE_TARGET)
248         (TPF_TRACE_EPILOGUE_TARGET, TPF_TRACE_PROLOGUE_SKIP_TARGET)
249         (TPF_TRACE_EPILOGUE_SKIP_TARGET): New macro definitions.
250         * config/s390/tpf.md ("prologue_tpf", "epilogue_tpf"): Add two new
251         operands for the check flag and the branch target.
252         * config/s390/tpf.opt ("mtpf-trace-hook-prologue-check")
253         ("mtpf-trace-hook-prologue-target")
254         ("mtpf-trace-hook-epilogue-check")
255         ("mtpf-trace-hook-epilogue-target", "mtpf-trace-skip"): New
256         options.
257         * doc/invoke.texi: Document -mtpf-trace-skip option. The other
258         options are for debugging purposes and will not be documented
259         here.
261 2020-03-04  Jakub Jelinek  <jakub@redhat.com>
263         PR debug/93888
264         * tree-inline.c (copy_decl_to_var): Copy DECL_BY_REFERENCE flag.
266         * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Add offseti
267         argument.  Change pd argument so that it can be modified.  Turn
268         constant non-CONSTRUCTOR store into non-constant if it is too large.
269         Adjust offset and size of CONSTRUCTOR or non-constant store to avoid
270         overflows.
271         (vn_walk_cb_data::vn_walk_cb_data, vn_reference_lookup_3): Adjust
272         callers.
274 2020-02-04  Richard Biener  <rguenther@suse.de>
276         PR tree-optimization/93964
277         * graphite-isl-ast-to-gimple.c
278         (gcc_expression_from_isl_ast_expr_id): Add intermediate
279         conversion for pointer to integer converts.
280         * graphite-scop-detection.c (assign_parameter_index_in_region):
281         Relax assert.
283 2020-03-04  Martin Liska  <mliska@suse.cz>
285         PR c/93886
286         PR c/93887
287         * doc/invoke.texi: Clarify --help=language and --help=common
288         interaction.
290 2020-03-04  Jakub Jelinek  <jakub@redhat.com>
292         PR tree-optimization/94001
293         * tree-tailcall.c (process_assignment): Before comparing op1 to
294         *ass_var, verify *ass_var is non-NULL.
296 2020-03-04  Kito Cheng  <kito.cheng@sifive.com>
298         PR target/93995
299         * config/riscv/riscv.c (riscv_emit_float_compare): Using NE to compare
300         the result of IOR.
302 2020-03-03  Dennis Zhang  <dennis.zhang@arm.com>
304         * config/arm/arm_bf16.h (vcvtah_f32_bf16, vcvth_bf16_f32): New.
305         * config/arm/arm_neon.h (vcvt_f32_bf16, vcvtq_low_f32_bf16): New.
306         (vcvtq_high_f32_bf16, vcvt_bf16_f32): New.
307         (vcvtq_low_bf16_f32, vcvtq_high_bf16_f32): New.
308         * config/arm/arm_neon_builtins.def (vbfcvt, vbfcvt_high): New entries.
309         (vbfcvtv4sf, vbfcvtv4sf_high): Likewise.
310         * config/arm/iterators.md (VBFCVT, VBFCVTM): New mode iterators.
311         (V_bf_low, V_bf_cvt_m): New mode attributes.
312         * config/arm/neon.md (neon_vbfcvtv4sf<VBFCVT:mode>): New.
313         (neon_vbfcvtv4sf_highv8bf, neon_vbfcvtsf): New.
314         (neon_vbfcvt<VBFCVT:mode>, neon_vbfcvt_highv8bf): New.
315         (neon_vbfcvtbf_cvtmode<mode>, neon_vbfcvtbf): New
316         * config/arm/unspecs.md (UNSPEC_BFCVT, UNSPEC_BFCVT_HIG): New.
318 2020-03-03  Jakub Jelinek  <jakub@redhat.com>
320         PR tree-optimization/93582
321         * tree-ssa-sccvn.h (vn_reference_lookup): Add mask argument.
322         * tree-ssa-sccvn.c (struct vn_walk_cb_data): Add mask and masked_result
323         members, initialize them in the constructor and if mask is non-NULL,
324         artificially push_partial_def {} for the portions of the mask that
325         contain zeros.
326         (vn_walk_cb_data::finish): If mask is non-NULL, set masked_result to
327         val and return (void *)-1.  Formatting fix.
328         (vn_reference_lookup_pieces): Adjust vn_walk_cb_data initialization.
329         Formatting fix.
330         (vn_reference_lookup): Add mask argument.  If non-NULL, don't call
331         fully_constant_vn_reference_p nor vn_reference_lookup_1 and return
332         data.mask_result.
333         (visit_nary_op): Handle BIT_AND_EXPR of a memory load and INTEGER_CST
334         mask.
335         (visit_stmt): Formatting fix.
337 2020-03-03  Richard Biener  <rguenther@suse.de>
339         PR tree-optimization/93946
340         * alias.h (refs_same_for_tbaa_p): Declare.
341         * alias.c (refs_same_for_tbaa_p): New function.
342         * tree-ssa-alias.c (ao_ref_alias_set): For a NULL ref return
343         zero.
344         * tree-ssa-scopedtables.h
345         (avail_exprs_stack::lookup_avail_expr): Add output argument
346         giving access to the hashtable entry.
347         * tree-ssa-scopedtables.c (avail_exprs_stack::lookup_avail_expr):
348         Likewise.
349         * tree-ssa-dom.c: Include alias.h.
350         (dom_opt_dom_walker::optimize_stmt): Validate TBAA state before
351         removing redundant store.
352         * tree-ssa-sccvn.h (vn_reference_s::base_set): New member.
353         (ao_ref_init_from_vn_reference): Adjust prototype.
354         (vn_reference_lookup_pieces): Likewise.
355         (vn_reference_insert_pieces): Likewise.
356         * tree-ssa-sccvn.c: Track base alias set in addition to alias
357         set everywhere.
358         (eliminate_dom_walker::eliminate_stmt): Also check base alias
359         set when removing redundant stores.
360         (visit_reference_op_store): Likewise.
361         * dse.c (record_store): Adjust valdity check for redundant
362         store removal.
364 2020-03-03  Jakub Jelinek  <jakub@redhat.com>
366         PR target/26877
367         * config/s390/s390.h (OPTION_DEFAULT_SPECS): Reorder.
369         PR rtl-optimization/94002
370         * explow.c (plus_constant): Punt if cst has VOIDmode and
371         get_pool_mode is different from mode.
373 2020-03-03  Claudiu Zissulescu  <claziss@synopsys.com>
375         * config/arc/arc.c (leigitimate_small_data_address_p): Check if an
376         address has an offset which fits the scalling constraint for a
377         load/store operation.
378         (legitimate_scaled_address_p): Update use
379         leigitimate_small_data_address_p.
380         (arc_print_operand): Likewise.
381         (arc_legitimate_address_p): Likewise.
382         (legitimate_small_data_address_p): Likewise.
384 2020-03-03  Claudiu Zissulescu  <claziss@synopsys.com>
386         * config/arc/arc.md (fmasf4_fpu): Use accl_operand predicate.
387         (fnmasf4_fpu): Likewise.
389 2020-03-03  Claudiu Zissulescu  <claziss@synopsys.com>
391         * config/arc/arc.md (adddi3): Early expand the 64bit operation into
392         32bit ops.
393         (subdi3): Likewise.
394         (adddi3_i): Remove pattern.
395         (subdi3_i): Likewise.
397 2020-03-03  Claudiu Zissulescu  <claziss@synopsys.com>
399         * config/arc/arc.md (eh_return): Add length info.
401 2020-03-02  David Malcolm  <dmalcolm@redhat.com>
403         * doc/invoke.texi (-fanalyzer-show-duplicate-count): New.
405 2020-03-02  David Malcolm  <dmalcolm@redhat.com>
407         * doc/invoke.texi (Static Analyzer Options): Add
408         -Wanalyzer-stale-setjmp-buffer to the list of options enabled
409         by -fanalyzer.
411 2020-03-02  UroÅ¡ Bizjak  <ubizjak@gmail.com>
413         PR target/93997
414         * config/i386/i386.md (movstrict<mode>): Allow only
415         registers with VALID_INT_MODE_P modes.
417 2020-03-02  Andrew Stubbs  <ams@codesourcery.com>
419         * config/gcn/gcn-valu.md (dpp_move<mode>): New.
420         (reduc_insn): Use 'U' and 'B' operand codes.
421         (reduc_<reduc_op>_scal_<mode>): Allow all types.
422         (reduc_<reduc_op>_scal_v64di): Delete.
423         (*<reduc_op>_dpp_shr_<mode>): Allow all 1reg types.
424         (*plus_carry_dpp_shr_v64si): Change to ...
425         (*plus_carry_dpp_shr_<mode>): ... this and allow all 1reg int types.
426         (mov_from_lane63_v64di): Change to ...
427         (mov_from_lane63_<mode>): ... this, and allow all 64-bit modes.
428         * config/gcn/gcn.c (gcn_expand_dpp_shr_insn): Increase buffer size.
429         Support UNSPEC_MOV_DPP_SHR output formats.
430         (gcn_expand_reduc_scalar): Add "use_moves" reductions.
431         Add "use_extends" reductions.
432         (print_operand_address): Add 'I' and 'U' codes.
433         * config/gcn/gcn.md (unspec): Add UNSPEC_MOV_DPP_SHR.
435 2020-03-02  Martin Liska  <mliska@suse.cz>
437         * lto-wrapper.c: Fix typo in comment about
438         C++ standard version.
440 2020-03-01  Martin Sebor  <msebor@redhat.com>
442         PR c++/92721
443         * calls.c (init_attr_rdwr_indices): Correctly handle attribute.
445 2020-03-01  Martin Sebor  <msebor@redhat.com>
447         PR middle-end/93829
448         * tree-ssa-strlen.c (count_nonzero_bytes): Set the size to that
449           of a pointer in the outermost ADDR_EXPRs.
451 2020-02-28  Jeff Law  <law@redhat.com>
453         * config/v850/v850.h (STATIC_CHAIN_REGNUM): Change to r19.
454         * config/v850/v850.c (v850_asm_trampoline_template): Update
455         accordingly.
457 2020-02-28  Michael Meissner  <meissner@linux.ibm.com>
459         PR target/93937
460         * config/rs6000/vsx.md (vsx_extract_<mode>_<VS_scalar>mode_var):
461         Delete insn.
463 2020-02-28  Martin Liska  <mliska@suse.cz>
465         PR other/93965
466         * configure.ac: Improve detection of ld_date by requiring
467         either two dashes or none.
468         * configure: Regenerate.
470 2020-02-28  Vladimir Makarov  <vmakarov@redhat.com>
472         PR rtl-optimization/93564
473         * ira-color.c (assign_hard_reg): Prefer smaller hard regno when we
474         do not honor reg alloc order.
476 2020-02-27  Joel Hutton  <Joel.Hutton@arm.com>
478         PR target/87612
479         * config/aarch64/aarch64.c (aarch64_override_options): Fix
480         misleading warning string.
482 2020-02-27  Martin Sebor  <msebor@redhat.com>
484         * doc/invoke.texi (-Wbuiltin-declaration-mismatch): Fix a typo.
486 2020-02-27  Michael Meissner  <meissner@linux.ibm.com>
488         PR target/93932
489         * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
490         Split the insn into two parts.  This insn only does variable
491         extract from a register.
492         (vsx_extract_<mode>_var_load, VSX_D iterator): New insn, do
493         variable extract from memory.
494         (vsx_extract_v4sf_var): Split the insn into two parts.  This insn
495         only does variable extract from a register.
496         (vsx_extract_v4sf_var_load): New insn, do variable extract from
497         memory.
498         (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Split the insn
499         into two parts.  This insn only does variable extract from a
500         register.
501         (vsx_extract_<mode>_var_load, VSX_EXTRACT_I iterator): New insn,
502         do variable extract from memory.
504 2020-02-27  Martin Jambor  <mjambor@suse.cz>
505             Feng Xue  <fxue@os.amperecomputing.com>
507         PR ipa/93707
508         * ipa-cp.c (same_node_or_its_all_contexts_clone_p): Replaced with
509         new function calls_same_node_or_its_all_contexts_clone_p.
510         (cgraph_edge_brings_value_p): Use it.
511         (cgraph_edge_brings_value_p): Likewise.
512         (self_recursive_pass_through_p): Return false if caller is a clone.
513         (self_recursive_agg_pass_through_p): Likewise.
515 2020-02-27  Jan Hubicka  <hubicka@ucw.cz>
517         PR middle-end/92152
518         * alias.c (ends_tbaa_access_path_p): Break out from ...
519         (component_uses_parent_alias_set_from): ... here.
520         * alias.h (ends_tbaa_access_path_p): Declare.
521         * tree-ssa-alias.c (access_path_may_continue_p): Break out from ...;
522         handle trailing arrays past end of tbaa access path.
523         (aliasing_component_refs_p): ... here; likewise.
524         (nonoverlapping_refs_since_match_p): Track TBAA segment of the access
525         path; disambiguate also past end of it.
526         (nonoverlapping_component_refs_p): Use only TBAA segment of the access
527         path.
529 2020-02-27  Mihail Ionescu  <mihail.ionescu@arm.com>
531         * (__ARM_NUM_LANES, __arm_lane, __arm_lane_q): Move to the
532         beginning of the file.
533         (vcreate_bf16, vcombine_bf16): New.
534         (vdup_n_bf16, vdupq_n_bf16): New.
535         (vdup_lane_bf16, vdup_laneq_bf16): New.
536         (vdupq_lane_bf16, vdupq_laneq_bf16): New.
537         (vduph_lane_bf16, vduph_laneq_bf16): New.
538         (vset_lane_bf16, vsetq_lane_bf16): New.
539         (vget_lane_bf16, vgetq_lane_bf16): New.
540         (vget_high_bf16, vget_low_bf16): New.
541         (vreinterpret_bf16_u8, vreinterpretq_bf16_u8): New.
542         (vreinterpret_bf16_u16, vreinterpretq_bf16_u16): New.
543         (vreinterpret_bf16_u32, vreinterpretq_bf16_u32): New.
544         (vreinterpret_bf16_u64, vreinterpretq_bf16_u64): New.
545         (vreinterpret_bf16_s8, vreinterpretq_bf16_s8): New.
546         (vreinterpret_bf16_s16, vreinterpretq_bf16_s16): New.
547         (vreinterpret_bf16_s32, vreinterpretq_bf16_s32): New.
548         (vreinterpret_bf16_s64, vreinterpretq_bf16_s64): New.
549         (vreinterpret_bf16_p8, vreinterpretq_bf16_p8): New.
550         (vreinterpret_bf16_p16, vreinterpretq_bf16_p16): New.
551         (vreinterpret_bf16_p64, vreinterpretq_bf16_p64): New.
552         (vreinterpret_bf16_f32, vreinterpretq_bf16_f32): New.
553         (vreinterpret_bf16_f64, vreinterpretq_bf16_f64): New.
554         (vreinterpretq_bf16_p128): New.
555         (vreinterpret_s8_bf16, vreinterpretq_s8_bf16): New.
556         (vreinterpret_s16_bf16, vreinterpretq_s16_bf16): New.
557         (vreinterpret_s32_bf16, vreinterpretq_s32_bf16): New.
558         (vreinterpret_s64_bf16, vreinterpretq_s64_bf16): New.
559         (vreinterpret_u8_bf16, vreinterpretq_u8_bf16): New.
560         (vreinterpret_u16_bf16, vreinterpretq_u16_bf16): New.
561         (vreinterpret_u32_bf16, vreinterpretq_u32_bf16): New.
562         (vreinterpret_u64_bf16, vreinterpretq_u64_bf16): New.
563         (vreinterpret_p8_bf16, vreinterpretq_p8_bf16): New.
564         (vreinterpret_p16_bf16, vreinterpretq_p16_bf16): New.
565         (vreinterpret_p64_bf16, vreinterpretq_p64_bf16): New.
566         (vreinterpret_f32_bf16, vreinterpretq_f32_bf16): New.
567         (vreinterpretq_p128_bf16): New.
568         * config/arm/arm_neon_builtins.def (VDX): Add V4BF.
569         (V_elem): Likewise.
570         (V_elem_l): Likewise.
571         (VD_LANE): Likewise.
572         (VQX) Add V8BF.
573         (V_DOUBLE): Likewise.
574         (VDQX): Add V4BF and V8BF.
575         (V_two_elem, V_three_elem, V_four_elem): Likewise.
576         (V_reg): Likewise.
577         (V_HALF): Likewise.
578         (V_double_vector_mode): Likewise.
579         (V_cmp_result): Likewise.
580         (V_uf_sclr): Likewise.
581         (V_sz_elem): Likewise.
582         (Is_d_reg): Likewise.
583         (V_mode_nunits): Likewise.
584         * config/arm/neon.md (neon_vdup_lane): Enable for BFloat16.
586 2020-02-27  Andrew Stubbs  <ams@codesourcery.com>
588         * config/gcn/gcn-valu.md (VEC_SUBDWORD_MODE): New mode iterator.
589         (<expander><mode>2<exec>): Change modes to VEC_ALL1REG_INT_MODE.
590         (<expander><mode>3<exec>): Likewise.
591         (<expander><mode>3): New.
592         (v<expander><mode>3): New.
593         (<expander><mode>3): New.
594         (<expander><mode>3<exec>): Rename to ...
595         (<expander>v64si3<exec>): ... this, and change modes to V64SI.
596         * config/gcn/gcn.md (mnemonic): Use '%B' for not.
598 2020-02-27  Alexandre Oliva <oliva@adacore.com>
600         * config/vx-common.h (NO_DOLLAR_IN_LABEL, NO_DOT_IN_LABEL): Leave
601         them alone on vx7.
603 2020-02-27  Richard Biener  <rguenther@suse.de>
605         PR tree-optimization/93508
606         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle _CHK like
607         non-_CHK variants.  Valueize their length arguments.
609 2020-02-27  Richard Biener  <rguenther@suse.de>
611         PR tree-optimization/93953
612         * tree-vect-slp.c (slp_copy_subtree): Avoid keeping a reference
613         to the hash-map entry.
615 2020-02-27  Andrew Stubbs  <ams@codesourcery.com>
617         * config/gcn/gcn.md (mov<mode>): Add transformations for BI subregs.
619 2020-02-27  Mark Williams  <mwilliams@fb.com>
621         * dwarf2out.c (file_name_acquire): Call remap_debug_filename.
622         * lto-opts.c (lto_write_options): Drop -fdebug-prefix-map,
623         -ffile-prefix-map and -fmacro-prefix-map.
624         * lto-streamer-out.c: Include file-prefix-map.h.
625         (lto_output_location): Remap the file part of locations.
627 2020-02-27  Jakub Jelinek  <jakub@redhat.com>
629         PR c/93949
630         * gimplify.c (gimplify_init_constructor): Don't promote readonly
631         DECL_REGISTER variables to TREE_STATIC.
633         PR tree-optimization/93582
634         PR tree-optimization/93945
635         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle memset with
636         non-zero INTEGER_CST second argument and ref->offset or ref->size
637         not a multiple of BITS_PER_UNIT.
639 2020-02-27  Jonathan Wakely  <jwakely@redhat.com>
641         * doc/install.texi (Binaries): Update description of BullFreeware.
643 2020-02-26  Sandra Loosemore  <sandra@codesourcery.com>
645         PR c++/90467
647         * doc/invoke.texi (Option Summary): Re-alphabetize warnings in
648         C++ Language Options, Warning Options, and Static Analyzer
649         Options lists.  Document negative form of options enabled by
650         default.  Move some things around to more accurately sort
651         warnings by category.
652         (C++ Dialect Options, Warning Options, Static Analyzer
653         Options): Document negative form of options when enabled by
654         default.  Move some things around to more accurately sort
655         warnings by category.  Add some missing index entries.
656         Light copy-editing.
658 2020-02-26  Carl Love  <cel@us.ibm.com>
660         PR target/91276
661         * doc/extend.texi (PowerPC AltiVec Built-in Functions available on
662         ISA 2.07): The builtin-function name __builtin_crypto_vpmsumb is only
663         for the vector unsigned short arguments.  It is also listed as the
664         name of the built-in for arguments vector unsigned short,
665         vector unsigned int and vector unsigned long long built-ins.  The
666         name of the builtins for these arguments should be:
667         __builtin_crypto_vpmsumh, __builtin_crypto_vpmsumw and
668         __builtin_crypto_vpmsumd respectively.
670 2020-02-26  Richard Biener  <rguenther@suse.de>
672         * tree-vect-slp.c (vect_print_slp_tree): Also dump ref count
673         and load permutation.
675 2020-02-26  Richard Sandiford  <richard.sandiford@arm.com>
677         PR middle-end/93843
678         * optabs-tree.c (supportable_convert_operation): Reject types with
679         scalar modes.
681 2020-02-26  David Malcolm  <dmalcolm@redhat.com>
683         * Makefile.in (ANALYZER_OBJS): Add analyzer/bar-chart.o.
685 2020-02-26  Jakub Jelinek  <jakub@redhat.com>
687         PR tree-optimization/93820
688         * gimple-ssa-store-merging.c (check_no_overlap): Change RHS_CODE
689         argument to ALL_INTEGER_CST_P boolean.
690         (imm_store_chain_info::try_coalesce_bswap): Adjust caller.
691         (imm_store_chain_info::coalesce_immediate_stores): Likewise.  Handle
692         adjacent INTEGER_CST store into merged_store->only_constants like
693         overlapping one.
695 2020-02-25  Jakub Jelinek  <jakub@redhat.com>
697         PR other/93912
698         * config/sh/sh.c (expand_cbranchdi4): Fix comment typo, probablity
699         -> probability.
700         * cfghooks.c (verify_flow_info): Likewise.
701         * predict.c (combine_predictions_for_bb): Likewise.
702         * bb-reorder.c (connect_better_edge_p): Likewise.  Fix comment typo,
703         sucessor -> successor.
704         (find_traces_1_round): Fix comment typo, destinarion -> destination.
705         * omp-expand.c (expand_oacc_for): Fix comment typo, sucessors ->
706         successors.
707         * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): Fix dump
708         message typo, sucessors -> successors.
710 2020-02-25  Martin Sebor  <msebor@redhat.com>
712         * doc/extend.texi (attribute access): Correct an example.
714 2020-02-25  Mihail Ionescu  <mihail.ionescu@arm.com>
716         * config/aarch64/aarch64-builtins.c (aarch64_scalar_builtin_types):
717         Add simd_bf.
718         (aarch64_init_simd_builtin_scalar_types): Register simd_bf.
719         (VAR15, VAR16): New.
720         * config/aarch64/iterators.md (VALLDIF): Enable for V4BF and V8BF.
721         (VD): Enable for V4BF.
722         (VDC): Likewise.
723         (VQ): Enable for V8BF.
724         (VQ2): Likewise.
725         (VQ_NO2E): Likewise.
726         (VDBL, Vdbl): Add V4BF.
727         (V_INT_EQUIV, v_int_equiv): Add V4BF and V8BF.
728         * config/aarch64/arm_neon.h (bfloat16x4x2_t): New typedef.
729         (bfloat16x8x2_t): Likewise.
730         (bfloat16x4x3_t): Likewise.
731         (bfloat16x8x3_t): Likewise.
732         (bfloat16x4x4_t): Likewise.
733         (bfloat16x8x4_t): Likewise.
734         (vcombine_bf16): New.
735         (vld1_bf16, vld1_bf16_x2): New.
736         (vld1_bf16_x3, vld1_bf16_x4): New.
737         (vld1q_bf16, vld1q_bf16_x2): New.
738         (vld1q_bf16_x3, vld1q_bf16_x4): New.
739         (vld1_lane_bf16): New.
740         (vld1q_lane_bf16): New.
741         (vld1_dup_bf16): New.
742         (vld1q_dup_bf16): New.
743         (vld2_bf16): New.
744         (vld2q_bf16): New.
745         (vld2_dup_bf16): New.
746         (vld2q_dup_bf16): New.
747         (vld3_bf16): New.
748         (vld3q_bf16): New.
749         (vld3_dup_bf16): New.
750         (vld3q_dup_bf16): New.
751         (vld4_bf16): New.
752         (vld4q_bf16): New.
753         (vld4_dup_bf16): New.
754         (vld4q_dup_bf16): New.
755         (vst1_bf16, vst1_bf16_x2): New.
756         (vst1_bf16_x3, vst1_bf16_x4): New.
757         (vst1q_bf16, vst1q_bf16_x2): New.
758         (vst1q_bf16_x3, vst1q_bf16_x4): New.
759         (vst1_lane_bf16): New.
760         (vst1q_lane_bf16): New.
761         (vst2_bf16): New.
762         (vst2q_bf16): New.
763         (vst3_bf16): New.
764         (vst3q_bf16): New.
765         (vst4_bf16): New.
766         (vst4q_bf16): New.
768 2020-02-25  Mihail Ionescu  <mihail.ionescu@arm.com>
770         * config/aarch64/iterators.md (VDQF_F16) Add V4BF and V8BF.
771         (VALL_F16): Likewise.
772         (VALLDI_F16): Likewise.
773         (Vtype): Likewise.
774         (Vetype): Likewise.
775         (vswap_width_name): Likewise.
776         (VSWAP_WIDTH): Likewise.
777         (Vel): Likewise.
778         (VEL): Likewise.
779         (q): Likewise.
780         * config/aarch64/arm_neon.h (vset_lane_bf16, vsetq_lane_bf16): New.
781         (vget_lane_bf16, vgetq_lane_bf16): New.
782         (vcreate_bf16): New.
783         (vdup_n_bf16, vdupq_n_bf16): New.
784         (vdup_lane_bf16, vdup_laneq_bf16): New.
785         (vdupq_lane_bf16, vdupq_laneq_bf16): New.
786         (vduph_lane_bf16, vduph_laneq_bf16): New.
787         (vreinterpret_bf16_u8, vreinterpretq_bf16_u8): New.
788         (vreinterpret_bf16_u16, vreinterpretq_bf16_u16): New.
789         (vreinterpret_bf16_u32, vreinterpretq_bf16_u32): New.
790         (vreinterpret_bf16_u64, vreinterpretq_bf16_u64): New.
791         (vreinterpret_bf16_s8, vreinterpretq_bf16_s8): New.
792         (vreinterpret_bf16_s16, vreinterpretq_bf16_s16): New.
793         (vreinterpret_bf16_s32, vreinterpretq_bf16_s32): New.
794         (vreinterpret_bf16_s64, vreinterpretq_bf16_s64): New.
795         (vreinterpret_bf16_p8, vreinterpretq_bf16_p8): New.
796         (vreinterpret_bf16_p16, vreinterpretq_bf16_p16): New.
797         (vreinterpret_bf16_p64, vreinterpretq_bf16_p64): New
798         (vreinterpret_bf16_f16, vreinterpretq_bf16_f16): New
799         (vreinterpret_bf16_f32, vreinterpretq_bf16_f32): New.
800         (vreinterpret_bf16_f64, vreinterpretq_bf16_f64): New.
801         (vreinterpretq_bf16_p128): New.
802         (vreinterpret_s8_bf16, vreinterpretq_s8_bf16): New.
803         (vreinterpret_s16_bf16, vreinterpretq_s16_bf16): New.
804         (vreinterpret_s32_bf16, vreinterpretq_s32_bf16): New.
805         (vreinterpret_s64_bf16, vreinterpretq_s64_bf16): New.
806         (vreinterpret_u8_bf16, vreinterpretq_u8_bf16): New.
807         (vreinterpret_u16_bf16, vreinterpretq_u16_bf16): New.
808         (vreinterpret_u32_bf16, vreinterpretq_u32_bf16): New.
809         (vreinterpret_u64_bf16, vreinterpretq_u64_bf16): New.
810         (vreinterpret_p8_bf16, vreinterpretq_p8_bf16): New.
811         (vreinterpret_p16_bf16, vreinterpretq_p16_bf16): New.
812         (vreinterpret_p64_bf16, vreinterpretq_p64_bf16): New.
813         (vreinterpret_f32_bf16, vreinterpretq_f32_bf16): New.
814         (vreinterpret_f64_bf16,vreinterpretq_f64_bf16): New.
815         (vreinterpret_f16_bf16,vreinterpretq_f16_bf16): New.
816         (vreinterpretq_p128_bf16): New.
818 2020-02-25  Dennis Zhang  <dennis.zhang@arm.com>
820         * config/arm/arm_neon.h (vbfdot_f32, vbfdotq_f32): New
821         (vbfdot_lane_f32, vbfdotq_laneq_f32): New.
822         (vbfdot_laneq_f32, vbfdotq_lane_f32): New.
823         * config/arm/arm_neon_builtins.def (vbfdot): New entry.
824         (vbfdot_lanev4bf, vbfdot_lanev8bf): Likewise.
825         * config/arm/iterators.md (VSF2BF): New attribute.
826         * config/arm/neon.md (neon_vbfdot<VCVTF:mode>): New entry.
827         (neon_vbfdot_lanev4bf<VCVTF:mode>): Likewise.
828         (neon_vbfdot_lanev8bf<VCVTF:mode>): Likewise.
830 2020-02-25  Christophe Lyon  <christophe.lyon@linaro.org>
832         * config/arm/arm.md (required_for_purecode): New attribute.
833         (enabled): Handle required_for_purecode.
834         * config/arm/thumb1.md (thumb1_movsi_insn): Add alternative to
835         work with -mpure-code.
837 2020-02-25  Jakub Jelinek  <jakub@redhat.com>
839         PR rtl-optimization/93908
840         * combine.c (find_split_point): For store into ZERO_EXTRACT, and src
841         with mask.
843 2019-02-25  Eric Botcazou  <ebotcazou@adacore.com>
845         * dwarf2out.c (dwarf2out_size_function): Run in early-DWARF mode.
847 2020-02-25  Roman Zhuykov  <zhroma@ispras.ru>
849         * doc/install.texi (--enable-checking): Adjust wording.
851 2020-02-25  Richard Biener  <rguenther@suse.de>
853         PR tree-optimization/93868
854         * tree-vect-slp.c (slp_copy_subtree): New function.
855         (vect_attempt_slp_rearrange_stmts): Copy the SLP tree before
856         re-arranging stmts in it.
858 2020-02-25  Jakub Jelinek  <jakub@redhat.com>
860         PR middle-end/93874
861         * passes.c (pass_manager::dump_passes): Create a cgraph node for the
862         dummy function and remove it at the end.
864         PR translation/93864
865         * config/lm32/lm32.c (lm32_setup_incoming_varargs): Fix comment typo
866         paramter -> parameter.
867         * config/aarch64/aarch64.c (aarch64_is_extend_from_extract): Likewise.
868         * ipa-prop.h (struct ipa_agg_replacement_value): Likewise.
870 2020-02-24  Roman Zhuykov  <zhroma@ispras.ru>
872         * doc/install.texi (--enable-checking): Properly document current
873         behavior.
874         (--enable-stage1-checking): Minor clarification about bootstrap.
876 2020-02-24  David Malcolm  <dmalcolm@redhat.com>
878         PR analyzer/93032
879         * doc/invoke.texi (-Wnanalyzer-tainted-array-index): Note that
880         -fanalyzer-checker=taint is also required.
881         (-fanalyzer-checker=): Note that providing this option enables the
882         given checker, and doing so may be required for checkers that are
883         disabled by default.
885 2020-02-24  David Malcolm  <dmalcolm@redhat.com>
887         * doc/invoke.texi (-fanalyzer-verbosity=): "2" only shows
888         significant control flow events; add a "3" which shows all
889         control flow events; the old "3" becomes "4".
891 2020-02-24  Jakub Jelinek  <jakub@redhat.com>
893         PR tree-optimization/93582
894         * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Consider
895         pd.offset and pd.size to be counted in bits rather than bytes, add
896         support for maxsizei that is not a multiple of BITS_PER_UNIT and
897         handle bitfield stores and loads.
898         (vn_reference_lookup_3): Don't call ranges_known_overlap_p with
899         uncomparable quantities - bytes vs. bits.  Allow push_partial_def
900         on offsets/sizes that aren't multiple of BITS_PER_UNIT and adjust
901         pd.offset/pd.size to be counted in bits rather than bytes.
902         Formatting fix.  Rename shadowed len variable to buflen.
904 2020-02-24  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
905             Kugan Vivekandarajah  <kugan.vivekanandarajah@linaro.org>
907         PR driver/47785
908         * gcc.c (putenv_COLLECT_AS_OPTIONS): New function.
909         (driver::main): Call putenv_COLLECT_AS_OPTIONS.
910         * opts-common.c (parse_options_from_collect_gcc_options): New function.
911         (prepend_xassembler_to_collect_as_options): Likewise.
912         * opts.h (parse_options_from_collect_gcc_options): Declare prototype.
913         (prepend_xassembler_to_collect_as_options): Likewise.
914         * lto-opts.c (lto_write_options): Stream assembler options
915         in COLLECT_AS_OPTIONS.
916         * lto-wrapper.c (xassembler_options_error): New static variable.
917         (get_options_from_collect_gcc_options): Move parsing options code to
918         parse_options_from_collect_gcc_options and call it.
919         (merge_and_complain): Validate -Xassembler options.
920         (append_compiler_options): Handle OPT_Xassembler.
