RISC-V: Improve caller-save code generation.
[official-gcc.git] / gcc / ChangeLog
blob0ae8c5442a0bfba4f5da79ee8f4a83d7f7a6a1a1
1 2020-02-08  Jim Wilson  <jimw@sifive.com>
3         PR target/93532
4         * config/riscv/riscv.h (HARD_REGNO_CALLER_SAVE_MODE): Define.
6 2020-02-08  Uroš Bizjak  <ubizjak@gmail.com>
7             Jakub Jelinek  <jakub@redhat.com>
9         PR target/65782
10         * config/i386/i386.h (CALL_USED_REGISTERS): Make
11         xmm16-xmm31 call-used even in 64-bit ms-abi.
13 2020-02-07  Dennis Zhang  <dennis.zhang@arm.com>
15         * config/aarch64/aarch64-simd-builtins.def (simd_smmla): New entry.
16         (simd_ummla, simd_usmmla): Likewise.
17         * config/aarch64/aarch64-simd.md (aarch64_simd_<sur>mmlav16qi): New.
18         * config/aarch64/arm_neon.h (vmmlaq_s32, vmmlaq_u32): New.
19         (vusmmlaq_s32): New.
21 2020-02-07  Richard Biener  <rguenther@suse.de>
23         PR middle-end/93519
24         * tree-inline.c (fold_marked_statements): Do a PRE walk,
25         skipping unreachable regions.
26         (optimize_inline_calls): Skip folding stmts when we didn't
27         inline.
29 2020-02-07  H.J. Lu  <hongjiu.lu@intel.com>
31         PR target/85667
32         * config/i386/i386.c (function_arg_ms_64): Add a type argument.
33         Don't return aggregates with only SFmode and DFmode in SSE
34         register.
35         (ix86_function_arg): Pass arg.type to function_arg_ms_64.
37 2020-02-07  Jakub Jelinek  <jakub@redhat.com>
39         PR target/93122
40         * config/rs6000/rs6000-logue.c
41         (rs6000_emit_probe_stack_range_stack_clash): Always use gen_add3_insn,
42         if it fails, move rs into end_addr and retry.  Add
43         REG_FRAME_RELATED_EXPR note whenever it returns more than one insn or
44         the insn pattern doesn't describe well what exactly happens to
45         dwarf2cfi.c.
47         PR target/93594
48         * config/i386/predicates.md (avx_identity_operand): Remove.
49         * config/i386/sse.md (*avx_vec_concat<mode>_1): Remove.
50         (avx_<castmode><avxsizesuffix>_<castmode>,
51         avx512f_<castmode><avxsizesuffix>_256<castmode>): Change patterns to
52         a VEC_CONCAT of the operand and UNSPEC_CAST.
53         (avx512f_<castmode><avxsizesuffix>_<castmode>): Change pattern to
54         a VEC_CONCAT of VEC_CONCAT of the operand and UNSPEC_CAST with
55         UNSPEC_CAST.
57         PR target/93611
58         * config/i386/i386.c (ix86_lea_outperforms): Make sure to clear
59         recog_data.insn if distance_non_agu_define changed it.
61 2020-02-06  Michael Meissner  <meissner@linux.ibm.com>
63         PR target/93569
64         * config/rs6000/rs6000.c (reg_to_non_prefixed): Before ISA 3.0
65         we only had X-FORM (reg+reg) addressing for vectors.  Also before
66         ISA 3.0, we only had X-FORM addressing for scalars in the
67         traditional Altivec registers.
69 2020-02-06  <zhongyunde@huawei.com>
70             Vladimir Makarov  <vmakarov@redhat.com>
72         PR rtl-optimization/93561
73         * lra-assigns.c (spill_for): Check that tested hard regno is not out of
74         hard register range.
76 2020-02-06  Richard Sandiford  <richard.sandiford@arm.com>
78         * config/aarch64/aarch64.md (aarch64_movk<mode>): Add a type
79         attribute.
81 2020-02-06  Segher Boessenkool  <segher@kernel.crashing.org>
83         * config/rs6000/rs6000.c (rs6000_emit_set_long_const): Handle the case
84         where the low and the high 32 bits are equal to each other specially,
85         with an rldimi instruction.
87 2020-02-06  Mihail Ionescu  <mihail.ionescu@arm.com>
89         * config/arm/arm-cpus.in: Set profile M for armv8.1-m.main.
91 2020-02-06  Mihail Ionescu  <mihail.ionescu@arm.com>
93         * config/arm/arm-tables.opt: Regenerate.
95 2020-02-06  Richard Sandiford  <richard.sandiford@arm.com>
97         PR target/87763
98         * config/aarch64/aarch64-protos.h (aarch64_movk_shift): Declare.
99         * config/aarch64/aarch64.c (aarch64_movk_shift): New function.
100         * config/aarch64/aarch64.md (aarch64_movk<mode>): New pattern.
102 2020-02-06  Richard Sandiford  <richard.sandiford@arm.com>
104         PR rtl-optimization/87763
105         * config/aarch64/aarch64.md (*ashiftsi_extvdi_bfiz): New pattern.
107 2020-02-06  Delia Burduv  <delia.burduv@arm.com>
109         * config/aarch64/aarch64-simd-builtins.def
110         (bfmlaq): New built-in function.
111         (bfmlalb): New built-in function.
112         (bfmlalt): New built-in function.
113         (bfmlalb_lane): New built-in function.
114         (bfmlalt_lane): New built-in function.
115         * config/aarch64/aarch64-simd.md
116         (aarch64_bfmmlaqv4sf): New pattern.
117         (aarch64_bfmlal<bt>v4sf): New pattern.
118         (aarch64_bfmlal<bt>_lane<q>v4sf): New pattern.
119         * config/aarch64/arm_neon.h (vbfmmlaq_f32): New intrinsic.
120         (vbfmlalbq_f32): New intrinsic.
121         (vbfmlaltq_f32): New intrinsic.
122         (vbfmlalbq_lane_f32): New intrinsic.
123         (vbfmlaltq_lane_f32): New intrinsic.
124         (vbfmlalbq_laneq_f32): New intrinsic.
125         (vbfmlaltq_laneq_f32): New intrinsic.
126         * config/aarch64/iterators.md (BF_MLA): New int iterator.
127         (bt): New int attribute.
129 2020-02-06  Uroš Bizjak  <ubizjak@gmail.com>
131         * config/i386/i386.md (*pushtf): Emit "#" instead of
132         calling gcc_unreachable in insn output.
133         (*pushxf): Ditto.
134         (*pushdf): Ditto.
135         (*pushsf_rex64): Ditto for alternatives other than 1.
136         (*pushsf): Ditto for alternatives other than 1.
138 2020-02-06  Martin Liska  <mliska@suse.cz>
140         PR gcov-profile/91971
141         PR gcov-profile/93466
142         * coverage.c (coverage_init): Revert mangling of
143         path into filename.  It can lead to huge filename length.
144         Creation of subfolders seem more natural.
146 2020-02-06  Stam Markianos-Wright  <stam.markianos-wright@arm.com>
148         PR target/93300
149         * config/arm/arm.c (arm_block_arith_comp_libfuncs_for_mode): New.
150         (arm_init_libfuncs): Add BFmode support to block spurious BF libfuncs.
151         Use arm_block_arith_comp_libfuncs_for_mode for HFmode.
153 2020-02-06  Jakub Jelinek  <jakub@redhat.com>
155         PR target/93594
156         * config/i386/predicates.md (avx_identity_operand): New predicate.
157         * config/i386/sse.md (*avx_vec_concat<mode>_1): New
158         define_insn_and_split.
160         PR libgomp/93515
161         * omp-low.c (use_pointer_for_field): For nested constructs, also
162         look for map clauses on target construct.
163         (scan_omp_1_stmt) <case GIMPLE_OMP_TARGET>: Bump temporarily
164         taskreg_nesting_level.
166         PR libgomp/93515
167         * gimplify.c (gimplify_scan_omp_clauses) <do_notice>: If adding
168         shared clause, call omp_notice_variable on outer context if any.
170 2020-02-05  Jason Merrill  <jason@redhat.com>
172         PR c++/92003
173         * symtab.c (symtab_node::nonzero_address): A DECL_COMDAT decl has
174         non-zero address even if weak and not yet defined.
176 2020-02-05  Martin Sebor  <msebor@redhat.com>
178         PR tree-optimization/92765
179         * gimple-fold.c (get_range_strlen_tree): Handle MEM_REF and PARM_DECL.
180         * tree-ssa-strlen.c (compute_string_length): Remove.
181         (determine_min_objsize): Remove.
182         (get_len_or_size): Add an argument.  Call get_range_strlen_dynamic.
183         Avoid using type size as the upper bound on string length.
184         (handle_builtin_string_cmp): Add an argument.  Adjust.
185         (strlen_check_and_optimize_call): Pass additional argument to
186         handle_builtin_string_cmp.
188 2020-02-05  Uroš Bizjak  <ubizjak@gmail.com>
190         * config/i386/i386.md (*pushdi2_rex64 peephole2): Remove.
191         (*pushdi2_rex64 peephole2): Unconditionally split after
192         epilogue_completed.
193         (*ashl<mode>3_doubleword): Ditto.
194         (*<shift_insn><mode>3_doubleword): Ditto.
196 2020-02-05  Michael Meissner  <meissner@linux.ibm.com>
198         PR target/93568
199         * config/rs6000/rs6000.c (get_vector_offset): Fix
201 2020-02-05  Andrew Stubbs  <ams@codesourcery.com>
203         * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Use / not space.
205 2020-02-05  David Malcolm  <dmalcolm@redhat.com>
207         * doc/analyzer.texi
208         (Special Functions for Debugging the Analyzer): Update description
209         of __analyzer_dump_exploded_nodes.
211 2020-02-05  Jakub Jelinek  <jakub@redhat.com>
213         PR target/92190
214         * config/i386/i386-features.c (ix86_add_reg_usage_to_vzeroupper): Only
215         include sets and not clobbers in the vzeroupper pattern.
216         * config/i386/sse.md (*avx_vzeroupper): Require in insn condition that
217         the parallel has 17 (64-bit) or 9 (32-bit) elts.
218         (*avx_vzeroupper_1): New define_insn_and_split.
220         PR target/92190
221         * recog.c (pass_split_after_reload::gate): For STACK_REGS targets,
222         don't run when !optimize.
223         (pass_split_before_regstack::gate): For STACK_REGS targets, run even
224         when !optimize.
226 2020-02-05  Richard Biener  <rguenther@suse.de>
228         PR middle-end/90648
229         * genmatch.c (dt_node::gen_kids_1): Emit number of argument
230         checks before matching calls.