921         (run_gcc): Append command line -Xassembler options to
922         collect_gcc_options.
923         * doc/invoke.texi: Add documentation about using Xassembler
924         options with LTO.
926 2020-02-24  Kito Cheng  <kito.cheng@sifive.com>
928         * config/riscv/riscv.c (riscv_emit_float_compare): Change the code gen
929         for LTGT.
930         (riscv_rtx_costs): Update cost model for LTGT.
932 2020-02-23  Vladimir Makarov  <vmakarov@redhat.com>
934         PR rtl-optimization/93564
935         * ira-color.c (struct update_cost_queue_elem): New member start.
936         (queue_update_cost, get_next_update_cost): Add new arg start.
937         (allocnos_conflict_p): New function.
938         (update_costs_from_allocno): Add new arg conflict_cost_update_p.
939         Add checking conflicts with allocnos_conflict_p.
940         (update_costs_from_prefs, restore_costs_from_copies): Adjust
941         update_costs_from_allocno calls.
942         (update_conflict_hard_regno_costs): Add checking conflicts with
943         allocnos_conflict_p.  Adjust calls of queue_update_cost and
944         get_next_update_cost.
945         (assign_hard_reg): Adjust calls of queue_update_cost.  Add
946         debugging print.
947         (bucket_allocno_compare_func): Restore previous version.
949 2020-02-21  John David Anglin  <danglin@gcc.gnu.org>
951         * gcc/config/pa/pa.c (pa_function_value): Fix check for word and
952         double-word size when handling aggregate return values.
953         * gcc/config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Fix to indicate
954         that homogeneous SFmode and DFmode aggregates are passed and returned
955         in general registers.
957 2020-02-21  Jakub Jelinek  <jakub@redhat.com>
959         PR translation/93759
960         * opts.c (print_filtered_help): Translate help before appending
961         messages to it rather than after that.
963 2020-02-19  Richard Sandiford  <richard.sandiford@arm.com>
965         PR rtl-optimization/PR92989
966         * lra-lives.c (process_bb_lives): Restore the original order
967         of the bb liveness update.  Call make_hard_regno_dead for each
968         register clobbered at the start of an EH receiver.
970 2020-02-18  Feng Xue  <fxue@os.amperecomputing.com>
972         PR ipa/93763
973         * ipa-cp.c (self_recursively_generated_p): Mark self-dependent value as
974         self-recursively generated.
976 2020-02-21  Iain Sandoe  <iain@sandoe.co.uk>
978         PR target/93860
979         * config/darwin-c.c (pop_field_alignment): Adjust quoting of
980         error string.
982 2020-02-21  Mihail Ionescu  <mihail.ionescu@arm.com>
984         * doc/sourcebuild.texi (arm_v8_1m_mve_ok):
985         Document new target supports option.
987 2020-02-21  Dennis Zhang  <dennis.zhang@arm.com>
989         * config/arm/arm_neon.h (vmmlaq_s32, vmmlaq_u32, vusmmlaq_s32): New.
990         * config/arm/arm_neon_builtins.def (smmla, ummla, usmmla): New.
991         * config/arm/iterators.md (MATMUL): New iterator.
992         (sup): Add UNSPEC_MATMUL_S, UNSPEC_MATMUL_U, and UNSPEC_MATMUL_US.
993         (mmla_sfx): New attribute.
994         * config/arm/neon.md (neon_<sup>mmlav16qi): New.
995         * config/arm/unspecs.md (UNSPEC_MATMUL_S, UNSPEC_MATMUL_U): New.
996         (UNSPEC_MATMUL_US): New.
998 2020-02-21  Mihail-Calin Ionescu  <mihail.ionescu@arm.com>
1000         * config/arm/arm.md: Prevent scalar shifts from being used when big
1001         endian is enabled.
1003 2020-02-21  Jan Hubicka  <hubicka@ucw.cz>
1004             Richard Biener  <rguenther@suse.de>
1006         PR tree-optimization/93586
1007         * tree-ssa-alias.c (nonoverlapping_array_refs_p): Finish array walk
1008         after mismatched array refs; do not sure type size information to
1009         recover from unmatched referneces with !flag_strict_aliasing_p.
1011 2020-02-21  Andrew Stubbs  <ams@codesourcery.com>
1013         * config/gcn/gcn-valu.md (gather_load<mode>): Rename to ...
1014         (gather_load<mode>v64si): ... this and set operand 2 to V64SI.
1015         (scatter_store<mode>): Rename to ...
1016         (scatter_store<mode>v64si): ... this and set operand 1 to V64SI.
1017         (scatter<mode>_exec): Delete. Move contents ...
1018         (mask_scatter_store<mode>): ... here, and rename that to ...
1019         (mask_gather_load<mode>v64si): ... this. Set operand 2 to V64SI.
1020         Remove mode conversion.
1021         (mask_gather_load<mode>): Rename to ...
1022         (mask_scatter_store<mode>v64si): ... this. Set operand 1 to V64SI.
1023         Remove mode conversion.
1024         * config/gcn/gcn.c (gcn_expand_scaled_offsets): Remove mode conversion.
1026 2020-02-21  Martin Jambor  <mjambor@suse.cz>
1028         PR tree-optimization/93845
1029         * tree-sra.c (verify_sra_access_forest): Only test access size of
1030         scalar types.
1032 2020-02-21  Andrew Stubbs  <ams@codesourcery.com>
1034         * config/gcn/gcn.c (gcn_hard_regno_mode_ok): Align VGPR pairs.
1035         * config/gcn/gcn-valu.md (addv64di3): Remove early-clobber.
1036         (addv64di3_exec): Likewise.
1037         (subv64di3): Likewise.
1038         (subv64di3_exec): Likewise.
1039         (addv64di3_zext): Likewise.
1040         (addv64di3_zext_exec): Likewise.
1041         (addv64di3_zext_dup): Likewise.
1042         (addv64di3_zext_dup_exec): Likewise.
1043         (addv64di3_zext_dup2): Likewise.
1044         (addv64di3_zext_dup2_exec): Likewise.
1045         (addv64di3_sext_dup2): Likewise.
1046         (addv64di3_sext_dup2_exec): Likewise.
1047         (<expander>v64di3): Likewise.
1048         (<expander>v64di3_exec): Likewise.
1049         (*<reduc_op>_dpp_shr_v64di): Likewise.
1050         (*plus_carry_dpp_shr_v64di): Likewise.
1051         * config/gcn/gcn.md (adddi3): Likewise.
1052         (addptrdi3): Likewise.
1053         (<expander>di3): Likewise.
1055 2020-02-21  Andrew Stubbs  <ams@codesourcery.com>
1057         * config/gcn/gcn-valu.md (vec_seriesv64di): Use gen_vec_duplicatev64di.
1059 2020-02-21  Richard Sandiford  <richard.sandiford@arm.com>
1061         * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Add SVE
1062         support.  Use aarch64_emit_mult instead of emitting multiplication
1063         instructions directly.
1064         * config/aarch64/aarch64-sve.md (sqrt<mode>2, rsqrt<mode>2)
1065         (@aarch64_rsqrte<mode>, @aarch64_rsqrts<mode>): New expanders.
1067 2020-02-21  Richard Sandiford  <richard.sandiford@arm.com>
1069         * config/aarch64/aarch64.c (aarch64_emit_mult): New function.
1070         (aarch64_emit_approx_div): Add SVE support.  Use aarch64_emit_mult
1071         instead of emitting multiplication instructions directly.
1072         * config/aarch64/iterators.md (SVE_COND_FP_BINARY_OPTAB): New iterator.
1073         * config/aarch64/aarch64-sve.md (div<mode>3, @aarch64_frecpe<mode>)
1074         (@aarch64_frecps<mode>): New expanders.
1076 2020-02-21  Richard Sandiford  <richard.sandiford@arm.com>
1078         * config/aarch64/aarch64-protos.h (AARCH64_APPROX_MODE): Operate
1079         on and produce uint64_ts rather than ints.
1080         (AARCH64_APPROX_NONE, AARCH64_APPROX_ALL): Change to uint64_ts.
1081         (cpu_approx_modes): Change the fields from unsigned int to uint64_t.
1083 2020-02-21  Richard Sandiford  <richard.sandiford@arm.com>
1085         * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Don't create
1086         an unused xmsk register when handling approximate rsqrt.
1088 2020-02-21  Richard Sandiford  <richard.sandiford@arm.com>
1090         * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Fix inverted
1091         flag_finite_math_only condition.
1093 2020-02-20  UroÅ¡ Bizjak  <ubizjak@gmail.com>
1095         PR target/93828
1096         * config/i386/mmx.md (*vec_extractv2sf_1): Match source operand
1097         to destination operand for shufps alternative.
1098         (*vec_extractv2si_1): Ditto.
1100 2020-02-20  Peter Bergner  <bergner@linux.ibm.com>
1102         PR target/93658
1103         * config/rs6000/rs6000.c (rs6000_legitimate_address_p): Handle VSX
1104         vector modes.
1106 2020-02-20  Martin Liska  <mliska@suse.cz>
1108         PR translation/93831
1109         * config/darwin.c (darwin_override_options): Change 64b to 64-bit mode.
1111 2020-02-20  Martin Liska  <mliska@suse.cz>
1113         PR translation/93830
1114         * common/config/avr/avr-common.c: Remote trailing "|".
1116 2020-02-19  Bernd Edlinger  <bernd.edlinger@hotmail.de>
1118         * collect2.c (maybe_run_lto_and_relink): Fix typo in
1119         comment.
1121 2020-02-19  Richard Sandiford  <richard.sandiford@arm.com>
1123         PR tree-optimization/93767
1124         * tree-vect-data-refs.c (vect_compile_time_alias): Remove the
1125         access-size bias from the offset calculations for negative strides.
1127 2020-02-19  Bernd Edlinger  <bernd.edlinger@hotmail.de>
1129         * collect2.c (c_file, o_file): Make const again.
1130         (ldout,lderrout, dump_ld_file): Remove.
1131         (tool_cleanup): Avoid calling not signal-safe functions.
1132         (maybe_run_lto_and_relink): Avoid possible signal handler
1133         access to unintialzed memory (lto_o_files).
1134         (main): Avoid leaking temp files in $TMPDIR.
1135         Initialize c_file/o_file with concat, which avoids exposing
1136         uninitialized memory to signal handler, which calls unlink(!).
1137         Avoid calling maybe_unlink when the main function returns,
1138         since the atexit handler is already doing this.
1139         * collect2.h (dump_ld_file, ldout, lderrout): Remove.
1141 2020-02-19  Martin Jambor  <mjambor@suse.cz>
1143         PR tree-optimization/93776
1144         * tree-sra.c (create_access): Do not create zero size accesses.
1145         (get_access_for_expr): Do not search for zero sized accesses.
1147 2020-02-19  Martin Jambor  <mjambor@suse.cz>
1149         PR tree-optimization/93667
1150         * tree-sra.c (scalarizable_type_p): Return false if record fields
1151         do not follow wach other.
1153 2020-01-21  Kito Cheng  <kito.cheng@sifive.com>
1155         * config/riscv/riscv.c (riscv_output_move) Using fmv.x.w/fmv.w.x
1156         rather than fmv.x.s/fmv.s.x.
1158 2020-02-18  James Greenhalgh  <james.greenhalgh@arm.com>
1160         * config/aarch64/aarch64-simd-builtins.def
1161         (intrinsic_vec_smult_lo_): New.
1162         (intrinsic_vec_umult_lo_): Likewise.
1163         (vec_widen_smult_hi_): Likewise.
1164         (vec_widen_umult_hi_): Likewise.
1165         * config/aarch64/aarch64-simd.md
1166         (aarch64_intrinsic_vec_<su>mult_lo_<mode>): New.
1167         * config/aarch64/arm_neon.h (vmull_high_s8): Use intrinsics.
1168         (vmull_high_s16): Likewise.
1169         (vmull_high_s32): Likewise.
1170         (vmull_high_u8): Likewise.
1171         (vmull_high_u16): Likewise.
1172         (vmull_high_u32): Likewise.
1173         (vmull_s8): Likewise.
1174         (vmull_s16): Likewise.
1175         (vmull_s32): Likewise.
1176         (vmull_u8): Likewise.
1177         (vmull_u16): Likewise.
1178         (vmull_u32): Likewise.
1180 2020-02-18  Martin Liska  <mliska@suse.cz>
1182         * value-prof.c (stream_out_histogram_value): Restore LTO PGO
1183         bootstrap by missing removal of invalid sanity check.
1185 2020-02-18  Martin Liska  <mliska@suse.cz>
1187         PR ipa/92518
1188         * ipa-icf-gimple.c (func_checker::compare_gimple_assign):
1189         Always compare LHS of gimple_assign.
1191 2020-02-18  Martin Liska  <mliska@suse.cz>
1193         PR ipa/93583
1194         * cgraph.c (cgraph_node::verify_node): Verify MALLOC attribute
1195         and return type of functions.
1196         * ipa-param-manipulation.c (ipa_param_adjustments::adjust_decl):
1197         Drop MALLOC attribute for void functions.
1198         * ipa-pure-const.c (funct_state_summary_t::duplicate): Drop
1199         malloc_state for a new VOID clone.
1201 2020-02-18  Martin Liska  <mliska@suse.cz>
1203         PR ipa/92924
1204         * common.opt: Add -fprofile-reproducibility.
1205         * doc/invoke.texi: Document it.
1206         * value-prof.c (dump_histogram_value):
1207         Document and support behavior for counters[0]
1208         being a negative value.
1209         (get_nth_most_common_value): Handle negative
1210         counters[0] in respect to flag_profile_reproducible.
1212 2020-02-18  Jakub Jelinek  <jakub@redhat.com>
1214         PR ipa/93797
1215         * cgraph.c (verify_speculative_call): Use speculative_id instead of
1216         speculative_uid in messages.  Remove trailing whitespace from error
1217         message.  Use num_speculative_call_targets instead of
1218         num_speculative_targets in a message.
1219         (cgraph_node::verify_node): Use call_stmt instead of cal_stmt in
1220         edge messages and stmt instead of cal_stmt in reference message.
1222         PR tree-optimization/93780
1223         * tree-ssa.c (non_rewritable_lvalue_p): Check valid_vector_subparts_p
1224         before calling build_vector_type.
1225         (execute_update_addresses_taken): Likewise.
1227         PR driver/93796
1228         * params.opt (-param=ipa-max-switch-predicate-bounds=): Fix help
1229         typo, functoin -> function.
1230         * tree.c (free_lang_data_in_decl): Fix comment typo,
1231         functoin -> function.
1232         * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
1234 2020-02-17  David Malcolm  <dmalcolm@redhat.com>
1236         * diagnostic.c (print_any_cwe): Don't call get_cwe_url if URLs
1237         won't be printed.
1238         (print_option_information): Don't call get_option_url if URLs
1239         won't be printed.
1241 2020-02-17  Alexandre Oliva  <oliva@adacore.com>
1243         * tree-emutls.c (new_emutls_decl, emutls_common_1): Complete
1244         handling of register_common-less targets.
1246 2020-02-17  Martin Liska  <mliska@suse.cz>
1248         PR ipa/93760
1249         * ipa-devirt.c (odr_types_equivalent_p): Fix grammar.
1251 2020-02-17  Martin Liska  <mliska@suse.cz>
1253         PR translation/93755
1254         * config/rs6000/rs6000.c (rs6000_option_override_internal):
1255         Fix double quotes.
1257 2020-02-17  Martin Liska  <mliska@suse.cz>
1259         PR other/93756
1260         * config/rx/elf.opt: Fix typo.
1262 2020-02-17  Richard Biener  <rguenther@suse.de>
1264         PR c/86134
1265         * opts-global.c (print_ignored_options): Use inform and
1266         amend message.
1268 2020-02-17  Jiufu Guo  <guojiufu@linux.ibm.com>
1270         PR target/93047
1271         * config/rs6000/rs6000.md (untyped_call): Add emit_clobber.
1273 2020-02-16  UroÅ¡ Bizjak  <ubizjak@gmail.com>
1275         PR target/93743
1276         * config/i386/i386.md (atan2xf3): Swap operands 1 and 2.
1277         (atan2<mode>3): Update operand order in the call to gen_atan2xf3.
1279 2020-02-15  Jason Merrill  <jason@redhat.com>
1281         * doc/invoke.texi (C Dialect Options): Add -std=c++20.
1283 2020-02-15  Jakub Jelinek  <jakub@redhat.com>
1285         PR tree-optimization/93744
1286         * match.pd (((m1 >/</>=/<= m2) * d -> (m1 >/</>=/<= m2) ? d : 0,
1287         A - ((A - B) & -(C cmp D)) -> (C cmp D) ? B : A,
1288         A + ((B - A) & -(C cmp D)) -> (C cmp D) ? B : A): For GENERIC, make
1289         sure @2 in the first and @1 in the other patterns has no side-effects.
1291 2020-02-15  David Malcolm  <dmalcolm@redhat.com>
1292             Bernd Edlinger  <bernd.edlinger@hotmail.de>
1294         PR 87488
1295         PR other/93168
1296         * config.in (DIAGNOSTICS_URLS_DEFAULT): New define.
1297         * configure.ac (--with-diagnostics-urls): New configuration
1298         option, based on --with-diagnostics-color.
1299         (DIAGNOSTICS_URLS_DEFAULT): New define.
1300         * config.h: Regenerate.
1301         * configure: Regenerate.
1302         * diagnostic.c (diagnostic_urls_init): Handle -1 for
1303         DIAGNOSTICS_URLS_DEFAULT from configure-time
1304         --with-diagnostics-urls=auto-if-env by querying for a GCC_URLS
1305         and TERM_URLS environment variable.
1306         * diagnostic-url.h (diagnostic_url_format): New enum type.
1307         (diagnostic_urls_enabled_p): rename to...
1308         (determine_url_format): ... this, and change return type.
1309         * diagnostic-color.c (parse_env_vars_for_urls): New helper function.
1310         (auto_enable_urls): Disable URLs on xfce4-terminal, gnome-terminal,
1311         the linux console, and mingw.
1312         (diagnostic_urls_enabled_p): rename to...
1313         (determine_url_format): ... this, and adjust.
1314         * pretty-print.h (pretty_printer::show_urls): rename to...
1315         (pretty_printer::url_format): ... this, and change to enum.
1316         * pretty-print.c (pretty_printer::pretty_printer,
1317         pp_begin_url, pp_end_url, test_urls): Adjust.
1318         * doc/install.texi (--with-diagnostics-urls): Document the new
1319         configuration option.
1320         (--with-diagnostics-color): Document the existing interaction
1321         with GCC_COLORS better.
1322         * doc/invoke.texi (-fdiagnostics-urls): Add GCC_URLS and TERM_URLS
1323         vindex reference.  Update description of defaults based on the above.
1324         (-fdiagnostics-color): Update description of how -fdiagnostics-color
1325         interacts with GCC_COLORS.
1327 2020-02-14  Eric Botcazou  <ebotcazou@adacore.com>
1329         PR target/93704
1330         * config/sparc/sparc.c (eligible_for_call_delay): Test HAVE_GNU_LD in
1331         conjunction with TARGET_GNU_TLS in early return.
1333 2020-02-14  Alexander Monakov  <amonakov@ispras.ru>
1335         * rtlanal.c (rtx_cost): Handle a SET up front. Avoid division if
1336         the mode is not wider than UNITS_PER_WORD.
1338 2020-02-14  Martin Jambor  <mjambor@suse.cz>
1340         PR tree-optimization/93516
1341         * tree-sra.c (propagate_subaccesses_from_rhs): Do not create
1342         access of the same type as the parent.
1343         (propagate_subaccesses_from_lhs): Likewise.
1345 2020-02-14 Hongtao Liu  <hongtao.liu@intel.com>
1347         PR target/93724
1348         * config/i386/avx512vbmi2intrin.h
1349         (_mm512_shrdi_epi16, _mm512_mask_shrdi_epi16,
1350         _mm512_maskz_shrdi_epi16, _mm512_shrdi_epi32,
1351         _mm512_mask_shrdi_epi32, _mm512_maskz_shrdi_epi32,
1352         _m512_shrdi_epi64, _m512_mask_shrdi_epi64,
1353         _m512_maskz_shrdi_epi64, _mm512_shldi_epi16,
1354         _mm512_mask_shldi_epi16, _mm512_maskz_shldi_epi16,
1355         _mm512_shldi_epi32, _mm512_mask_shldi_epi32,
1356         _mm512_maskz_shldi_epi32, _mm512_shldi_epi64,
1357         _mm512_mask_shldi_epi64, _mm512_maskz_shldi_epi64): Fix typo
1358         of lacking a closing parenthesis.
1359         * config/i386/avx512vbmi2vlintrin.h
1360         (_mm256_shrdi_epi16, _mm256_mask_shrdi_epi16,
1361         _mm256_maskz_shrdi_epi16, _mm256_shrdi_epi32,
1362         _mm256_mask_shrdi_epi32, _mm256_maskz_shrdi_epi32,
1363         _m256_shrdi_epi64, _m256_mask_shrdi_epi64,
1364         _m256_maskz_shrdi_epi64, _mm256_shldi_epi16,
1365         _mm256_mask_shldi_epi16, _mm256_maskz_shldi_epi16,
1366         _mm256_shldi_epi32, _mm256_mask_shldi_epi32,
1367         _mm256_maskz_shldi_epi32, _mm256_shldi_epi64,
1368         _mm256_mask_shldi_epi64, _mm256_maskz_shldi_epi64,
1369         _mm_shrdi_epi16, _mm_mask_shrdi_epi16,
1370         _mm_maskz_shrdi_epi16, _mm_shrdi_epi32,
1371         _mm_mask_shrdi_epi32, _mm_maskz_shrdi_epi32,
1372         _mm_shrdi_epi64, _mm_mask_shrdi_epi64,
1373         _m_maskz_shrdi_epi64, _mm_shldi_epi16,
1374         _mm_mask_shldi_epi16, _mm_maskz_shldi_epi16,
1375         _mm_shldi_epi32, _mm_mask_shldi_epi32,
1376         _mm_maskz_shldi_epi32, _mm_shldi_epi64,
1377         _mm_mask_shldi_epi64, _mm_maskz_shldi_epi64): Ditto.
1379 2020-02-13  H.J. Lu  <hongjiu.lu@intel.com>
1381         PR target/93656
1382         * config/i386/i386.c (ix86_trampoline_init): Skip ENDBR32 at
1383         the target function entry.
1385 2020-02-13  Claudiu Zissulescu  <claziss@synopsys.com>
1387         * common/config/arc/arc-common.c (arc_option_optimization_table):
1388         Disable if-conversion step when optimized for size.
1390 2020-02-13  Claudiu Zissulescu  <claziss@synopsys.com>
1392         * config/arc/arc.c (arc_conditional_register_usage): R0-R3 and
1393         R12-R15 are always in ARCOMPACT16_REGS register class.
1394         * config/arc/arc.opt (mq-class): Deprecate.
1395         * config/arc/constraint.md ("q"): Remove dependency on mq-class
1396         option.
1397         * doc/invoke.texi (mq-class): Update text.
1398         * common/config/arc/arc-common.c (arc_option_optimization_table):
1399         Update list.
1401 2020-02-13  Claudiu Zissulescu  <claziss@synopsys.com>
1403         * config/arc/arc.c (arc_insn_cost): New function.
1404         (TARGET_INSN_COST): Define.
1405         * config/arc/arc.md (cost): New attribute.
1406         (add_n): Use arc_nonmemory_operand.
1407         (ashlsi3_insn): Likewise, also update constraints.
1408         (ashrsi3_insn): Likewise.
1409         (rotrsi3): Likewise.
1410         (add_shift): Likewise.
1411         * config/arc/predicates.md (arc_nonmemory_operand): New predicate.
1413 2020-02-13  Claudiu Zissulescu  <claziss@synopsys.com>
1415         * config/arc/arc.md (mulsidi_600): Correctly select mlo/mhi
1416         registers.
1417         (umulsidi_600): Likewise.
1419 2020-02-13  Jakub Jelinek  <jakub@redhat.com>
1421         PR target/93696
1422         * config/i386/avx512bitalgintrin.h (_mm512_mask_popcnt_epi8,
1423         _mm512_mask_popcnt_epi16, _mm256_mask_popcnt_epi8,
1424         _mm256_mask_popcnt_epi16, _mm_mask_popcnt_epi8,
1425         _mm_mask_popcnt_epi16): Rename __B argument to __A and __A to __W,
1426         pass __A to the builtin followed by __W instead of __A followed by
1427         __B.
1428         * config/i386/avx512vpopcntdqintrin.h (_mm512_mask_popcnt_epi32,
1429         _mm512_mask_popcnt_epi64): Likewise.
1430         * config/i386/avx512vpopcntdqvlintrin.h (_mm_mask_popcnt_epi32,
1431         _mm256_mask_popcnt_epi32, _mm_mask_popcnt_epi64,
1432         _mm256_mask_popcnt_epi64): Likewise.
1434         PR tree-optimization/93582
1435         * fold-const.h (shift_bytes_in_array_left,
1436         shift_bytes_in_array_right): Declare.
1437         * fold-const.c (shift_bytes_in_array_left,
1438         shift_bytes_in_array_right): New function, moved from
1439         gimple-ssa-store-merging.c, no longer static.
1440         * gimple-ssa-store-merging.c (shift_bytes_in_array): Move
1441         to gimple-ssa-store-merging.c and rename to shift_bytes_in_array_left.
1442         (shift_bytes_in_array_right): Move to gimple-ssa-store-merging.c.
1443         (encode_tree_to_bitpos): Use shift_bytes_in_array_left instead of
1444         shift_bytes_in_array.
1445         (verify_shift_bytes_in_array): Rename to ...
1446         (verify_shift_bytes_in_array_left): ... this.  Use
1447         shift_bytes_in_array_left instead of shift_bytes_in_array.
1448         (store_merging_c_tests): Call verify_shift_bytes_in_array_left
1449         instead of verify_shift_bytes_in_array.
1450         * tree-ssa-sccvn.c (vn_reference_lookup_3): For native_encode_expr
1451         / native_interpret_expr where the store covers all needed bits,
1452         punt on PDP-endian, otherwise allow all involved offsets and sizes
1453         not to be byte-aligned.
1455         PR target/93673
1456         * config/i386/sse.md (k<code><mode>): Drop mode from last operand and
1457         use const_0_to_255_operand predicate instead of immediate_operand.
1458         (avx512dq_fpclass<mode><mask_scalar_merge_name>,
1459         avx512dq_vmfpclass<mode><mask_scalar_merge_name>,
1460         vgf2p8affineinvqb_<mode><mask_name>,
1461         vgf2p8affineqb_<mode><mask_name>): Drop mode from
1462         const_0_to_255_operand predicated operands.
1464 2020-02-12  Jeff Law  <law@redhat.com>
1466         * config/h8300/h8300.md (comparison shortening peepholes): Use
1467         a mode iterator to merge the HImode and SImode peepholes.
1469 2020-02-12  Jakub Jelinek  <jakub@redhat.com>
1471         PR middle-end/93663
1472         * real.c (is_even): Make static.  Function comment fix.
1473         (is_halfway_below): Make static, don't assert R is not inf/nan,
1474         instead return false for those.  Small formatting fixes.
1476 2020-02-12  Martin Sebor  <msebor@redhat.com>
1478         PR middle-end/93646
1479         * tree-ssa-strlen.c (handle_builtin_stxncpy): Rename...
1480         (handle_builtin_stxncpy_strncat): ...to this.  Change first argument.
1481         Issue only -Wstringop-overflow strncat, never -Wstringop-truncation.
1482         (strlen_check_and_optimize_call): Adjust callee name.
1484 2020-02-12  Jeff Law  <law@redhat.com>
1486         * config/h8300/h8300.md (comparison shortening peepholes): Drop
1487         (and (xor)) variant.  Combine other two into single peephole.
1489 2020-02-12  Wilco Dijkstra  <wdijkstr@arm.com>
1491         PR rtl-optimization/93565
1492         * config/aarch64/aarch64.c (aarch64_rtx_costs): Add CTZ costs.
1494 2020-02-12  Wilco Dijkstra  <wdijkstr@arm.com>
1496         * config/aarch64/aarch64-simd.md
1497         (aarch64_zero_extend<GPI:mode>_reduc_plus_<VDQV_E:mode>): New pattern.
1498         * config/aarch64/aarch64.md (popcount<mode>2): Use it instead of
1499         generating separate ADDV and zero_extend patterns.
1500         * config/aarch64/iterators.md (VDQV_E): New iterator.
1502 2020-02-12  Jeff Law  <law@redhat.com>
1504         * config/h8300/h8300.md (cpymemsi, movmd): Remove dead patterns,
1505         expanders, splits, etc.
1506         (movmd_internal_<mode>, movmd splitter, movstr, movsd): Likewise.
1507         (stpcpy_internal_<mode>, stpcpy splitter): Likewise.
1508         (peepholes to convert QI/HI mode pushes to SI mode pushes): Likewise.
1509         * config/h8300/h8300.c (h8300_swap_into_er6): Remove unused function.
1510         (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise
1511         * config/h8300/h8300-protos.h (h8300_swap_into_er6): Remove unused
1512         function prototype.
1513         (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise.
1515 2020-02-12  Jakub Jelinek  <jakub@redhat.com>
1517         PR target/93670
1518         * config/i386/sse.md (VI48F_256_DQ): New mode iterator.
1519         (avx512vl_vextractf128<mode>): Use it instead of VI48F_256.  Remove
1520         TARGET_AVX512DQ from condition.
1521         (vec_extract_lo_<mode><mask_name>): Use <mask_avx512dq_condition>
1522         instead of <mask_mode512bit_condition> in condition.  If
1523         TARGET_AVX512DQ is false, emit vextract*64x4 instead of
1524         vextract*32x8.
1525         (vec_extract_lo_<mode><mask_name>): Drop <mask_avx512dq_condition>
1526         from condition.
1528 2020-02-12  Kewen Lin  <linkw@gcc.gnu.org>
1530         PR target/91052
1531         * ira.c (combine_and_move_insns): Skip multiple_sets def_insn.
1533 2020-02-12  Segher Boessenkool  <segher@kernel.crashing.org>
1535         * config/rs6000/rs6000.c (rs6000_debug_print_mode): Don't use sizeof
1536         where strlen is more legible.
1537         (rs6000_builtin_vectorized_libmass): Ditto.
1538         (rs6000_print_options_internal): Ditto.
1540 2020-02-11  Martin Sebor  <msebor@redhat.com>
1542         PR tree-optimization/93683
1543         * tree-ssa-alias.c (stmt_kills_ref_p): Avoid using LHS when not set.
1545 2020-02-11  Michael Meissner  <meissner@linux.ibm.com>
1547         * config/rs6000/predicates.md (cint34_operand): Rename the
1548         -mprefixed-addr option to be -mprefixed.
1549         * config/rs6000/rs6000-cpus.def (ISA_FUTURE_MASKS_SERVER): Rename
1550         the -mprefixed-addr option to be -mprefixed.
1551         (OTHER_FUTURE_MASKS): Likewise.
1552         (POWERPC_MASKS): Likewise.
1553         * config/rs6000/rs6000.c (rs6000_option_override_internal): Rename
1554         the -mprefixed-addr option to be -mprefixed.  Change error
1555         messages to refer to -mprefixed.
1556         (num_insns_constant_gpr): Rename the -mprefixed-addr option to be
1557         -mprefixed.
1558         (rs6000_legitimate_offset_address_p): Likewise.
1559         (rs6000_mode_dependent_address): Likewise.
1560         (rs6000_opt_masks): Change the spelling of "-mprefixed-addr" to be
1561         "-mprefixed" for target attributes and pragmas.
1562         (address_to_insn_form): Rename the -mprefixed-addr option to be
1563         -mprefixed.
1564         (rs6000_adjust_insn_length): Likewise.
1565         * config/rs6000/rs6000.h (FINAL_PRESCAN_INSN): Rename the
1566         -mprefixed-addr option to be -mprefixed.
1567         (ASM_OUTPUT_OPCODE): Likewise.
1568         * config/rs6000/rs6000.md (prefixed insn attribute): Rename the
1569         -mprefixed-addr option to be -mprefixed.
1570         * config/rs6000/rs6000.opt (-mprefixed): Rename the
1571         -mprefixed-addr option to be prefixed.  Change the option from
1572         being undocumented to being documented.
1573         * doc/invoke.texi (RS/6000 and PowerPC Options): Document the
1574         -mprefixed option.  Update the -mpcrel documentation to mention
1575         -mprefixed.
1577 2020-02-11  Hans-Peter Nilsson  <hp@axis.com>
1579         * ira-conflicts.c (print_hard_reg_set): Correct output for sets
1580         including FIRST_PSEUDO_REGISTER - 1.
1581         * ira-color.c (print_hard_reg_set): Ditto.
1583 2020-02-11  Stam Markianos-Wright  <stam.markianos-wright@arm.com>
1585         * config/arm/arm-builtins.c (enum arm_type_qualifiers): 
1586         (USTERNOP_QUALIFIERS): New define.
1587         (USMAC_LANE_QUADTUP_QUALIFIERS): New define.
1588         (SUMAC_LANE_QUADTUP_QUALIFIERS): New define.
1589         (arm_expand_builtin_args): Add case ARG_BUILTIN_LANE_QUADTUP_INDEX.
1590         (arm_expand_builtin_1): Add qualifier_lane_quadtup_index.
1591         * config/arm/arm_neon.h (vusdot_s32): New.
1592         (vusdot_lane_s32): New.
1593         (vusdotq_lane_s32): New.
1594         (vsudot_lane_s32): New.
1595         (vsudotq_lane_s32): New.
1596         * config/arm/arm_neon_builtins.def (usdot, usdot_lane,sudot_lane): New.
1597         * config/arm/iterators.md (DOTPROD_I8MM): New.
1598         (sup, opsuffix): Add <us/su>.
1599         * config/arm/neon.md (neon_usdot, <us/su>dot_lane: New.
1600         * config/arm/unspecs.md (UNSPEC_DOT_US, UNSPEC_DOT_SU): New.
1602 2020-02-11  Richard Biener  <rguenther@suse.de>
1604         PR tree-optimization/93661
1605         PR tree-optimization/93662
1606         * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly guard
1607         tree_to_poly_int64.
1608         * tree-sra.c (get_access_for_expr): Likewise.
1610 2020-02-10  Jakub Jelinek  <jakub@redhat.com>
1612         PR target/93637
1613         * config/i386/sse.md (VI_256_AVX2): New mode iterator.
1614         (vcond_mask_<mode><sseintvecmodelower>): Use it instead of VI_256.
1615         Change condition from TARGET_AVX2 to TARGET_AVX.
1617 2020-02-10  Iain Sandoe  <iain@sandoe.co.uk>
1619         PR other/93641
1620         * config/darwin-c.c (darwin_cfstring_ref_p): Fix up last
1621         argument of strncmp.
1623 2020-02-10  Hans-Peter Nilsson  <hp@axis.com>
1625         Try to generate zero-based comparisons.
1626         * config/cris/cris.c (cris_reduce_compare): New function.
1627         * config/cris/cris-protos.h  (cris_reduce_compare): Add prototype.
1628         * config/cris/cris.md ("cbranch<mode>4", "cbranchdi4", "cstoredi4")
1629         (cstore<mode>4"): Apply cris_reduce_compare in expanders.
1631 2020-02-10  Richard Earnshaw  <rearnsha@arm.com>
1633         PR target/91913
1634         * config/arm/arm.md (movsi_compare0): Allow SP as a source register
1635         in Thumb state and also as a destination in Arm state.  Add T16
1636         variants.
1638 2020-02-10  Hans-Peter Nilsson  <hp@axis.com>
1640         * md.texi (Define Subst): Match closing paren in example.
1642 2020-02-10  Jakub Jelinek  <jakub@redhat.com>
1644         PR target/58218
1645         PR other/93641
1646         * config/i386/i386.c (x86_64_elf_section_type_flags): Fix up last
1647         arguments of strncmp.
1649 2020-02-10  Feng Xue  <fxue@os.amperecomputing.com>
1651         PR ipa/93203
1652         * ipa-cp.c (ipcp_lattice::add_value): Add source with same call edge
1653         but different source value.
1654         (adjust_callers_for_value_intersection): New function.
1655         (gather_edges_for_value): Adjust order of callers to let a
1656         non-self-recursive caller be the first element.
1657         (self_recursive_pass_through_p): Add a new parameter "simple", and
1658         check generalized self-recursive pass-through jump function.
1659         (self_recursive_agg_pass_through_p): Likewise.
1660         (find_more_scalar_values_for_callers_subset): Compute value from
1661         pass-through jump function for self-recursive.
1662         (intersect_with_plats): Cleanup previous implementation code for value
1663         itersection with self-recursive call edge.
1664         (intersect_with_agg_replacements): Likewise.
1665         (intersect_aggregates_with_edge): Deduce value from pass-through jump
1666         function for self-recursive call edge.  Cleanup previous implementation
1667         code for value intersection with self-recursive call edge.
1668         (decide_whether_version_node): Remove dead callers and adjust order
1669         to let a non-self-recursive caller be the first element.
1671 2020-02-09  UroÅ¡ Bizjak  <ubizjak@gmail.com>
1673         * recog.c: Move pass_split_before_sched2 code in front of
1674         pass_split_before_regstack.
1675         (pass_data_split_before_sched2): Rename pass to split3 from split4.
1676         (pass_data_split_before_regstack): Rename pass to split4 from split3.
1677         (rest_of_handle_split_before_sched2): Remove.
1678         (pass_split_before_sched2::execute): Unconditionally call
1679         split_all_insns.
1680         (enable_split_before_sched2): New function.
1681         (pass_split_before_sched2::gate): Use enable_split_before_sched2.
1682         (pass_split_before_regstack::gate): Ditto.
1683         * config/nds32/nds32.c (nds32_split_double_word_load_store_p):
1684         Update name check for renamed split4 pass.
1685         * config/sh/sh.c (register_sh_passes): Update pass insertion
1686         point for renamed split4 pass.
1688 2020-02-09  Jakub Jelinek  <jakub@redhat.com>
1690         * gimplify.c (gimplify_adjust_omp_clauses_1): Promote
1691         DECL_IN_CONSTANT_POOL variables into "omp declare target" to avoid
1692         copying them around between host and target.
1694 2020-02-08  Andrew Pinski  <apinski@marvell.com>
1696         PR target/91927
1697         * config/aarch64/aarch64-simd.md (movmisalign<mode>): Check
1698         STRICT_ALIGNMENT also.
1700 2020-02-08  Jim Wilson  <jimw@sifive.com>
1702         PR target/93532
1703         * config/riscv/riscv.h (HARD_REGNO_CALLER_SAVE_MODE): Define.
1705 2020-02-08  UroÅ¡ Bizjak  <ubizjak@gmail.com>
1706             Jakub Jelinek  <jakub@redhat.com>
1708         PR target/65782
1709         * config/i386/i386.h (CALL_USED_REGISTERS): Make
1710         xmm16-xmm31 call-used even in 64-bit ms-abi.
1712 2020-02-07  Dennis Zhang  <dennis.zhang@arm.com>
1714         * config/aarch64/aarch64-simd-builtins.def (simd_smmla): New entry.
1715         (simd_ummla, simd_usmmla): Likewise.
1716         * config/aarch64/aarch64-simd.md (aarch64_simd_<sur>mmlav16qi): New.
1717         * config/aarch64/arm_neon.h (vmmlaq_s32, vmmlaq_u32): New.
1718         (vusmmlaq_s32): New.
1720 2020-02-07  Richard Biener  <rguenther@suse.de>
1722         PR middle-end/93519
1723         * tree-inline.c (fold_marked_statements): Do a PRE walk,
1724         skipping unreachable regions.
1725         (optimize_inline_calls): Skip folding stmts when we didn't
1726         inline.
1728 2020-02-07  H.J. Lu  <hongjiu.lu@intel.com>
1730         PR target/85667
1731         * config/i386/i386.c (function_arg_ms_64): Add a type argument.
1732         Don't return aggregates with only SFmode and DFmode in SSE
1733         register.
1734         (ix86_function_arg): Pass arg.type to function_arg_ms_64.
1736 2020-02-07  Jakub Jelinek  <jakub@redhat.com>
1738         PR target/93122
1739         * config/rs6000/rs6000-logue.c
1740         (rs6000_emit_probe_stack_range_stack_clash): Always use gen_add3_insn,
1741         if it fails, move rs into end_addr and retry.  Add
1742         REG_FRAME_RELATED_EXPR note whenever it returns more than one insn or
1743         the insn pattern doesn't describe well what exactly happens to
1744         dwarf2cfi.c.
1746         PR target/93594
1747         * config/i386/predicates.md (avx_identity_operand): Remove.
1748         * config/i386/sse.md (*avx_vec_concat<mode>_1): Remove.
1749         (avx_<castmode><avxsizesuffix>_<castmode>,
1750         avx512f_<castmode><avxsizesuffix>_256<castmode>): Change patterns to
1751         a VEC_CONCAT of the operand and UNSPEC_CAST.
1752         (avx512f_<castmode><avxsizesuffix>_<castmode>): Change pattern to
1753         a VEC_CONCAT of VEC_CONCAT of the operand and UNSPEC_CAST with
1754         UNSPEC_CAST.
1756         PR target/93611
1757         * config/i386/i386.c (ix86_lea_outperforms): Make sure to clear
1758         recog_data.insn if distance_non_agu_define changed it.
1760 2020-02-06  Michael Meissner  <meissner@linux.ibm.com>
1762         PR target/93569
1763         * config/rs6000/rs6000.c (reg_to_non_prefixed): Before ISA 3.0
1764         we only had X-FORM (reg+reg) addressing for vectors.  Also before
1765         ISA 3.0, we only had X-FORM addressing for scalars in the
1766         traditional Altivec registers.
1768 2020-02-06  <zhongyunde@huawei.com>
1769             Vladimir Makarov  <vmakarov@redhat.com>
1771         PR rtl-optimization/93561
1772         * lra-assigns.c (spill_for): Check that tested hard regno is not out of
1773         hard register range.
1775 2020-02-06  Richard Sandiford  <richard.sandiford@arm.com>
1777         * config/aarch64/aarch64.md (aarch64_movk<mode>): Add a type
1778         attribute.
1780 2020-02-06  Segher Boessenkool  <segher@kernel.crashing.org>
1782         * config/rs6000/rs6000.c (rs6000_emit_set_long_const): Handle the case
1783         where the low and the high 32 bits are equal to each other specially,
1784         with an rldimi instruction.
1786 2020-02-06  Mihail Ionescu  <mihail.ionescu@arm.com>
1788         * config/arm/arm-cpus.in: Set profile M for armv8.1-m.main.
1790 2020-02-06  Mihail Ionescu  <mihail.ionescu@arm.com>
1792         * config/arm/arm-tables.opt: Regenerate.
1794 2020-02-06  Richard Sandiford  <richard.sandiford@arm.com>
1796         PR target/87763
1797         * config/aarch64/aarch64-protos.h (aarch64_movk_shift): Declare.
1798         * config/aarch64/aarch64.c (aarch64_movk_shift): New function.
1799         * config/aarch64/aarch64.md (aarch64_movk<mode>): New pattern.
1801 2020-02-06  Richard Sandiford  <richard.sandiford@arm.com>
1803         PR rtl-optimization/87763
1804         * config/aarch64/aarch64.md (*ashiftsi_extvdi_bfiz): New pattern.
1806 2020-02-06  Delia Burduv  <delia.burduv@arm.com>
1808         * config/aarch64/aarch64-simd-builtins.def
1809         (bfmlaq): New built-in function.
1810         (bfmlalb): New built-in function.
1811         (bfmlalt): New built-in function.
1812         (bfmlalb_lane): New built-in function.
1813         (bfmlalt_lane): New built-in function.
1814         * config/aarch64/aarch64-simd.md
1815         (aarch64_bfmmlaqv4sf): New pattern.
1816         (aarch64_bfmlal<bt>v4sf): New pattern.
1817         (aarch64_bfmlal<bt>_lane<q>v4sf): New pattern.
1818         * config/aarch64/arm_neon.h (vbfmmlaq_f32): New intrinsic.
1819         (vbfmlalbq_f32): New intrinsic.
1820         (vbfmlaltq_f32): New intrinsic.
1821         (vbfmlalbq_lane_f32): New intrinsic.
1822         (vbfmlaltq_lane_f32): New intrinsic.
1823         (vbfmlalbq_laneq_f32): New intrinsic.
1824         (vbfmlaltq_laneq_f32): New intrinsic.
1825         * config/aarch64/iterators.md (BF_MLA): New int iterator.
1826         (bt): New int attribute.
1828 2020-02-06  UroÅ¡ Bizjak  <ubizjak@gmail.com>
1830         * config/i386/i386.md (*pushtf): Emit "#" instead of
1831         calling gcc_unreachable in insn output.
1832         (*pushxf): Ditto.
1833         (*pushdf): Ditto.
1834         (*pushsf_rex64): Ditto for alternatives other than 1.
1835         (*pushsf): Ditto for alternatives other than 1.
1837 2020-02-06  Martin Liska  <mliska@suse.cz>
1839         PR gcov-profile/91971
1840         PR gcov-profile/93466
1841         * coverage.c (coverage_init): Revert mangling of
1842         path into filename.  It can lead to huge filename length.
1843         Creation of subfolders seem more natural.
1845 2020-02-06  Stam Markianos-Wright  <stam.markianos-wright@arm.com>
1847         PR target/93300
1848         * config/arm/arm.c (arm_block_arith_comp_libfuncs_for_mode): New.
1849         (arm_init_libfuncs): Add BFmode support to block spurious BF libfuncs.
1850         Use arm_block_arith_comp_libfuncs_for_mode for HFmode.
1852 2020-02-06  Jakub Jelinek  <jakub@redhat.com>
1854         PR target/93594
1855         * config/i386/predicates.md (avx_identity_operand): New predicate.
1856         * config/i386/sse.md (*avx_vec_concat<mode>_1): New
1857         define_insn_and_split.
1859         PR libgomp/93515
1860         * omp-low.c (use_pointer_for_field): For nested constructs, also
1861         look for map clauses on target construct.
1862         (scan_omp_1_stmt) <case GIMPLE_OMP_TARGET>: Bump temporarily
1863         taskreg_nesting_level.
1865         PR libgomp/93515
1866         * gimplify.c (gimplify_scan_omp_clauses) <do_notice>: If adding
1867         shared clause, call omp_notice_variable on outer context if any.
1869 2020-02-05  Jason Merrill  <jason@redhat.com>
1871         PR c++/92003
1872         * symtab.c (symtab_node::nonzero_address): A DECL_COMDAT decl has
1873         non-zero address even if weak and not yet defined.
1875 2020-02-05  Martin Sebor  <msebor@redhat.com>
1877         PR tree-optimization/92765
1878         * gimple-fold.c (get_range_strlen_tree): Handle MEM_REF and PARM_DECL.
1879         * tree-ssa-strlen.c (compute_string_length): Remove.
1880         (determine_min_objsize): Remove.
1881         (get_len_or_size): Add an argument.  Call get_range_strlen_dynamic.
1882         Avoid using type size as the upper bound on string length.
1883         (handle_builtin_string_cmp): Add an argument.  Adjust.
1884         (strlen_check_and_optimize_call): Pass additional argument to
1885         handle_builtin_string_cmp.
1887 2020-02-05  UroÅ¡ Bizjak  <ubizjak@gmail.com>
1889         * config/i386/i386.md (*pushdi2_rex64 peephole2): Remove.
1890         (*pushdi2_rex64 peephole2): Unconditionally split after
1891         epilogue_completed.
1892         (*ashl<mode>3_doubleword): Ditto.
1893         (*<shift_insn><mode>3_doubleword): Ditto.
1895 2020-02-05  Michael Meissner  <meissner@linux.ibm.com>
1897         PR target/93568
1898         * config/rs6000/rs6000.c (get_vector_offset): Fix
1900 2020-02-05  Andrew Stubbs  <ams@codesourcery.com>
1902         * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Use / not space.
1904 2020-02-05  David Malcolm  <dmalcolm@redhat.com>
1906         * doc/analyzer.texi
1907         (Special Functions for Debugging the Analyzer): Update description
1908         of __analyzer_dump_exploded_nodes.
1910 2020-02-05  Jakub Jelinek  <jakub@redhat.com>
1912         PR target/92190
1913         * config/i386/i386-features.c (ix86_add_reg_usage_to_vzeroupper): Only
1914         include sets and not clobbers in the vzeroupper pattern.
1915         * config/i386/sse.md (*avx_vzeroupper): Require in insn condition that
1916         the parallel has 17 (64-bit) or 9 (32-bit) elts.
1917         (*avx_vzeroupper_1): New define_insn_and_split.
1919         PR target/92190
1920         * recog.c (pass_split_after_reload::gate): For STACK_REGS targets,
1921         don't run when !optimize.
1922         (pass_split_before_regstack::gate): For STACK_REGS targets, run even
1923         when !optimize.
1925 2020-02-05  Richard Biener  <rguenther@suse.de>
1927         PR middle-end/90648
1928         * genmatch.c (dt_node::gen_kids_1): Emit number of argument
1929         checks before matching calls.
1931 2020-02-05  Jakub Jelinek  <jakub@redhat.com>
1933         * tree-ssa-alias.c (aliasing_matching_component_refs_p): Fix up
1934         function comment typo.
1936         PR middle-end/93555
1937         * omp-simd-clone.c (expand_simd_clones): If simd_clone_mangle or
1938         simd_clone_create failed when i == 0, adjust clone->nargs by
1939         clone->inbranch.
1941 2020-02-05  Martin Liska  <mliska@suse.cz>
1943         PR c++/92717
1944         * doc/invoke.texi: Document that one should
1945         not combine ASLR and -fpch.
1947 2020-02-04  Richard Biener  <rguenther@suse.de>
1949         PR tree-optimization/93538
1950         * match.pd (addr EQ/NE ptr): Amend to handle &ptr->x EQ/NE ptr.
1952 2020-02-04  Richard Biener  <rguenther@suse.de>
1954         PR tree-optimization/91123
1955         * tree-ssa-sccvn.c (vn_walk_cb_data::finish): New method.
1956         (vn_walk_cb_data::last_vuse): New member.
1957         (vn_walk_cb_data::saved_operands): Likewsie.
1958         (vn_walk_cb_data::~vn_walk_cb_data): Release saved_operands.
1959         (vn_walk_cb_data::push_partial_def): Use finish.
1960         (vn_reference_lookup_2): Update last_vuse and use finish if
1961         we've saved operands.
1962         (vn_reference_lookup_3): Use finish and update calls to
1963         push_partial_defs everywhere.  When translating through
1964         memcpy or aggregate copies save off operands and alias-set.
1965         (eliminate_dom_walker::eliminate_stmt): Restore VN_WALKREWRITE
1966         operation for redundant store removal.
1968 2020-02-04  Richard Biener  <rguenther@suse.de>
1970         PR tree-optimization/92819
1971         * tree-ssa-forwprop.c (simplify_vector_constructor): Avoid
1972         generating more stmts than before.
1974 2020-02-04  Martin Liska  <mliska@suse.cz>
1976         * config/arm/arm.c (arm_gen_far_branch): Move the function
1977         outside of selftests.
1979 2020-02-03  Michael Meissner  <meissner@linux.ibm.com>
1981         * config/rs6000/rs6000.c (adjust_vec_address_pcrel): New helper
1982         function to adjust PC-relative vector addresses.
1983         (rs6000_adjust_vec_address): Call adjust_vec_address_pcrel to
1984         handle vectors with PC-relative addresses.
1986 2020-02-03  Michael Meissner  <meissner@linux.ibm.com>
1988         * config/rs6000/rs6000.c (reg_to_non_prefixed): Add forward
1989         reference.
1990         (hard_reg_and_mode_to_addr_mask): Delete.
1991         (rs6000_adjust_vec_address): If the original vector address
1992         was REG+REG or REG+OFFSET and the element is not zero, do the add
1993         of the elements in the original address before adding the offset
1994         for the vector element.  Use address_to_insn_form to validate the
1995         address using the register being loaded, rather than guessing
1996         whether the address is a DS-FORM or DQ-FORM address.
1998 2020-02-03  Michael Meissner  <meissner@linux.ibm.com>
2000         * config/rs6000/rs6000.c (get_vector_offset): New helper function
2001         to calculate the offset in memory from the start of a vector of a
2002         particular element.  Add code to keep the element number in
2003         bounds if the element number is variable.
2004         (rs6000_adjust_vec_address): Move calculation of offset of the
2005         vector element to get_vector_offset.
2006         (rs6000_split_vec_extract_var): Do not do the initial AND of
2007         element here, move the code to get_vector_offset.
2009 2020-02-03  Michael Meissner  <meissner@linux.ibm.com>
2011         * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add some
2012         gcc_asserts.
2014 2020-02-03  Segher Boessenkool  <segher@kernel.crashing.org>
2016         * config/rs6000/constraints.md: Improve documentation.
2018 2020-02-03  Richard Earnshaw  <rearnsha@arm.com>
2020         PR target/93548
2021         * config/arm/t-arm: ($(srcdir)/config/arm/arm-tune.md)
2022         ($(srcdir)/config/arm/arm-tables.opt): Use move-if-change.
2024 2020-02-03  Andrew Stubbs  <ams@codesourcery.com>
2026         * config.gcc: Remove "carrizo" support.
2027         * config/gcn/gcn-opts.h (processor_type): Likewise.
2028         * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Likewise.
2029         * config/gcn/gcn.opt (gpu_type): Likewise.
2030         * config/gcn/t-omp-device: Likewise.
2032 2020-02-03  Stam Markianos-Wright  <stam.markianos-wright@arm.com>
2034         PR target/91816
2035         * config/arm/arm-protos.h: New function arm_gen_far_branch prototype.
2036         * config/arm/arm.c (arm_gen_far_branch): New function
2037         arm_gen_far_branch.
2038         * config/arm/arm.md: Update b<cond> for Thumb2 range checks.
2040 2020-02-03  Julian Brown  <julian@codesourcery.com>
2041             Tobias Burnus  <tobias@codesourcery.com>
2043         * doc/invoke.texi: Update mention of OpenACC version to 2.6.
2045 2020-02-03  Jakub Jelinek  <jakub@redhat.com>
2047         PR target/93533
2048         * config/s390/s390.md (popcounthi2_z196): Fix up expander to emit
2049         valid RTL to sum up the lowest and second lowest bytes of the popcnt
2050         result.
2052 2020-02-02  Vladimir Makarov  <vmakarov@redhat.com>
2054         PR rtl-optimization/91333
2055         * ira-color.c (struct allocno_color_data): Add member
2056         hard_reg_prefs.
2057         (init_allocno_threads): Set the member up.
2058         (bucket_allocno_compare_func): Add compare hard reg
2059         prefs.
2061 2020-01-31  Sandra Loosemore  <sandra@codesourcery.com>
2063         nios2: Support for GOT-relative DW_EH_PE_datarel encoding.
2065         * configure.ac [nios2-*-*]: Check HAVE_AS_NIOS2_GOTOFF_RELOCATION.
2066         * config.in: Regenerated.
2067         * configure: Regenerated.
2068         * config/nios2/nios2.h (ASM_PREFERRED_EH_DATA_FORMAT): Fix handling
2069         for PIC when HAVE_AS_NIOS2_GOTOFF_RELOCATION.
2070         (ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX): New.
2072 2020-02-01  Andrew Burgess  <andrew.burgess@embecosm.com>
2074         * configure: Regenerate.
2076 2020-01-31  Vladimir Makarov  <vmakarov@redhat.com>
2078         PR rtl-optimization/91333
2079         * ira-color.c (bucket_allocno_compare_func): Move conflict hard
2080         reg preferences comparison up.
2082 2020-01-31  Richard Sandiford  <richard.sandiford@arm.com>
2084         * config/aarch64/aarch64.h (TARGET_SVE_BF16): New macro.
2085         * config/aarch64/aarch64-sve-builtins-sve2.h (svcvtnt): Move to
2086         aarch64-sve-builtins-base.h.
2087         * config/aarch64/aarch64-sve-builtins-sve2.cc (svcvtnt): Move to
2088         aarch64-sve-builtins-base.cc.
2089         * config/aarch64/aarch64-sve-builtins-base.h (svbfdot, svbfdot_lane)
2090         (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
2091         (svcvtnt): Declare.
2092         * config/aarch64/aarch64-sve-builtins-base.cc (svbfdot, svbfdot_lane)
2093         (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
2094         (svcvtnt): New functions.
2095         * config/aarch64/aarch64-sve-builtins-base.def (svbfdot, svbfdot_lane)
2096         (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
2097         (svcvtnt): New functions.
2098         (svcvt): Add a form that converts f32 to bf16.
2099         * config/aarch64/aarch64-sve-builtins-shapes.h (ternary_bfloat)
2100         (ternary_bfloat_lane, ternary_bfloat_lanex2, ternary_bfloat_opt_n):
2101         Declare.
2102         * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_element_type):
2103         Treat B as bfloat16_t.
2104         (ternary_bfloat_lane_base): New class.
2105         (ternary_bfloat_def): Likewise.
2106         (ternary_bfloat): New shape.
2107         (ternary_bfloat_lane_def): New class.
2108         (ternary_bfloat_lane): New shape.
2109         (ternary_bfloat_lanex2_def): New class.
2110         (ternary_bfloat_lanex2): New shape.
2111         (ternary_bfloat_opt_n_def): New class.
2112         (ternary_bfloat_opt_n): New shape.
2113         * config/aarch64/aarch64-sve-builtins.cc (TYPES_cvt_bfloat): New macro.
2114         * config/aarch64/aarch64-sve.md (@aarch64_sve_<sve_fp_op>vnx4sf)
2115         (@aarch64_sve_<sve_fp_op>_lanevnx4sf): New patterns.
2116         (@aarch64_sve_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>)
2117         (@cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
2118         (*cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
2119         (@aarch64_sve_cvtnt<VNx8BF_ONLY:mode>): Likewise.
2120         * config/aarch64/aarch64-sve2.md (@aarch64_sve2_cvtnt<mode>): Key
2121         the pattern off the narrow mode instead of the wider one.
2122         * config/aarch64/iterators.md (VNx8BF_ONLY): New mode iterator.
2123         (UNSPEC_BFMLALB, UNSPEC_BFMLALT, UNSPEC_BFMMLA): New unspecs.
2124         (sve_fp_op): Handle them.
2125         (SVE_BFLOAT_TERNARY_LONG): New int itertor.
2126         (SVE_BFLOAT_TERNARY_LONG_LANE): Likewise.
2128 2020-01-31  Richard Sandiford  <richard.sandiford@arm.com>
2130         * config/aarch64/arm_sve.h: Include arm_bf16.h.
2131         * config/aarch64/aarch64-modes.def (BF): Move definition before
2132         VECTOR_MODES.  Remove separate VECTOR_MODES for V4BF and V8BF.
2133         (SVE_MODES): Handle BF modes.
2134         * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
2135         BF modes.
2136         (aarch64_full_sve_mode): Likewise.
2137         * config/aarch64/iterators.md (SVE_STRUCT): Add VNx16BF, VNx24BF
2138         and VNx32BF.
2139         (SVE_FULL, SVE_FULL_HSD, SVE_ALL): Add VNx8BF.
2140         (Vetype, Vesize, Vctype, VEL, Vel, VEL_INT, V128, v128, vwcore)
2141         (V_INT_EQUIV, v_int_equiv, V_FP_EQUIV, v_fp_equiv, vector_count)
2142         (insn_length, VSINGLE, vsingle, VPRED, vpred, VDOUBLE): Handle the
2143         new SVE BF modes.
2144         * config/aarch64/aarch64-sve-builtins.h (TYPE_bfloat): New
2145         type_class_index.
2146         * config/aarch64/aarch64-sve-builtins.cc (TYPES_all_arith): New macro.
2147         (TYPES_all_data): Add bf16.
2148         (TYPES_reinterpret1, TYPES_reinterpret): Likewise.
2149         (register_tuple_type): Increase buffer size.
2150         * config/aarch64/aarch64-sve-builtins.def (svbfloat16_t): New type.
2151         (bf16): New type suffix.
2152         * config/aarch64/aarch64-sve-builtins-base.def (svabd, svadd, svaddv)
2153         (svcmpeq, svcmpge, svcmpgt, svcmple, svcmplt, svcmpne, svmad, svmax)
2154         (svmaxv, svmin, svminv, svmla, svmls, svmsb, svmul, svsub, svsubr):
2155         Change type from all_data to all_arith.
2156         * config/aarch64/aarch64-sve-builtins-sve2.def (svaddp, svmaxp)
2157         (svminp): Likewise.
2159 2020-01-31  Dennis Zhang  <dennis.zhang@arm.com>
2160             Matthew Malcomson  <matthew.malcomson@arm.com>
2161             Richard Sandiford  <richard.sandiford@arm.com>
2163         * doc/invoke.texi (f32mm): Document new AArch64 -march= extension.
2164         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Define
2165         __ARM_FEATURE_SVE_MATMUL_INT8, __ARM_FEATURE_SVE_MATMUL_FP32 and
2166         __ARM_FEATURE_SVE_MATMUL_FP64 as appropriate.  Don't define
2167         __ARM_FEATURE_MATMUL_FP64.
2168         * config/aarch64/aarch64-option-extensions.def (fp, simd, fp16)
2169         (sve): Add AARCH64_FL_F32MM to the list of extensions that should
2170         be disabled at the same time.
2171         (f32mm): New extension.
2172         * config/aarch64/aarch64.h (AARCH64_FL_F32MM): New macro.
2173         (AARCH64_FL_F64MM): Bump to the next bit up.
2174         (AARCH64_ISA_F32MM, TARGET_SVE_I8MM, TARGET_F32MM, TARGET_SVE_F32MM)
2175         (TARGET_SVE_F64MM): New macros.
2176         * config/aarch64/iterators.md (SVE_MATMULF): New mode iterator.
2177         (UNSPEC_FMMLA, UNSPEC_SMATMUL, UNSPEC_UMATMUL, UNSPEC_USMATMUL)
2178         (UNSPEC_TRN1Q, UNSPEC_TRN2Q, UNSPEC_UZP1Q, UNSPEC_UZP2Q, UNSPEC_ZIP1Q)
2179         (UNSPEC_ZIP2Q): New unspeccs.
2180         (DOTPROD_US_ONLY, PERMUTEQ, MATMUL, FMMLA): New int iterators.
2181         (optab, sur, perm_insn): Handle the new unspecs.
2182         (sve_fp_op): Handle UNSPEC_FMMLA.  Resort.
2183         * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use
2184         TARGET_SVE_F64MM instead of separate tests.
2185         (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod<vsi2qi>): New pattern.
2186         (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod_lane<vsi2qi>): Likewise.
2187         (@aarch64_sve_add_<MATMUL:optab><vsi2qi>): Likewise.
2188         (@aarch64_sve_<FMMLA:sve_fp_op><mode>): Likewise.
2189         (@aarch64_sve_<PERMUTEQ:optab><mode>): Likewise.
2190         * config/aarch64/aarch64-sve-builtins.cc (TYPES_s_float): New macro.
2191         (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): Use it.
2192         (TYPES_s_signed): New macro.
2193         (TYPES_s_integer): Use it.
2194         (TYPES_d_float): New macro.
2195         (TYPES_d_data): Use it.
2196         * config/aarch64/aarch64-sve-builtins-shapes.h (mmla): Declare.
2197         (ternary_intq_uintq_lane, ternary_intq_uintq_opt_n, ternary_uintq_intq)
2198         (ternary_uintq_intq_lane, ternary_uintq_intq_opt_n): Likewise.
2199         * config/aarch64/aarch64-sve-builtins-shapes.cc (mmla_def): New class.
2200         (svmmla): New shape.
2201         (ternary_resize2_opt_n_base): Add TYPE_CLASS2 and TYPE_CLASS3
2202         template parameters.
2203         (ternary_resize2_lane_base): Likewise.
2204         (ternary_resize2_base): New class.
2205         (ternary_qq_lane_base): Likewise.
2206         (ternary_intq_uintq_lane_def): Likewise.
2207         (ternary_intq_uintq_lane): New shape.
2208         (ternary_intq_uintq_opt_n_def): New class
2209         (ternary_intq_uintq_opt_n): New shape.
2210         (ternary_qq_lane_def): Inherit from ternary_qq_lane_base.
2211         (ternary_uintq_intq_def): New class.
2212         (ternary_uintq_intq): New shape.
2213         (ternary_uintq_intq_lane_def): New class.
2214         (ternary_uintq_intq_lane): New shape.
2215         (ternary_uintq_intq_opt_n_def): New class.
2216         (ternary_uintq_intq_opt_n): New shape.
2217         * config/aarch64/aarch64-sve-builtins-base.h (svmmla, svsudot)
2218         (svsudot_lane, svtrn1q, svtrn2q, svusdot, svusdot_lane, svusmmla)
2219         (svuzp1q, svuzp2q, svzip1q, svzip2q): Declare.
2220         * config/aarch64/aarch64-sve-builtins-base.cc (svdot_lane_impl):
2221         Generalize to...
2222         (svdotprod_lane_impl): ...this new class.
2223         (svmmla_impl, svusdot_impl): New classes.
2224         (svdot_lane): Update to use svdotprod_lane_impl.
2225         (svmmla, svsudot, svsudot_lane, svtrn1q, svtrn2q, svusdot)
2226         (svusdot_lane, svusmmla, svuzp1q, svuzp2q, svzip1q, svzip2q): New
2227         functions.
2228         * config/aarch64/aarch64-sve-builtins-base.def (svmmla): New base
2229         function, with no types defined.
2230         (svmmla, svusmmla, svsudot, svsudot_lane, svusdot, svusdot_lane): New
2231         AARCH64_FL_I8MM functions.
2232         (svmmla): New AARCH64_FL_F32MM function.
2233         (svld1ro): Depend only on AARCH64_FL_F64MM, not on AARCH64_FL_V8_6.
2234         (svmmla, svtrn1q, svtrn2q, svuz1q, svuz2q, svzip1q, svzip2q): New
2235         AARCH64_FL_F64MM function.
2236         (REQUIRED_EXTENSIONS):
2238 2020-01-31  Andrew Stubbs  <ams@codesourcery.com>
2240         * config/gcn/gcn-valu.md (addv64di3_exec): Allow one '0' in each
2241         alternative only.