232 2020-02-05  Jakub Jelinek  <jakub@redhat.com>
234         * tree-ssa-alias.c (aliasing_matching_component_refs_p): Fix up
235         function comment typo.
237         PR middle-end/93555
238         * omp-simd-clone.c (expand_simd_clones): If simd_clone_mangle or
239         simd_clone_create failed when i == 0, adjust clone->nargs by
240         clone->inbranch.
242 2020-02-05  Martin Liska  <mliska@suse.cz>
244         PR c++/92717
245         * doc/invoke.texi: Document that one should
246         not combine ASLR and -fpch.
248 2020-02-04  Richard Biener  <rguenther@suse.de>
250         PR tree-optimization/93538
251         * match.pd (addr EQ/NE ptr): Amend to handle &ptr->x EQ/NE ptr.
253 2020-02-04  Richard Biener  <rguenther@suse.de>
255         PR tree-optimization/91123
256         * tree-ssa-sccvn.c (vn_walk_cb_data::finish): New method.
257         (vn_walk_cb_data::last_vuse): New member.
258         (vn_walk_cb_data::saved_operands): Likewsie.
259         (vn_walk_cb_data::~vn_walk_cb_data): Release saved_operands.
260         (vn_walk_cb_data::push_partial_def): Use finish.
261         (vn_reference_lookup_2): Update last_vuse and use finish if
262         we've saved operands.
263         (vn_reference_lookup_3): Use finish and update calls to
264         push_partial_defs everywhere.  When translating through
265         memcpy or aggregate copies save off operands and alias-set.
266         (eliminate_dom_walker::eliminate_stmt): Restore VN_WALKREWRITE
267         operation for redundant store removal.
269 2020-02-04  Richard Biener  <rguenther@suse.de>
271         PR tree-optimization/92819
272         * tree-ssa-forwprop.c (simplify_vector_constructor): Avoid
273         generating more stmts than before.
275 2020-02-04  Martin Liska  <mliska@suse.cz>
277         * config/arm/arm.c (arm_gen_far_branch): Move the function
278         outside of selftests.
280 2020-02-03  Michael Meissner  <meissner@linux.ibm.com>
282         * config/rs6000/rs6000.c (adjust_vec_address_pcrel): New helper
283         function to adjust PC-relative vector addresses.
284         (rs6000_adjust_vec_address): Call adjust_vec_address_pcrel to
285         handle vectors with PC-relative addresses.
287 2020-02-03  Michael Meissner  <meissner@linux.ibm.com>
289         * config/rs6000/rs6000.c (reg_to_non_prefixed): Add forward
290         reference.
291         (hard_reg_and_mode_to_addr_mask): Delete.
292         (rs6000_adjust_vec_address): If the original vector address
293         was REG+REG or REG+OFFSET and the element is not zero, do the add
294         of the elements in the original address before adding the offset
295         for the vector element.  Use address_to_insn_form to validate the
296         address using the register being loaded, rather than guessing
297         whether the address is a DS-FORM or DQ-FORM address.
299 2020-02-03  Michael Meissner  <meissner@linux.ibm.com>
301         * config/rs6000/rs6000.c (get_vector_offset): New helper function
302         to calculate the offset in memory from the start of a vector of a
303         particular element.  Add code to keep the element number in
304         bounds if the element number is variable.
305         (rs6000_adjust_vec_address): Move calculation of offset of the
306         vector element to get_vector_offset.
307         (rs6000_split_vec_extract_var): Do not do the initial AND of
308         element here, move the code to get_vector_offset.
310 2020-02-03  Michael Meissner  <meissner@linux.ibm.com>
312         * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add some
313         gcc_asserts.
315 2020-02-03  Segher Boessenkool  <segher@kernel.crashing.org>
317         * config/rs6000/constraints.md: Improve documentation.
319 2020-02-03  Richard Earnshaw  <rearnsha@arm.com>
321         PR target/93548
322         * config/arm/t-arm: ($(srcdir)/config/arm/arm-tune.md)
323         ($(srcdir)/config/arm/arm-tables.opt): Use move-if-change.
325 2020-02-03  Andrew Stubbs  <ams@codesourcery.com>
327         * config.gcc: Remove "carrizo" support.
328         * config/gcn/gcn-opts.h (processor_type): Likewise.
329         * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Likewise.
330         * config/gcn/gcn.opt (gpu_type): Likewise.
331         * config/gcn/t-omp-device: Likewise.
333 2020-02-03  Stam Markianos-Wright  <stam.markianos-wright@arm.com>
335         PR target/91816
336         * config/arm/arm-protos.h: New function arm_gen_far_branch prototype.
337         * config/arm/arm.c (arm_gen_far_branch): New function
338         arm_gen_far_branch.
339         * config/arm/arm.md: Update b<cond> for Thumb2 range checks.
341 2020-02-03  Julian Brown  <julian@codesourcery.com>
342             Tobias Burnus  <tobias@codesourcery.com>
344         * doc/invoke.texi: Update mention of OpenACC version to 2.6.
346 2020-02-03  Jakub Jelinek  <jakub@redhat.com>
348         PR target/93533
349         * config/s390/s390.md (popcounthi2_z196): Fix up expander to emit
350         valid RTL to sum up the lowest and second lowest bytes of the popcnt
351         result.
353 2020-02-02  Vladimir Makarov  <vmakarov@redhat.com>
355         PR rtl-optimization/91333
356         * ira-color.c (struct allocno_color_data): Add member
357         hard_reg_prefs.
358         (init_allocno_threads): Set the member up.
359         (bucket_allocno_compare_func): Add compare hard reg
360         prefs.
362 2020-01-31  Sandra Loosemore  <sandra@codesourcery.com>
364         nios2: Support for GOT-relative DW_EH_PE_datarel encoding.
366         * configure.ac [nios2-*-*]: Check HAVE_AS_NIOS2_GOTOFF_RELOCATION.
367         * config.in: Regenerated.
368         * configure: Regenerated.
369         * config/nios2/nios2.h (ASM_PREFERRED_EH_DATA_FORMAT): Fix handling
370         for PIC when HAVE_AS_NIOS2_GOTOFF_RELOCATION.
371         (ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX): New.
373 2020-02-01  Andrew Burgess  <andrew.burgess@embecosm.com>
375         * configure: Regenerate.
377 2020-01-31  Vladimir Makarov  <vmakarov@redhat.com>
379         PR rtl-optimization/91333
380         * ira-color.c (bucket_allocno_compare_func): Move conflict hard
381         reg preferences comparison up.
383 2020-01-31  Richard Sandiford  <richard.sandiford@arm.com>
385         * config/aarch64/aarch64.h (TARGET_SVE_BF16): New macro.
386         * config/aarch64/aarch64-sve-builtins-sve2.h (svcvtnt): Move to
387         aarch64-sve-builtins-base.h.
388         * config/aarch64/aarch64-sve-builtins-sve2.cc (svcvtnt): Move to
389         aarch64-sve-builtins-base.cc.
390         * config/aarch64/aarch64-sve-builtins-base.h (svbfdot, svbfdot_lane)
391         (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
392         (svcvtnt): Declare.
393         * config/aarch64/aarch64-sve-builtins-base.cc (svbfdot, svbfdot_lane)
394         (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
395         (svcvtnt): New functions.
396         * config/aarch64/aarch64-sve-builtins-base.def (svbfdot, svbfdot_lane)
397         (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
398         (svcvtnt): New functions.
399         (svcvt): Add a form that converts f32 to bf16.
400         * config/aarch64/aarch64-sve-builtins-shapes.h (ternary_bfloat)
401         (ternary_bfloat_lane, ternary_bfloat_lanex2, ternary_bfloat_opt_n):
402         Declare.
403         * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_element_type):
404         Treat B as bfloat16_t.
405         (ternary_bfloat_lane_base): New class.
406         (ternary_bfloat_def): Likewise.
407         (ternary_bfloat): New shape.
408         (ternary_bfloat_lane_def): New class.
409         (ternary_bfloat_lane): New shape.
410         (ternary_bfloat_lanex2_def): New class.
411         (ternary_bfloat_lanex2): New shape.
412         (ternary_bfloat_opt_n_def): New class.
413         (ternary_bfloat_opt_n): New shape.
414         * config/aarch64/aarch64-sve-builtins.cc (TYPES_cvt_bfloat): New macro.
415         * config/aarch64/aarch64-sve.md (@aarch64_sve_<sve_fp_op>vnx4sf)
416         (@aarch64_sve_<sve_fp_op>_lanevnx4sf): New patterns.
417         (@aarch64_sve_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>)
418         (@cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
419         (*cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
420         (@aarch64_sve_cvtnt<VNx8BF_ONLY:mode>): Likewise.
421         * config/aarch64/aarch64-sve2.md (@aarch64_sve2_cvtnt<mode>): Key
422         the pattern off the narrow mode instead of the wider one.
423         * config/aarch64/iterators.md (VNx8BF_ONLY): New mode iterator.
424         (UNSPEC_BFMLALB, UNSPEC_BFMLALT, UNSPEC_BFMMLA): New unspecs.
425         (sve_fp_op): Handle them.
426         (SVE_BFLOAT_TERNARY_LONG): New int itertor.
427         (SVE_BFLOAT_TERNARY_LONG_LANE): Likewise.
429 2020-01-31  Richard Sandiford  <richard.sandiford@arm.com>
431         * config/aarch64/arm_sve.h: Include arm_bf16.h.
432         * config/aarch64/aarch64-modes.def (BF): Move definition before
433         VECTOR_MODES.  Remove separate VECTOR_MODES for V4BF and V8BF.
434         (SVE_MODES): Handle BF modes.
435         * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
436         BF modes.
437         (aarch64_full_sve_mode): Likewise.
438         * config/aarch64/iterators.md (SVE_STRUCT): Add VNx16BF, VNx24BF
439         and VNx32BF.
440         (SVE_FULL, SVE_FULL_HSD, SVE_ALL): Add VNx8BF.
441         (Vetype, Vesize, Vctype, VEL, Vel, VEL_INT, V128, v128, vwcore)
442         (V_INT_EQUIV, v_int_equiv, V_FP_EQUIV, v_fp_equiv, vector_count)
443         (insn_length, VSINGLE, vsingle, VPRED, vpred, VDOUBLE): Handle the
444         new SVE BF modes.
445         * config/aarch64/aarch64-sve-builtins.h (TYPE_bfloat): New
446         type_class_index.
447         * config/aarch64/aarch64-sve-builtins.cc (TYPES_all_arith): New macro.
448         (TYPES_all_data): Add bf16.
449         (TYPES_reinterpret1, TYPES_reinterpret): Likewise.