2243 2020-01-31  UroÅ¡ Bizjak  <ubizjak@gmail.com>
2245         * config/i386/i386.md (*movoi_internal_avx): Do not check for
2246         TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.  Remove MODE_V8SF handling.
2247         (*movti_internal): Do not check for
2248         TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
2249         (*movtf_internal): Move check for TARGET_SSE2 and size optimization
2250         just after check for TARGET_AVX.
2251         (*movdf_internal): Ditto.
2252         * config/i386/mmx.md (*mov<mode>_internal): Do not check for
2253         TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
2254         * config/i386/sse.md (mov<mode>_internal): Only check
2255         TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL with V2DFmode.  Move check
2256         for TARGET_SSE2 and size optimization just after check for TARGET_AVX.
2257         (<sse>_andnot<mode>3<mask_name>): Move check for
2258         TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL after check for TARGET_AVX.
2259         (<code><mode>3<mask_name>): Ditto.
2260         (*andnot<mode>3): Ditto.
2261         (*andnottf3): Ditto.
2262         (*<code><mode>3): Ditto.
2263         (*<code>tf3): Ditto.
2264         (*andnot<VI:mode>3): Remove
2265         TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL handling.
2266         (<mask_codefor><code><VI48_AVX_AVX512F:mode>3<mask_name>): Ditto.
2267         (*<code><VI12_AVX_AVX512F:mode>3): Ditto.
2268         (sse4_1_blendv<ssemodesuffix>): Ditto.
2269         * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL):
2270         Explain that tune applies to 128bit instructions only.
2272 2020-01-31  Kwok Cheung Yeung  <kcy@codesourcery.com>
2274         * config/gcn/mkoffload.c (process_asm): Add sgpr_count and vgpr_count
2275         to definition of hsa_kernel_description.  Parse assembly to find SGPR
2276         and VGPR count of kernel and store in hsa_kernel_description.
2278 2020-01-31  Tamar Christina  <tamar.christina@arm.com>
2280         PR rtl-optimization/91838
2281         * simplify-rtx.c (simplify_binary_operation_1): Update LSHIFTRT case
2282         to truncate if allowed or reject combination.
2284 2020-01-31  Andrew Stubbs  <ams@codesourcery.com>
2286         * tree-ssa-loop-ivopts.c (get_iv): Use sizetype for zero-step.
2287         (find_inv_vars_cb): Likewise.
2289 2020-01-31  David Malcolm  <dmalcolm@redhat.com>
2291         * calls.c (special_function_p): Split out the check for DECL_NAME
2292         being non-NULL and fndecl being extern at file scope into a
2293         new maybe_special_function_p and call it.  Drop check for fndecl
2294         being non-NULL that was after a usage of DECL_NAME (fndecl).
2295         * tree.h (maybe_special_function_p): New inline function.
2297 2020-01-30  Andrew Stubbs  <ams@codesourcery.com>
2299         * config/gcn/gcn-valu.md (gather<mode>_exec): Move contents ...
2300         (mask_gather_load<mode>): ... here, and zero-initialize the
2301         destination.
2302         (maskload<mode>di): Zero-initialize the destination.
2303         * config/gcn/gcn.c:
2305 2020-01-30  David Malcolm  <dmalcolm@redhat.com>
2307         PR analyzer/93356
2308         * doc/analyzer.texi (Limitations): Note that constraints on
2309         floating-point values are currently ignored.
2311 2020-01-30  Jakub Jelinek  <jakub@redhat.com>
2313         PR lto/93384
2314         * symtab.c (symtab_node::noninterposable_alias): If localalias
2315         already exists, but is not usable, append numbers after it until
2316         a unique name is found.  Formatting fix.
2318         PR middle-end/93505
2319         * combine.c (simplify_comparison) <case ROTATE>: Punt on out of range
2320         rotate counts.
2322 2020-01-30  Andrew Stubbs  <ams@codesourcery.com>
2324         * config/gcn/gcn.c (print_operand): Handle LTGT.
2325         * config/gcn/predicates.md (gcn_fp_compare_operator): Allow ltgt.
2327 2020-01-30  Richard Biener  <rguenther@suse.de>
2329         * tree-pretty-print.c (dump_generic_node): Wrap VECTOR_CST
2330         and CONSTRUCTOR in _Literal (type) with TDF_GIMPLE.
2332 2020-01-30  John David Anglin  <danglin@gcc.gnu.org>
2334         * config/pa/pa.c (pa_elf_select_rtx_section): Place function pointers
2335         without a DECL in .data.rel.ro.local.
2337 2020-01-30  Jakub Jelinek  <jakub@redhat.com>
2339         PR target/93494
2340         * config/arm/arm.md (uaddvdi4): Actually emit what gen_uaddvsi4
2341         returned.
2343         PR target/91824
2344         * config/i386/sse.md
2345         (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext): Renamed to ...
2346         (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext): ... this.  Use
2347         any_extend code iterator instead of always zero_extend.
2348         (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_lt): Renamed to ...
2349         (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_lt): ... this.
2350         Use any_extend code iterator instead of always zero_extend.
2351         (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_shift): Renamed to ...
2352         (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_shift): ... this.
2353         Use any_extend code iterator instead of always zero_extend.
2354         (*sse2_pmovmskb_ext): New define_insn.
2355         (*sse2_pmovmskb_ext_lt): New define_insn_and_split.
2357         PR target/91824
2358         * config/i386/i386.md (*popcountsi2_zext): New define_insn_and_split.
2359         (*popcountsi2_zext_falsedep): New define_insn.
2361 2020-01-30  Dragan Mladjenovic  <dmladjenovic@wavecomp.com>
2363         * config.in: Regenerated.
2364         * configure: Regenerated.
2366 2020-01-29  Tobias Burnus  <tobias@codesourcery.com>
2368         PR bootstrap/93409
2369         * config/gcn/gcn-hsa.h (ASM_SPEC): Add -mattr=-code-object-v3 as
2370         LLVM's assembler changed the default in version 9.
2372 2020-01-24  Jeff Law  <law@redhat.com>
2374         PR tree-optimization/89689
2375         * builtins.def (BUILT_IN_OBJECT_SIZE): Make it const rather than pure.
2377 2020-01-29  Richard Sandiford  <richard.sandiford@arm.com>
2379         Revert:
2381         2020-01-28  Richard Sandiford  <richard.sandiford@arm.com>
2383         PR rtl-optimization/87763
2384         * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
2385         simplification to handle subregs as well as bare regs.
2386         * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
2388 2020-01-29  Joel Hutton  <Joel.Hutton@arm.com>
2390         PR target/93221
2391         * ira.c (ira): Revert use of simplified LRA algorithm.
2393 2020-01-29  Martin Jambor  <mjambor@suse.cz>
2395         PR tree-optimization/92706
2396         * tree-sra.c (struct access): Fields first_link, last_link,
2397         next_queued and grp_queued renamed to first_rhs_link, last_rhs_link,
2398         next_rhs_queued and grp_rhs_queued respectively, new fields
2399         first_lhs_link, last_lhs_link, next_lhs_queued and grp_lhs_queued.
2400         (struct assign_link): Field next renamed to next_rhs, new field
2401         next_lhs.  Updated comment.
2402         (work_queue_head): Renamed to rhs_work_queue_head.
2403         (lhs_work_queue_head): New variable.
2404         (add_link_to_lhs): New function.
2405         (relink_to_new_repr): Also relink LHS lists.
2406         (add_access_to_work_queue): Renamed to add_access_to_rhs_work_queue.
2407         (add_access_to_lhs_work_queue): New function.
2408         (pop_access_from_work_queue): Renamed to
2409         pop_access_from_rhs_work_queue.
2410         (pop_access_from_lhs_work_queue): New function.
2411         (build_accesses_from_assign): Also add links to LHS lists and to LHS
2412         work_queue.
2413         (child_would_conflict_in_lacc): Renamed to
2414         child_would_conflict_in_acc.  Adjusted parameter names.
2415         (create_artificial_child_access): New parameter set_grp_read, use it.
2416         (subtree_mark_written_and_enqueue): Renamed to
2417         subtree_mark_written_and_rhs_enqueue.
2418         (propagate_subaccesses_across_link): Renamed to
2419         propagate_subaccesses_from_rhs.
2420         (propagate_subaccesses_from_lhs): New function.
2421         (propagate_all_subaccesses): Also propagate subaccesses from LHSs to
2422         RHSs.
2424 2020-01-29  Martin Jambor  <mjambor@suse.cz>
2426         PR tree-optimization/92706
2427         * tree-sra.c (struct access): Adjust comment of
2428         grp_total_scalarization.
2429         (find_access_in_subtree): Look for single children spanning an entire
2430         access.
2431         (scalarizable_type_p): Allow register accesses, adjust callers.
2432         (completely_scalarize): Remove function.
2433         (scalarize_elem): Likewise.
2434         (create_total_scalarization_access): Likewise.
2435         (sort_and_splice_var_accesses): Do not track total scalarization
2436         flags.
2437         (analyze_access_subtree): New parameter totally, adjust to new meaning
2438         of grp_total_scalarization.
2439         (analyze_access_trees): Pass new parameter to analyze_access_subtree.
2440         (can_totally_scalarize_forest_p): New function.
2441         (create_total_scalarization_access): Likewise.
2442         (create_total_access_and_reshape): Likewise.
2443         (total_should_skip_creating_access): Likewise.
2444         (totally_scalarize_subtree): Likewise.
2445         (analyze_all_variable_accesses): Perform total scalarization after
2446         subaccess propagation using the new functions above.
2447         (initialize_constant_pool_replacements): Output initializers by
2448         traversing the access tree.
2450 2020-01-29  Martin Jambor  <mjambor@suse.cz>
2452         * tree-sra.c (verify_sra_access_forest): New function.
2453         (verify_all_sra_access_forests): Likewise.
2454         (create_artificial_child_access): Set parent.
2455         (analyze_all_variable_accesses): Call the verifier.
2457 2020-01-28  Jan Hubicka  <hubicka@ucw.cz>
2459         * cgraph.c (cgraph_edge::resolve_speculation): Only lookup direct edge
2460         if called on indirect edge.
2461         (cgraph_edge::redirect_call_stmt_to_callee): Lookup indirect edge of
2462         speculative call if needed.
2464 2020-01-29  Richard Biener  <rguenther@suse.de>
2466         PR tree-optimization/93428
2467         * tree-vect-slp.c (vect_build_slp_tree_2): Compute the load
2468         permutation when the load node is created.
2469         (vect_analyze_slp_instance): Re-use it here.
2471 2020-01-28  Jan Hubicka  <hubicka@ucw.cz>
2473         * ipa-prop.c (update_indirect_edges_after_inlining): Fix warning.
2475 2020-01-28  Vladimir Makarov  <vmakarov@redhat.com>
2477         PR rtl-optimization/93272
2478         * ira-lives.c (process_out_of_region_eh_regs): New function.
2479         (process_bb_node_lives): Call it.
2481 2020-01-28  Jan Hubicka  <hubicka@ucw.cz>
2483         * coverage.c (read_counts_file): Make error message lowercase.
2485 2020-01-28  Jan Hubicka  <hubicka@ucw.cz>
2487         * profile-count.c (profile_quality_display_names): Fix ordering.
2489 2020-01-28  Jan Hubicka  <hubicka@ucw.cz>
2491         PR lto/93318    
2492         * cgraph.c (cgraph_add_edge_to_call_site_hash): Update call site
2493         hash only when edge is first within the sequence.
2494         (cgraph_edge::set_call_stmt): Update handling of speculative calls.
2495         (symbol_table::create_edge): Do not set target_prob.
2496         (cgraph_edge::remove_caller): Watch for speculative calls when updating
2497         the call site hash.
2498         (cgraph_edge::make_speculative): Drop target_prob parameter.
2499         (cgraph_edge::speculative_call_info): Remove.
2500         (cgraph_edge::first_speculative_call_target): New member function.
2501         (update_call_stmt_hash_for_removing_direct_edge): New function.
2502         (cgraph_edge::resolve_speculation): Rewrite to new API.
2503         (cgraph_edge::speculative_call_for_target): New member function.
2504         (cgraph_edge::make_direct): Rewrite to new API; fix handling of
2505         multiple speculation targets.
2506         (cgraph_edge::redirect_call_stmt_to_callee): Likewise; fix updating
2507         of profile.
2508         (verify_speculative_call): Verify that targets form an interval.
2509         * cgraph.h (cgraph_edge::speculative_call_info): Remove.
2510         (cgraph_edge::first_speculative_call_target): New member function.
2511         (cgraph_edge::next_speculative_call_target): New member function.
2512         (cgraph_edge::speculative_call_target_ref): New member function.
2513         (cgraph_edge;:speculative_call_indirect_edge): New member funtion.
2514         (cgraph_edge): Remove target_prob.
2515         * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
2516         Fix handling of speculative calls.
2517         * ipa-devirt.c (ipa_devirt): Fix handling of speculative cals.
2518         * ipa-fnsummary.c (analyze_function_body): Likewise.
2519         * ipa-inline.c (speculation_useful_p): Use new speculative call API.
2520         * ipa-profile.c (dump_histogram): Fix formating.
2521         (ipa_profile_generate_summary): Watch for overflows.
2522         (ipa_profile): Do not require probablity to be 1/2; update to new API.
2523         * ipa-prop.c (ipa_make_edge_direct_to_target): Update to new API.
2524         (update_indirect_edges_after_inlining): Update to new API.
2525         * ipa-utils.c (ipa_merge_profiles): Rewrite merging of speculative call
2526         profiles.
2527         * profile-count.h: (profile_probability::adjusted): New.
2528         * tree-inline.c (copy_bb): Update to new speculative call API; fix
2529         updating of profile.
2530         * value-prof.c (gimple_ic_transform): Rename to ...
2531         (dump_ic_profile): ... this one; update dumping.
2532         (stream_in_histogram_value): Fix formating.
2533         (gimple_value_profile_transformations): Update.
2535 2020-01-28  H.J. Lu  <hongjiu.lu@intel.com>
2537         PR target/91461
2538         * config/i386/i386.md (*movoi_internal_avx): Remove
2539         TARGET_SSE_TYPELESS_STORES check.
2540         (*movti_internal): Prefer TARGET_AVX over
2541         TARGET_SSE_TYPELESS_STORES.
2542         (*movtf_internal): Likewise.
2543         * config/i386/sse.md (mov<mode>_internal): Prefer TARGET_AVX over
2544         TARGET_SSE_TYPELESS_STORES.  Remove "<MODE_SIZE> == 16" check
2545         from TARGET_SSE_TYPELESS_STORES.
2547 2020-01-28  David Malcolm  <dmalcolm@redhat.com>
2549         * diagnostic-core.h (warning_at): Rename overload to...
2550         (warning_meta): ...this.
2551         (emit_diagnostic_valist): Delete decl of overload taking
2552         diagnostic_metadata.
2553         * diagnostic.c (emit_diagnostic_valist): Likewise for defn.
2554         (warning_at): Rename overload taking diagnostic_metadata to...
2555         (warning_meta): ...this.
2557 2020-01-28  Richard Biener  <rguenther@suse.de>
2559         PR tree-optimization/93439
2560         * tree-parloops.c (create_loop_fn): Move clique bookkeeping...
2561         * tree-cfg.c (move_sese_region_to_fn): ... here.
2562         (verify_types_in_gimple_reference): Verify used cliques are
2563         tracked.
2565 2020-01-28  H.J. Lu  <hongjiu.lu@intel.com>
2567         PR target/91399
2568         * config/i386/i386-options.c (set_ix86_tune_features): Add an
2569         argument of a pointer to struct gcc_options and pass it to
2570         parse_mtune_ctrl_str.
2571         (ix86_function_specific_restore): Pass opts to
2572         set_ix86_tune_features.
2573         (ix86_option_override_internal): Likewise.
2574         (parse_mtune_ctrl_str): Add an argument of a pointer to struct
2575         gcc_options and use it for x_ix86_tune_ctrl_string.
2577 2020-01-28  Richard Sandiford  <richard.sandiford@arm.com>
2579         PR rtl-optimization/87763
2580         * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
2581         simplification to handle subregs as well as bare regs.
2582         * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
2584 2020-01-28  Richard Sandiford  <richard.sandiford@arm.com>
2586         * tree-vect-loop.c (vectorizable_reduction): Fail gracefully
2587         for reduction chains that (now) include a call.
2589 2020-01-28  Richard Sandiford  <richard.sandiford@arm.com>
2591         PR tree-optimization/92822
2592         * tree-ssa-forwprop.c (simplify_vector_constructor): When filling
2593         out the don't-care elements of a vector whose significant elements
2594         are duplicates, make the don't-care elements duplicates too.
2596 2020-01-28  Richard Sandiford  <richard.sandiford@arm.com>
2598         PR tree-optimization/93434
2599         * tree-predcom.c (split_data_refs_to_components): Record which
2600         components have had aliasing loads removed.  Prevent store-store
2601         commoning for all such components.
2603 2020-01-28  Jakub Jelinek  <jakub@redhat.com>
2605         PR target/93418
2606         * config/i386/i386.c (ix86_fold_builtin) <do_shift>: If mask is not
2607         -1 or is_vshift is true, use new_vector with number of elts npatterns
2608         rather than new_unary_operation.
2610         PR tree-optimization/93454
2611         * gimple-fold.c (fold_array_ctor_reference): Perform
2612         elt_size.to_uhwi () just once, instead of calling it in every
2613         iteration.  Punt if that value is above size of the temporary
2614         buffer.  Decrease third native_encode_expr argument when
2615         bufoff + elt_sz is above size of buf.
2617 2020-01-27  Joseph Myers  <joseph@codesourcery.com>
2619         * config/mips/mips.c (mips_declare_object_name)
2620         [USE_GNU_UNIQUE_OBJECT]: Support use of gnu_unique_object.
2622 2020-01-27  Martin Liska  <mliska@suse.cz>
2624         PR gcov-profile/93403
2625         * tree-profile.c (gimple_init_gcov_profiler): Generate
2626         both __gcov_indirect_call_profiler_v4 and
2627         __gcov_indirect_call_profiler_v4_atomic.
2629 2020-01-27  Richard Sandiford  <richard.sandiford@arm.com>
2631         PR target/92822
2632         * config/aarch64/aarch64-simd.md (aarch64_get_half<mode>): New
2633         expander.
2634         (@aarch64_split_simd_mov<mode>): Use it.
2635         (aarch64_simd_mov_from_<mode>low): Add a GPR alternative.
2636         Leave the vec_extract patterns to handle 2-element vectors.
2637         (aarch64_simd_mov_from_<mode>high): Likewise.
2638         (vec_extract<VQMOV_NO2E:mode><Vhalf>): New expander.
2639         (vec_extractv2dfv1df): Likewise.
2641 2020-01-27  Richard Sandiford  <richard.sandiford@arm.com>
2643         * config/aarch64/aarch64.c (aarch64_if_then_else_costs): Match
2644         jump conditions for *compare_condjump<GPI:mode>.
2646 2020-01-27  David Malcolm  <dmalcolm@redhat.com>
2648         PR analyzer/93276
2649         * digraph.cc (test_edge::test_edge): Specify template for base
2650         class initializer.
2652 2020-01-27  Claudiu Zissulescu  <claziss@synopsys.com>
2654         * config/arc/arc.c (arc_rtx_costs): Update mul64 cost.
2656 2020-01-27  Claudiu Zissulescu  <claziss@synopsys.com>
2658         * config/arc/arc-protos.h (gen_mlo): Remove.
2659         (gen_mhi): Likewise.
2660         * config/arc/arc.c (AUX_MULHI): Define.
2661         (arc_must_save_reister): Special handling for r58/59.
2662         (arc_compute_frame_size): Consider mlo/mhi registers.
2663         (arc_save_callee_saves): Emit fp/sp move only when emit_move
2664         paramter is true.
2665         (arc_conditional_register_usage): Remove TARGET_BIG_ENDIAN from
2666         mlo/mhi name selection.
2667         (arc_restore_callee_saves): Don't early restore blink when ISR.
2668         (arc_expand_prologue): Add mlo/mhi saving.
2669         (arc_expand_epilogue): Add mlo/mhi restoring.
2670         (gen_mlo): Remove.
2671         (gen_mhi): Remove.
2672         * config/arc/arc.h (DBX_REGISTER_NUMBER): Correct register
2673         numbering when MUL64 option is used.
2674         (DWARF2_FRAME_REG_OUT): Define.
2675         * config/arc/arc.md (arc600_stall): New pattern.
2676         (VUNSPEC_ARC_ARC600_STALL): Define.
2677         (mulsi64): Use correct mlo/mhi registers.
2678         (mulsi_600): Clean it up.
2679         * config/arc/predicates.md (mlo_operand): Remove any dependency on
2680         TARGET_BIG_ENDIAN.
2681         (mhi_operand): Likewise.
2683 2020-01-27  Claudiu Zissulescu  <claziss@synopsys.com>
2684             Petro Karashchenko  <petro.karashchenko@ring.com>
2686         * config/arc/arc.c (arc_is_uncached_mem_p): Check struct
2687         attributes if needed.
2688         (prepare_move_operands): Generate special unspec instruction for
2689         direct access.
2690         (arc_isuncached_mem_p): Propagate uncached attribute to each
2691         structure member.
2692         * config/arc/arc.md (VUNSPEC_ARC_LDDI): Define.
2693         (VUNSPEC_ARC_STDI): Likewise.
2694         (ALLI): New mode iterator.
2695         (mALLI): New mode attribute.
2696         (lddi): New instruction pattern.
2697         (stdi): Likewise.
2698         (stdidi_split): Split instruction for architectures which are not
2699         supporting ll64 option.
2700         (lddidi_split): Likewise.
2702 2020-01-27  Richard Sandiford  <richard.sandiford@arm.com>
2704         PR rtl-optimization/92989
2705         * lra-lives.c (process_bb_lives): Update the live-in set before
2706         processing additional clobbers.
2708 2020-01-27  Richard Sandiford  <richard.sandiford@arm.com>
2710         PR rtl-optimization/93170
2711         * cselib.c (cselib_invalidate_regno_val): New function, split out
2712         from...
2713         (cselib_invalidate_regno): ...here.
2714         (cselib_invalidated_by_call_p): New function.
2715         (cselib_process_insn): Iterate over all the hard-register entries in
2716         REG_VALUES and invalidate any that cross call-clobbered registers.
2718 2020-01-27  Richard Sandiford  <richard.sandiford@arm.com>
2720         * dojump.c (split_comparison): Use HONOR_NANS rather than
2721         HONOR_SNANS when splitting LTGT.
2723 2020-01-27  Martin Liska  <mliska@suse.cz>
2725         PR driver/91220
2726         * opts.c (print_filtered_help): Exclude language-specific
2727         options from --help=common unless enabled in all FEs.
2729 2020-01-27  Martin Liska  <mliska@suse.cz>
2731         * opts.c (print_help): Exclude params from
2732         all except --help=param.
2734 2020-01-27  Martin Liska  <mliska@suse.cz>
2736         PR target/93274
2737         * config/i386/i386-features.c (make_resolver_func):
2738         Align the code with ppc64 target implementation.
2739         Do not generate a unique name for resolver function.
2741 2020-01-27  Richard Biener  <rguenther@suse.de>
2743         PR tree-optimization/93397
2744         * tree-vect-slp.c (vect_analyze_slp_instance): Delay
2745         converted reduction chain SLP graph adjustment.
2747 2020-01-26  Marek Polacek  <polacek@redhat.com>
2749         PR sanitizer/93436
2750         * sanopt.c (sanitize_rewrite_addressable_params): Avoid crash on
2751         null DECL_NAME.
2753 2020-01-26  Jason Merrill  <jason@redhat.com>
2755         PR c++/92601
2756         * tree.c (verify_type_variant): Only verify TYPE_NEEDS_CONSTRUCTING
2757         of complete types.
2759 2020-01-26  Darius Galis  <darius.galis@cyberthorstudios.com>
2761         * config/rx/rx.md (setmemsi): Added rx_allow_string_insns constraint
2762         (rx_setmem): Likewise.
2764 2020-01-26  Jakub Jelinek  <jakub@redhat.com>
2766         PR target/93412
2767         * config/i386/i386.md (*addv<dwi>4_doubleword, *subv<dwi>4_doubleword):
2768         Use nonimmediate_operand instead of x86_64_hilo_general_operand and
2769         drop <di> from constraint of last operand.
2771         PR target/93430
2772         * config/i386/sse.md (*avx_vperm_broadcast_<mode>): Disallow for
2773         TARGET_AVX2 and V4DFmode not in the split condition, but in the
2774         pattern condition, though allow { 0, 0, 0, 0 } broadcast always.
2776 2020-01-25  Feng Xue  <fxue@os.amperecomputing.com>
2778         PR ipa/93166
2779         * ipa-cp.c (get_info_about_necessary_edges): Remove value
2780         check assertion.
2782 2020-01-24  Jeff Law  <law@redhat.com>
2784         PR tree-optimization/92788
2785         * tree-ssa-threadedge.c (thread_across_edge): Check EDGE_COMPLEX
2786         not EDGE_ABNORMAL.
2788 2020-01-24  Jakub Jelinek  <jakub@redhat.com>
2790         PR target/93395
2791         * config/i386/sse.md (*avx_vperm_broadcast_v4sf,
2792         *avx_vperm_broadcast_<mode>,
2793         <sse2_avx_avx512f>_vpermil<mode><mask_name>,
2794         *<sse2_avx_avx512f>_vpermilp<mode><mask_name>):
2795         Move before avx2_perm<mode>/avx512f_perm<mode>.
2797         PR target/93376
2798         * simplify-rtx.c (simplify_const_unary_operation,
2799         simplify_const_binary_operation): Punt for mode precision above
2800         MAX_BITSIZE_MODE_ANY_INT.
2802 2020-01-24  Andrew Pinski  <apinski@marvell.com>
2804         * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): Change
2805         alu.shift_reg to 0.
2807 2020-01-24  Jeff Law  <law@redhat.com>
2809         PR target/13721
2810         * config/h8300/h8300.c (h8300_print_operand): Only call byte_reg
2811         for REGs.  Call output_operand_lossage to get more reasonable
2812         diagnostics.
2814 2020-01-24  Andrew Stubbs  <ams@codesourcery.com>
2816         * config/gcn/gcn-valu.md (vec_cmp<mode>di): Use
2817         gcn_fp_compare_operator.
2818         (vec_cmpu<mode>di): Use gcn_compare_operator.
2819         (vec_cmp<u>v64qidi): Use gcn_compare_operator.
2820         (vec_cmp<mode>di_exec): Use gcn_fp_compare_operator.
2821         (vec_cmpu<mode>di_exec): Use gcn_compare_operator.
2822         (vec_cmp<u>v64qidi_exec): Use gcn_compare_operator.
2823         (vec_cmp<mode>di_dup): Use gcn_fp_compare_operator.
2824         (vec_cmp<mode>di_dup_exec): Use gcn_fp_compare_operator.
2825         (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): Use
2826         gcn_fp_compare_operator.
2827         (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): Use
2828         gcn_fp_compare_operator.
2829         (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): Use
2830         gcn_fp_compare_operator.
2831         (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): Use
2832         gcn_fp_compare_operator.
2834 2020-01-24  Maciej W. Rozycki  <macro@wdc.com>
2836         * doc/install.texi (Cross-Compiler-Specific Options): Document
2837         `--with-toolexeclibdir' option.
2839 2020-01-24  Hans-Peter Nilsson  <hp@axis.com>
2841         * target.def (flags_regnum): Also mention effect on delay slot filling.
2842         * doc/tm.texi: Regenerate.
2844 2020-01-23  Jeff Law  <law@redhat.com>
2846         PR translation/90162
2847         * config/h8300/h8300.c (h8300_option_override): Fix diagnostic text.
2849 2020-01-23  Mikael Tillenius  <mti-1@tillenius.com>
2851         PR target/92269
2852         * config/h8300/h8300.h (FUNCTION_PROFILER): Fix emission of
2853         profiling label
2855 2020-01-23  Jakub Jelinek  <jakub@redhat.com>
2857         PR rtl-optimization/93402
2858         * postreload.c (reload_combine_recognize_pattern): Don't try to adjust
2859         USE insns.
2861 2020-01-23  Dragan Mladjenovic  <dmladjenovic@wavecomp.com>
2863         * config.in: Regenerated.
2864         * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to 1
2865         for TARGET_LIBC_GNUSTACK.
2866         * configure: Regenerated.
2867         * configure.ac: Define TARGET_LIBC_GNUSTACK if glibc version is
2868         found to be 2.31 or greater.
2870 2020-01-23  Dragan Mladjenovic  <dmladjenovic@wavecomp.com>
2872         * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to
2873         TARGET_SOFT_FLOAT.
2874         * config/mips/mips.c (TARGET_ASM_FILE_END): Define to ...
2875         (mips_asm_file_end): New function. Delegate to
2876         file_end_indicate_exec_stack if NEED_INDICATE_EXEC_STACK is true.
2877         * config/mips/mips.h (NEED_INDICATE_EXEC_STACK): Define to 0.
2879 2020-01-23  Jakub Jelinek  <jakub@redhat.com>
2881         PR target/93376
2882         * config/i386/i386-modes.def (POImode): New mode.
2883         (MAX_BITSIZE_MODE_ANY_INT): Change from 128 to 160.
2884         * config/i386/i386.md (DPWI): New mode attribute.
2885         (addv<mode>4, subv<mode>4): Use <DPWI> instead of <DWI>.
2886         (QWI): Rename to...
2887         (QPWI): ... this.  Use POI instead of OI for TImode.
2888         (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1,
2889         *subv<dwi>4_doubleword, *subv<dwi>4_doubleword_1): Use <QPWI>
2890         instead of <QWI>.
2892 2020-01-23  Richard Sandiford  <richard.sandiford@arm.com>
2894         PR target/93341
2895         * config/aarch64/aarch64.md (UNSPEC_SPECULATION_TRACKER_REV): New
2896         unspec.
2897         (speculation_tracker_rev): New pattern.
2898         * config/aarch64/aarch64-speculation.cc (aarch64_do_track_speculation):
2899         Use speculation_tracker_rev to track the inverse condition.
2901 2020-01-23  Richard Biener  <rguenther@suse.de>
2903         PR tree-optimization/93381
2904         * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Take
2905         alias-set of the def as argument and record the first one.
2906         (vn_walk_cb_data::first_set): New member.
2907         (vn_reference_lookup_3): Pass the alias-set of the current def
2908         to push_partial_def.  Fix alias-set used in the aggregate copy
2909         case.
2910         (vn_reference_lookup): Consistently set *last_vuse_ptr.
2911         * real.c (clear_significand_below): Fix out-of-bound access.
2913 2020-01-23  Jakub Jelinek  <jakub@redhat.com>
2915         PR target/93346
2916         * config/i386/i386.md (*bmi2_bzhi_<mode>3_2, *bmi2_bzhi_<mode>3_3):
2917         New define_insn patterns.
2919 2020-01-23  Richard Sandiford  <richard.sandiford@arm.com>
2921         * doc/sourcebuild.texi (check-function-bodies): Add an
2922         optional target/xfail selector.
2924 2020-01-23  Richard Sandiford  <richard.sandiford@arm.com>
2926         PR rtl-optimization/93124
2927         * auto-inc-dec.c (merge_in_block): Don't add auto inc/decs to
2928         bare USE and CLOBBER insns.
2930 2020-01-22  Andrew Pinski  <apinski@marvell.com>
2932         * config/arc/arc.c (output_short_suffix): Check insn for nullness.
2934 2020-01-22  David Malcolm  <dmalcolm@redhat.com>
2936         PR analyzer/93307
2937         * gdbinit.in (break-on-saved-diagnostic): Update for move of
2938         diagnostic_manager into "ana" namespace.
2939         * selftest-run-tests.c (selftest::run_tests): Update for move of
2940         selftest::run_analyzer_selftests to
2941         ana::selftest::run_analyzer_selftests.
2943 2020-01-22  Richard Sandiford  <richard.sandiford@arm.com>
2945         * cfgexpand.c (union_stack_vars): Update the size.
2947 2020-01-22  Richard Biener  <rguenther@suse.de>
2949         PR tree-optimization/93381
2950         * tree-ssa-structalias.c (find_func_aliases): Assume offsetting
2951         throughout, handle all conversions the same.
2953 2020-01-22  Jakub Jelinek  <jakub@redhat.com>
2955         PR target/93335
2956         * config/aarch64/aarch64.c (aarch64_expand_subvti): Only use
2957         gen_subdi3_compare1_imm if low_in2 satisfies aarch64_plus_immediate
2958         predicate, not whenever it is CONST_INT.  Otherwise, force_reg it.
2959         Call force_reg on high_in2 unconditionally.
2961 2020-01-22  Martin Liska  <mliska@suse.cz>
2963         PR tree-optimization/92924
2964         * profile.c (compute_value_histograms): Divide
2965         all counter values.
2967 2020-01-22  Jakub Jelinek  <jakub@redhat.com>
2969         PR target/91298
2970         * output.h (assemble_name_resolve): Declare.
2971         * varasm.c (assemble_name_resolve): New function.
2972         (assemble_name): Use it.