450         (register_tuple_type): Increase buffer size.
451         * config/aarch64/aarch64-sve-builtins.def (svbfloat16_t): New type.
452         (bf16): New type suffix.
453         * config/aarch64/aarch64-sve-builtins-base.def (svabd, svadd, svaddv)
454         (svcmpeq, svcmpge, svcmpgt, svcmple, svcmplt, svcmpne, svmad, svmax)
455         (svmaxv, svmin, svminv, svmla, svmls, svmsb, svmul, svsub, svsubr):
456         Change type from all_data to all_arith.
457         * config/aarch64/aarch64-sve-builtins-sve2.def (svaddp, svmaxp)
458         (svminp): Likewise.
460 2020-01-31  Dennis Zhang  <dennis.zhang@arm.com>
461             Matthew Malcomson  <matthew.malcomson@arm.com>
462             Richard Sandiford  <richard.sandiford@arm.com>
464         * doc/invoke.texi (f32mm): Document new AArch64 -march= extension.
465         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Define
466         __ARM_FEATURE_SVE_MATMUL_INT8, __ARM_FEATURE_SVE_MATMUL_FP32 and
467         __ARM_FEATURE_SVE_MATMUL_FP64 as appropriate.  Don't define
468         __ARM_FEATURE_MATMUL_FP64.
469         * config/aarch64/aarch64-option-extensions.def (fp, simd, fp16)
470         (sve): Add AARCH64_FL_F32MM to the list of extensions that should
471         be disabled at the same time.
472         (f32mm): New extension.
473         * config/aarch64/aarch64.h (AARCH64_FL_F32MM): New macro.
474         (AARCH64_FL_F64MM): Bump to the next bit up.
475         (AARCH64_ISA_F32MM, TARGET_SVE_I8MM, TARGET_F32MM, TARGET_SVE_F32MM)
476         (TARGET_SVE_F64MM): New macros.
477         * config/aarch64/iterators.md (SVE_MATMULF): New mode iterator.
478         (UNSPEC_FMMLA, UNSPEC_SMATMUL, UNSPEC_UMATMUL, UNSPEC_USMATMUL)
479         (UNSPEC_TRN1Q, UNSPEC_TRN2Q, UNSPEC_UZP1Q, UNSPEC_UZP2Q, UNSPEC_ZIP1Q)
480         (UNSPEC_ZIP2Q): New unspeccs.
481         (DOTPROD_US_ONLY, PERMUTEQ, MATMUL, FMMLA): New int iterators.
482         (optab, sur, perm_insn): Handle the new unspecs.
483         (sve_fp_op): Handle UNSPEC_FMMLA.  Resort.
484         * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use
485         TARGET_SVE_F64MM instead of separate tests.
486         (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod<vsi2qi>): New pattern.
487         (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod_lane<vsi2qi>): Likewise.
488         (@aarch64_sve_add_<MATMUL:optab><vsi2qi>): Likewise.
489         (@aarch64_sve_<FMMLA:sve_fp_op><mode>): Likewise.
490         (@aarch64_sve_<PERMUTEQ:optab><mode>): Likewise.
491         * config/aarch64/aarch64-sve-builtins.cc (TYPES_s_float): New macro.
492         (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): Use it.
493         (TYPES_s_signed): New macro.
494         (TYPES_s_integer): Use it.
495         (TYPES_d_float): New macro.
496         (TYPES_d_data): Use it.
497         * config/aarch64/aarch64-sve-builtins-shapes.h (mmla): Declare.
498         (ternary_intq_uintq_lane, ternary_intq_uintq_opt_n, ternary_uintq_intq)
499         (ternary_uintq_intq_lane, ternary_uintq_intq_opt_n): Likewise.
500         * config/aarch64/aarch64-sve-builtins-shapes.cc (mmla_def): New class.
501         (svmmla): New shape.
502         (ternary_resize2_opt_n_base): Add TYPE_CLASS2 and TYPE_CLASS3
503         template parameters.
504         (ternary_resize2_lane_base): Likewise.
505         (ternary_resize2_base): New class.
506         (ternary_qq_lane_base): Likewise.
507         (ternary_intq_uintq_lane_def): Likewise.
508         (ternary_intq_uintq_lane): New shape.
509         (ternary_intq_uintq_opt_n_def): New class
510         (ternary_intq_uintq_opt_n): New shape.
511         (ternary_qq_lane_def): Inherit from ternary_qq_lane_base.
512         (ternary_uintq_intq_def): New class.
513         (ternary_uintq_intq): New shape.
514         (ternary_uintq_intq_lane_def): New class.
515         (ternary_uintq_intq_lane): New shape.
516         (ternary_uintq_intq_opt_n_def): New class.
517         (ternary_uintq_intq_opt_n): New shape.
518         * config/aarch64/aarch64-sve-builtins-base.h (svmmla, svsudot)
519         (svsudot_lane, svtrn1q, svtrn2q, svusdot, svusdot_lane, svusmmla)
520         (svuzp1q, svuzp2q, svzip1q, svzip2q): Declare.
521         * config/aarch64/aarch64-sve-builtins-base.cc (svdot_lane_impl):
522         Generalize to...
523         (svdotprod_lane_impl): ...this new class.
524         (svmmla_impl, svusdot_impl): New classes.
525         (svdot_lane): Update to use svdotprod_lane_impl.
526         (svmmla, svsudot, svsudot_lane, svtrn1q, svtrn2q, svusdot)
527         (svusdot_lane, svusmmla, svuzp1q, svuzp2q, svzip1q, svzip2q): New
528         functions.
529         * config/aarch64/aarch64-sve-builtins-base.def (svmmla): New base
530         function, with no types defined.
531         (svmmla, svusmmla, svsudot, svsudot_lane, svusdot, svusdot_lane): New
532         AARCH64_FL_I8MM functions.
533         (svmmla): New AARCH64_FL_F32MM function.
534         (svld1ro): Depend only on AARCH64_FL_F64MM, not on AARCH64_FL_V8_6.
535         (svmmla, svtrn1q, svtrn2q, svuz1q, svuz2q, svzip1q, svzip2q): New
536         AARCH64_FL_F64MM function.
537         (REQUIRED_EXTENSIONS):
539 2020-01-31  Andrew Stubbs  <ams@codesourcery.com>
541         * config/gcn/gcn-valu.md (addv64di3_exec): Allow one '0' in each
542         alternative only.
544 2020-01-31  Uroš Bizjak  <ubizjak@gmail.com>
546         * config/i386/i386.md (*movoi_internal_avx): Do not check for
547         TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.  Remove MODE_V8SF handling.
548         (*movti_internal): Do not check for
549         TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
550         (*movtf_internal): Move check for TARGET_SSE2 and size optimization
551         just after check for TARGET_AVX.
552         (*movdf_internal): Ditto.
553         * config/i386/mmx.md (*mov<mode>_internal): Do not check for
554         TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
555         * config/i386/sse.md (mov<mode>_internal): Only check
556         TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL with V2DFmode.  Move check
557         for TARGET_SSE2 and size optimization just after check for TARGET_AVX.
558         (<sse>_andnot<mode>3<mask_name>): Move check for
559         TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL after check for TARGET_AVX.
560         (<code><mode>3<mask_name>): Ditto.
561         (*andnot<mode>3): Ditto.
562         (*andnottf3): Ditto.
563         (*<code><mode>3): Ditto.
564         (*<code>tf3): Ditto.
565         (*andnot<VI:mode>3): Remove
566         TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL handling.
567         (<mask_codefor><code><VI48_AVX_AVX512F:mode>3<mask_name>): Ditto.
568         (*<code><VI12_AVX_AVX512F:mode>3): Ditto.
569         (sse4_1_blendv<ssemodesuffix>): Ditto.
570         * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL):
571         Explain that tune applies to 128bit instructions only.
573 2020-01-31  Kwok Cheung Yeung  <kcy@codesourcery.com>
575         * config/gcn/mkoffload.c (process_asm): Add sgpr_count and vgpr_count
576         to definition of hsa_kernel_description.  Parse assembly to find SGPR
577         and VGPR count of kernel and store in hsa_kernel_description.
579 2020-01-31  Tamar Christina  <tamar.christina@arm.com>
581         PR rtl-optimization/91838
582         * simplify-rtx.c (simplify_binary_operation_1): Update LSHIFTRT case
583         to truncate if allowed or reject combination.
585 2020-01-31  Andrew Stubbs  <ams@codesourcery.com>
587         * tree-ssa-loop-ivopts.c (get_iv): Use sizetype for zero-step.
588         (find_inv_vars_cb): Likewise.
590 2020-01-31  David Malcolm  <dmalcolm@redhat.com>
592         * calls.c (special_function_p): Split out the check for DECL_NAME
593         being non-NULL and fndecl being extern at file scope into a
594         new maybe_special_function_p and call it.  Drop check for fndecl
595         being non-NULL that was after a usage of DECL_NAME (fndecl).
596         * tree.h (maybe_special_function_p): New inline function.
598 2020-01-30  Andrew Stubbs  <ams@codesourcery.com>
600         * config/gcn/gcn-valu.md (gather<mode>_exec): Move contents ...
601         (mask_gather_load<mode>): ... here, and zero-initialize the
602         destination.
603         (maskload<mode>di): Zero-initialize the destination.
604         * config/gcn/gcn.c:
606 2020-01-30  David Malcolm  <dmalcolm@redhat.com>
608         PR analyzer/93356
609         * doc/analyzer.texi (Limitations): Note that constraints on
610         floating-point values are currently ignored.
612 2020-01-30  Jakub Jelinek  <jakub@redhat.com>
614         PR lto/93384
615         * symtab.c (symtab_node::noninterposable_alias): If localalias
616         already exists, but is not usable, append numbers after it until
617         a unique name is found.  Formatting fix.
619         PR middle-end/93505
620         * combine.c (simplify_comparison) <case ROTATE>: Punt on out of range
621         rotate counts.
623 2020-01-30  Andrew Stubbs  <ams@codesourcery.com>
625         * config/gcn/gcn.c (print_operand): Handle LTGT.
626         * config/gcn/predicates.md (gcn_fp_compare_operator): Allow ltgt.
628 2020-01-30  Richard Biener  <rguenther@suse.de>
630         * tree-pretty-print.c (dump_generic_node): Wrap VECTOR_CST
631         and CONSTRUCTOR in _Literal (type) with TDF_GIMPLE.
633 2020-01-30  John David Anglin  <danglin@gcc.gnu.org>
635         * config/pa/pa.c (pa_elf_select_rtx_section): Place function pointers
636         without a DECL in .data.rel.ro.local.