2973         * config/i386/i386.h (ASM_OUTPUT_SYMBOL_REF): Define.
2975 2020-01-22  Joseph Myers  <joseph@codesourcery.com>
2977         * doc/sourcebuild.texi (Texinfo Manuals, Front End): Refer to
2978         update_web_docs_git instead of update_web_docs_svn.
2980 2020-01-21  Andrew Pinski  <apinski@marvell.com>
2982         PR target/9311
2983         * config/aarch64/aarch64.md (tlsgd_small_<mode>): Have operand 0
2984         as PTR mode. Have operand 1 as being modeless, it can be P mode.
2985         (*tlsgd_small_<mode>): Likewise.
2986         * config/aarch64/aarch64.c (aarch64_load_symref_appropriately)
2987         <case SYMBOL_SMALL_TLSGD>: Call gen_tlsgd_small_* with a ptr_mode
2988         register.  Convert that register back to dest using convert_mode.
2990 2020-01-21  Jim Wilson  <jimw@sifive.com>
2992         * config/riscv/riscv-sr.c (riscv_sr_match_prologue): Use INTVAL
2993         instead of XINT.
2995 2020-01-21  H.J. Lu  <hongjiu.lu@intel.com>
2996             Uros Bizjak    <ubizjak@gmail.com>
2998         PR target/93319
2999         * config/i386/i386.c (ix86_tls_module_base): Replace Pmode
3000         with ptr_mode.
3001         (legitimize_tls_address): Do GNU2 TLS address computation in
3002         ptr_mode and zero-extend result to Pmode.
3003         *  config/i386/i386.md (@tls_dynamic_gnu2_64_<mode>): Replace
3004         :P with :PTR and Pmode with ptr_mode.
3005         (*tls_dynamic_gnu2_lea_64_<mode>): Likewise.
3006         (*tls_dynamic_gnu2_call_64_<mode>): Likewise.
3007         (*tls_dynamic_gnu2_combine_64_<mode>): Likewise.
3009 2020-01-21  Jakub Jelinek  <jakub@redhat.com>
3011         PR target/93333
3012         * config/riscv/riscv.c (riscv_rtx_costs) <case ZERO_EXTRACT>: Verify
3013         the last two operands are CONST_INT_P before using them as such.
3015 2020-01-21  Richard Sandiford  <richard.sandiford@arm.com>
3017         * config/aarch64/aarch64-sve-builtins.def: Use get_typenode_from_name
3018         to get the integer element types.
3020 2020-01-21  Richard Sandiford  <richard.sandiford@arm.com>
3022         * config/aarch64/aarch64-sve-builtins.h
3023         (function_expander::convert_to_pmode): Declare.
3024         * config/aarch64/aarch64-sve-builtins.cc
3025         (function_expander::convert_to_pmode): New function.
3026         (function_expander::get_contiguous_base): Use it.
3027         (function_expander::prepare_gather_address_operands): Likewise.
3028         * config/aarch64/aarch64-sve-builtins-sve2.cc
3029         (svwhilerw_svwhilewr_impl::expand): Likewise.
3031 2020-01-21  Szabolcs Nagy  <szabolcs.nagy@arm.com>
3033         PR target/92424
3034         * config/aarch64/aarch64.c (aarch64_declare_function_name): Set
3035         cfun->machine->label_is_assembled.
3036         (aarch64_print_patchable_function_entry): New.
3037         (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY): Define.
3038         * config/aarch64/aarch64.h (struct machine_function): New field,
3039         label_is_assembled.
3041 2020-01-21  David Malcolm  <dmalcolm@redhat.com>
3043         PR ipa/93315
3044         * ipa-profile.c (ipa_profile): Delete call_sums and set it to
3045         NULL on exit.
3047 2020-01-18  Jan Hubicka  <hubicka@ucw.cz>
3049         PR lto/93318    
3050         * cgraph.c (cgraph_edge::resolve_speculation,
3051         cgraph_edge::redirect_call_stmt_to_callee): Fix update of
3052         call_stmt_site_hash.
3054 2020-01-21  Martin Liska  <mliska@suse.cz>
3056         * config/rs6000/rs6000.c (common_mode_defined): Remove
3057         unused variable.
3059 2020-01-21  Richard Biener  <rguenther@suse.de>
3061         PR tree-optimization/92328
3062         * tree-ssa-sccvn.c (vn_reference_lookup_3): Preserve
3063         type when value-numbering same-sized store by inserting a
3064         VIEW_CONVERT_EXPR.
3065         (eliminate_dom_walker::eliminate_stmt): When eliminating
3066         a redundant store handle bit-reinterpretation of the same value.
3068 2020-01-21  Andrew Pinski  <apinski@marvel.com>
3070         PR tree-opt/93321
3071         * tree-into-ssa.c (prepare_block_for_update_1): Split out
3072         from ...
3073         (prepare_block_for_update): This.  Use a worklist instead of
3074         recursing.
3076 2020-01-21  Mihail-Calin Ionescu  <mihail.ionescu@arm.com>
3078         * gcc/config/arm/arm.c (clear_operation_p):
3079         Initialise last_regno, skip first iteration
3080         based on the first_set value and use ints instead
3081         of the unnecessary HOST_WIDE_INTs.
3083 2020-01-21  Jakub Jelinek  <jakub@redhat.com>
3085         PR target/93073
3086         * config/rs6000/rs6000.c (rs6000_emit_cmove): If using fsel, punt for
3087         compare_mode other than SFmode or DFmode.
3089 2020-01-21  Kito Cheng  <kito.cheng@sifive.com>
3091         PR target/93304
3092         * config/riscv/riscv-protos.h (riscv_hard_regno_rename_ok): New.
3093         * config/riscv/riscv.c (riscv_hard_regno_rename_ok): New.
3094         * config/riscv/riscv.h (HARD_REGNO_RENAME_OK): Defined.
3096 2020-01-20  Wilco Dijkstra  <wdijkstr@arm.com>
3098         * config/aarch64/aarch64.c (neoversen1_tunings): Set jump_align to 4.
3100 2020-01-20  Andrew Pinski  <apinski@marvell.com>
3102         PR middle-end/93242
3103         * targhooks.c (default_print_patchable_function_entry): Use
3104         output_asm_insn to emit the nop instruction.
3106 2020-01-20  Fangrui Song  <maskray@google.com>
3108         PR middle-end/93194
3109         * targhooks.c (default_print_patchable_function_entry): Align to
3110         POINTER_SIZE.
3112 2020-01-20  H.J. Lu  <hongjiu.lu@intel.com>
3114         PR target/93319
3115         * config/i386/i386.c (legitimize_tls_address): Pass Pmode to
3116         gen_tls_dynamic_gnu2_64.  Compute GNU2 TLS address in ptr_mode.
3117         * config/i386/i386.md (tls_dynamic_gnu2_64): Renamed to ...
3118         (@tls_dynamic_gnu2_64_<mode>): This.  Replace DI with P.
3119         (*tls_dynamic_gnu2_lea_64): Renamed to ...
3120         (*tls_dynamic_gnu2_lea_64_<mode>): This.  Replace DI with P.
3121         Remove the {q} suffix from lea.
3122         (*tls_dynamic_gnu2_call_64): Renamed to ...
3123         (*tls_dynamic_gnu2_call_64_<mode>): This.  Replace DI with P.
3124         (*tls_dynamic_gnu2_combine_64): Renamed to ...
3125         (*tls_dynamic_gnu2_combine_64_<mode>): This.  Replace DI with P.
3126         Pass Pmode to gen_tls_dynamic_gnu2_64.
3128 2020-01-20  Wilco Dijkstra  <wdijkstr@arm.com>
3130         * config/aarch64/aarch64.h (SLOW_BYTE_ACCESS): Set to 1.
3132 2020-01-20  Richard Sandiford  <richard.sandiford@arm.com>
3134         * config/aarch64/aarch64-sve-builtins-base.cc
3135         (svld1ro_impl::memory_vector_mode): Remove parameter name.
3137 2020-01-20  Richard Biener  <rguenther@suse.de>
3139         PR debug/92763
3140         * dwarf2out.c (prune_unused_types): Unconditionally mark
3141         called function DIEs.
3143 2020-01-20  Martin Liska  <mliska@suse.cz>
3145         PR tree-optimization/93199
3146         * tree-eh.c (struct leh_state): Add
3147         new field outer_non_cleanup.
3148         (cleanup_is_dead_in): Pass leh_state instead
3149         of eh_region.  Add a checking that state->outer_non_cleanup
3150         points to outer non-clean up region.
3151         (lower_try_finally): Record outer_non_cleanup
3152         for this_state.
3153         (lower_catch): Likewise.
3154         (lower_eh_filter): Likewise.
3155         (lower_eh_must_not_throw): Likewise.
3156         (lower_cleanup): Likewise.
3158 2020-01-20  Richard Biener  <rguenther@suse.de>
3160         PR tree-optimization/93094
3161         * tree-vectorizer.h (vect_loop_versioning): Adjust.
3162         (vect_transform_loop): Likewise.
3163         * tree-vectorizer.c (try_vectorize_loop_1): Pass down
3164         loop_vectorized_call to vect_transform_loop.
3165         * tree-vect-loop.c (vect_transform_loop): Pass down
3166         loop_vectorized_call to vect_loop_versioning.
3167         * tree-vect-loop-manip.c (vect_loop_versioning): Use
3168         the earlier discovered loop_vectorized_call.
3170 2020-01-19  Eric S. Raymond <esr@thyrsus.com>
3172         * doc/contribute.texi: Update for SVN -> Git transition.
3173         * doc/install.texi: Likewise.
3175 2020-01-18  Jan Hubicka  <hubicka@ucw.cz>
3177         PR lto/93318
3178         * cgraph.c (cgraph_edge::make_speculative): Increase number of
3179         speculative targets.
3180         (verify_speculative_call): New function
3181         (cgraph_node::verify_node): Use it.
3182         * ipa-profile.c (ipa_profile): Fix formating; do not set number of
3183         speculations.
3185 2020-01-18  Jan Hubicka  <hubicka@ucw.cz>
3187         PR lto/93318
3188         * cgraph.c (cgraph_edge::resolve_speculation): Fix foramting.
3189         (cgraph_edge::make_direct): Remove all indirect targets.
3190         (cgraph_edge::redirect_call_stmt_to_callee): Use make_direct..
3191         (cgraph_node::verify_node): Verify that only one call_stmt or
3192         lto_stmt_uid is set.
3193         * cgraphclones.c (cgraph_edge::clone): Set only one call_stmt or
3194         lto_stmt_uid.
3195         * lto-cgraph.c (lto_output_edge): Simplify streaming of stmt.
3196         (lto_output_ref): Simplify streaming of stmt.
3197         * lto-streamer-in.c (fixup_call_stmt_edges_1): Clear lto_stmt_uid.
3199 2020-01-18  Tamar Christina  <tamar.christina@arm.com>
3201         * config/aarch64/aarch64-sve-builtins-base.cc (memory_vector_mode):
3202         Mark parameter unused.
3204 2020-01-18  Hans-Peter Nilsson  <hp@axis.com>
3206         * config.gcc <obsolete targets>: Add crisv32-*-* and cris-*-linux*
3208 2019-01-18  Gerald Pfeifer  <gerald@pfeifer.com>
3210         * varpool.c (ctor_useable_for_folding_p): Fix grammar.
3212 2020-01-18  Iain Sandoe  <iain@sandoe.co.uk>
3214         * Makefile.in: Add coroutine-passes.o.
3215         * builtin-types.def (BT_CONST_SIZE): New.
3216         (BT_FN_BOOL_PTR): New.
3217         (BT_FN_PTR_PTR_CONST_SIZE_BOOL): New.
3218         * builtins.def (DEF_COROUTINE_BUILTIN): New.
3219         * coroutine-builtins.def: New file.
3220         * coroutine-passes.cc: New file.
3221         * function.h (struct GTY function): Add a bit to indicate that the
3222         function is a coroutine component.
3223         * internal-fn.c (expand_CO_FRAME): New.
3224         (expand_CO_YIELD): New.
3225         (expand_CO_SUSPN): New.
3226         (expand_CO_ACTOR): New.
3227         * internal-fn.def (CO_ACTOR): New.
3228         (CO_YIELD): New.
3229         (CO_SUSPN): New.
3230         (CO_FRAME): New.
3231         * passes.def: Add pass_coroutine_lower_builtins,
3232         pass_coroutine_early_expand_ifns.
3233         * tree-pass.h (make_pass_coroutine_lower_builtins): New.
3234         (make_pass_coroutine_early_expand_ifns): New.
3235         * doc/invoke.texi: Document the fcoroutines command line
3236         switch.
3238 2020-01-18  Jakub Jelinek  <jakub@redhat.com>
3240         * config/arm/vfp.md (*clear_vfp_multiple): Remove unused variable.
3242         PR target/93312
3243         * config/arm/arm.c (clear_operation_p): Don't use REGNO until
3244         after checking the argument is a REG.  Don't use REGNO (reg)
3245         again to set last_regno, reuse regno variable instead.
3247 2020-01-17  David Malcolm  <dmalcolm@redhat.com>
3249         * doc/analyzer.texi (Limitations): Add note about NaN.
3251 2020-01-17  Mihail-Calin Ionescu  <mihail.ionescu@arm.com>
3252             Sudakshina Das  <sudi.das@arm.com>
3254         * config/arm/arm.md (ashldi3): Generate thumb2_lsll for both reg
3255         and valid immediate.
3256         (ashrdi3): Generate thumb2_asrl for both reg and valid immediate.
3257         (lshrdi3): Generate thumb2_lsrl for valid immediates.
3258         * config/arm/constraints.md (Pg): New.
3259         * config/arm/predicates.md (long_shift_imm): New.
3260         (arm_reg_or_long_shift_imm): Likewise.
3261         * config/arm/thumb2.md (thumb2_asrl): New immediate alternative.
3262         (thumb2_lsll): Likewise.
3263         (thumb2_lsrl): New.
3265 2020-01-17  Mihail-Calin Ionescu  <mihail.ionescu@arm.com>
3266             Sudakshina Das  <sudi.das@arm.com>
3268         * config/arm/arm.md (ashldi3): Generate thumb2_lsll for TARGET_HAVE_MVE.
3269         (ashrdi3): Generate thumb2_asrl for TARGET_HAVE_MVE.
3270         * config/arm/arm.c (arm_hard_regno_mode_ok): Allocate even odd
3271         register pairs for doubleword quantities for ARMv8.1M-Mainline.
3272         * config/arm/thumb2.md (thumb2_asrl): New.
3273         (thumb2_lsll): Likewise.
3275 2020-01-17  Jakub Jelinek  <jakub@redhat.com>
3277         * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear): Remove
3278         unused variable.
3280 2020-01-17  Alexander Monakov  <amonakov@ispras.ru>
3282         * gdbinit.in (help-gcc-hooks): New command.
3283         (pp, pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, ptc, pdn, ptn, pdd, prc,
3284         pi, pbm, pel, trt): Take $arg0 instead of $ if supplied. Update
3285         documentation.
3287 2020-01-17  Matthew Malcomson  <matthew.malcomson@arm.com>
3289         * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use the
3290         correct target macro.
3292 2020-01-17  Matthew Malcomson  <matthew.malcomson@arm.com>
3294         * config/aarch64/aarch64-protos.h
3295         (aarch64_sve_ld1ro_operand_p): New.
3296         * config/aarch64/aarch64-sve-builtins-base.cc
3297         (class load_replicate): New.
3298         (class svld1ro_impl): New.
3299         (class svld1rq_impl): Change to inherit from load_replicate.
3300         (svld1ro): New sve intrinsic function base.
3301         * config/aarch64/aarch64-sve-builtins-base.def (svld1ro):
3302         New DEF_SVE_FUNCTION.
3303         * config/aarch64/aarch64-sve-builtins-base.h
3304         (svld1ro): New decl.
3305         * config/aarch64/aarch64-sve-builtins.cc
3306         (function_expander::add_mem_operand): Modify assert to allow
3307         OImode.
3308         * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): New
3309         pattern.
3310         * config/aarch64/aarch64.c
3311         (aarch64_sve_ld1rq_operand_p): Implement in terms of ...
3312         (aarch64_sve_ld1rq_ld1ro_operand_p): This.
3313         (aarch64_sve_ld1ro_operand_p): New.
3314         * config/aarch64/aarch64.md (UNSPEC_LD1RO): New unspec.
3315         * config/aarch64/constraints.md (UOb,UOh,UOw,UOd): New.
3316         * config/aarch64/predicates.md
3317         (aarch64_sve_ld1ro_operand_{b,h,w,d}): New.
3319 2020-01-17  Matthew Malcomson  <matthew.malcomson@arm.com>
3321         * config/aarch64/aarch64-c.c (_ARM_FEATURE_MATMUL_FLOAT64):
3322         Introduce this ACLE specified predefined macro.
3323         * config/aarch64/aarch64-option-extensions.def (f64mm): New.
3324         (fp): Disabling this disables f64mm.
3325         (simd): Disabling this disables f64mm.
3326         (fp16): Disabling this disables f64mm.
3327         (sve): Disabling this disables f64mm.
3328         * config/aarch64/aarch64.h (AARCH64_FL_F64MM): New.
3329         (AARCH64_ISA_F64MM): New.
3330         (TARGET_F64MM): New.
3331         * doc/invoke.texi (f64mm): Document new option.
3333 2020-01-17  Wilco Dijkstra  <wdijkstr@arm.com>
3335         * config/aarch64/aarch64.c (generic_tunings): Add branch fusion.
3336         (neoversen1_tunings): Likewise.
3338 2020-01-17  Wilco Dijkstra  <wdijkstr@arm.com>
3340         PR target/92692
3341         * config/aarch64/aarch64.c (aarch64_split_compare_and_swap)
3342         Add assert to ensure prolog has been emitted.
3343         (aarch64_split_atomic_op): Likewise.
3344         * config/aarch64/atomics.md (aarch64_compare_and_swap<mode>)
3345         Use epilogue_completed rather than reload_completed.
3346         (aarch64_atomic_exchange<mode>): Likewise.
3347         (aarch64_atomic_<atomic_optab><mode>): Likewise.
3348         (atomic_nand<mode>): Likewise.
3349         (aarch64_atomic_fetch_<atomic_optab><mode>): Likewise.
3350         (atomic_fetch_nand<mode>): Likewise.
3351         (aarch64_atomic_<atomic_optab>_fetch<mode>): Likewise.
3352         (atomic_nand_fetch<mode>): Likewise.
3354 2020-01-17  Richard Sandiford  <richard.sandiford@arm.com>
3356         PR target/93133
3357         * config/aarch64/aarch64.h (REVERSIBLE_CC_MODE): Return false
3358         for FP modes.
3359         (REVERSE_CONDITION): Delete.
3360         * config/aarch64/iterators.md (CC_ONLY): New mode iterator.
3361         (CCFP_CCFPE): Likewise.
3362         (e): New mode attribute.
3363         * config/aarch64/aarch64.md (ccmp<GPI:mode>): Rename to...
3364         (@ccmp<CC_ONLY:mode><GPI:mode>): ...this, using CC_ONLY instead of CC.
3365         (fccmp<GPF:mode>, fccmpe<GPF:mode>): Merge into...
3366         (@ccmp<CCFP_CCFPE:mode><GPF:mode>): ...this combined pattern.
3367         (@ccmp<CC_ONLY:mode><GPI:mode>_rev): New pattern.
3368         (@ccmp<CCFP_CCFPE:mode><GPF:mode>_rev): Likewise.
3369         * config/aarch64/aarch64.c (aarch64_gen_compare_reg): Update
3370         name of generator from gen_ccmpdi to gen_ccmpccdi.
3371         (aarch64_gen_ccmp_next): Use code_for_ccmp.  If we want to reverse
3372         the previous comparison but aren't able to, use the new ccmp_rev
3373         patterns instead.
3375 2020-01-17  Richard Sandiford  <richard.sandiford@arm.com>
3377         * gimplify.c (gimplify_return_expr): Use poly_int_tree_p rather
3378         than testing directly for INTEGER_CST.
3379         (gimplify_target_expr, gimplify_omp_depend): Likewise.
3381 2020-01-17  Jakub Jelinek  <jakub@redhat.com>
3383         PR tree-optimization/93292
3384         * tree-vect-stmts.c (vectorizable_comparison): Punt also if
3385         get_vectype_for_scalar_type returns NULL.
3387 2020-01-16  Jan Hubicka  <hubicka@ucw.cz>
3389         * params.opt (-param=max-predicted-iterations): Increase range from 0.
3390         * predict.c (estimate_loops): Add 1 to param_max_predicted_iterations.
3392 2020-01-16  Jan Hubicka  <hubicka@ucw.cz>
3394         * ipa-fnsummary.c (estimate_calls_size_and_time): Fix formating of
3395         dump.
3396         * params.opt: (max-predicted-iterations): Set bounds.
3397         * predict.c (real_almost_one, real_br_prob_base,
3398         real_inv_br_prob_base, real_one_half, real_bb_freq_max): Remove.
3399         (propagate_freq): Add max_cyclic_prob parameter; cap cyclic
3400         probabilities; do not truncate to reg_br_prob_bases.
3401         (estimate_loops_at_level): Pass max_cyclic_prob.
3402         (estimate_loops): Compute max_cyclic_prob.
3403         (estimate_bb_frequencies): Do not initialize real_*; update calculation
3404         of back edge prob.
3405         * profile-count.c (profile_probability::to_sreal): New.
3406         * profile-count.h (class sreal): Move up in file.
3407         (profile_probability::to_sreal): Declare.
3409 2020-01-16  Stam Markianos-Wright  <stam.markianos-wright@arm.com>
3411         * config/arm/arm.c
3412         (arm_invalid_conversion): New function for target hook.
3413         (arm_invalid_unary_op): New function for target hook.
3414         (arm_invalid_binary_op): New function for target hook.
3416 2020-01-16  Stam Markianos-Wright  <stam.markianos-wright@arm.com>
3418         * config.gcc: Add arm_bf16.h.
3419         * config/arm/arm-builtins.c (arm_mangle_builtin_type): Fix comment.
3420         (arm_simd_builtin_std_type): Add BFmode.
3421         (arm_init_simd_builtin_types): Define element types for vector types.
3422         (arm_init_bf16_types): New function.
3423         (arm_init_builtins): Add arm_init_bf16_types function call.
3424         * config/arm/arm-modes.def: Add BFmode and V4BF, V8BF vector modes.
3425         * config/arm/arm-simd-builtin-types.def: Add V4BF, V8BF.
3426         * config/arm/arm.c (aapcs_vfp_sub_candidate):  Add BFmode.
3427         (arm_hard_regno_mode_ok): Add BFmode and tidy up statements.
3428         (arm_vector_mode_supported_p): Add V4BF, V8BF.
3429         (arm_mangle_type):  Add __bf16.
3430         * config/arm/arm.h: Add V4BF, V8BF to VALID_NEON_DREG_MODE, 
3431         VALID_NEON_QREG_MODE respectively. Add export arm_bf16_type_node,
3432         arm_bf16_ptr_type_node.
3433         * config/arm/arm.md: Add BFmode to movhf expand, mov pattern and
3434         define_split between ARM registers.
3435         * config/arm/arm_bf16.h: New file.
3436         * config/arm/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
3437         * config/arm/iterators.md: (ANY64_BF, VDXMOV, VHFBF, HFBF, fporbf): New.
3438         (VQXMOV): Add V8BF.
3439         * config/arm/neon.md: Add BF vector types to movhf NEON move patterns.
3440         * config/arm/vfp.md: Add BFmode to movhf patterns.
3442 2020-01-16  Mihail Ionescu  <mihail.ionescu@arm.com>
3443             Andre Vieira  <andre.simoesdiasvieira@arm.com>
3445         * config/arm/arm-cpus.in (mve, mve_float): New features.
3446         (dsp, mve, mve.fp): New options.
3447         * config/arm/arm.h (TARGET_HAVE_MVE, TARGET_HAVE_MVE_FLOAT): Define.
3448         * config/arm/t-rmprofile: Map v8.1-M multilibs to v8-M.
3449         * doc/invoke.texi: Document the armv8.1-m mve and dps options.
3451 2020-01-16  Mihail-Calin Ionescu  <mihail.ionescu@arm.com>
3452             Thomas Preud'homme  <thomas.preudhomme@arm.com>
3454         * config/arm/arm-cpus.in (ARMv8_1m_main): Redefine as an extension to
3455         Armv8-M Mainline.
3456         * config/arm/arm.c (arm_options_perform_arch_sanity_checks): Remove
3457         error for using -mcmse when targeting Armv8.1-M Mainline.
3459 2020-01-16  Mihail-Calin Ionescu  <mihail.ionescu@arm.com>
3460             Thomas Preud'homme  <thomas.preudhomme@arm.com>
3462         * config/arm/arm.md (nonsecure_call_internal): Do not force memory
3463         address in r4 when targeting Armv8.1-M Mainline.
3464         (nonsecure_call_value_internal): Likewise.
3465         * config/arm/thumb2.md (nonsecure_call_reg_thumb2): Make memory address
3466         a register match_operand again.  Emit BLXNS when targeting
3467         Armv8.1-M Mainline.
3468         (nonsecure_call_value_reg_thumb2): Likewise.
3470 2020-01-16  Mihail-Calin Ionescu  <mihail.ionescu@arm.com>
3471             Thomas Preud'homme  <thomas.preudhomme@arm.com>
3473         * config/arm/arm.c (arm_add_cfa_adjust_cfa_note): Declare early.
3474         (cmse_nonsecure_call_inline_register_clear): Define new lazy_fpclear
3475         variable as true when floating-point ABI is not hard.  Replace
3476         check against TARGET_HARD_FLOAT_ABI by checks against lazy_fpclear.
3477         Generate VLSTM and VLLDM instruction respectively before and
3478         after a function call to cmse_nonsecure_call function.
3479         * config/arm/unspecs.md (VUNSPEC_VLSTM): Define unspec.
3480         (VUNSPEC_VLLDM): Likewise.
3481         * config/arm/vfp.md (lazy_store_multiple_insn): New define_insn.
3482         (lazy_load_multiple_insn): Likewise.
3484 2020-01-16  Mihail-Calin Ionescu  <mihail.ionescu@arm.com>
3485             Thomas Preud'homme  <thomas.preudhomme@arm.com>
3487         * config/arm/arm.c (vfp_emit_fstmd): Declare early.
3488         (arm_emit_vfp_multi_reg_pop): Likewise.
3489         (cmse_nonsecure_call_inline_register_clear): Abstract number of VFP
3490         registers to clear in max_fp_regno.  Emit VPUSH and VPOP to save and
3491         restore callee-saved VFP registers.
3493 2020-01-16  Mihail-Calin Ionescu  <mihail.ionescu@arm.com>
3494             Thomas Preud'homme  <thomas.preudhomme@arm.com>
3496         * config/arm/arm.c (arm_emit_multi_reg_pop): Declare early.
3497         (cmse_nonsecure_call_clear_caller_saved): Rename into ...
3498         (cmse_nonsecure_call_inline_register_clear): This.  Save and clear
3499         callee-saved GPRs as well as clear ip register before doing a nonsecure
3500         call then restore callee-saved GPRs after it when targeting
3501         Armv8.1-M Mainline.
3502         (arm_reorg): Adapt to function rename.
3504 2020-01-16  Mihail-Calin Ionescu  <mihail.ionescu@arm.com>
3505             Thomas Preud'homme  <thomas.preudhomme@arm.com>
3507         * config/arm/arm-protos.h (clear_operation_p): Adapt prototype.
3508         * config/arm/arm.c (clear_operation_p): Extend to be able to check a
3509         clear_vfp_multiple pattern based on a new vfp parameter.
3510         (cmse_clear_registers): Generate VSCCLRM to clear VFP registers when
3511         targeting Armv8.1-M Mainline.
3512         (cmse_nonsecure_entry_clear_before_return): Clear VFP registers
3513         unconditionally when targeting Armv8.1-M Mainline architecture.  Check
3514         whether VFP registers are available before looking call_used_regs for a
3515         VFP register.
3516         * config/arm/predicates.md (clear_multiple_operation): Adapt to change
3517         of prototype of clear_operation_p.
3518         (clear_vfp_multiple_operation): New predicate.
3519         * config/arm/unspecs.md (VUNSPEC_VSCCLRM_VPR): New volatile unspec.
3520         * config/arm/vfp.md (clear_vfp_multiple): New define_insn.
3522 2020-01-16  Mihail-Calin Ionescu  <mihail.ionescu@arm.com>
3523             Thomas Preud'homme  <thomas.preudhomme@arm.com>
3525         * config/arm/arm-protos.h (clear_operation_p): Declare.
3526         * config/arm/arm.c (clear_operation_p): New function.
3527         (cmse_clear_registers): Generate clear_multiple instruction pattern if
3528         targeting Armv8.1-M Mainline or successor.
3529         (output_return_instruction): Only output APSR register clearing if
3530         Armv8.1-M Mainline instructions not available.
3531         (thumb_exit): Likewise.
3532         * config/arm/predicates.md (clear_multiple_operation): New predicate.
3533         * config/arm/thumb2.md (clear_apsr): New define_insn.
3534         (clear_multiple): Likewise.
3535         * config/arm/unspecs.md (VUNSPEC_CLRM_APSR): New volatile unspec.
3537 2020-01-16  Mihail-Calin Ionescu  <mihail.ionescu@arm.com>
3538             Thomas Preud'homme  <thomas.preudhomme@arm.com>
3540         * config/arm/arm.c (fp_sysreg_names): Declare and define.
3541         (use_return_insn): Also return false for Armv8.1-M Mainline.
3542         (output_return_instruction): Skip FPSCR clearing if Armv8.1-M
3543         Mainline instructions are available.
3544         (arm_compute_frame_layout): Allocate space in frame for FPCXTNS
3545         when targeting Armv8.1-M Mainline Security Extensions.
3546         (arm_expand_prologue): Save FPCXTNS if this is an Armv8.1-M
3547         Mainline entry function.
3548         (cmse_nonsecure_entry_clear_before_return): Clear IP and r4 if
3549         targeting Armv8.1-M Mainline or successor.
3550         (arm_expand_epilogue): Fix indentation of caller-saved register
3551         clearing.  Restore FPCXTNS if this is an Armv8.1-M Mainline
3552         entry function.
3553         * config/arm/arm.h (TARGET_HAVE_FP_CMSE): New macro.
3554         (FP_SYSREGS): Likewise.
3555         (enum vfp_sysregs_encoding): Define enum.
3556         (fp_sysreg_names): Declare.
3557         * config/arm/unspecs.md (VUNSPEC_VSTR_VLDR): New volatile unspec.
3558         * config/arm/vfp.md (push_fpsysreg_insn): New define_insn.
3559         (pop_fpsysreg_insn): Likewise.
3561 2020-01-16  Mihail-Calin Ionescu  <mihail.ionescu@arm.com>
3562             Thomas Preud'homme  <thomas.preudhomme@arm.com>
3564         * config/arm/arm-cpus.in (armv8_1m_main): New feature.
3565         (ARMv4, ARMv4t, ARMv5t, ARMv5te, ARMv5tej, ARMv6, ARMv6j, ARMv6k,
3566         ARMv6z, ARMv6kz, ARMv6zk, ARMv6t2, ARMv6m, ARMv7, ARMv7a, ARMv7ve,
3567         ARMv7r, ARMv7m, ARMv7em, ARMv8a, ARMv8_1a, ARMv8_2a, ARMv8_3a,
3568         ARMv8_4a, ARMv8_5a, ARMv8m_base, ARMv8m_main, ARMv8r): Reindent.
3569         (ARMv8_1m_main): New feature group.
3570         (armv8.1-m.main): New architecture.
3571         * config/arm/arm-tables.opt: Regenerate.
3572         * config/arm/arm.c (arm_arch8_1m_main): Define and default initialize.
3573         (arm_option_reconfigure_globals): Initialize arm_arch8_1m_main.
3574         (arm_options_perform_arch_sanity_checks): Error out when targeting
3575         Armv8.1-M Mainline Security Extensions.
3576         * config/arm/arm.h (arm_arch8_1m_main): Declare.
3578 2020-01-16  Stam Markianos-Wright  <stam.markianos-wright@arm.com>
3580         * config/aarch64/aarch64-simd-builtins.def (aarch64_bfdot,
3581         aarch64_bfdot_lane, aarch64_bfdot_laneq): New.
3582         * config/aarch64/aarch64-simd.md (aarch64_bfdot, aarch64_bfdot_lane,
3583         aarch64_bfdot_laneq): New.
3584         * config/aarch64/arm_bf16.h (vbfdot_f32, vbfdotq_f32,
3585         vbfdot_lane_f32, vbfdotq_lane_f32, vbfdot_laneq_f32,
3586         vbfdotq_laneq_f32): New.
3587         * config/aarch64/iterators.md (UNSPEC_BFDOT, Vbfdottype,
3588         VBFMLA_W, VBF): New.
3589         (isquadop): Add V4BF, V8BF.
3591 2020-01-16  Stam Markianos-Wright  <stam.markianos-wright@arm.com>
3593         * config/aarch64/aarch64-builtins.c: (enum aarch64_type_qualifiers):
3594         New qualifier_lane_quadtup_index, TYPES_TERNOP_SSUS,
3595         TYPES_QUADOPSSUS_LANE_QUADTUP, TYPES_QUADOPSSSU_LANE_QUADTUP.
3596         (aarch64_simd_expand_args): Add case SIMD_ARG_LANE_QUADTUP_INDEX.