638 2020-01-30  Jakub Jelinek  <jakub@redhat.com>
640         PR target/93494
641         * config/arm/arm.md (uaddvdi4): Actually emit what gen_uaddvsi4
642         returned.
644         PR target/91824
645         * config/i386/sse.md
646         (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext): Renamed to ...
647         (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext): ... this.  Use
648         any_extend code iterator instead of always zero_extend.
649         (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_lt): Renamed to ...
650         (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_lt): ... this.
651         Use any_extend code iterator instead of always zero_extend.
652         (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_shift): Renamed to ...
653         (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_shift): ... this.
654         Use any_extend code iterator instead of always zero_extend.
655         (*sse2_pmovmskb_ext): New define_insn.
656         (*sse2_pmovmskb_ext_lt): New define_insn_and_split.
658         PR target/91824
659         * config/i386/i386.md (*popcountsi2_zext): New define_insn_and_split.
660         (*popcountsi2_zext_falsedep): New define_insn.
662 2020-01-30  Dragan Mladjenovic  <dmladjenovic@wavecomp.com>
664         * config.in: Regenerated.
665         * configure: Regenerated.
667 2020-01-29  Tobias Burnus  <tobias@codesourcery.com>
669         PR bootstrap/93409
670         * config/gcn/gcn-hsa.h (ASM_SPEC): Add -mattr=-code-object-v3 as
671         LLVM's assembler changed the default in version 9.
673 2020-01-24  Jeff Law  <law@redhat.com>
675         PR tree-optimization/89689
676         * builtins.def (BUILT_IN_OBJECT_SIZE): Make it const rather than pure.
678 2020-01-29  Richard Sandiford  <richard.sandiford@arm.com>
680         Revert:
682         2020-01-28  Richard Sandiford  <richard.sandiford@arm.com>
684         PR rtl-optimization/87763
685         * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
686         simplification to handle subregs as well as bare regs.
687         * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
689 2020-01-29  Joel Hutton  <Joel.Hutton@arm.com>
691         PR target/93221
692         * ira.c (ira): Revert use of simplified LRA algorithm.
694 2020-01-29  Martin Jambor  <mjambor@suse.cz>
696         PR tree-optimization/92706
697         * tree-sra.c (struct access): Fields first_link, last_link,
698         next_queued and grp_queued renamed to first_rhs_link, last_rhs_link,
699         next_rhs_queued and grp_rhs_queued respectively, new fields
700         first_lhs_link, last_lhs_link, next_lhs_queued and grp_lhs_queued.
701         (struct assign_link): Field next renamed to next_rhs, new field
702         next_lhs.  Updated comment.
703         (work_queue_head): Renamed to rhs_work_queue_head.
704         (lhs_work_queue_head): New variable.
705         (add_link_to_lhs): New function.
706         (relink_to_new_repr): Also relink LHS lists.
707         (add_access_to_work_queue): Renamed to add_access_to_rhs_work_queue.
708         (add_access_to_lhs_work_queue): New function.
709         (pop_access_from_work_queue): Renamed to
710         pop_access_from_rhs_work_queue.
711         (pop_access_from_lhs_work_queue): New function.
712         (build_accesses_from_assign): Also add links to LHS lists and to LHS
713         work_queue.
714         (child_would_conflict_in_lacc): Renamed to
715         child_would_conflict_in_acc.  Adjusted parameter names.
716         (create_artificial_child_access): New parameter set_grp_read, use it.
717         (subtree_mark_written_and_enqueue): Renamed to
718         subtree_mark_written_and_rhs_enqueue.
719         (propagate_subaccesses_across_link): Renamed to
720         propagate_subaccesses_from_rhs.
721         (propagate_subaccesses_from_lhs): New function.
722         (propagate_all_subaccesses): Also propagate subaccesses from LHSs to
723         RHSs.
725 2020-01-29  Martin Jambor  <mjambor@suse.cz>
727         PR tree-optimization/92706
728         * tree-sra.c (struct access): Adjust comment of
729         grp_total_scalarization.
730         (find_access_in_subtree): Look for single children spanning an entire
731         access.
732         (scalarizable_type_p): Allow register accesses, adjust callers.
733         (completely_scalarize): Remove function.
734         (scalarize_elem): Likewise.
735         (create_total_scalarization_access): Likewise.
736         (sort_and_splice_var_accesses): Do not track total scalarization
737         flags.
738         (analyze_access_subtree): New parameter totally, adjust to new meaning
739         of grp_total_scalarization.
740         (analyze_access_trees): Pass new parameter to analyze_access_subtree.
741         (can_totally_scalarize_forest_p): New function.
742         (create_total_scalarization_access): Likewise.
743         (create_total_access_and_reshape): Likewise.
744         (total_should_skip_creating_access): Likewise.
745         (totally_scalarize_subtree): Likewise.
746         (analyze_all_variable_accesses): Perform total scalarization after
747         subaccess propagation using the new functions above.
748         (initialize_constant_pool_replacements): Output initializers by
749         traversing the access tree.
751 2020-01-29  Martin Jambor  <mjambor@suse.cz>
753         * tree-sra.c (verify_sra_access_forest): New function.
754         (verify_all_sra_access_forests): Likewise.
755         (create_artificial_child_access): Set parent.
756         (analyze_all_variable_accesses): Call the verifier.
758 2020-01-28  Jan Hubicka  <hubicka@ucw.cz>
760         * cgraph.c (cgraph_edge::resolve_speculation): Only lookup direct edge
761         if called on indirect edge.
762         (cgraph_edge::redirect_call_stmt_to_callee): Lookup indirect edge of
763         speculative call if needed.
765 2020-01-29  Richard Biener  <rguenther@suse.de>
767         PR tree-optimization/93428
768         * tree-vect-slp.c (vect_build_slp_tree_2): Compute the load
769         permutation when the load node is created.
770         (vect_analyze_slp_instance): Re-use it here.
772 2020-01-28  Jan Hubicka  <hubicka@ucw.cz>
774         * ipa-prop.c (update_indirect_edges_after_inlining): Fix warning.
776 2020-01-28  Vladimir Makarov  <vmakarov@redhat.com>
778         PR rtl-optimization/93272
779         * ira-lives.c (process_out_of_region_eh_regs): New function.
780         (process_bb_node_lives): Call it.
782 2020-01-28  Jan Hubicka  <hubicka@ucw.cz>
784         * coverage.c (read_counts_file): Make error message lowercase.
786 2020-01-28  Jan Hubicka  <hubicka@ucw.cz>
788         * profile-count.c (profile_quality_display_names): Fix ordering.
790 2020-01-28  Jan Hubicka  <hubicka@ucw.cz>
792         PR lto/93318    
793         * cgraph.c (cgraph_add_edge_to_call_site_hash): Update call site
794         hash only when edge is first within the sequence.
795         (cgraph_edge::set_call_stmt): Update handling of speculative calls.
796         (symbol_table::create_edge): Do not set target_prob.
797         (cgraph_edge::remove_caller): Watch for speculative calls when updating
798         the call site hash.
799         (cgraph_edge::make_speculative): Drop target_prob parameter.
800         (cgraph_edge::speculative_call_info): Remove.
801         (cgraph_edge::first_speculative_call_target): New member function.
802         (update_call_stmt_hash_for_removing_direct_edge): New function.
803         (cgraph_edge::resolve_speculation): Rewrite to new API.
804         (cgraph_edge::speculative_call_for_target): New member function.
805         (cgraph_edge::make_direct): Rewrite to new API; fix handling of
806         multiple speculation targets.
807         (cgraph_edge::redirect_call_stmt_to_callee): Likewise; fix updating
808         of profile.
809         (verify_speculative_call): Verify that targets form an interval.
810         * cgraph.h (cgraph_edge::speculative_call_info): Remove.
811         (cgraph_edge::first_speculative_call_target): New member function.
812         (cgraph_edge::next_speculative_call_target): New member function.
813         (cgraph_edge::speculative_call_target_ref): New member function.
814         (cgraph_edge;:speculative_call_indirect_edge): New member funtion.
815         (cgraph_edge): Remove target_prob.
816         * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
817         Fix handling of speculative calls.
818         * ipa-devirt.c (ipa_devirt): Fix handling of speculative cals.
819         * ipa-fnsummary.c (analyze_function_body): Likewise.
820         * ipa-inline.c (speculation_useful_p): Use new speculative call API.
821         * ipa-profile.c (dump_histogram): Fix formating.
822         (ipa_profile_generate_summary): Watch for overflows.
823         (ipa_profile): Do not require probablity to be 1/2; update to new API.
824         * ipa-prop.c (ipa_make_edge_direct_to_target): Update to new API.
825         (update_indirect_edges_after_inlining): Update to new API.
826         * ipa-utils.c (ipa_merge_profiles): Rewrite merging of speculative call
827         profiles.
828         * profile-count.h: (profile_probability::adjusted): New.
829         * tree-inline.c (copy_bb): Update to new speculative call API; fix
830         updating of profile.
831         * value-prof.c (gimple_ic_transform): Rename to ...
832         (dump_ic_profile): ... this one; update dumping.
833         (stream_in_histogram_value): Fix formating.
834         (gimple_value_profile_transformations): Update.
836 2020-01-28  H.J. Lu  <hongjiu.lu@intel.com>
838         PR target/91461
839         * config/i386/i386.md (*movoi_internal_avx): Remove
840         TARGET_SSE_TYPELESS_STORES check.
841         (*movti_internal): Prefer TARGET_AVX over
842         TARGET_SSE_TYPELESS_STORES.
843         (*movtf_internal): Likewise.
844         * config/i386/sse.md (mov<mode>_internal): Prefer TARGET_AVX over
845         TARGET_SSE_TYPELESS_STORES.  Remove "<MODE_SIZE> == 16" check
846         from TARGET_SSE_TYPELESS_STORES.
848 2020-01-28  David Malcolm  <dmalcolm@redhat.com>
850         * diagnostic-core.h (warning_at): Rename overload to...
851         (warning_meta): ...this.
852         (emit_diagnostic_valist): Delete decl of overload taking
853         diagnostic_metadata.
854         * diagnostic.c (emit_diagnostic_valist): Likewise for defn.
855         (warning_at): Rename overload taking diagnostic_metadata to...
856         (warning_meta): ...this.
858 2020-01-28  Richard Biener  <rguenther@suse.de>
860         PR tree-optimization/93439
861         * tree-parloops.c (create_loop_fn): Move clique bookkeeping...
862         * tree-cfg.c (move_sese_region_to_fn): ... here.
863         (verify_types_in_gimple_reference): Verify used cliques are
864         tracked.