3597         (aarch64_simd_expand_builtin): Add qualifier_lane_quadtup_index.
3598         * config/aarch64/aarch64-simd-builtins.def (usdot, usdot_lane,
3599         usdot_laneq, sudot_lane,sudot_laneq): New.
3600         * config/aarch64/aarch64-simd.md (aarch64_usdot): New.
3601         (aarch64_<sur>dot_lane): New.
3602         * config/aarch64/arm_neon.h (vusdot_s32): New.
3603         (vusdotq_s32): New.
3604         (vusdot_lane_s32): New.
3605         (vsudot_lane_s32): New.
3606         * config/aarch64/iterators.md (DOTPROD_I8MM): New iterator.
3607         (UNSPEC_USDOT, UNSPEC_SUDOT): New unspecs.
3609 2020-01-16  Martin Liska  <mliska@suse.cz>
3611         * value-prof.c (dump_histogram_value): Fix
3612         obvious spacing issue.
3614 2020-01-16  Andrew Pinski  <apinski@marvell.com>
3616         * tree-ssa-sccvn.c(vn_reference_lookup_3): Check lhs for
3617         !storage_order_barrier_p.
3619 2020-01-16  Andrew Pinski  <apinski@marvell.com>
3621         * sched-int.h (_dep): Add unused bit-field field for the padding.
3622         * sched-deps.c (init_dep_1): Init unused field.
3624 2020-01-16  Andrew Pinski  <apinski@marvell.com>
3626         * optabs.h (create_expand_operand): Initialize target field also.
3628 2020-01-16  Andre Vieira  <andre.simoesdiasvieira@arm.com>
3630         PR tree-optimization/92429
3631         * tree-ssa-loop-niter.h (simplify_replace_tree): Add parameter.
3632         * tree-ssa-loop-niter.c (simplify_replace_tree): Add parameter to
3633         control folding.
3634         * tree-vect-loop.c (update_epilogue_vinfo): Do not fold when replacing
3635         tree.
3637 2020-01-16  Richard Sandiford  <richard.sandiford@arm.com>
3639         * config/aarch64/aarch64.c (aarch64_split_sve_subreg_move): Apply
3640         aarch64_sve_int_mode to each mode.
3642 2020-01-15  David Malcolm  <dmalcolm@redhat.com>
3644         * doc/analyzer.texi (Overview): Add note about
3645         -fdump-ipa-analyzer.
3647 2020-01-15  Wilco Dijkstra  <wdijkstr@arm.com>
3649         PR tree-optimization/93231
3650         * tree-ssa-forwprop.c (optimize_count_trailing_zeroes): Check
3651         input_type is unsigned.  Use tree_to_shwi for shift constant.
3652         Check CST_STRING element size is CHAR_TYPE_SIZE bits.
3653         (simplify_count_trailing_zeroes): Add test to handle known non-zero
3654         inputs more efficiently.
3656 2020-01-15  UroÅ¡ Bizjak  <ubizjak@gmail.com>
3658         * config/i386/i386.md (*movsf_internal): Do not require
3659         SSE2 ISA for alternatives 14 and 15.
3661 2020-01-15  Richard Biener  <rguenther@suse.de>
3663         PR middle-end/93273
3664         * tree-eh.c (sink_clobbers): If we already visited the destination
3665         block do not defer insertion.
3666         (pass_lower_eh_dispatch::execute): Maintain BB_VISITED for
3667         the purpose of defered insertion.
3669 2020-01-15  Jakub Jelinek  <jakub@redhat.com>
3671         * BASE-VER: Bump to 10.0.1.
3673 2020-01-15  Richard Sandiford  <richard.sandiford@arm.com>
3675         PR tree-optimization/93247
3676         * tree-vect-loop.c (update_epilogue_loop_vinfo): Check the access
3677         type of the stmt that we're going to vectorize.
3679 2020-01-15  Richard Sandiford  <richard.sandiford@arm.com>
3681         * tree-vect-slp.c (vectorize_slp_instance_root_stmt): Use a
3682         VIEW_CONVERT_EXPR if the vectorized constructor has a diffeent
3683         type from the lhs.
3685 2020-01-15  Martin Liska  <mliska@suse.cz>
3687         * ipa-profile.c (ipa_profile_read_edge_summary): Do not allow
3688         2 calls of streamer_read_hwi in a function call.
3690 2020-01-15  Richard Biener  <rguenther@suse.de>
3692         * alias.c (record_alias_subset): Avoid redundant work when
3693         subset is already recorded.
3695 2020-01-14  David Malcolm  <dmalcolm@redhat.com>
3697         * doc/invoke.texi (-fdiagnostics-show-cwe): Add note that some of
3698         the analyzer options provide CWE identifiers.
3700 2020-01-14  David Malcolm  <dmalcolm@redhat.com>
3702         * tree-diagnostic-path.cc (path_summary::event_range::print):
3703         When testing for UNKNOWN_LOCATION, look through ad-hoc wrappers
3704         using get_pure_location.
3706 2020-01-15  Jakub Jelinek  <jakub@redhat.com>
3708         PR tree-optimization/93262
3709         * tree-ssa-dse.c (maybe_trim_memstar_call): For *_chk builtins,
3710         perform head trimming only if the last argument is constant,
3711         either all ones, or larger or equal to head trim, in the latter
3712         case decrease the last argument by head_trim.
3714         PR tree-optimization/93249
3715         * tree-ssa-dse.c: Include builtins.h and gimple-fold.h.
3716         (maybe_trim_memstar_call): Move head_trim and tail_trim vars to
3717         function body scope, reindent.  For BUILTIN_IN_STRNCPY*, don't
3718         perform head trim unless we can prove there are no '\0' chars
3719         from the source among the first head_trim chars.
3721 2020-01-14  David Malcolm  <dmalcolm@redhat.com>
3723         * Makefile.in (ANALYZER_OBJS): Add analyzer/function-set.o.
3725 2020-01-15  Jakub Jelinek  <jakub@redhat.com>
3727         PR target/93009
3728         * config/i386/sse.md
3729         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1,
3730         *<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1,
3731         *<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1,
3732         *<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Use
3733         just a single alternative instead of two, make operands 1 and 2
3734         commutative.
3736 2020-01-14  Jan Hubicka  <hubicka@ucw.cz>
3738         PR lto/91576
3739         * ipa-devirt.c (odr_types_equivalent_p): Compare TREE_ADDRESSABLE and
3740         TYPE_MODE.
3742 2020-01-14  David Malcolm  <dmalcolm@redhat.com>
3744         * Makefile.in (lang_opt_files): Add analyzer.opt.
3745         (ANALYZER_OBJS): New.
3746         (OBJS): Add digraph.o, graphviz.o, ordered-hash-map-tests.o,
3747         tristate.o and ANALYZER_OBJS.
3748         (TEXI_GCCINT_FILES): Add analyzer.texi.
3749         * common.opt (-fanalyzer): New driver option.
3750         * config.in: Regenerate.
3751         * configure: Regenerate.
3752         * configure.ac (--disable-analyzer, ENABLE_ANALYZER): New option.
3753         (gccdepdir): Also create depdir for "analyzer" subdir.
3754         * digraph.cc: New file.
3755         * digraph.h: New file.
3756         * doc/analyzer.texi: New file.
3757         * doc/gccint.texi ("Static Analyzer") New menu item.
3758         (analyzer.texi): Include it.
3759         * doc/invoke.texi ("Static Analyzer Options"): New list and new section.
3760         ("Warning Options"): Add static analysis warnings to the list.
3761         (-Wno-analyzer-double-fclose): New option.
3762         (-Wno-analyzer-double-free): New option.
3763         (-Wno-analyzer-exposure-through-output-file): New option.
3764         (-Wno-analyzer-file-leak): New option.
3765         (-Wno-analyzer-free-of-non-heap): New option.
3766         (-Wno-analyzer-malloc-leak): New option.
3767         (-Wno-analyzer-possible-null-argument): New option.
3768         (-Wno-analyzer-possible-null-dereference): New option.
3769         (-Wno-analyzer-null-argument): New option.
3770         (-Wno-analyzer-null-dereference): New option.
3771         (-Wno-analyzer-stale-setjmp-buffer): New option.
3772         (-Wno-analyzer-tainted-array-index): New option.
3773         (-Wno-analyzer-use-after-free): New option.
3774         (-Wno-analyzer-use-of-pointer-in-stale-stack-frame): New option.
3775         (-Wno-analyzer-use-of-uninitialized-value): New option.
3776         (-Wanalyzer-too-complex): New option.
3777         (-fanalyzer-call-summaries): New warning.
3778         (-fanalyzer-checker=): New warning.
3779         (-fanalyzer-fine-grained): New warning.
3780         (-fno-analyzer-state-merge): New warning.
3781         (-fno-analyzer-state-purge): New warning.
3782         (-fanalyzer-transitivity): New warning.
3783         (-fanalyzer-verbose-edges): New warning.
3784         (-fanalyzer-verbose-state-changes): New warning.
3785         (-fanalyzer-verbosity=): New warning.
3786         (-fdump-analyzer): New warning.
3787         (-fdump-analyzer-callgraph): New warning.
3788         (-fdump-analyzer-exploded-graph): New warning.
3789         (-fdump-analyzer-exploded-nodes): New warning.
3790         (-fdump-analyzer-exploded-nodes-2): New warning.
3791         (-fdump-analyzer-exploded-nodes-3): New warning.
3792         (-fdump-analyzer-supergraph): New warning.
3793         * doc/sourcebuild.texi (dg-require-dot): New.
3794         (dg-check-dot): New.
3795         * gdbinit.in (break-on-saved-diagnostic): New command.
3796         * graphviz.cc: New file.
3797         * graphviz.h: New file.
3798         * ordered-hash-map-tests.cc: New file.
3799         * ordered-hash-map.h: New file.
3800         * passes.def (pass_analyzer): Add before
3801         pass_ipa_whole_program_visibility.
3802         * selftest-run-tests.c (selftest::run_tests): Call
3803         selftest::ordered_hash_map_tests_cc_tests.
3804         * selftest.h (selftest::ordered_hash_map_tests_cc_tests): New
3805         decl.
3806         * shortest-paths.h: New file.
3807         * timevar.def (TV_ANALYZER): New timevar.
3808         (TV_ANALYZER_SUPERGRAPH): Likewise.
3809         (TV_ANALYZER_STATE_PURGE): Likewise.
3810         (TV_ANALYZER_PLAN): Likewise.
3811         (TV_ANALYZER_SCC): Likewise.
3812         (TV_ANALYZER_WORKLIST): Likewise.
3813         (TV_ANALYZER_DUMP): Likewise.
3814         (TV_ANALYZER_DIAGNOSTICS): Likewise.
3815         (TV_ANALYZER_SHORTEST_PATHS): Likewise.
3816         * tree-pass.h (make_pass_analyzer): New decl.
3817         * tristate.cc: New file.
3818         * tristate.h: New file.
3820 2020-01-14  UroÅ¡ Bizjak  <ubizjak@gmail.com>
3822         PR target/93254
3823         * config/i386/i386.md (*movsf_internal): Require SSE2 ISA for
3824         alternatives 9 and 10.
3826 2020-01-14  David Malcolm  <dmalcolm@redhat.com>
3828         * attribs.c (excl_hash_traits::empty_zero_p): New static constant.
3829         * gcov.c (function_start_pair_hash::empty_zero_p): Likewise.
3830         * graphite.c (struct sese_scev_hash::empty_zero_p): Likewise.
3831         * hash-map-tests.c (selftest::test_nonzero_empty_key): New selftest.
3832         (selftest::hash_map_tests_c_tests): Call it.
3833         * hash-map-traits.h (simple_hashmap_traits::empty_zero_p):
3834         New static constant, using the value of = H::empty_zero_p.
3835         (unbounded_hashmap_traits::empty_zero_p): Likewise, using the value
3836         from default_hash_traits <Value>.
3837         * hash-map.h (hash_map::empty_zero_p): Likewise, using the value
3838         from Traits.
3839         * hash-set-tests.c (value_hash_traits::empty_zero_p): Likewise.
3840         * hash-table.h (hash_table::alloc_entries): Guard the loop of
3841         calls to mark_empty with !Descriptor::empty_zero_p.
3842         (hash_table::empty_slow): Conditionalize the memset call with a
3843         check that Descriptor::empty_zero_p; otherwise, loop through the
3844         entries calling mark_empty on them.
3845         * hash-traits.h (int_hash::empty_zero_p): New static constant.
3846         (pointer_hash::empty_zero_p): Likewise.
3847         (pair_hash::empty_zero_p): Likewise.
3848         * ipa-devirt.c (default_hash_traits <type_pair>::empty_zero_p):
3849         Likewise.
3850         * ipa-prop.c (ipa_bit_ggc_hash_traits::empty_zero_p): Likewise.
3851         (ipa_vr_ggc_hash_traits::empty_zero_p): Likewise.
3852         * profile.c (location_triplet_hash::empty_zero_p): Likewise.
3853         * sanopt.c (sanopt_tree_triplet_hash::empty_zero_p): Likewise.
3854         (sanopt_tree_couple_hash::empty_zero_p): Likewise.
3855         * tree-hasher.h (int_tree_hasher::empty_zero_p): Likewise.
3856         * tree-ssa-sccvn.c (vn_ssa_aux_hasher::empty_zero_p): Likewise.
3857         * tree-vect-slp.c (bst_traits::empty_zero_p): Likewise.
3858         * tree-vectorizer.h
3859         (default_hash_traits<scalar_cond_masked_key>::empty_zero_p):
3860         Likewise.
3862 2020-01-14  Kewen Lin  <linkw@gcc.gnu.org>
3864         * cfgloopanal.c (average_num_loop_insns): Free bbs when early return,
3865         fix typo on return value.
3867 2020-01-14  Xiong Hu Luo  <luoxhu@linux.ibm.com>
3869         PR ipa/69678
3870         * cgraph.c (symbol_table::create_edge): Init speculative_id and
3871         target_prob.
3872         (cgraph_edge::make_speculative): Add param for setting speculative_id
3873         and target_prob.
3874         (cgraph_edge::speculative_call_info): Update comments and find reference
3875         by speculative_id for multiple indirect targets.
3876         (cgraph_edge::resolve_speculation): Decrease the speculations
3877         for indirect edge, drop it's speculative if not direct target
3878         left. Update comments.
3879         (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
3880         (cgraph_node::dump): Print num_speculative_call_targets.
3881         (cgraph_node::verify_node): Don't report error if speculative
3882         edge not include statement.
3883         (cgraph_edge::num_speculative_call_targets_p): New function.
3884         * cgraph.h (int common_target_id): Remove.
3885         (int common_target_probability): Remove.
3886         (num_speculative_call_targets): New variable.
3887         (make_speculative): Add param for setting speculative_id.
3888         (cgraph_edge::num_speculative_call_targets_p): New declare.
3889         (target_prob): New variable.
3890         (speculative_id): New variable.
3891         * ipa-fnsummary.c (analyze_function_body): Create and duplicate
3892           call summaries for multiple speculative call targets.
3893         * cgraphclones.c (cgraph_node::create_clone): Clone speculative_id.
3894         * ipa-profile.c (struct speculative_call_target): New struct.
3895         (class speculative_call_summary): New class.
3896         (class speculative_call_summaries): New class.
3897         (call_sums): New variable.
3898         (ipa_profile_generate_summary): Generate indirect multiple targets summaries.
3899         (ipa_profile_write_edge_summary): New function.
3900         (ipa_profile_write_summary): Stream out indirect multiple targets summaries.
3901         (ipa_profile_dump_all_summaries): New function.
3902         (ipa_profile_read_edge_summary): New function.
3903         (ipa_profile_read_summary_section): New function.
3904         (ipa_profile_read_summary): Stream in indirect multiple targets summaries.
3905         (ipa_profile): Generate num_speculative_call_targets from
3906         profile summaries.
3907         * ipa-ref.h (speculative_id): New variable.
3908         * ipa-utils.c (ipa_merge_profiles): Update with target_prob.
3909         * lto-cgraph.c (lto_output_edge): Remove indirect common_target_id and
3910         common_target_probability.   Stream out speculative_id and
3911         num_speculative_call_targets.
3912         (input_edge): Likewise.
3913         * predict.c (dump_prediction): Remove edges count assert to be
3914         precise.
3915         * symtab.c (symtab_node::create_reference): Init speculative_id.
3916         (symtab_node::clone_references): Clone speculative_id.
3917         (symtab_node::clone_referring): Clone speculative_id.
3918         (symtab_node::clone_reference): Clone speculative_id.
3919         (symtab_node::clear_stmts_in_references): Clear speculative_id.
3920         * tree-inline.c (copy_bb): Duplicate all the speculative edges
3921         if indirect call contains multiple speculative targets.
3922         * value-prof.h  (check_ic_target): Remove.
3923         * value-prof.c  (gimple_value_profile_transformations):
3924         Use void function gimple_ic_transform.
3925         * value-prof.c  (gimple_ic_transform): Handle topn case.
3926         Fix comment typos.  Change it to a void function.
3928 2020-01-13  Andrew Pinski  <apinski@marvell.com>
3930         * config/aarch64/aarch64-cores.def (octeontx2): New define.
3931         (octeontx2t98): New define.
3932         (octeontx2t96): New define.
3933         (octeontx2t93): New define.
3934         (octeontx2f95): New define.
3935         (octeontx2f95n): New define.
3936         (octeontx2f95mm): New define.
3937         * config/aarch64/aarch64-tune.md: Regenerate.
3938         * doc/invoke.texi (-mcpu=): Document the new cpu types.
3940 2020-01-13  Jason Merrill  <jason@redhat.com>
3942         PR c++/33799 - destroy return value if local cleanup throws.
3943         * gimplify.c (gimplify_return_expr): Handle COMPOUND_EXPR.
3945 2020-01-13  Martin Liska  <mliska@suse.cz>
3947         * ipa-cp.c (get_max_overall_size): Use newly
3948         renamed param param_ipa_cp_unit_growth.
3949         * params.opt: Remove legacy param name.
3951 2020-01-13  Martin Sebor  <msebor@redhat.com>
3953         PR tree-optimization/93213
3954         * tree-ssa-strlen.c (handle_store): Only allow single-byte nul-over-nul
3955         stores to be eliminated.
3957 2020-01-13  Martin Liska  <mliska@suse.cz>
3959         * opts.c (print_help): Do not print CL_PARAM
3960         and CL_WARNING for CL_OPTIMIZATION.
3962 2020-01-13  Jonathan Wakely  <jwakely@redhat.com>
3964         PR driver/92757
3965         * doc/invoke.texi (Warning Options): Add caveat about some warnings
3966         depending on optimization settings.
3968 2020-01-13  Jakub Jelinek  <jakub@redhat.com>
3970         PR tree-optimization/90838
3971         * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
3972         SCALAR_INT_TYPE_MODE directly in CTZ_DEFINED_VALUE_AT_ZERO macro
3973         argument rather than to initialize temporary for targets that
3974         don't use the mode argument at all.  Initialize ctzval to avoid
3975         warning at -O0.
3977 2020-01-10  Thomas Schwinge  <thomas@codesourcery.com>
3979         * tree.h (OMP_CLAUSE_USE_DEVICE_PTR_IF_PRESENT): New definition.
3980         * tree-core.h: Document it.
3981         * gimplify.c (gimplify_omp_workshare): Set it.
3982         * omp-low.c (lower_omp_target): Use it.
3983         * tree-pretty-print.c (dump_omp_clause): Print it.
3985         * omp-low.c (lower_omp_target) <OMP_CLAUSE_USE_DEVICE_PTR etc.>:
3986         Assert that for OpenACC we always have 'GOMP_MAP_USE_DEVICE_PTR'.
3988 2020-01-10  David Malcolm  <dmalcolm@redhat.com>
3990         * Makefile.in (OBJS): Add tree-diagnostic-path.o.
3991         * common.opt (fdiagnostics-path-format=): New option.
3992         (diagnostic_path_format): New enum.
3993         (fdiagnostics-show-path-depths): New option.
3994         * coretypes.h (diagnostic_event_id_t): New forward decl.
3995         * diagnostic-color.c (color_dict): Add "path".
3996         * diagnostic-event-id.h: New file.
3997         * diagnostic-format-json.cc (json_from_expanded_location): Make
3998         non-static.
3999         (json_end_diagnostic): Call context->make_json_for_path if it
4000         exists and the diagnostic has a path.
4001         (diagnostic_output_format_init): Clear context->print_path.
4002         * diagnostic-path.h: New file.
4003         * diagnostic-show-locus.c (colorizer::set_range): Special-case
4004         when printing a run of events in a diagnostic_path so that they
4005         all get the same color.
4006         (layout::m_diagnostic_path_p): New field.
4007         (layout::layout): Initialize it.
4008         (layout::print_any_labels): Don't colorize the label text for an
4009         event in a diagnostic_path.
4010         (gcc_rich_location::add_location_if_nearby): Add
4011         "restrict_to_current_line_spans" and "label" params.  Pass the
4012         former to layout.maybe_add_location_range; pass the latter
4013         when calling add_range.
4014         * diagnostic.c: Include "diagnostic-path.h".
4015         (diagnostic_initialize): Initialize context->path_format and
4016         context->show_path_depths.
4017         (diagnostic_show_any_path): New function.
4018         (diagnostic_path::interprocedural_p): New function.
4019         (diagnostic_report_diagnostic): Call diagnostic_show_any_path.
4020         (simple_diagnostic_path::num_events): New function.
4021         (simple_diagnostic_path::get_event): New function.
4022         (simple_diagnostic_path::add_event): New function.
4023         (simple_diagnostic_event::simple_diagnostic_event): New ctor.
4024         (simple_diagnostic_event::~simple_diagnostic_event): New dtor.
4025         (debug): New overload taking a diagnostic_path *.
4026         * diagnostic.def (DK_DIAGNOSTIC_PATH): New.
4027         * diagnostic.h (enum diagnostic_path_format): New enum.
4028         (json::value): New forward decl.
4029         (diagnostic_context::path_format): New field.
4030         (diagnostic_context::show_path_depths): New field.
4031         (diagnostic_context::print_path): New callback field.
4032         (diagnostic_context::make_json_for_path): New callback field.
4033         (diagnostic_show_any_path): New decl.
4034         (json_from_expanded_location): New decl.
4035         * doc/invoke.texi (-fdiagnostics-path-format=): New option.
4036         (-fdiagnostics-show-path-depths): New option.
4037         (-fdiagnostics-color): Add "path" to description of default
4038         GCC_COLORS; describe it.
4039         (-fdiagnostics-format=json): Document how diagnostic paths are
4040         represented in the JSON output format.
4041         * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
4042         Add optional params "restrict_to_current_line_spans" and "label".
4043         * opts.c (common_handle_option): Handle
4044         OPT_fdiagnostics_path_format_ and
4045         OPT_fdiagnostics_show_path_depths.
4046         * pretty-print.c: Include "diagnostic-event-id.h".
4047         (pp_format): Implement "%@" format code for printing
4048         diagnostic_event_id_t *.
4049         (selftest::test_pp_format): Add tests for "%@".
4050         * selftest-run-tests.c (selftest::run_tests): Call
4051         selftest::tree_diagnostic_path_cc_tests.
4052         * selftest.h (selftest::tree_diagnostic_path_cc_tests): New decl.
4053         * toplev.c (general_init): Initialize global_dc->path_format and
4054         global_dc->show_path_depths.
4055         * tree-diagnostic-path.cc: New file.
4056         * tree-diagnostic.c (maybe_unwind_expanded_macro_loc): Make
4057         non-static.  Drop "diagnostic" param in favor of storing the
4058         original value of "where" and re-using it.
4059         (virt_loc_aware_diagnostic_finalizer): Update for dropped param of
4060         maybe_unwind_expanded_macro_loc.
4061         (tree_diagnostics_defaults): Initialize context->print_path and
4062         context->make_json_for_path.
4063         * tree-diagnostic.h (default_tree_diagnostic_path_printer): New
4064         decl.
4065         (default_tree_make_json_for_path): New decl.
4066         (maybe_unwind_expanded_macro_loc): New decl.
4068 2020-01-10  Jakub Jelinek  <jakub@redhat.com>
4070         PR tree-optimization/93210
4071         * fold-const.h (native_encode_initializer,
4072         can_native_interpret_type_p): Declare.
4073         * fold-const.c (native_encode_string): Fix up handling with off != -1,
4074         simplify.
4075         (native_encode_initializer): New function, moved from dwarf2out.c.
4076         Adjust to native_encode_expr compatible arguments, including dry-run
4077         and partial extraction modes.  Don't handle STRING_CST.
4078         (can_native_interpret_type_p): No longer static.
4079         * gimple-fold.c (fold_ctor_reference): For native_encode_expr, verify
4080         offset / BITS_PER_UNIT fits into int and don't call it if
4081         can_native_interpret_type_p fails.  If suboff is NULL and for
4082         CONSTRUCTOR fold_{,non}array_ctor_reference returns NULL, retry with
4083         native_encode_initializer.
4084         (fold_const_aggregate_ref_1): Formatting fix.
4085         * dwarf2out.c (native_encode_initializer): Moved to fold-const.c.
4086         (tree_add_const_value_attribute): Adjust caller.
4088         PR tree-optimization/90838
4089         * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
4090         SCALAR_INT_TYPE_MODE instead of TYPE_MODE as operand of
4091         CTZ_DEFINED_VALUE_AT_ZERO.
4093 2020-01-10  Vladimir Makarov  <vmakarov@redhat.com>
4095         PR inline-asm/93027
4096         * lra-constraints.c (match_reload): Permit input operands have the
4097         same mode as output while other input operands have a different
4098         mode.
4100 2020-01-10  Wilco Dijkstra  <wdijkstr@arm.com>
4102         PR tree-optimization/90838
4103         * tree-ssa-forwprop.c (check_ctz_array): Add new function.
4104         (check_ctz_string): Likewise.
4105         (optimize_count_trailing_zeroes): Likewise.
4106         (simplify_count_trailing_zeroes): Likewise.
4107         (pass_forwprop::execute): Try ctz simplification.
4108         * match.pd: Add matching for ctz idioms.
4110 2020-01-10  Stam Markianos-Wright  <stam.markianos-wright@arm.com>
4112         * config/aarch64/aarch64.c (aarch64_invalid_conversion): New function
4113         for target hook.
4114         (aarch64_invalid_unary_op): New function for target hook.
4115         (aarch64_invalid_binary_op): New function for target hook.
4117 2020-01-10  Stam Markianos-Wright  <stam.markianos-wright@arm.com>
4119         * config.gcc: Add arm_bf16.h.
4120         * config/aarch64/aarch64-builtins.c
4121         (aarch64_simd_builtin_std_type): Add BFmode.
4122         (aarch64_init_simd_builtin_types): Define element types for vector
4123         types.
4124         (aarch64_init_bf16_types): New function.
4125         (aarch64_general_init_builtins): Add arm_init_bf16_types function call.
4126         * config/aarch64/aarch64-modes.def: Add BFmode and V4BF, V8BF vector
4127         modes.
4128         * config/aarch64/aarch64-simd-builtin-types.def: Add BF SIMD types.
4129         * config/aarch64/aarch64-simd.md: Add BF vector types to NEON move
4130         patterns.
4131         * config/aarch64/aarch64.h (AARCH64_VALID_SIMD_DREG_MODE): Add V4BF.
4132         (AARCH64_VALID_SIMD_QREG_MODE): Add V8BF.
4133         * config/aarch64/aarch64.c
4134         (aarch64_classify_vector_mode): Add support for BF types.
4135         (aarch64_gimplify_va_arg_expr): Add support for BF types.
4136         (aarch64_vq_mode): Add support for BF types.
4137         (aarch64_simd_container_mode): Add support for BF types.
4138         (aarch64_mangle_type): Add support for BF scalar type.
4139         * config/aarch64/aarch64.md: Add BFmode to movhf pattern.
4140         * config/aarch64/arm_bf16.h: New file.
4141         * config/aarch64/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
4142         * config/aarch64/iterators.md: Add BF types to mode attributes.
4143         (HFBF, GPF_TF_F16_MOV, VDMOV, VQMOV, VQMOV_NO2Em VALL_F16MOV): New.
4145 2020-01-10  Jason Merrill  <jason@redhat.com>
4147         PR c++/93173 - incorrect tree sharing.
4148         * gimplify.c (copy_if_shared): No longer static.
4149         * gimplify.h: Declare it.
4151 2020-01-10  Richard Sandiford  <richard.sandiford@arm.com>
4153         * doc/invoke.texi (-msve-vector-bits=): Document that
4154         -msve-vector-bits=128 now generates VL-specific code for
4155         little-endian targets.
4156         * config/aarch64/aarch64-sve-builtins.cc (register_builtin_types): Use
4157         build_vector_type_for_mode to construct the data vector types.
4158         * config/aarch64/aarch64.c (aarch64_convert_sve_vector_bits): Generate
4159         VL-specific code for -msve-vector-bits=128 on little-endian targets.
4160         (aarch64_simd_container_mode): Always prefer Advanced SIMD modes
4161         for 128-bit vectors.
4163 2020-01-10  Richard Sandiford  <richard.sandiford@arm.com>
4165         * config/aarch64/aarch64.c (aarch64_evpc_sel): Fix gen_vcond_mask
4166         invocation.
4168 2020-01-10  Richard Sandiford  <richard.sandiford@arm.com>
4170         * config/aarch64/aarch64-builtins.c
4171         (aarch64_builtin_vectorized_function): Check for specific vector modes,
4172         rather than checking the number of elements and the element mode.
4174 2020-01-10  Richard Sandiford  <richard.sandiford@arm.com>
4176         * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
4177         get_related_vectype_for_scalar_type rather than build_vector_type
4178         to create the index type for a conditional reduction.
4180 2020-01-10  Richard Sandiford  <richard.sandiford@arm.com>
4182         * tree-vect-loop.c (update_epilogue_loop_vinfo): Update DR_REF
4183         for any type of gather or scatter, including strided accesses.
4185 2020-01-10  Andre Vieira  <andre.simoesdiasvieira@arm.com>
4187         * tree-vectorizer.h (get_dr_vinfo_offset): Add missing function
4188          comment.
4190 2020-01-10  Andre Vieira  <andre.simoesdiasvieira@arm.com>
4192         * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref): Use
4193         get_dr_vinfo_offset
4194         * tree-vect-loop.c (update_epilogue_loop_vinfo):  Remove orig_drs_init
4195         parameter and its use to reset DR_OFFSET's.
4196         (vect_transform_loop): Remove orig_drs_init argument.
4197         * tree-vect-loop-manip.c (vect_update_init_of_dr): Update the offset
4198         member of dr_vec_info rather than the offset of the associated
4199         data_reference's innermost_loop_behavior.
4200         (vect_update_init_of_dr): Pass dr_vec_info instead of data_reference.
4201         (vect_do_peeling): Remove orig_drs_init parameter and its construction.
4202         * tree-vect-stmts.c (check_scan_store): Replace use of DR_OFFSET with
4203         get_dr_vinfo_offset.
4204         (vectorizable_store): Likewise.
4205         (vectorizable_load): Likewise.
4207 2020-01-10  Richard Biener  <rguenther@suse.de>
4209         * gimple-ssa-store-merging
4210         (pass_store_merging::terminate_all_aliasing_chains): Cache alias info.
4212 2020-01-10  Martin Liska  <mliska@suse.cz>
4214         PR ipa/93217
4215         * ipa-inline-analysis.c (offline_size): Make proper parenthesis
4216         encapsulation that was there before r280040.
4218 2020-01-10  Richard Biener  <rguenther@suse.de>
4220         PR middle-end/93199
4221         * tree-eh.c (sink_clobbers): Move clobbers to out-of-IL
4222         sequences to avoid walking them again for secondary opportunities.
4223         (pass_lower_eh_dispatch::execute): Instead actually insert
4224         them here.
4226 2020-01-10  Richard Biener  <rguenther@suse.de>
4228         PR middle-end/93199
4229         * tree-eh.c (redirect_eh_edge_1): Avoid some work if possible.
4230         (cleanup_all_empty_eh): Walk landing pads in reverse order to
4231         avoid quadraticness.
4233 2020-01-10  Martin Jambor  <mjambor@suse.cz>
4235         * params.opt (param_ipa_sra_max_replacements): Mark as Optimization.
4236         * ipa-sra.c (pull_accesses_from_callee): New parameter caller, use it
4237         to get param_ipa_sra_max_replacements.
4238         (param_splitting_across_edge): Pass the caller to
4239         pull_accesses_from_callee.
4241 2020-01-10  Martin Jambor  <mjambor@suse.cz>
4243         * params.opt (param_ipcp_unit_growth): Mark as Optimization.
4244         * ipa-cp.c (max_new_size): Removed.
4245         (orig_overall_size): New variable.
4246         (get_max_overall_size): New function.
4247         (estimate_local_effects): Use it.  Adjust dump.
4248         (decide_about_value): Likewise.
4249         (ipcp_propagate_stage): Do not calculate max_new_size, just store
4250         orig_overall_size.  Adjust dump.