866 2020-01-28  H.J. Lu  <hongjiu.lu@intel.com>
868         PR target/91399
869         * config/i386/i386-options.c (set_ix86_tune_features): Add an
870         argument of a pointer to struct gcc_options and pass it to
871         parse_mtune_ctrl_str.
872         (ix86_function_specific_restore): Pass opts to
873         set_ix86_tune_features.
874         (ix86_option_override_internal): Likewise.
875         (parse_mtune_ctrl_str): Add an argument of a pointer to struct
876         gcc_options and use it for x_ix86_tune_ctrl_string.
878 2020-01-28  Richard Sandiford  <richard.sandiford@arm.com>
880         PR rtl-optimization/87763
881         * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
882         simplification to handle subregs as well as bare regs.
883         * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
885 2020-01-28  Richard Sandiford  <richard.sandiford@arm.com>
887         * tree-vect-loop.c (vectorizable_reduction): Fail gracefully
888         for reduction chains that (now) include a call.
890 2020-01-28  Richard Sandiford  <richard.sandiford@arm.com>
892         PR tree-optimization/92822
893         * tree-ssa-forwprop.c (simplify_vector_constructor): When filling
894         out the don't-care elements of a vector whose significant elements
895         are duplicates, make the don't-care elements duplicates too.
897 2020-01-28  Richard Sandiford  <richard.sandiford@arm.com>
899         PR tree-optimization/93434
900         * tree-predcom.c (split_data_refs_to_components): Record which
901         components have had aliasing loads removed.  Prevent store-store
902         commoning for all such components.
904 2020-01-28  Jakub Jelinek  <jakub@redhat.com>
906         PR target/93418
907         * config/i386/i386.c (ix86_fold_builtin) <do_shift>: If mask is not
908         -1 or is_vshift is true, use new_vector with number of elts npatterns
909         rather than new_unary_operation.
911         PR tree-optimization/93454
912         * gimple-fold.c (fold_array_ctor_reference): Perform
913         elt_size.to_uhwi () just once, instead of calling it in every
914         iteration.  Punt if that value is above size of the temporary
915         buffer.  Decrease third native_encode_expr argument when
916         bufoff + elt_sz is above size of buf.
918 2020-01-27  Joseph Myers  <joseph@codesourcery.com>
920         * config/mips/mips.c (mips_declare_object_name)
921         [USE_GNU_UNIQUE_OBJECT]: Support use of gnu_unique_object.
923 2020-01-27  Martin Liska  <mliska@suse.cz>
925         PR gcov-profile/93403
926         * tree-profile.c (gimple_init_gcov_profiler): Generate
927         both __gcov_indirect_call_profiler_v4 and
928         __gcov_indirect_call_profiler_v4_atomic.
930 2020-01-27  Richard Sandiford  <richard.sandiford@arm.com>
932         PR target/92822
933         * config/aarch64/aarch64-simd.md (aarch64_get_half<mode>): New
934         expander.
935         (@aarch64_split_simd_mov<mode>): Use it.
936         (aarch64_simd_mov_from_<mode>low): Add a GPR alternative.
937         Leave the vec_extract patterns to handle 2-element vectors.
938         (aarch64_simd_mov_from_<mode>high): Likewise.
939         (vec_extract<VQMOV_NO2E:mode><Vhalf>): New expander.
940         (vec_extractv2dfv1df): Likewise.
942 2020-01-27  Richard Sandiford  <richard.sandiford@arm.com>
944         * config/aarch64/aarch64.c (aarch64_if_then_else_costs): Match
945         jump conditions for *compare_condjump<GPI:mode>.
947 2020-01-27  David Malcolm  <dmalcolm@redhat.com>
949         PR analyzer/93276
950         * digraph.cc (test_edge::test_edge): Specify template for base
951         class initializer.
953 2020-01-27  Claudiu Zissulescu  <claziss@synopsys.com>
955         * config/arc/arc.c (arc_rtx_costs): Update mul64 cost.
957 2020-01-27  Claudiu Zissulescu  <claziss@synopsys.com>
959         * config/arc/arc-protos.h (gen_mlo): Remove.
960         (gen_mhi): Likewise.
961         * config/arc/arc.c (AUX_MULHI): Define.
962         (arc_must_save_reister): Special handling for r58/59.
963         (arc_compute_frame_size): Consider mlo/mhi registers.
964         (arc_save_callee_saves): Emit fp/sp move only when emit_move
965         paramter is true.
966         (arc_conditional_register_usage): Remove TARGET_BIG_ENDIAN from
967         mlo/mhi name selection.
968         (arc_restore_callee_saves): Don't early restore blink when ISR.
969         (arc_expand_prologue): Add mlo/mhi saving.
970         (arc_expand_epilogue): Add mlo/mhi restoring.
971         (gen_mlo): Remove.
972         (gen_mhi): Remove.
973         * config/arc/arc.h (DBX_REGISTER_NUMBER): Correct register
974         numbering when MUL64 option is used.
975         (DWARF2_FRAME_REG_OUT): Define.
976         * config/arc/arc.md (arc600_stall): New pattern.
977         (VUNSPEC_ARC_ARC600_STALL): Define.
978         (mulsi64): Use correct mlo/mhi registers.
979         (mulsi_600): Clean it up.
980         * config/arc/predicates.md (mlo_operand): Remove any dependency on
981         TARGET_BIG_ENDIAN.
982         (mhi_operand): Likewise.
984 2020-01-27  Claudiu Zissulescu  <claziss@synopsys.com>
985             Petro Karashchenko  <petro.karashchenko@ring.com>
987         * config/arc/arc.c (arc_is_uncached_mem_p): Check struct
988         attributes if needed.
989         (prepare_move_operands): Generate special unspec instruction for
990         direct access.
991         (arc_isuncached_mem_p): Propagate uncached attribute to each
992         structure member.
993         * config/arc/arc.md (VUNSPEC_ARC_LDDI): Define.
994         (VUNSPEC_ARC_STDI): Likewise.
995         (ALLI): New mode iterator.
996         (mALLI): New mode attribute.
997         (lddi): New instruction pattern.
998         (stdi): Likewise.
999         (stdidi_split): Split instruction for architectures which are not
1000         supporting ll64 option.
1001         (lddidi_split): Likewise.
1003 2020-01-27  Richard Sandiford  <richard.sandiford@arm.com>
1005         PR rtl-optimization/92989
1006         * lra-lives.c (process_bb_lives): Update the live-in set before
1007         processing additional clobbers.
1009 2020-01-27  Richard Sandiford  <richard.sandiford@arm.com>
1011         PR rtl-optimization/93170
1012         * cselib.c (cselib_invalidate_regno_val): New function, split out
1013         from...
1014         (cselib_invalidate_regno): ...here.
1015         (cselib_invalidated_by_call_p): New function.
1016         (cselib_process_insn): Iterate over all the hard-register entries in
1017         REG_VALUES and invalidate any that cross call-clobbered registers.
1019 2020-01-27  Richard Sandiford  <richard.sandiford@arm.com>
1021         * dojump.c (split_comparison): Use HONOR_NANS rather than
1022         HONOR_SNANS when splitting LTGT.
1024 2020-01-27  Martin Liska  <mliska@suse.cz>
1026         PR driver/91220
1027         * opts.c (print_filtered_help): Exclude language-specific
1028         options from --help=common unless enabled in all FEs.
1030 2020-01-27  Martin Liska  <mliska@suse.cz>
1032         * opts.c (print_help): Exclude params from
1033         all except --help=param.
1035 2020-01-27  Martin Liska  <mliska@suse.cz>
1037         PR target/93274
1038         * config/i386/i386-features.c (make_resolver_func):
1039         Align the code with ppc64 target implementation.
1040         Do not generate a unique name for resolver function.
1042 2020-01-27  Richard Biener  <rguenther@suse.de>
1044         PR tree-optimization/93397
1045         * tree-vect-slp.c (vect_analyze_slp_instance): Delay
1046         converted reduction chain SLP graph adjustment.
1048 2020-01-26  Marek Polacek  <polacek@redhat.com>
1050         PR sanitizer/93436
1051         * sanopt.c (sanitize_rewrite_addressable_params): Avoid crash on
1052         null DECL_NAME.
1054 2020-01-26  Jason Merrill  <jason@redhat.com>
1056         PR c++/92601
1057         * tree.c (verify_type_variant): Only verify TYPE_NEEDS_CONSTRUCTING
1058         of complete types.
1060 2020-01-26  Darius Galis  <darius.galis@cyberthorstudios.com>
1062         * config/rx/rx.md (setmemsi): Added rx_allow_string_insns constraint
1063         (rx_setmem): Likewise.
1065 2020-01-26  Jakub Jelinek  <jakub@redhat.com>
1067         PR target/93412
1068         * config/i386/i386.md (*addv<dwi>4_doubleword, *subv<dwi>4_doubleword):
1069         Use nonimmediate_operand instead of x86_64_hilo_general_operand and
1070         drop <di> from constraint of last operand.
1072         PR target/93430
1073         * config/i386/sse.md (*avx_vperm_broadcast_<mode>): Disallow for
1074         TARGET_AVX2 and V4DFmode not in the split condition, but in the
1075         pattern condition, though allow { 0, 0, 0, 0 } broadcast always.
1077 2020-01-25  Feng Xue  <fxue@os.amperecomputing.com>
1079         PR ipa/93166
1080         * ipa-cp.c (get_info_about_necessary_edges): Remove value
1081         check assertion.
1083 2020-01-24  Jeff Law  <law@redhat.com>
1085         PR tree-optimization/92788
1086         * tree-ssa-threadedge.c (thread_across_edge): Check EDGE_COMPLEX
1087         not EDGE_ABNORMAL.
1089 2020-01-24  Jakub Jelinek  <jakub@redhat.com>
1091         PR target/93395
1092         * config/i386/sse.md (*avx_vperm_broadcast_v4sf,
1093         *avx_vperm_broadcast_<mode>,
1094         <sse2_avx_avx512f>_vpermil<mode><mask_name>,
1095         *<sse2_avx_avx512f>_vpermilp<mode><mask_name>):
1096         Move before avx2_perm<mode>/avx512f_perm<mode>.
1098         PR target/93376
1099         * simplify-rtx.c (simplify_const_unary_operation,
1100         simplify_const_binary_operation): Punt for mode precision above
1101         MAX_BITSIZE_MODE_ANY_INT.
1103 2020-01-24  Andrew Pinski  <apinski@marvell.com>
1105         * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): Change
1106         alu.shift_reg to 0.
1108 2020-01-24  Jeff Law  <law@redhat.com>
1110         PR target/13721
1111         * config/h8300/h8300.c (h8300_print_operand): Only call byte_reg
1112         for REGs.  Call output_operand_lossage to get more reasonable
1113         diagnostics.
1115 2020-01-24  Andrew Stubbs  <ams@codesourcery.com>
1117         * config/gcn/gcn-valu.md (vec_cmp<mode>di): Use
1118         gcn_fp_compare_operator.
1119         (vec_cmpu<mode>di): Use gcn_compare_operator.
1120         (vec_cmp<u>v64qidi): Use gcn_compare_operator.
1121         (vec_cmp<mode>di_exec): Use gcn_fp_compare_operator.
1122         (vec_cmpu<mode>di_exec): Use gcn_compare_operator.
1123         (vec_cmp<u>v64qidi_exec): Use gcn_compare_operator.
1124         (vec_cmp<mode>di_dup): Use gcn_fp_compare_operator.
1125         (vec_cmp<mode>di_dup_exec): Use gcn_fp_compare_operator.
1126         (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): Use
1127         gcn_fp_compare_operator.
1128         (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): Use
1129         gcn_fp_compare_operator.
1130         (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): Use
1131         gcn_fp_compare_operator.
1132         (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): Use
1133         gcn_fp_compare_operator.
1135 2020-01-24  Maciej W. Rozycki  <macro@wdc.com>
1137         * doc/install.texi (Cross-Compiler-Specific Options): Document
1138         `--with-toolexeclibdir' option.
1140 2020-01-24  Hans-Peter Nilsson  <hp@axis.com>
1142         * target.def (flags_regnum): Also mention effect on delay slot filling.
1143         * doc/tm.texi: Regenerate.
1145 2020-01-23  Jeff Law  <law@redhat.com>
1147         PR translation/90162
1148         * config/h8300/h8300.c (h8300_option_override): Fix diagnostic text.
1150 2020-01-23  Mikael Tillenius  <mti-1@tillenius.com>
1152         PR target/92269
1153         * config/h8300/h8300.h (FUNCTION_PROFILER): Fix emission of
1154         profiling label
1156 2020-01-23  Jakub Jelinek  <jakub@redhat.com>
1158         PR rtl-optimization/93402
1159         * postreload.c (reload_combine_recognize_pattern): Don't try to adjust
1160         USE insns.
1162 2020-01-23  Dragan Mladjenovic  <dmladjenovic@wavecomp.com>
1164         * config.in: Regenerated.
1165         * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to 1
1166         for TARGET_LIBC_GNUSTACK.
1167         * configure: Regenerated.
1168         * configure.ac: Define TARGET_LIBC_GNUSTACK if glibc version is
1169         found to be 2.31 or greater.
1171 2020-01-23  Dragan Mladjenovic  <dmladjenovic@wavecomp.com>
1173         * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to
1174         TARGET_SOFT_FLOAT.
1175         * config/mips/mips.c (TARGET_ASM_FILE_END): Define to ...
1176         (mips_asm_file_end): New function. Delegate to
1177         file_end_indicate_exec_stack if NEED_INDICATE_EXEC_STACK is true.
1178         * config/mips/mips.h (NEED_INDICATE_EXEC_STACK): Define to 0.
1180 2020-01-23  Jakub Jelinek  <jakub@redhat.com>
1182         PR target/93376
1183         * config/i386/i386-modes.def (POImode): New mode.
1184         (MAX_BITSIZE_MODE_ANY_INT): Change from 128 to 160.
1185         * config/i386/i386.md (DPWI): New mode attribute.
1186         (addv<mode>4, subv<mode>4): Use <DPWI> instead of <DWI>.
1187         (QWI): Rename to...
1188         (QPWI): ... this.  Use POI instead of OI for TImode.
1189         (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1,
1190         *subv<dwi>4_doubleword, *subv<dwi>4_doubleword_1): Use <QPWI>
1191         instead of <QWI>.
1193 2020-01-23  Richard Sandiford  <richard.sandiford@arm.com>
1195         PR target/93341
1196         * config/aarch64/aarch64.md (UNSPEC_SPECULATION_TRACKER_REV): New
1197         unspec.
1198         (speculation_tracker_rev): New pattern.
1199         * config/aarch64/aarch64-speculation.cc (aarch64_do_track_speculation):
1200         Use speculation_tracker_rev to track the inverse condition.
1202 2020-01-23  Richard Biener  <rguenther@suse.de>
1204         PR tree-optimization/93381
1205         * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Take
1206         alias-set of the def as argument and record the first one.
1207         (vn_walk_cb_data::first_set): New member.
1208         (vn_reference_lookup_3): Pass the alias-set of the current def
1209         to push_partial_def.  Fix alias-set used in the aggregate copy
1210         case.
1211         (vn_reference_lookup): Consistently set *last_vuse_ptr.
1212         * real.c (clear_significand_below): Fix out-of-bound access.
1214 2020-01-23  Jakub Jelinek  <jakub@redhat.com>
1216         PR target/93346
1217         * config/i386/i386.md (*bmi2_bzhi_<mode>3_2, *bmi2_bzhi_<mode>3_3):
1218         New define_insn patterns.
1220 2020-01-23  Richard Sandiford  <richard.sandiford@arm.com>
1222         * doc/sourcebuild.texi (check-function-bodies): Add an
1223         optional target/xfail selector.
1225 2020-01-23  Richard Sandiford  <richard.sandiford@arm.com>
1227         PR rtl-optimization/93124
1228         * auto-inc-dec.c (merge_in_block): Don't add auto inc/decs to
1229         bare USE and CLOBBER insns.
1231 2020-01-22  Andrew Pinski  <apinski@marvell.com>
1233         * config/arc/arc.c (output_short_suffix): Check insn for nullness.
1235 2020-01-22  David Malcolm  <dmalcolm@redhat.com>
1237         PR analyzer/93307
1238         * gdbinit.in (break-on-saved-diagnostic): Update for move of
1239         diagnostic_manager into "ana" namespace.
1240         * selftest-run-tests.c (selftest::run_tests): Update for move of
1241         selftest::run_analyzer_selftests to
1242         ana::selftest::run_analyzer_selftests.
1244 2020-01-22  Richard Sandiford  <richard.sandiford@arm.com>
1246         * cfgexpand.c (union_stack_vars): Update the size.
1248 2020-01-22  Richard Biener  <rguenther@suse.de>
1250         PR tree-optimization/93381
1251         * tree-ssa-structalias.c (find_func_aliases): Assume offsetting
1252         throughout, handle all conversions the same.
1254 2020-01-22  Jakub Jelinek  <jakub@redhat.com>
1256         PR target/93335
1257         * config/aarch64/aarch64.c (aarch64_expand_subvti): Only use
1258         gen_subdi3_compare1_imm if low_in2 satisfies aarch64_plus_immediate
1259         predicate, not whenever it is CONST_INT.  Otherwise, force_reg it.
1260         Call force_reg on high_in2 unconditionally.
1262 2020-01-22  Martin Liska  <mliska@suse.cz>
1264         PR tree-optimization/92924
1265         * profile.c (compute_value_histograms): Divide
1266         all counter values.
1268 2020-01-22  Jakub Jelinek  <jakub@redhat.com>
1270         PR target/91298
1271         * output.h (assemble_name_resolve): Declare.
1272         * varasm.c (assemble_name_resolve): New function.
1273         (assemble_name): Use it.
1274         * config/i386/i386.h (ASM_OUTPUT_SYMBOL_REF): Define.
1276 2020-01-22  Joseph Myers  <joseph@codesourcery.com>
1278         * doc/sourcebuild.texi (Texinfo Manuals, Front End): Refer to
1279         update_web_docs_git instead of update_web_docs_svn.
1281 2020-01-21  Andrew Pinski  <apinski@marvell.com>
1283         PR target/9311
1284         * config/aarch64/aarch64.md (tlsgd_small_<mode>): Have operand 0
1285         as PTR mode. Have operand 1 as being modeless, it can be P mode.
1286         (*tlsgd_small_<mode>): Likewise.
1287         * config/aarch64/aarch64.c (aarch64_load_symref_appropriately)
1288         <case SYMBOL_SMALL_TLSGD>: Call gen_tlsgd_small_* with a ptr_mode
1289         register.  Convert that register back to dest using convert_mode.
1291 2020-01-21  Jim Wilson  <jimw@sifive.com>
1293         * config/riscv/riscv-sr.c (riscv_sr_match_prologue): Use INTVAL
1294         instead of XINT.
1296 2020-01-21  H.J. Lu  <hongjiu.lu@intel.com>
1297             Uros Bizjak    <ubizjak@gmail.com>
1299         PR target/93319
1300         * config/i386/i386.c (ix86_tls_module_base): Replace Pmode
1301         with ptr_mode.
1302         (legitimize_tls_address): Do GNU2 TLS address computation in
1303         ptr_mode and zero-extend result to Pmode.
1304         *  config/i386/i386.md (@tls_dynamic_gnu2_64_<mode>): Replace
1305         :P with :PTR and Pmode with ptr_mode.
1306         (*tls_dynamic_gnu2_lea_64_<mode>): Likewise.
1307         (*tls_dynamic_gnu2_call_64_<mode>): Likewise.
1308         (*tls_dynamic_gnu2_combine_64_<mode>): Likewise.
1310 2020-01-21  Jakub Jelinek  <jakub@redhat.com>
1312         PR target/93333
1313         * config/riscv/riscv.c (riscv_rtx_costs) <case ZERO_EXTRACT>: Verify
1314         the last two operands are CONST_INT_P before using them as such.
1316 2020-01-21  Richard Sandiford  <richard.sandiford@arm.com>
1318         * config/aarch64/aarch64-sve-builtins.def: Use get_typenode_from_name
1319         to get the integer element types.
1321 2020-01-21  Richard Sandiford  <richard.sandiford@arm.com>
1323         * config/aarch64/aarch64-sve-builtins.h
1324         (function_expander::convert_to_pmode): Declare.
1325         * config/aarch64/aarch64-sve-builtins.cc
1326         (function_expander::convert_to_pmode): New function.
1327         (function_expander::get_contiguous_base): Use it.
1328         (function_expander::prepare_gather_address_operands): Likewise.
1329         * config/aarch64/aarch64-sve-builtins-sve2.cc
1330         (svwhilerw_svwhilewr_impl::expand): Likewise.
1332 2020-01-21  Szabolcs Nagy  <szabolcs.nagy@arm.com>
1334         PR target/92424
1335         * config/aarch64/aarch64.c (aarch64_declare_function_name): Set
1336         cfun->machine->label_is_assembled.
1337         (aarch64_print_patchable_function_entry): New.
1338         (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY): Define.
1339         * config/aarch64/aarch64.h (struct machine_function): New field,
1340         label_is_assembled.