4251         (ipa_cp_c_finalize): Clear orig_overall_size instead of max_new_size.
4253 2020-01-10  Martin Jambor  <mjambor@suse.cz>
4255         * params.opt (param_ipa_max_agg_items): Mark as Optimization
4256         * ipa-cp.c (merge_agg_lats_step): New parameter max_agg_items, use
4257         instead of param_ipa_max_agg_items.
4258         (merge_aggregate_lattices): Extract param_ipa_max_agg_items from
4259         optimization info for the callee.
4261 2020-01-09  Kwok Cheung Yeung  <kcy@codesourcery.com>
4263         * lto-streamer-in.c (input_function): Remove streamed-in inline debug
4264         markers if debug_inline_points is false.
4266 2020-01-09  Richard Sandiford  <richard.sandiford@arm.com>
4268         * config.gcc (aarch64*-*-*): Add aarch64-sve-builtins-sve2.o to
4269         extra_objs.
4270         * config/aarch64/t-aarch64 (aarch64-sve-builtins.o): Depend on
4271         aarch64-sve-builtins-base.def, aarch64-sve-builtins-sve2.def and
4272         aarch64-sve-builtins-sve2.h.
4273         (aarch64-sve-builtins-sve2.o): New rule.
4274         * config/aarch64/aarch64.h (AARCH64_ISA_SVE2_AES): New macro.
4275         (AARCH64_ISA_SVE2_BITPERM, AARCH64_ISA_SVE2_SHA3): Likewise.
4276         (AARCH64_ISA_SVE2_SM4, TARGET_SVE2_AES, TARGET_SVE2_BITPERM): Likewise.
4277         (TARGET_SVE2_SHA, TARGET_SVE2_SM4): Likewise.
4278         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
4279         TARGET_SVE2_AES, TARGET_SVE2_BITPERM, TARGET_SVE2_SHA3 and
4280         TARGET_SVE2_SM4.
4281         * config/aarch64/aarch64-sve.md: Update comments with SVE2
4282         instructions that are handled here.
4283         (@cond_asrd<mode>): Generalize to...
4284         (@cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>): ...this.
4285         (*cond_asrd<mode>_2): Generalize to...
4286         (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_2): ...this.
4287         (*cond_asrd<mode>_z): Generalize to...
4288         (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_z): ...this.
4289         * config/aarch64/aarch64.md (UNSPEC_LDNT1_GATHER): New unspec.
4290         (UNSPEC_STNT1_SCATTER, UNSPEC_WHILEGE, UNSPEC_WHILEGT): Likewise.
4291         (UNSPEC_WHILEHI, UNSPEC_WHILEHS): Likewise.
4292         * config/aarch64/aarch64-sve2.md (@aarch64_gather_ldnt<mode>): New
4293         pattern.
4294         (@aarch64_gather_ldnt_<ANY_EXTEND:optab><SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
4295         (@aarch64_scatter_stnt<mode>): Likewise.
4296         (@aarch64_scatter_stnt_<SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
4297         (@aarch64_mul_lane_<mode>): Likewise.
4298         (@aarch64_sve_suqadd<mode>_const): Likewise.
4299         (*<sur>h<addsub><mode>): Generalize to...
4300         (@aarch64_pred_<SVE2_COND_INT_BINARY_REV:sve_int_op><mode>): ...this
4301         new pattern.
4302         (@cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>): New expander.
4303         (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_2): New pattern.
4304         (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_3): Likewise.
4305         (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_any): Likewise.
4306         (*cond_<SVE2_COND_INT_BINARY_NOREV:sve_int_op><mode>_z): Likewise.
4307         (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op><mode>):: Likewise.
4308         (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op>_lane_<mode>): Likewise.
4309         (@aarch64_pred_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): Likewise.
4310         (@cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): New expander.
4311         (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_2): New pattern.
4312         (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_3): Likewise.
4313         (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_any): Likewise.
4314         (@aarch64_sve_<SVE2_INT_TERNARY:sve_int_op><mode>): Likewise.
4315         (@aarch64_sve_<SVE2_INT_TERNARY_LANE:sve_int_op>_lane_<mode>)
4316         (@aarch64_sve_add_mul_lane_<mode>): Likewise.
4317         (@aarch64_sve_sub_mul_lane_<mode>): Likewise.
4318         (@aarch64_sve2_xar<mode>): Likewise.
4319         (@aarch64_sve2_bcax<mode>): Likewise.
4320         (*aarch64_sve2_eor3<mode>): Rename to...
4321         (@aarch64_sve2_eor3<mode>): ...this.
4322         (@aarch64_sve2_bsl<mode>): New expander.
4323         (@aarch64_sve2_nbsl<mode>): Likewise.
4324         (@aarch64_sve2_bsl1n<mode>): Likewise.
4325         (@aarch64_sve2_bsl2n<mode>): Likewise.
4326         (@aarch64_sve_add_<SHIFTRT:sve_int_op><mode>): Likewise.
4327         (*aarch64_sve2_sra<mode>): Add MOVPRFX support.
4328         (@aarch64_sve_add_<VRSHR_N:sve_int_op><mode>): New pattern.
4329         (@aarch64_sve_<SVE2_INT_SHIFT_INSERT:sve_int_op><mode>): Likewise.
4330         (@aarch64_sve2_<USMAX:su>aba<mode>): New expander.
4331         (*aarch64_sve2_<USMAX:su>aba<mode>): New pattern.
4332         (@aarch64_sve_<SVE2_INT_BINARY_WIDE:sve_int_op><mode>): Likewise.
4333         (<su>mull<bt><Vwide>): Generalize to...
4334         (@aarch64_sve_<SVE2_INT_BINARY_LONG:sve_int_op><mode>): ...this new
4335         pattern.
4336         (@aarch64_sve_<SVE2_INT_BINARY_LONG_lANE:sve_int_op>_lane_<mode>)
4337         (@aarch64_sve_<SVE2_INT_SHIFT_IMM_LONG:sve_int_op><mode>)
4338         (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG:sve_int_op><mode>)
4339         (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
4340         (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG:sve_int_op><mode>)
4341         (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
4342         (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG:sve_int_op><mode>)
4343         (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
4344         (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG:sve_int_op><mode>)
4345         (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
4346         (@aarch64_sve_<SVE2_FP_TERNARY_LONG:sve_fp_op><mode>): New patterns.
4347         (@aarch64_<SVE2_FP_TERNARY_LONG_LANE:sve_fp_op>_lane_<mode>)
4348         (@aarch64_sve_<SVE2_INT_UNARY_NARROWB:sve_int_op><mode>): Likewise.
4349         (@aarch64_sve_<SVE2_INT_UNARY_NARROWT:sve_int_op><mode>): Likewise.
4350         (@aarch64_sve_<SVE2_INT_BINARY_NARROWB:sve_int_op><mode>): Likewise.
4351         (@aarch64_sve_<SVE2_INT_BINARY_NARROWT:sve_int_op><mode>): Likewise.
4352         (<SHRNB:r>shrnb<mode>): Generalize to...
4353         (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWB:sve_int_op><mode>): ...this
4354         new pattern.
4355         (<SHRNT:r>shrnt<mode>): Generalize to...
4356         (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWT:sve_int_op><mode>): ...this
4357         new pattern.
4358         (@aarch64_pred_<SVE2_INT_BINARY_PAIR:sve_int_op><mode>): New pattern.
4359         (@aarch64_pred_<SVE2_FP_BINARY_PAIR:sve_fp_op><mode>): Likewise.
4360         (@cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>): New expander.
4361         (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_2): New pattern.
4362         (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_z): Likewise.
4363         (@aarch64_sve_<SVE2_INT_CADD:optab><mode>): Likewise.
4364         (@aarch64_sve_<SVE2_INT_CMLA:optab><mode>): Likewise.
4365         (@aarch64_<SVE2_INT_CMLA:optab>_lane_<mode>): Likewise.
4366         (@aarch64_sve_<SVE2_INT_CDOT:optab><mode>): Likewise.
4367         (@aarch64_<SVE2_INT_CDOT:optab>_lane_<mode>): Likewise.
4368         (@aarch64_pred_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): Likewise.
4369         (@cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New expander.
4370         (*cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New pattern.
4371         (@aarch64_sve2_cvtnt<mode>): Likewise.
4372         (@aarch64_pred_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): Likewise.
4373         (@cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): New expander.
4374         (*cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>_any): New pattern.
4375         (@aarch64_sve2_cvtxnt<mode>): Likewise.
4376         (@aarch64_pred_<SVE2_U32_UNARY:sve_int_op><mode>): Likewise.
4377         (@cond_<SVE2_U32_UNARY:sve_int_op><mode>): New expander.
4378         (*cond_<SVE2_U32_UNARY:sve_int_op><mode>): New pattern.
4379         (@aarch64_pred_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): Likewise.
4380         (@cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New expander.
4381         (*cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New pattern.
4382         (@aarch64_sve2_pmul<mode>): Likewise.
4383         (@aarch64_sve_<SVE2_PMULL:optab><mode>): Likewise.
4384         (@aarch64_sve_<SVE2_PMULL_PAIR:optab><mode>): Likewise.
4385         (@aarch64_sve2_tbl2<mode>): Likewise.
4386         (@aarch64_sve2_tbx<mode>): Likewise.
4387         (@aarch64_sve_<SVE2_INT_BITPERM:sve_int_op><mode>): Likewise.
4388         (@aarch64_sve2_histcnt<mode>): Likewise.
4389         (@aarch64_sve2_histseg<mode>): Likewise.
4390         (@aarch64_pred_<SVE2_MATCH:sve_int_op><mode>): Likewise.
4391         (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_cc): Likewise.
4392         (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_ptest): Likewise.
4393         (aarch64_sve2_aes<CRYPTO_AES:aes_op>): Likewise.
4394         (aarch64_sve2_aes<CRYPTO_AESMC:aesmc_op>): Likewise.
4395         (*aarch64_sve2_aese_fused, *aarch64_sve2_aesd_fused): Likewise.
4396         (aarch64_sve2_rax1, aarch64_sve2_sm4e, aarch64_sve2_sm4ekey): Likewise.
4397         (<su>mulh<r>s<mode>3): Update after above pattern name changes.
4398         * config/aarch64/iterators.md (VNx16QI_ONLY, VNx4SF_ONLY)
4399         (SVE_STRUCT2, SVE_FULL_BHI, SVE_FULL_HSI, SVE_FULL_HDI)
4400         (SVE2_PMULL_PAIR_I): New mode iterators.
4401         (UNSPEC_ADCLB, UNSPEC_ADCLT, UNSPEC_ADDHNB, UNSPEC_ADDHNT, UNSPEC_BDEP)
4402         (UNSPEC_BEXT, UNSPEC_BGRP, UNSPEC_CADD90, UNSPEC_CADD270, UNSPEC_CDOT)
4403         (UNSPEC_CDOT90, UNSPEC_CDOT180, UNSPEC_CDOT270, UNSPEC_CMLA)
4404         (UNSPEC_CMLA90, UNSPEC_CMLA180, UNSPEC_CMLA270, UNSPEC_COND_FCVTLT)
4405         (UNSPEC_COND_FCVTNT, UNSPEC_COND_FCVTX, UNSPEC_COND_FCVTXNT)
4406         (UNSPEC_COND_FLOGB, UNSPEC_EORBT, UNSPEC_EORTB, UNSPEC_FADDP)
4407         (UNSPEC_FMAXP, UNSPEC_FMAXNMP, UNSPEC_FMLALB, UNSPEC_FMLALT)
4408         (UNSPEC_FMLSLB, UNSPEC_FMLSLT, UNSPEC_FMINP, UNSPEC_FMINNMP)
4409         (UNSPEC_HISTCNT, UNSPEC_HISTSEG, UNSPEC_MATCH, UNSPEC_NMATCH)
4410         (UNSPEC_PMULLB, UNSPEC_PMULLB_PAIR, UNSPEC_PMULLT, UNSPEC_PMULLT_PAIR)
4411         (UNSPEC_RADDHNB, UNSPEC_RADDHNT, UNSPEC_RSUBHNB, UNSPEC_RSUBHNT)
4412         (UNSPEC_SLI, UNSPEC_SRI, UNSPEC_SABDLB, UNSPEC_SABDLT, UNSPEC_SADDLB)
4413         (UNSPEC_SADDLBT, UNSPEC_SADDLT, UNSPEC_SADDWB, UNSPEC_SADDWT)
4414         (UNSPEC_SBCLB, UNSPEC_SBCLT, UNSPEC_SMAXP, UNSPEC_SMINP)
4415         (UNSPEC_SQCADD90, UNSPEC_SQCADD270, UNSPEC_SQDMULLB, UNSPEC_SQDMULLBT)
4416         (UNSPEC_SQDMULLT, UNSPEC_SQRDCMLAH, UNSPEC_SQRDCMLAH90)
4417         (UNSPEC_SQRDCMLAH180, UNSPEC_SQRDCMLAH270, UNSPEC_SQRSHRNB)
4418         (UNSPEC_SQRSHRNT, UNSPEC_SQRSHRUNB, UNSPEC_SQRSHRUNT, UNSPEC_SQSHRNB)
4419         (UNSPEC_SQSHRNT, UNSPEC_SQSHRUNB, UNSPEC_SQSHRUNT, UNSPEC_SQXTNB)
4420         (UNSPEC_SQXTNT, UNSPEC_SQXTUNB, UNSPEC_SQXTUNT, UNSPEC_SSHLLB)
4421         (UNSPEC_SSHLLT, UNSPEC_SSUBLB, UNSPEC_SSUBLBT, UNSPEC_SSUBLT)
4422         (UNSPEC_SSUBLTB, UNSPEC_SSUBWB, UNSPEC_SSUBWT, UNSPEC_SUBHNB)
4423         (UNSPEC_SUBHNT, UNSPEC_TBL2, UNSPEC_UABDLB, UNSPEC_UABDLT)
4424         (UNSPEC_UADDLB, UNSPEC_UADDLT, UNSPEC_UADDWB, UNSPEC_UADDWT)
4425         (UNSPEC_UMAXP, UNSPEC_UMINP, UNSPEC_UQRSHRNB, UNSPEC_UQRSHRNT)
4426         (UNSPEC_UQSHRNB, UNSPEC_UQSHRNT, UNSPEC_UQXTNB, UNSPEC_UQXTNT)
4427         (UNSPEC_USHLLB, UNSPEC_USHLLT, UNSPEC_USUBLB, UNSPEC_USUBLT)
4428         (UNSPEC_USUBWB, UNSPEC_USUBWT): New unspecs.
4429         (UNSPEC_SMULLB, UNSPEC_SMULLT, UNSPEC_UMULLB, UNSPEC_UMULLT)
4430         (UNSPEC_SMULHS, UNSPEC_SMULHRS, UNSPEC_UMULHS, UNSPEC_UMULHRS)
4431         (UNSPEC_RSHRNB, UNSPEC_RSHRNT, UNSPEC_SHRNB, UNSPEC_SHRNT): Move
4432         further down file.
4433         (VNARROW, Ventype): New mode attributes.
4434         (Vewtype): Handle VNx2DI.  Fix typo in comment.
4435         (VDOUBLE): New mode attribute.
4436         (sve_lane_con): Handle VNx8HI.
4437         (SVE_INT_UNARY): Include ss_abs and ss_neg for TARGET_SVE2.
4438         (SVE_INT_BINARY): Likewise ss_plus, us_plus, ss_minus and us_minus.
4439         (sve_int_op, sve_int_op_rev): Handle the above codes.
4440         (sve_pred_int_rhs2_operand): Likewise.
4441         (MULLBT, SHRNB, SHRNT): Delete.
4442         (SVE_INT_SHIFT_IMM): New int iterator.
4443         (SVE_WHILE): Add UNSPEC_WHILEGE, UNSPEC_WHILEGT, UNSPEC_WHILEHI
4444         and UNSPEC_WHILEHS for TARGET_SVE2.
4445         (SVE2_U32_UNARY, SVE2_INT_UNARY_NARROWB, SVE2_INT_UNARY_NARROWT)
4446         (SVE2_INT_BINARY, SVE2_INT_BINARY_LANE, SVE2_INT_BINARY_LONG)
4447         (SVE2_INT_BINARY_LONG_LANE, SVE2_INT_BINARY_NARROWB)
4448         (SVE2_INT_BINARY_NARROWT, SVE2_INT_BINARY_PAIR, SVE2_FP_BINARY_PAIR)
4449         (SVE2_INT_BINARY_PAIR_LONG, SVE2_INT_BINARY_WIDE): New int iterators.
4450         (SVE2_INT_SHIFT_IMM_LONG, SVE2_INT_SHIFT_IMM_NARROWB): Likewise.
4451         (SVE2_INT_SHIFT_IMM_NARROWT, SVE2_INT_SHIFT_INSERT, SVE2_INT_CADD)
4452         (SVE2_INT_BITPERM, SVE2_INT_TERNARY, SVE2_INT_TERNARY_LANE): Likewise.
4453         (SVE2_FP_TERNARY_LONG, SVE2_FP_TERNARY_LONG_LANE, SVE2_INT_CMLA)
4454         (SVE2_INT_CDOT, SVE2_INT_ADD_BINARY_LONG, SVE2_INT_QADD_BINARY_LONG)
4455         (SVE2_INT_SUB_BINARY_LONG, SVE2_INT_QSUB_BINARY_LONG): Likewise.
4456         (SVE2_INT_ADD_BINARY_LONG_LANE, SVE2_INT_QADD_BINARY_LONG_LANE)
4457         (SVE2_INT_SUB_BINARY_LONG_LANE, SVE2_INT_QSUB_BINARY_LONG_LANE)
4458         (SVE2_COND_INT_UNARY_FP, SVE2_COND_FP_UNARY_LONG): Likewise.
4459         (SVE2_COND_FP_UNARY_NARROWB, SVE2_COND_INT_BINARY): Likewise.
4460         (SVE2_COND_INT_BINARY_NOREV, SVE2_COND_INT_BINARY_REV): Likewise.
4461         (SVE2_COND_INT_SHIFT, SVE2_MATCH, SVE2_PMULL): Likewise.
4462         (optab): Handle the new unspecs.
4463         (su, r): Remove entries for UNSPEC_SHRNB, UNSPEC_SHRNT, UNSPEC_RSHRNB
4464         and UNSPEC_RSHRNT.
4465         (lr): Handle the new unspecs.
4466         (bt): Delete.
4467         (cmp_op, while_optab_cmp, sve_int_op): Handle the new unspecs.
4468         (sve_int_op_rev, sve_int_add_op, sve_int_qadd_op, sve_int_sub_op)
4469         (sve_int_qsub_op): New int attributes.
4470         (sve_fp_op, rot): Handle the new unspecs.
4471         * config/aarch64/aarch64-sve-builtins.h
4472         (function_resolver::require_matching_pointer_type): Declare.
4473         (function_resolver::resolve_unary): Add an optional boolean argument.
4474         (function_resolver::finish_opt_n_resolution): Add an optional
4475         type_suffix_index argument.
4476         (gimple_folder::redirect_call): Declare.
4477         (gimple_expander::prepare_gather_address_operands): Add an optional
4478         bool parameter.
4479         * config/aarch64/aarch64-sve-builtins.cc: Include
4480         aarch64-sve-builtins-sve2.h.
4481         (TYPES_b_unsigned, TYPES_b_integer, TYPES_bh_integer): New macros.
4482         (TYPES_bs_unsigned, TYPES_hs_signed, TYPES_hs_integer): Likewise.
4483         (TYPES_hd_unsigned, TYPES_hsd_signed): Likewise.
4484         (TYPES_hsd_integer): Use TYPES_hsd_signed.
4485         (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): New macros.
4486         (TYPES_s_unsigned): Likewise.
4487         (TYPES_s_integer): Use TYPES_s_unsigned.
4488         (TYPES_sd_signed, TYPES_sd_unsigned): New macros.
4489         (TYPES_sd_integer): Use them.
4490         (TYPES_d_unsigned): New macro.
4491         (TYPES_d_integer): Use it.
4492         (TYPES_d_data, TYPES_cvt_long, TYPES_cvt_narrow_s): New macros.
4493         (TYPES_cvt_narrow): Likewise.
4494         (DEF_SVE_TYPES_ARRAY): Include the new types macros above.
4495         (preds_mx): New variable.
4496         (function_builder::add_overloaded_function): Allow the new feature
4497         set to be more restrictive than the original one.
4498         (function_resolver::infer_pointer_type): Remove qualifiers from
4499         the pointer type before printing it.
4500         (function_resolver::require_matching_pointer_type): New function.
4501         (function_resolver::resolve_sv_displacement): Handle functions
4502         that don't support 32-bit vector indices or svint32_t vector offsets.
4503         (function_resolver::finish_opt_n_resolution): Take the inferred type
4504         as a separate argument.
4505         (function_resolver::resolve_unary): Optionally treat all forms in
4506         the same way as normal merging functions.
4507         (gimple_folder::redirect_call): New function.
4508         (function_expander::prepare_gather_address_operands): Add an argument
4509         that says whether scaled forms are available.  If they aren't,
4510         handle scaling of vector indices and don't add the extension and
4511         scaling operands.
4512         (function_expander::map_to_unspecs): If aarch64_sve isn't available,
4513         fall back to using cond_* instead.
4514         * config/aarch64/aarch64-sve-builtins-functions.h (rtx_code_function):
4515         Split out the member variables into...
4516         (rtx_code_function_base): ...this new base class.
4517         (rtx_code_function_rotated): Inherit rtx_code_function_base.
4518         (unspec_based_function): Split out the member variables into...
4519         (unspec_based_function_base): ...this new base class.
4520         (unspec_based_function_rotated): Inherit unspec_based_function_base.
4521         (unspec_based_function_exact_insn): New class.
4522         (unspec_based_add_function, unspec_based_add_lane_function)
4523         (unspec_based_lane_function, unspec_based_pred_function)
4524         (unspec_based_qadd_function, unspec_based_qadd_lane_function)
4525         (unspec_based_qsub_function, unspec_based_qsub_lane_function)
4526         (unspec_based_sub_function, unspec_based_sub_lane_function): New
4527         typedefs.
4528         (unspec_based_fused_function): New class.
4529         (unspec_based_mla_function, unspec_based_mls_function): New typedefs.
4530         (unspec_based_fused_lane_function): New class.
4531         (unspec_based_mla_lane_function, unspec_based_mls_lane_function): New
4532         typedefs.
4533         (CODE_FOR_MODE1): New macro.
4534         (fixed_insn_function): New class.
4535         (while_comparison): Likewise.
4536         * config/aarch64/aarch64-sve-builtins-shapes.h (binary_long_lane)
4537         (binary_long_opt_n, binary_narrowb_opt_n, binary_narrowt_opt_n)
4538         (binary_to_uint, binary_wide, binary_wide_opt_n, compare, compare_ptr)
4539         (load_ext_gather_index_restricted, load_ext_gather_offset_restricted)
4540         (load_gather_sv_restricted, shift_left_imm_long): Declare.
4541         (shift_left_imm_to_uint, shift_right_imm_narrowb): Likewise.
4542         (shift_right_imm_narrowt, shift_right_imm_narrowb_to_uint): Likewise.
4543         (shift_right_imm_narrowt_to_uint, store_scatter_index_restricted)
4544         (store_scatter_offset_restricted, tbl_tuple, ternary_long_lane)
4545         (ternary_long_opt_n, ternary_qq_lane_rotate, ternary_qq_rotate)
4546         (ternary_shift_left_imm, ternary_shift_right_imm, ternary_uint)
4547         (unary_convert_narrowt, unary_long, unary_narrowb, unary_narrowt)
4548         (unary_narrowb_to_uint, unary_narrowt_to_uint, unary_to_int): Likewise.
4549         * config/aarch64/aarch64-sve-builtins-shapes.cc (apply_predication):
4550         Also add an initial argument for unary_convert_narrowt, regardless
4551         of the predication type.
4552         (build_32_64): Allow loads and stores to specify MODE_none.
4553         (build_sv_index64, build_sv_uint_offset): New functions.
4554         (long_type_suffix): New function.
4555         (binary_imm_narrowb_base, binary_imm_narrowt_base): New classes.
4556         (binary_imm_long_base, load_gather_sv_base): Likewise.
4557         (shift_right_imm_narrow_wrapper, ternary_shift_imm_base): Likewise.
4558         (ternary_resize2_opt_n_base, ternary_resize2_lane_base): Likewise.
4559         (unary_narrowb_base, unary_narrowt_base): Likewise.
4560         (binary_long_lane_def, binary_long_lane): New shape.
4561         (binary_long_opt_n_def, binary_long_opt_n): Likewise.
4562         (binary_narrowb_opt_n_def, binary_narrowb_opt_n): Likewise.
4563         (binary_narrowt_opt_n_def, binary_narrowt_opt_n): Likewise.
4564         (binary_to_uint_def, binary_to_uint): Likewise.
4565         (binary_wide_def, binary_wide): Likewise.
4566         (binary_wide_opt_n_def, binary_wide_opt_n): Likewise.
4567         (compare_def, compare): Likewise.
4568         (compare_ptr_def, compare_ptr): Likewise.
4569         (load_ext_gather_index_restricted_def,
4570         load_ext_gather_index_restricted): Likewise.
4571         (load_ext_gather_offset_restricted_def,
4572         load_ext_gather_offset_restricted): Likewise.
4573         (load_gather_sv_def): Inherit from load_gather_sv_base.
4574         (load_gather_sv_restricted_def, load_gather_sv_restricted): New shape.
4575         (shift_left_imm_def, shift_left_imm): Likewise.
4576         (shift_left_imm_long_def, shift_left_imm_long): Likewise.
4577         (shift_left_imm_to_uint_def, shift_left_imm_to_uint): Likewise.
4578         (store_scatter_index_restricted_def,
4579         store_scatter_index_restricted): Likewise.
4580         (store_scatter_offset_restricted_def,
4581         store_scatter_offset_restricted): Likewise.
4582         (tbl_tuple_def, tbl_tuple): Likewise.
4583         (ternary_long_lane_def, ternary_long_lane): Likewise.
4584         (ternary_long_opt_n_def, ternary_long_opt_n): Likewise.
4585         (ternary_qq_lane_def): Inherit from ternary_resize2_lane_base.
4586         (ternary_qq_lane_rotate_def, ternary_qq_lane_rotate): New shape
4587         (ternary_qq_opt_n_def): Inherit from ternary_resize2_opt_n_base.
4588         (ternary_qq_rotate_def, ternary_qq_rotate): New shape.
4589         (ternary_shift_left_imm_def, ternary_shift_left_imm): Likewise.
4590         (ternary_shift_right_imm_def, ternary_shift_right_imm): Likewise.
4591         (ternary_uint_def, ternary_uint): Likewise.
4592         (unary_convert): Fix typo in comment.
4593         (unary_convert_narrowt_def, unary_convert_narrowt): New shape.
4594         (unary_long_def, unary_long): Likewise.
4595         (unary_narrowb_def, unary_narrowb): Likewise.
4596         (unary_narrowt_def, unary_narrowt): Likewise.
4597         (unary_narrowb_to_uint_def, unary_narrowb_to_uint): Likewise.
4598         (unary_narrowt_to_uint_def, unary_narrowt_to_uint): Likewise.
4599         (unary_to_int_def, unary_to_int): Likewise.
4600         * config/aarch64/aarch64-sve-builtins-base.cc (unspec_cmla)
4601         (unspec_fcmla, unspec_cond_fcmla, expand_mla_mls_lane): New functions.
4602         (svasrd_impl): Delete.
4603         (svcadd_impl::expand): Handle integer operations too.
4604         (svcmla_impl::expand, svcmla_lane::expand): Likewise, using the
4605         new functions to derive the unspec numbers.
4606         (svmla_svmls_lane_impl): Replace with...
4607         (svmla_lane_impl, svmls_lane_impl): ...these new classes.  Handle
4608         integer operations too.
4609         (svwhile_impl): Rename to...
4610         (svwhilelx_impl): ...this and inherit from while_comparison.
4611         (svasrd): Use unspec_based_function.
4612         (svmla_lane): Use svmla_lane_impl.
4613         (svmls_lane): Use svmls_lane_impl.
4614         (svrecpe, svrsqrte): Handle unsigned integer operations too.
4615         (svwhilele, svwhilelt): Use svwhilelx_impl.
4616         * config/aarch64/aarch64-sve-builtins-sve2.h: New file.
4617         * config/aarch64/aarch64-sve-builtins-sve2.cc: Likewise.
4618         * config/aarch64/aarch64-sve-builtins-sve2.def: Likewise.
4619         * config/aarch64/aarch64-sve-builtins.def: Include
4620         aarch64-sve-builtins-sve2.def.
4622 2020-01-09  Richard Sandiford  <richard.sandiford@arm.com>
4624         * config/aarch64/aarch64-protos.h (aarch64_sve_arith_immediate_p)
4625         (aarch64_sve_sqadd_sqsub_immediate_p): Add a machine_mode argument.
4626         * config/aarch64/aarch64.c (aarch64_sve_arith_immediate_p)
4627         (aarch64_sve_sqadd_sqsub_immediate_p): Likewise.  Handle scalar
4628         immediates as well as vector ones.
4629         * config/aarch64/predicates.md (aarch64_sve_arith_immediate)
4630         (aarch64_sve_sub_arith_immediate, aarch64_sve_qadd_immediate)
4631         (aarch64_sve_qsub_immediate): Update calls accordingly.
4633 2020-01-09  Richard Sandiford  <richard.sandiford@arm.com>
4635         * config/aarch64/aarch64-sve2.md: Add banner comments.
4636         (<su>mulh<r>s<mode>3): Move further up file.
4637         (<su>mull<bt><Vwide>, <r>shrnb<mode>, <r>shrnt<mode>)
4638         (*aarch64_sve2_sra<mode>): Move further down file.
4639         * config/aarch64/t-aarch64 (s-check-sve-md): Check aarch64-sve2.md too.
4641 2020-01-09  Richard Sandiford  <richard.sandiford@arm.com>
4643         * config/aarch64/iterators.md (SVE_WHILE): Add UNSPEC_WHILERW
4644         and UNSPEC_WHILEWR.
4645         (while_optab_cmp): Handle them.
4646         * config/aarch64/aarch64-sve.md
4647         (*while_<while_optab_cmp><GPI:mode><PRED_ALL:mode>_ptest): Make public
4648         and add a "@" marker.
4649         * config/aarch64/aarch64-sve2.md (check_<raw_war>_ptrs<mode>): Use it
4650         instead of gen_aarch64_sve2_while_ptest.
4651         (@aarch64_sve2_while<cmp_op><GPI:mode><PRED_ALL:mode>_ptest): Delete.
4653 2020-01-09  Richard Sandiford  <richard.sandiford@arm.com>
4655         * config/aarch64/aarch64.md (UNSPEC_WHILE_LE): Rename to...
4656         (UNSPEC_WHILELE): ...this.
4657         (UNSPEC_WHILE_LO): Rename to...
4658         (UNSPEC_WHILELO): ...this.
4659         (UNSPEC_WHILE_LS): Rename to...
4660         (UNSPEC_WHILELS): ...this.
4661         (UNSPEC_WHILE_LT): Rename to...
4662         (UNSPEC_WHILELT): ...this.
4663         * config/aarch64/iterators.md (SVE_WHILE): Update accordingly.
4664         (cmp_op, while_optab_cmp): Likewise.
4665         * config/aarch64/aarch64.c (aarch64_sve_move_pred_via_while): Likewise.
4666         * config/aarch64/aarch64-sve-builtins-base.cc (svwhilele): Likewise.
4667         (svwhilelt): Likewise.
4669 2020-01-09  Richard Sandiford  <richard.sandiford@arm.com>
4671         * config/aarch64/aarch64-sve-builtins-shapes.h (unary_count): Delete.
4672         (unary_to_uint): Define.
4673         * config/aarch64/aarch64-sve-builtins-shapes.cc (unary_count_def)
4674         (unary_count): Rename to...
4675         (unary_to_uint_def, unary_to_uint): ...this.
4676         * config/aarch64/aarch64-sve-builtins-base.def: Update accordingly.
4678 2020-01-09  Richard Sandiford  <richard.sandiford@arm.com>
4680         * config/aarch64/aarch64-sve-builtins-functions.h
4681         (code_for_mode_function): New class.
4682         (CODE_FOR_MODE0, QUIET_CODE_FOR_MODE0): New macros.
4683         * config/aarch64/aarch64-sve-builtins-base.cc (svcompact_impl)
4684         (svext_impl, svmul_lane_impl, svsplice_impl, svtmad_impl): Delete.
4685         (svcompact, svext, svsplice): Use QUIET_CODE_FOR_MODE0.
4686         (svmul_lane, svtmad): Use CODE_FOR_MODE0.
4688 2020-01-09  Richard Sandiford  <richard.sandiford@arm.com>
4690         * config/aarch64/iterators.md (addsub): New code attribute.
4691         * config/aarch64/aarch64-simd.md (aarch64_<su_optab><optab><mode>):
4692         Re-express as...
4693         (aarch64_<su_optab>q<addsub><mode>): ...this, making the same change
4694         in the asm string and attributes.  Fix indentation.
4695         * config/aarch64/aarch64-sve.md (@aarch64_<su_optab><optab><mode>):
4696         Re-express as...
4697         (@aarch64_sve_<optab><mode>): ...this.
4698         * config/aarch64/aarch64-sve-builtins.h
4699         (function_expander::expand_signed_unpred_op): Delete.
4700         * config/aarch64/aarch64-sve-builtins.cc
4701         (function_expander::expand_signed_unpred_op): Likewise.
4702         (function_expander::map_to_rtx_codes): If the optab isn't defined,
4703         try using code_for_aarch64_sve instead.
4704         * config/aarch64/aarch64-sve-builtins-base.cc (svqadd_impl): Delete.
4705         (svqsub_impl): Likewise.
4706         (svqadd, svqsub): Use rtx_code_function instead.
4708 2020-01-09  Richard Sandiford  <richard.sandiford@arm.com>
4710         * config/aarch64/iterators.md (SRHSUB, URHSUB): Delete.
4711         (HADDSUB, sur, addsub): Remove them.
4713 2020-01-09  Richard Sandiford  <richard.sandiford@arm.com>
4715         * tree-nrv.c (pass_return_slot::execute): Handle all internal
4716         functions the same way, rather than singling out those that
4717         aren't mapped directly to optabs.
4719 2020-01-09  Richard Sandiford  <richard.sandiford@arm.com>
4721         * target.def (compatible_vector_types_p): New target hook.
4722         * hooks.h (hook_bool_const_tree_const_tree_true): Declare.
4723         * hooks.c (hook_bool_const_tree_const_tree_true): New function.
4724         * doc/tm.texi.in (TARGET_COMPATIBLE_VECTOR_TYPES_P): New hook.
4725         * doc/tm.texi: Regenerate.
4726         * gimple-expr.c: Include target.h.
4727         (useless_type_conversion_p): Use targetm.compatible_vector_types_p.
4728         * config/aarch64/aarch64.c (aarch64_compatible_vector_types_p): New
4729         function.
4730         (TARGET_COMPATIBLE_VECTOR_TYPES_P): Define.
4731         * config/aarch64/aarch64-sve-builtins.cc (gimple_folder::convert_pred):
4732         Use the original predicate if it already has a suitable type.
4734 2020-01-09  Martin Jambor  <mjambor@suse.cz>
4736         * cgraph.h (cgraph_edge): Make remove, set_call_stmt, make_direct,
4737         resolve_speculation and redirect_call_stmt_to_callee static.  Change
4738         return type of set_call_stmt to cgraph_edge *.
4739         * auto-profile.c (afdo_indirect_call): Adjust call to
4740         redirect_call_stmt_to_callee.
4741         * cgraph.c (cgraph_edge::set_call_stmt): Make return cgraph-edge *,
4742         make the this pointer explicit, adjust self-recursive calls and the
4743         call top make_direct.  Return the resulting edge.
4744         (cgraph_edge::remove): Make this pointer explicit.
4745         (cgraph_edge::resolve_speculation): Likewise, adjust call to remove.
4746         (cgraph_edge::make_direct): Likewise, adjust call to
4747         resolve_speculation.
4748         (cgraph_edge::redirect_call_stmt_to_callee): Likewise, also adjust
4749         call to set_call_stmt.
4750         (cgraph_update_edges_for_call_stmt_node): Update call to
4751         set_call_stmt and remove.
4752         * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
4753         Renamed edge to master_edge.  Adjusted calls to set_call_stmt.
4754         (cgraph_node::create_edge_including_clones): Moved "first" definition
4755         of edge to the block where it was used.  Adjusted calls to
4756         set_call_stmt.
4757         (cgraph_node::remove_symbol_and_inline_clones): Adjust call to
4758         cgraph_edge::remove.
4759         * cgraphunit.c (walk_polymorphic_call_targets): Adjusted calls to
4760         make_direct and redirect_call_stmt_to_callee.
4761         * ipa-fnsummary.c (redirect_to_unreachable): Adjust calls to
4762         resolve_speculation and make_direct.
4763         * ipa-inline-transform.c (inline_transform): Adjust call to
4764         redirect_call_stmt_to_callee.
4765         (check_speculations_1):: Adjust call to resolve_speculation.
4766         * ipa-inline.c (resolve_noninline_speculation): Adjust call to
4767         resolve-speculation.
4768         (inline_small_functions): Adjust call to resolve_speculation.
4769         (ipa_inline): Likewise.
4770         * ipa-prop.c (ipa_make_edge_direct_to_target): Adjust call to
4771         make_direct.
4772         * ipa-visibility.c (function_and_variable_visibility): Make iteration
4773         safe with regards to edge removal, adjust calls to
4774         redirect_call_stmt_to_callee.
4775         * ipa.c (walk_polymorphic_call_targets): Adjust calls to make_direct
4776         and redirect_call_stmt_to_callee.
4777         * multiple_target.c (create_dispatcher_calls): Adjust call to
4778         redirect_call_stmt_to_callee
4779         (redirect_to_specific_clone): Likewise.
4780         * tree-cfgcleanup.c (delete_unreachable_blocks_update_callgraph):
4781         Adjust calls to cgraph_edge::remove.
4782         * tree-inline.c (copy_bb): Adjust call to set_call_stmt.
4783         (redirect_all_calls): Adjust call to redirect_call_stmt_to_callee.
4784         (expand_call_inline): Adjust call to cgraph_edge::remove.
4786 2020-01-09  Martin Liska  <mliska@suse.cz>
4788         * params.opt: Set Optimization for
4789         param_max_speculative_devirt_maydefs.
4791 2020-01-09  Martin Sebor  <msebor@redhat.com>
4793         PR middle-end/93200
4794         PR fortran/92956
4795         * builtins.c (compute_objsize): Avoid handling MEM_REFs of vector type.
4797 2020-01-09  Martin Liska  <mliska@suse.cz>
4799         * auto-profile.c (auto_profile): Use opt_for_fn
4800         for a parameter.
4801         * ipa-cp.c (ipcp_lattice::add_value): Likewise.
4802         (propagate_vals_across_arith_jfunc): Likewise.
4803         (hint_time_bonus): Likewise.
4804         (incorporate_penalties): Likewise.
4805         (good_cloning_opportunity_p): Likewise.
4806         (perform_estimation_of_a_value): Likewise.
4807         (estimate_local_effects): Likewise.
4808         (ipcp_propagate_stage): Likewise.
4809         * ipa-fnsummary.c (decompose_param_expr): Likewise.
4810         (set_switch_stmt_execution_predicate): Likewise.
4811         (analyze_function_body): Likewise.
4812         * ipa-inline-analysis.c (offline_size): Likewise.
4813         * ipa-inline.c (early_inliner): Likewise.
4814         * ipa-prop.c (ipa_analyze_node): Likewise.
4815         (ipcp_transform_function): Likewise.
4816         * ipa-sra.c (process_scan_results): Likewise.
4817         (ipa_sra_summarize_function): Likewise.
4818         * params.opt: Rename ipcp-unit-growth to
4819         ipa-cp-unit-growth.  Add Optimization for various
4820         IPA-related parameters.
4822 2020-01-09  Richard Biener  <rguenther@suse.de>
4824         PR middle-end/93054
4825         * gimplify.c (gimplify_expr): Deal with NOP definitions.
4827 2020-01-09  Richard Biener  <rguenther@suse.de>
4829         PR tree-optimization/93040
4830         * gimple-ssa-store-merging.c (find_bswap_or_nop): Raise search limit.
4832 2020-01-09  Georg-Johann Lay  <avr@gjlay.de>
4834         * common/config/avr/avr-common.c (avr_option_optimization_table)
4835         [OPT_LEVELS_1_PLUS]: Set -fsplit-wide-types-early.
4837 2020-01-09  Martin Liska  <mliska@suse.cz>
4839         * cgraphclones.c (symbol_table::materialize_all_clones):
4840         Use cgraph_node::dump_name.
4842 2020-01-09  Jakub Jelinek  <jakub@redhat.com>
4844         PR inline-asm/93202
4845         * config/riscv/riscv.c (riscv_print_operand_reloc): Use
4846         output_operand_lossage instead of gcc_unreachable.
4847         * doc/md.texi (riscv f constraint): Fix typo.
4849         PR target/93141
4850         * config/i386/i386.md (subv<mode>4): Use SWIDWI iterator instead of
4851         SWI.  Use <general_hilo_operand> instead of <general_operand>.  Use
4852         CONST_SCALAR_INT_P instead of CONST_INT_P.
4853         (*subv<mode>4_1): Rename to ...
4854         (subv<mode>4_1): ... this.
4855         (*subv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
4856         define_insn_and_split patterns.
4857         (*subv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
4858         patterns.
4860 2020-01-08  David Malcolm  <dmalcolm@redhat.com>
4862         * vec.c (class selftest::count_dtor): New class.
4863         (selftest::test_auto_delete_vec): New test.
4864         (selftest::vec_c_tests): Call it.
4865         * vec.h (class auto_delete_vec): New class template.
4866         (auto_delete_vec<T>::~auto_delete_vec): New dtor.
4868 2020-01-08  David Malcolm  <dmalcolm@redhat.com>
4870         * sbitmap.h (auto_sbitmap): Add operator const_sbitmap.
4872 2020-01-08  Jim Wilson  <jimw@sifive.com>
4874         * config/riscv/riscv.c (riscv_legitimize_tls_address): Ifdef out
4875         use of TLS_MODEL_LOCAL_EXEC when not pic.
4877 2020-01-08  David Malcolm  <dmalcolm@redhat.com>
4879         * hash-map-tests.c (selftest::test_map_of_strings_to_int): Fix
4880         memory leak.
4882 2020-01-08  Jakub Jelinek  <jakub@redhat.com>
4884         PR target/93187
4885         * config/i386/i386.md (*stack_protect_set_2_<mode> peephole2,
4886         *stack_protect_set_3 peephole2): Also check that the second
4887         insns source is general_operand.
4889         PR target/93174
4890         * config/i386/i386.md (addcarry<mode>_0): Use nonimmediate_operand
4891         predicate for output operand instead of register_operand.
4892         (addcarry<mode>, addcarry<mode>_1): Likewise.  Add alternative with
4893         memory destination and non-memory operands[2].
4895 2020-01-08  Martin Liska  <mliska@suse.cz>
4897         * cgraph.c (cgraph_node::dump): Use ::dump_name or
4898         ::dump_asm_name instead of (::name or ::asm_name).
4899         * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
4900         * cgraphunit.c (walk_polymorphic_call_targets): Likewise.
4901         (analyze_functions): Likewise.
4902         (expand_all_functions): Likewise.
4903         * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
4904         (propagate_bits_across_jump_function): Likewise.
4905         (dump_profile_updates): Likewise.
4906         (ipcp_store_bits_results): Likewise.
4907         (ipcp_store_vr_results): Likewise.
4908         * ipa-devirt.c (dump_targets): Likewise.
4909         * ipa-fnsummary.c (analyze_function_body): Likewise.
4910         * ipa-hsa.c (check_warn_node_versionable): Likewise.
4911         (process_hsa_functions): Likewise.
4912         * ipa-icf.c (sem_item_optimizer::merge_classes): Likewise.
4913         (set_alias_uids): Likewise.
4914         * ipa-inline-transform.c (save_inline_function_body): Likewise.
4915         * ipa-inline.c (recursive_inlining): Likewise.
4916         (inline_to_all_callers_1): Likewise.
4917         (ipa_inline): Likewise.
4918         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
4919         (ipa_propagate_frequency): Likewise.
4920         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
4921         (remove_described_reference): Likewise.
4922         * ipa-pure-const.c (worse_state): Likewise.
4923         (check_retval_uses): Likewise.
4924         (analyze_function): Likewise.
4925         (propagate_pure_const): Likewise.
4926         (propagate_nothrow): Likewise.
4927         (dump_malloc_lattice): Likewise.
4928         (propagate_malloc): Likewise.
4929         (pass_local_pure_const::execute): Likewise.
4930         * ipa-visibility.c (optimize_weakref): Likewise.
4931         (function_and_variable_visibility): Likewise.
4932         * ipa.c (symbol_table::remove_unreachable_nodes): Likewise.
4933         (ipa_discover_variable_flags): Likewise.
4934         * lto-streamer-out.c (output_function): Likewise.
4935         (output_constructor): Likewise.
4936         * tree-inline.c (copy_bb): Likewise.
4937         * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
4938         * varpool.c (symbol_table::remove_unreferenced_decls): Likewise.
4940 2020-01-08  Richard Biener  <rguenther@suse.de>
4942         PR middle-end/93199
4943         * tree-eh.c (sink_clobbers): Update virtual operands for
4944         the first and last stmt only.  Add a dry-run capability.
4945         (pass_lower_eh_dispatch::execute): Perform clobber sinking
4946         after CFG manipulations and in RPO order to catch all
4947         secondary opportunities reliably.
4949 2020-01-08  Georg-Johann Lay  <avr@gjlay.de>
4951         PR target/93182
4952         * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
4954 2019-01-08  Richard Biener  <rguenther@suse.de>
4956         PR middle-end/93199
4957         * gimple-fold.c (rewrite_to_defined_overflow): Mark stmt modified.
4958         * tree-ssa-loop-im.c (move_computations_worker): Properly adjust
4959         virtual operand, also updating SSA use.
4960         * gimple-loop-interchange.cc (loop_cand::undo_simple_reduction):
4961         Update stmt after resetting virtual operand.
4962         (tree_loop_interchange::move_code_to_inner_loop): Likewise.
4963         * gimple-iterator.c (gsi_remove): When not removing the stmt
4964         permanently do not delink immediate uses or mark the stmt modified.
4966 2020-01-08  Martin Liska  <mliska@suse.cz>
4968         * ipa-fnsummary.c (dump_ipa_call_summary): Use symtab_node::dump_name.
4969         (ipa_call_context::estimate_size_and_time): Likewise.
4970         (inline_analyze_function): Likewise.
4972 2020-01-08  Martin Liska  <mliska@suse.cz>
4974         * cgraph.c (cgraph_node::dump): Use systematically
4975         dump_asm_name.
4977 2020-01-08  Georg-Johann Lay  <avr@gjlay.de>
4979         Add -nodevicespecs option for avr.
4981         PR target/93182
4982         * config/avr/avr.opt (-nodevicespecs): New driver option.
4983         * config/avr/driver-avr.c (avr_devicespecs_file): Only issue
4984         "-specs=device-specs/..." if that option is not set.
4985         * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
4987 2020-01-08  Georg-Johann Lay  <avr@gjlay.de>
4989         Implement 64-bit double functions for avr.
4991         PR target/92055
4992         * config.gcc (tm_defines) [target=avr]: Support --with-libf7,
4993         --with-double-comparison.
4994         * doc/install.texi: Document them.
4995         * config/avr/avr-c.c (avr_cpu_cpp_builtins)
4996         <WITH_LIBF7_LIBGCC, WITH_LIBF7_MATH, WITH_LIBF7_MATH_SYMBOLS>
4997         <WITH_DOUBLE_COMPARISON>: New built-in defines.
4998         * doc/invoke.texi (AVR Built-in Macros): Document them.
4999         * config/avr/avr-protos.h (avr_float_lib_compare_returns_bool): New.
5000         * config/avr/avr.c (avr_float_lib_compare_returns_bool): New function.
5001         * config/avr/avr.h (FLOAT_LIB_COMPARE_RETURNS_BOOL): New macro.
5003 2020-01-08  Richard Earnshaw  <rearnsha@arm.com>
5005         PR target/93188
5006         * config/arm/t-multilib (MULTILIB_MATCHES): Add rules to match
5007         armv7-a{+mp,+sec,+mp+sec} to appropriate armv7 multilib variants
5008         when only building rm-profile multilibs.
5010 2020-01-08  Feng Xue  <fxue@os.amperecomputing.com>
5012         PR ipa/93084
5013         * ipa-cp.c (self_recursively_generated_p): Find matched aggregate
5014         lattice for a value to check.
5015         (propagate_vals_across_arith_jfunc): Add an assertion to ensure
5016         finite propagation in self-recursive scc.
5018 2020-01-08  Luo Xiong Hu  <luoxhu@linux.ibm.com>
5020         * ipa-inline.c (caller_growth_limits): Restore the AND.
5022 2020-01-07  Andrew Stubbs  <ams@codesourcery.com>
5024         * config/gcn/gcn-valu.md (VEC_1REG_INT_ALT): Delete iterator.
5025         (VEC_ALLREG_ALT): New iterator.
5026         (VEC_ALLREG_INT_MODE): New iterator.
5027         (VCMP_MODE): New iterator.
5028         (VCMP_MODE_INT): New iterator.
5029         (vec_cmpu<mode>di): Use VCMP_MODE_INT.
5030         (vec_cmp<u>v64qidi): New define_expand.
5031         (vec_cmp<mode>di_exec): Use VCMP_MODE.
5032         (vec_cmpu<mode>di_exec): New define_expand.
5033         (vec_cmp<u>v64qidi_exec): New define_expand.
5034         (vec_cmp<mode>di_dup): Use VCMP_MODE.
5035         (vec_cmp<mode>di_dup_exec): Use VCMP_MODE.
5036         (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>): Rename ...
5037         (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): ... to this.
5038         (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>_exec): Rename ...
5039         (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): ... to this.
5040         (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>): Rename ...
5041         (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): ... to this.
5042         (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>_exec): Rename ...
5043         (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): ... to
5044         this.
5045         * config/gcn/gcn.c (print_operand): Fix 8 and 16 bit suffixes.
5046         * config/gcn/gcn.md (expander): Add sign_extend and zero_extend.
5048 2020-01-07  Andrew Stubbs  <ams@codesourcery.com>
5050         * config/gcn/constraints.md (DA): Update description and match.
5051         (DB): Likewise.
5052         (Db): New constraint.
5053         * config/gcn/gcn-protos.h (gcn_inline_constant64_p): Add second
5054         parameter.
5055         * config/gcn/gcn.c (gcn_inline_constant64_p): Add 'mixed' parameter.
5056         Implement 'Db' mixed immediate type.
5057         * config/gcn/gcn-valu.md (addcv64si3<exec_vcc>): Rework constraints.
5058         (addcv64si3_dup<exec_vcc>): Delete.
5059         (subcv64si3<exec_vcc>): Rework constraints.
5060         (addv64di3): Rework constraints.
5061         (addv64di3_exec): Rework constraints.
5062         (subv64di3): Rework constraints.
5063         (addv64di3_dup): Delete.
5064         (addv64di3_dup_exec): Delete.
5065         (addv64di3_zext): Rework constraints.
5066         (addv64di3_zext_exec): Rework constraints.
5067         (addv64di3_zext_dup): Rework constraints.
5068         (addv64di3_zext_dup_exec): Rework constraints.
5069         (addv64di3_zext_dup2): Rework constraints.
5070         (addv64di3_zext_dup2_exec): Rework constraints.
5071         (addv64di3_sext_dup2): Rework constraints.
5072         (addv64di3_sext_dup2_exec): Rework constraints.
5074 2020-01-07  Andre Vieira  <andre.simoesdiasvieira@arm.com>
5076         * doc/sourcebuild.texi (arm_little_endian, arm_nothumb): Documented
5077         existing target checks.
5079 2020-01-07  Richard Biener  <rguenther@suse.de>
5081         * doc/install.texi: Bump minimal supported MPC version.
5083 2020-01-07  Richard Sandiford  <richard.sandiford@arm.com>
5085         * langhooks-def.h (lhd_simulate_enum_decl): Declare.
5086         (LANG_HOOKS_SIMULATE_ENUM_DECL): Use it.
5087         * langhooks.c: Include stor-layout.h.
5088         (lhd_simulate_enum_decl): New function.
5089         * config/aarch64/aarch64-sve-builtins.cc (init_builtins): Call
5090         handle_arm_sve_h for the LTO frontend.
5091         (register_vector_type): Cope with null returns from pushdecl.
5093 2020-01-07  Richard Sandiford  <richard.sandiford@arm.com>
5095         * config/aarch64/aarch64-protos.h (aarch64_sve::svbool_type_p)
5096         (aarch64_sve::nvectors_if_data_type): Replace with...
5097         (aarch64_sve::builtin_type_p): ...this.
5098         * config/aarch64/aarch64-sve-builtins.cc: Include attribs.h.
5099         (find_vector_type): Delete.
5100         (add_sve_type_attribute): New function.
5101         (lookup_sve_type_attribute): Likewise.
5102         (register_builtin_types): Add an "SVE type" attribute to each type.
5103         (register_tuple_type): Likewise.
5104         (svbool_type_p, nvectors_if_data_type): Delete.
5105         (mangle_builtin_type): Use lookup_sve_type_attribute.
5106         (builtin_type_p): Likewise.  Add an overload that returns the
5107         number of constituent vector and predicate registers.
5108         * config/aarch64/aarch64.c (aarch64_sve_argument_p): Delete.
5109         (aarch64_returns_value_in_sve_regs_p): Use aarch64_sve::builtin_type_p
5110         instead of aarch64_sve_argument_p.
5111         (aarch64_takes_arguments_in_sve_regs_p): Likewise.
5112         (aarch64_pass_by_reference): Likewise.
5113         (aarch64_function_value_1): Likewise.
5114         (aarch64_return_in_memory): Likewise.
5115         (aarch64_layout_arg): Likewise.
5117 2020-01-07  Jakub Jelinek  <jakub@redhat.com>
5119         PR tree-optimization/93156
5120         * tree-ssa-ccp.c (bit_value_binop): For x * x note that the second
5121         least significant bit is always clear.
5123         PR tree-optimization/93118
5124         * match.pd ((x >> c) << c -> x & (-1<<c)): Add nop_convert?.  Add new
5125         simplifier with two intermediate conversions.
5127 2020-01-07  Martin Liska  <mliska@suse.cz>
5129         * params.opt: Add Optimization for various parameters.
5131 2020-01-07  Martin Liska  <mliska@suse.cz>
5133         PR ipa/83411
5134         * doc/extend.texi: Explain cloning for target_clone
5135         attribute.
5137 2020-01-07  Martin Liska  <mliska@suse.cz>
5139         PR tree-optimization/92860
5140         * common.opt: Make in Optimization option
5141         as it is affected by -O0, which is an Optimization
5142         option.
5143         * tree-inline.c (tree_inlinable_function_p):
5144         Use opt_for_fn for warn_inline.
5145         (expand_call_inline): Likewise.
5147 2020-01-07  Martin Liska  <mliska@suse.cz>
5149         PR tree-optimization/92860
5150         * common.opt: Make flag_ree as optimization
5151         attribute. 
5153 2020-01-07  Martin Liska  <mliska@suse.cz>
5155         PR optimization/92860
5156         * params.opt: Mark param_min_crossjump_insns with Optimization
5157         keyword.
5159 2020-01-07  Luo Xiong Hu  <luoxhu@linux.ibm.com>
5161         * ipa-inline-analysis.c (estimate_growth): Fix typo.
5162         * ipa-inline.c (caller_growth_limits): Use OR instead of AND.
5164 2020-01-06  Michael Meissner  <meissner@linux.ibm.com>
5166         * config/rs6000/rs6000.c (hard_reg_and_mode_to_addr_mask): New
5167         helper function to return the valid addressing formats for a given
5168         hard register and mode.
5169         (rs6000_adjust_vec_address): Call hard_reg_and_mode_to_addr_mask.
5171         * config/rs6000/constraints.md (Q constraint): Update
5172         documentation.
5173         * doc/md.texi (RS/6000 constraints): Update 'Q' cosntraint
5174         documentation.
5176         * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
5177         Use 'Q' for doing vector extract from memory.
5178         (vsx_extract_v4sf_var): Use 'Q' for doing vector extract from
5179         memory.
5180         (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Use 'Q' for
5181         doing vector extract from memory.
5182         (vsx_extract_<mode>_<VS_scalar>mode_var): Use 'Q' for doing vector
5183         extract from memory.
5185         * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add support
5186         for the offset being 34-bits when -mcpu=future is used.
5188 2020-01-06  John David Anglin  <danglin@gcc.gnu.org>
5190         * config/pa/pa.md: Revert change to use ordered_comparison_operator
5191         instead of cmpib_comparison_operator in cmpib patterns.
5192         * config/pa/predicates.md (cmpib_comparison_operator): Revert removal
5193         of cmpib_comparison_operator.  Revise comment.
5195 2020-01-06  Richard Sandiford  <richard.sandiford@arm.com>
5197         * tree-vect-slp.c (vect_build_slp_tree_1): Require all shifts
5198         in an IFN_DIV_POW2 node to be equal.
5200 2020-01-06  Richard Sandiford  <richard.sandiford@arm.com>
5202         * tree-vect-stmts.c (vect_check_load_store_mask): Rename to...
5203         (vect_check_scalar_mask): ...this.
5204         (vectorizable_store, vectorizable_load): Update call accordingly.
5205         (vectorizable_call): Use vect_check_scalar_mask to check the mask
5206         argument in calls to conditional internal functions.
5208 2020-01-06  Andrew Stubbs  <ams@codesourcery.com>
5210         * config/gcn/gcn-valu.md (subv64di3): Use separate alternatives for
5211         '0' matching inputs.
5212         (subv64di3_exec): Likewise.
5214 2020-01-06  Bryan Stenson  <bryan@siliconvortex.com>
5216         * config/mips/mips.c (vr4130_align_insns): Fix typo.
5217         * doc/md.texi (movstr): Likewise.
5219 2020-01-06  Andrew Stubbs  <ams@codesourcery.com>
5221         * config/gcn/gcn-valu.md (vec_extract<mode><scalar_mode>): Add early
5222         clobber.
5224 2020-01-06  Richard Sandiford  <richard.sandiford@arm.com>
5226         * config/aarch64/t-aarch64 ($(srcdir)/config/aarch64/aarch64-tune.md):
5227         Depend on...
5228         (s-aarch64-tune-md): ...this new stamp file.  Pipe the new contents
5229         to a temporary file and use move-if-change to update the real
5230         file where necessary.
5232 2020-01-06  Richard Sandiford  <richard.sandiford@arm.com>
5234         * config/aarch64/aarch64-sve.md (@aarch64_sel_dup<mode>): Use Upl
5235         rather than Upa for CPY /M.
5237 2020-01-06  Andrew Stubbs  <ams@codesourcery.com>
5239         * config/gcn/gcn.c (gcn_inline_constant_p): Allow 64 as an inline
5240         immediate.
5242 2020-01-06  Martin Liska  <mliska@suse.cz>
5244     PR tree-optimization/92860
5245     * params.opt: Mark param_max_combine_insns with Optimization
5246     keyword. 
5248 2020-01-05  Jakub Jelinek  <jakub@redhat.com>
5250         PR target/93141
5251         * config/i386/i386.md (SWIDWI): New mode iterator.
5252         (DWI, dwi): Add TImode variants.
5253         (addv<mode>4): Use SWIDWI iterator instead of SWI.  Use
5254         <general_hilo_operand> instead of <general_operand>.  Use
5255         CONST_SCALAR_INT_P instead of CONST_INT_P.
5256         (*addv<mode>4_1): Rename to ...
5257         (addv<mode>4_1): ... this.
5258         (QWI): New mode attribute.
5259         (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
5260         define_insn_and_split patterns.
5261         (*addv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
5262         patterns.
5263         (uaddv<mode>4): Use SWIDWI iterator instead of SWI.  Use
5264         <general_hilo_operand> instead of <general_operand>.
5265         (*addcarry<mode>_1): New define_insn.
5266         (*add<dwi>3_doubleword_cc_overflow_1): New define_insn_and_split.
5268 2020-01-03  Konstantin Kharlamov  <Hi-Angel@yandex.ru>
5270         * gdbinit.in (pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, pdd, pbs, pbm):
5271         Use "call" instead of "set".
5273 2020-01-03  Martin Jambor  <mjambor@suse.cz>
5275         PR ipa/92917
5276         * ipa-cp.c (print_all_lattices): Skip functions without info.
5278 2020-01-03  Jakub Jelinek  <jakub@redhat.com>
5280         PR target/93089
5281         * config/i386/i386-options.c (ix86_simd_clone_adjust): If
5282         TARGET_PREFER_AVX128, use prefer-vector-width=256 for 'c' and 'd'
5283         simd clones.  If TARGET_PREFER_AVX256, use prefer-vector-width=512
5284         for 'e' simd clones.
5286         PR target/93089
5287         * config/i386/i386.opt (x_prefer_vector_width_type): Remove TargetSave
5288         entry.
5289         (mprefer-vector-width=): Add Save.
5290         * config/i386/i386-options.c (ix86_target_string): Add PVW argument, print
5291         -mprefer-vector-width= if non-zero.  Fix up -mfpmath= comment.
5292         (ix86_debug_options, ix86_function_specific_print): Adjust
5293         ix86_target_string callers.
5294         (ix86_valid_target_attribute_inner_p): Handle prefer-vector-width=.
5295         (ix86_valid_target_attribute_tree): Likewise.
5296         * config/i386/i386-options.h (ix86_target_string): Add PVW argument.
5297         * config/i386/i386-expand.c (ix86_expand_builtin): Adjust
5298         ix86_target_string caller.
5300         PR target/93110
5301         * config/i386/i386.md (abs<mode>2): Use expand_simple_binop instead of
5302         emitting ASHIFTRT, XOR and MINUS by hand.  Use gen_int_mode with QImode
5303         instead of gen_int_shift_amount + convert_modes.
5305         PR rtl-optimization/93088
5306         * loop-iv.c (find_single_def_src): Punt after looking through
5307         128 reg copies for regs with single definitions.  Move definitions
5308         to first uses.
5310 2020-01-02  Dennis Zhang  <dennis.zhang@arm.com>
5312         * config/arm/arm-c.c (arm_cpu_builtins): Define
5313         __ARM_FEATURE_MATMUL_INT8, __ARM_FEATURE_BF16_VECTOR_ARITHMETIC,
5314         __ARM_FEATURE_BF16_SCALAR_ARITHMETIC, and
5315         __ARM_BF16_FORMAT_ALTERNATIVE when enabled.
5316         * config/arm/arm-cpus.in (armv8_6, i8mm, bf16): New features.
5317         * config/arm/arm-tables.opt: Regenerated.
5318         * config/arm/arm.c (arm_option_reconfigure_globals): Initialize
5319         arm_arch_i8mm and arm_arch_bf16 when enabled.
5320         * config/arm/arm.h (TARGET_I8MM): New macro.
5321         (TARGET_BF16_FP, TARGET_BF16_SIMD): Likewise.
5322         * config/arm/t-aprofile: Add matching rules for -march=armv8.6-a.
5323         * config/arm/t-arm-elf (all_v8_archs): Add armv8.6-a.
5324         * config/arm/t-multilib: Add matching rules for -march=armv8.6-a.
5325         (v8_6_a_simd_variants): New.
5326         (v8_*_a_simd_variants): Add i8mm and bf16.
5327         * doc/invoke.texi (armv8.6-a, i8mm, bf16): Document new options.
5329 2020-01-02  Jakub Jelinek  <jakub@redhat.com>
5331         PR ipa/93087
5332         * predict.c (compute_function_frequency): Don't call
5333         warn_function_cold on functions that already have cold attribute.
5335 2020-01-01  John David Anglin  <danglin@gcc.gnu.org>
5337         PR target/67834
5338         * config/pa/pa.c (pa_elf_select_rtx_section): New.  Put references to
5339         COMDAT group function labels in .data.rel.ro.local section.
5340         * config/pa/pa32-linux.h (TARGET_ASM_SELECT_RTX_SECTION): Define.
5342         PR target/93111
5343         * config/pa/pa.md (scc): Use ordered_comparison_operator instead of
5344         comparison_operator in B and S integer comparisons.  Likewise, use
5345         ordered_comparison_operator instead of cmpib_comparison_operator in
5346         cmpib patterns.
5347         * config/pa/predicates.md (cmpib_comparison_operator): Remove.
5349 2020-01-01  Jakub Jelinek  <jakub@redhat.com>
5351         Update copyright years.
5353         * gcc.c (process_command): Update copyright notice dates.
5354         * gcov-dump.c (print_version): Ditto.
5355         * gcov.c (print_version): Ditto.
5356         * gcov-tool.c (print_version): Ditto.
5357         * gengtype.c (create_file): Ditto.
5358         * doc/cpp.texi: Bump @copying's copyright year.
5359         * doc/cppinternals.texi: Ditto.
5360         * doc/gcc.texi: Ditto.
5361         * doc/gccint.texi: Ditto.
5362         * doc/gcov.texi: Ditto.
5363         * doc/install.texi: Ditto.
5364         * doc/invoke.texi: Ditto.
5366 2020-01-01  Jan Hubicka  <hubicka@ucw.cz>
5368         * ipa.c (walk_polymorphic_call_targets): Fix updating of overall
5369         summary.
5371 2020-01-01  Jakub Jelinek  <jakub@redhat.com>
5373         PR tree-optimization/93098
5374         * match.pd (popcount): For shift amounts, use integer_onep
5375         or wi::to_widest () == cst instead of tree_to_uhwi () == cst
5376         tests.  Make sure that precision is power of two larger than or equal
5377         to 16.  Ensure shift is never negative.  Use HOST_WIDE_INT_UC macro
5378         instead of ULL suffixed constants.  Formatting fixes.
5380 Copyright (C) 2020 Free Software Foundation, Inc.
5382 Copying and distribution of this file, with or without modification,
5383 are permitted in any medium without royalty provided the copyright
5384 notice and this notice are preserved.