1342 2020-01-21  David Malcolm  <dmalcolm@redhat.com>
1344         PR ipa/93315
1345         * ipa-profile.c (ipa_profile): Delete call_sums and set it to
1346         NULL on exit.
1348 2020-01-18  Jan Hubicka  <hubicka@ucw.cz>
1350         PR lto/93318    
1351         * cgraph.c (cgraph_edge::resolve_speculation,
1352         cgraph_edge::redirect_call_stmt_to_callee): Fix update of
1353         call_stmt_site_hash.
1355 2020-01-21  Martin Liska  <mliska@suse.cz>
1357         * config/rs6000/rs6000.c (common_mode_defined): Remove
1358         unused variable.
1360 2020-01-21  Richard Biener  <rguenther@suse.de>
1362         PR tree-optimization/92328
1363         * tree-ssa-sccvn.c (vn_reference_lookup_3): Preserve
1364         type when value-numbering same-sized store by inserting a
1365         VIEW_CONVERT_EXPR.
1366         (eliminate_dom_walker::eliminate_stmt): When eliminating
1367         a redundant store handle bit-reinterpretation of the same value.
1369 2020-01-21  Andrew Pinski  <apinski@marvel.com>
1371         PR tree-opt/93321
1372         * tree-into-ssa.c (prepare_block_for_update_1): Split out
1373         from ...
1374         (prepare_block_for_update): This.  Use a worklist instead of
1375         recursing.
1377 2020-01-21  Mihail-Calin Ionescu  <mihail.ionescu@arm.com>
1379         * gcc/config/arm/arm.c (clear_operation_p):
1380         Initialise last_regno, skip first iteration
1381         based on the first_set value and use ints instead
1382         of the unnecessary HOST_WIDE_INTs.
1384 2020-01-21  Jakub Jelinek  <jakub@redhat.com>
1386         PR target/93073
1387         * config/rs6000/rs6000.c (rs6000_emit_cmove): If using fsel, punt for
1388         compare_mode other than SFmode or DFmode.
1390 2020-01-21  Kito Cheng  <kito.cheng@sifive.com>
1392         PR target/93304
1393         * config/riscv/riscv-protos.h (riscv_hard_regno_rename_ok): New.
1394         * config/riscv/riscv.c (riscv_hard_regno_rename_ok): New.
1395         * config/riscv/riscv.h (HARD_REGNO_RENAME_OK): Defined.
1397 2020-01-20  Wilco Dijkstra  <wdijkstr@arm.com>
1399         * config/aarch64/aarch64.c (neoversen1_tunings): Set jump_align to 4.
1401 2020-01-20  Andrew Pinski  <apinski@marvell.com>
1403         PR middle-end/93242
1404         * targhooks.c (default_print_patchable_function_entry): Use
1405         output_asm_insn to emit the nop instruction.
1407 2020-01-20  Fangrui Song  <maskray@google.com>
1409         PR middle-end/93194
1410         * targhooks.c (default_print_patchable_function_entry): Align to
1411         POINTER_SIZE.
1413 2020-01-20  H.J. Lu  <hongjiu.lu@intel.com>
1415         PR target/93319
1416         * config/i386/i386.c (legitimize_tls_address): Pass Pmode to
1417         gen_tls_dynamic_gnu2_64.  Compute GNU2 TLS address in ptr_mode.
1418         * config/i386/i386.md (tls_dynamic_gnu2_64): Renamed to ...
1419         (@tls_dynamic_gnu2_64_<mode>): This.  Replace DI with P.
1420         (*tls_dynamic_gnu2_lea_64): Renamed to ...
1421         (*tls_dynamic_gnu2_lea_64_<mode>): This.  Replace DI with P.
1422         Remove the {q} suffix from lea.
1423         (*tls_dynamic_gnu2_call_64): Renamed to ...
1424         (*tls_dynamic_gnu2_call_64_<mode>): This.  Replace DI with P.
1425         (*tls_dynamic_gnu2_combine_64): Renamed to ...
1426         (*tls_dynamic_gnu2_combine_64_<mode>): This.  Replace DI with P.
1427         Pass Pmode to gen_tls_dynamic_gnu2_64.
1429 2020-01-20  Wilco Dijkstra  <wdijkstr@arm.com>
1431         * config/aarch64/aarch64.h (SLOW_BYTE_ACCESS): Set to 1.
1433 2020-01-20  Richard Sandiford  <richard.sandiford@arm.com>
1435         * config/aarch64/aarch64-sve-builtins-base.cc
1436         (svld1ro_impl::memory_vector_mode): Remove parameter name.
1438 2020-01-20  Richard Biener  <rguenther@suse.de>
1440         PR debug/92763
1441         * dwarf2out.c (prune_unused_types): Unconditionally mark
1442         called function DIEs.
1444 2020-01-20  Martin Liska  <mliska@suse.cz>
1446         PR tree-optimization/93199
1447         * tree-eh.c (struct leh_state): Add
1448         new field outer_non_cleanup.
1449         (cleanup_is_dead_in): Pass leh_state instead
1450         of eh_region.  Add a checking that state->outer_non_cleanup
1451         points to outer non-clean up region.
1452         (lower_try_finally): Record outer_non_cleanup
1453         for this_state.
1454         (lower_catch): Likewise.
1455         (lower_eh_filter): Likewise.
1456         (lower_eh_must_not_throw): Likewise.
1457         (lower_cleanup): Likewise.
1459 2020-01-20  Richard Biener  <rguenther@suse.de>
1461         PR tree-optimization/93094
1462         * tree-vectorizer.h (vect_loop_versioning): Adjust.
1463         (vect_transform_loop): Likewise.
1464         * tree-vectorizer.c (try_vectorize_loop_1): Pass down
1465         loop_vectorized_call to vect_transform_loop.
1466         * tree-vect-loop.c (vect_transform_loop): Pass down
1467         loop_vectorized_call to vect_loop_versioning.
1468         * tree-vect-loop-manip.c (vect_loop_versioning): Use
1469         the earlier discovered loop_vectorized_call.
1471 2020-01-19  Eric S. Raymond <esr@thyrsus.com>
1473         * doc/contribute.texi: Update for SVN -> Git transition.
1474         * doc/install.texi: Likewise.
1476 2020-01-18  Jan Hubicka  <hubicka@ucw.cz>
1478         PR lto/93318
1479         * cgraph.c (cgraph_edge::make_speculative): Increase number of
1480         speculative targets.
1481         (verify_speculative_call): New function
1482         (cgraph_node::verify_node): Use it.
1483         * ipa-profile.c (ipa_profile): Fix formating; do not set number of
1484         speculations.
1486 2020-01-18  Jan Hubicka  <hubicka@ucw.cz>
1488         PR lto/93318
1489         * cgraph.c (cgraph_edge::resolve_speculation): Fix foramting.
1490         (cgraph_edge::make_direct): Remove all indirect targets.
1491         (cgraph_edge::redirect_call_stmt_to_callee): Use make_direct..
1492         (cgraph_node::verify_node): Verify that only one call_stmt or
1493         lto_stmt_uid is set.
1494         * cgraphclones.c (cgraph_edge::clone): Set only one call_stmt or
1495         lto_stmt_uid.
1496         * lto-cgraph.c (lto_output_edge): Simplify streaming of stmt.
1497         (lto_output_ref): Simplify streaming of stmt.
1498         * lto-streamer-in.c (fixup_call_stmt_edges_1): Clear lto_stmt_uid.
1500 2020-01-18  Tamar Christina  <tamar.christina@arm.com>
1502         * config/aarch64/aarch64-sve-builtins-base.cc (memory_vector_mode):
1503         Mark parameter unused.
1505 2020-01-18  Hans-Peter Nilsson  <hp@axis.com>
1507         * config.gcc <obsolete targets>: Add crisv32-*-* and cris-*-linux*
1509 2019-01-18  Gerald Pfeifer  <gerald@pfeifer.com>
1511         * varpool.c (ctor_useable_for_folding_p): Fix grammar.
1513 2020-01-18  Iain Sandoe  <iain@sandoe.co.uk>
1515         * Makefile.in: Add coroutine-passes.o.
1516         * builtin-types.def (BT_CONST_SIZE): New.
1517         (BT_FN_BOOL_PTR): New.
1518         (BT_FN_PTR_PTR_CONST_SIZE_BOOL): New.
1519         * builtins.def (DEF_COROUTINE_BUILTIN): New.
1520         * coroutine-builtins.def: New file.
1521         * coroutine-passes.cc: New file.
1522         * function.h (struct GTY function): Add a bit to indicate that the
1523         function is a coroutine component.
1524         * internal-fn.c (expand_CO_FRAME): New.
1525         (expand_CO_YIELD): New.
1526         (expand_CO_SUSPN): New.
1527         (expand_CO_ACTOR): New.
1528         * internal-fn.def (CO_ACTOR): New.
1529         (CO_YIELD): New.
1530         (CO_SUSPN): New.
1531         (CO_FRAME): New.
1532         * passes.def: Add pass_coroutine_lower_builtins,
1533         pass_coroutine_early_expand_ifns.
1534         * tree-pass.h (make_pass_coroutine_lower_builtins): New.
1535         (make_pass_coroutine_early_expand_ifns): New.
1536         * doc/invoke.texi: Document the fcoroutines command line
1537         switch.
1539 2020-01-18  Jakub Jelinek  <jakub@redhat.com>
1541         * config/arm/vfp.md (*clear_vfp_multiple): Remove unused variable.
1543         PR target/93312
1544         * config/arm/arm.c (clear_operation_p): Don't use REGNO until
1545         after checking the argument is a REG.  Don't use REGNO (reg)
1546         again to set last_regno, reuse regno variable instead.
1548 2020-01-17  David Malcolm  <dmalcolm@redhat.com>
1550         * doc/analyzer.texi (Limitations): Add note about NaN.
1552 2020-01-17  Mihail-Calin Ionescu  <mihail.ionescu@arm.com>
1553             Sudakshina Das  <sudi.das@arm.com>
1555         * config/arm/arm.md (ashldi3): Generate thumb2_lsll for both reg
1556         and valid immediate.
1557         (ashrdi3): Generate thumb2_asrl for both reg and valid immediate.
1558         (lshrdi3): Generate thumb2_lsrl for valid immediates.
1559         * config/arm/constraints.md (Pg): New.
1560         * config/arm/predicates.md (long_shift_imm): New.
1561         (arm_reg_or_long_shift_imm): Likewise.
1562         * config/arm/thumb2.md (thumb2_asrl): New immediate alternative.
1563         (thumb2_lsll): Likewise.
1564         (thumb2_lsrl): New.
1566 2020-01-17  Mihail-Calin Ionescu  <mihail.ionescu@arm.com>
1567             Sudakshina Das  <sudi.das@arm.com>
1569         * config/arm/arm.md (ashldi3): Generate thumb2_lsll for TARGET_HAVE_MVE.
1570         (ashrdi3): Generate thumb2_asrl for TARGET_HAVE_MVE.
1571         * config/arm/arm.c (arm_hard_regno_mode_ok): Allocate even odd
1572         register pairs for doubleword quantities for ARMv8.1M-Mainline.
1573         * config/arm/thumb2.md (thumb2_asrl): New.
1574         (thumb2_lsll): Likewise.
1576 2020-01-17  Jakub Jelinek  <jakub@redhat.com>
1578         * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear): Remove
1579         unused variable.
1581 2020-01-17  Alexander Monakov  <amonakov@ispras.ru>
1583         * gdbinit.in (help-gcc-hooks): New command.
1584         (pp, pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, ptc, pdn, ptn, pdd, prc,
1585         pi, pbm, pel, trt): Take $arg0 instead of $ if supplied. Update
1586         documentation.
1588 2020-01-17  Matthew Malcomson  <matthew.malcomson@arm.com>
1590         * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use the
1591         correct target macro.
1593 2020-01-17  Matthew Malcomson  <matthew.malcomson@arm.com>
1595         * config/aarch64/aarch64-protos.h
1596         (aarch64_sve_ld1ro_operand_p): New.
1597         * config/aarch64/aarch64-sve-builtins-base.cc
1598         (class load_replicate): New.
1599         (class svld1ro_impl): New.
1600         (class svld1rq_impl): Change to inherit from load_replicate.
1601         (svld1ro): New sve intrinsic function base.
1602         * config/aarch64/aarch64-sve-builtins-base.def (svld1ro):
1603         New DEF_SVE_FUNCTION.
1604         * config/aarch64/aarch64-sve-builtins-base.h
1605         (svld1ro): New decl.
1606         * config/aarch64/aarch64-sve-builtins.cc
1607         (function_expander::add_mem_operand): Modify assert to allow
1608         OImode.
1609         * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): New
1610         pattern.
1611         * config/aarch64/aarch64.c
1612         (aarch64_sve_ld1rq_operand_p): Implement in terms of ...
1613         (aarch64_sve_ld1rq_ld1ro_operand_p): This.
1614         (aarch64_sve_ld1ro_operand_p): New.
1615         * config/aarch64/aarch64.md (UNSPEC_LD1RO): New unspec.
1616         * config/aarch64/constraints.md (UOb,UOh,UOw,UOd): New.
1617         * config/aarch64/predicates.md
1618         (aarch64_sve_ld1ro_operand_{b,h,w,d}): New.
1620 2020-01-17  Matthew Malcomson  <matthew.malcomson@arm.com>
1622         * config/aarch64/aarch64-c.c (_ARM_FEATURE_MATMUL_FLOAT64):
1623         Introduce this ACLE specified predefined macro.
1624         * config/aarch64/aarch64-option-extensions.def (f64mm): New.
1625         (fp): Disabling this disables f64mm.
1626         (simd): Disabling this disables f64mm.
1627         (fp16): Disabling this disables f64mm.
1628         (sve): Disabling this disables f64mm.
1629         * config/aarch64/aarch64.h (AARCH64_FL_F64MM): New.
1630         (AARCH64_ISA_F64MM): New.
1631         (TARGET_F64MM): New.
1632         * doc/invoke.texi (f64mm): Document new option.
1634 2020-01-17  Wilco Dijkstra  <wdijkstr@arm.com>
1636         * config/aarch64/aarch64.c (generic_tunings): Add branch fusion.
1637         (neoversen1_tunings): Likewise.
1639 2020-01-17  Wilco Dijkstra  <wdijkstr@arm.com>
1641         PR target/92692
1642         * config/aarch64/aarch64.c (aarch64_split_compare_and_swap)
1643         Add assert to ensure prolog has been emitted.
1644         (aarch64_split_atomic_op): Likewise.
1645         * config/aarch64/atomics.md (aarch64_compare_and_swap<mode>)
1646         Use epilogue_completed rather than reload_completed.
1647         (aarch64_atomic_exchange<mode>): Likewise.
1648         (aarch64_atomic_<atomic_optab><mode>): Likewise.
1649         (atomic_nand<mode>): Likewise.
1650         (aarch64_atomic_fetch_<atomic_optab><mode>): Likewise.
1651         (atomic_fetch_nand<mode>): Likewise.
1652         (aarch64_atomic_<atomic_optab>_fetch<mode>): Likewise.
1653         (atomic_nand_fetch<mode>): Likewise.
1655 2020-01-17  Richard Sandiford  <richard.sandiford@arm.com>
1657         PR target/93133
1658         * config/aarch64/aarch64.h (REVERSIBLE_CC_MODE): Return false
1659         for FP modes.
1660         (REVERSE_CONDITION): Delete.
1661         * config/aarch64/iterators.md (CC_ONLY): New mode iterator.
1662         (CCFP_CCFPE): Likewise.
1663         (e): New mode attribute.
1664         * config/aarch64/aarch64.md (ccmp<GPI:mode>): Rename to...
1665         (@ccmp<CC_ONLY:mode><GPI:mode>): ...this, using CC_ONLY instead of CC.
1666         (fccmp<GPF:mode>, fccmpe<GPF:mode>): Merge into...
1667         (@ccmp<CCFP_CCFPE:mode><GPF:mode>): ...this combined pattern.
1668         (@ccmp<CC_ONLY:mode><GPI:mode>_rev): New pattern.
1669         (@ccmp<CCFP_CCFPE:mode><GPF:mode>_rev): Likewise.
1670         * config/aarch64/aarch64.c (aarch64_gen_compare_reg): Update
1671         name of generator from gen_ccmpdi to gen_ccmpccdi.
1672         (aarch64_gen_ccmp_next): Use code_for_ccmp.  If we want to reverse
1673         the previous comparison but aren't able to, use the new ccmp_rev
1674         patterns instead.
1676 2020-01-17  Richard Sandiford  <richard.sandiford@arm.com>
1678         * gimplify.c (gimplify_return_expr): Use poly_int_tree_p rather
1679         than testing directly for INTEGER_CST.
1680         (gimplify_target_expr, gimplify_omp_depend): Likewise.
1682 2020-01-17  Jakub Jelinek  <jakub@redhat.com>
1684         PR tree-optimization/93292
1685         * tree-vect-stmts.c (vectorizable_comparison): Punt also if
1686         get_vectype_for_scalar_type returns NULL.
1688 2020-01-16  Jan Hubicka  <hubicka@ucw.cz>
1690         * params.opt (-param=max-predicted-iterations): Increase range from 0.
1691         * predict.c (estimate_loops): Add 1 to param_max_predicted_iterations.
1693 2020-01-16  Jan Hubicka  <hubicka@ucw.cz>
1695         * ipa-fnsummary.c (estimate_calls_size_and_time): Fix formating of
1696         dump.
1697         * params.opt: (max-predicted-iterations): Set bounds.
1698         * predict.c (real_almost_one, real_br_prob_base,
1699         real_inv_br_prob_base, real_one_half, real_bb_freq_max): Remove.
1700         (propagate_freq): Add max_cyclic_prob parameter; cap cyclic
1701         probabilities; do not truncate to reg_br_prob_bases.
1702         (estimate_loops_at_level): Pass max_cyclic_prob.
1703         (estimate_loops): Compute max_cyclic_prob.
1704         (estimate_bb_frequencies): Do not initialize real_*; update calculation
1705         of back edge prob.
1706         * profile-count.c (profile_probability::to_sreal): New.
1707         * profile-count.h (class sreal): Move up in file.
1708         (profile_probability::to_sreal): Declare.
1710 2020-01-16  Stam Markianos-Wright  <stam.markianos-wright@arm.com>
1712         * config/arm/arm.c
1713         (arm_invalid_conversion): New function for target hook.
1714         (arm_invalid_unary_op): New function for target hook.
1715         (arm_invalid_binary_op): New function for target hook.
1717 2020-01-16  Stam Markianos-Wright  <stam.markianos-wright@arm.com>
1719         * config.gcc: Add arm_bf16.h.
1720         * config/arm/arm-builtins.c (arm_mangle_builtin_type): Fix comment.
1721         (arm_simd_builtin_std_type): Add BFmode.
1722         (arm_init_simd_builtin_types): Define element types for vector types.
1723         (arm_init_bf16_types): New function.
1724         (arm_init_builtins): Add arm_init_bf16_types function call.
1725         * config/arm/arm-modes.def: Add BFmode and V4BF, V8BF vector modes.
1726         * config/arm/arm-simd-builtin-types.def: Add V4BF, V8BF.
1727         * config/arm/arm.c (aapcs_vfp_sub_candidate):  Add BFmode.
1728         (arm_hard_regno_mode_ok): Add BFmode and tidy up statements.
1729         (arm_vector_mode_supported_p): Add V4BF, V8BF.
1730         (arm_mangle_type):  Add __bf16.
1731         * config/arm/arm.h: Add V4BF, V8BF to VALID_NEON_DREG_MODE, 
1732         VALID_NEON_QREG_MODE respectively. Add export arm_bf16_type_node,
1733         arm_bf16_ptr_type_node.
1734         * config/arm/arm.md: Add BFmode to movhf expand, mov pattern and
1735         define_split between ARM registers.
1736         * config/arm/arm_bf16.h: New file.
1737         * config/arm/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
1738         * config/arm/iterators.md: (ANY64_BF, VDXMOV, VHFBF, HFBF, fporbf): New.
1739         (VQXMOV): Add V8BF.
1740         * config/arm/neon.md: Add BF vector types to movhf NEON move patterns.
1741         * config/arm/vfp.md: Add BFmode to movhf patterns.
1743 2020-01-16  Mihail Ionescu  <mihail.ionescu@arm.com>
1744             Andre Vieira  <andre.simoesdiasvieira@arm.com>
1746         * config/arm/arm-cpus.in (mve, mve_float): New features.
1747         (dsp, mve, mve.fp): New options.
1748         * config/arm/arm.h (TARGET_HAVE_MVE, TARGET_HAVE_MVE_FLOAT): Define.
1749         * config/arm/t-rmprofile: Map v8.1-M multilibs to v8-M.
1750         * doc/invoke.texi: Document the armv8.1-m mve and dps options.
1752 2020-01-16  Mihail-Calin Ionescu  <mihail.ionescu@arm.com>
1753             Thomas Preud'homme  <thomas.preudhomme@arm.com>
1755         * config/arm/arm-cpus.in (ARMv8_1m_main): Redefine as an extension to
1756         Armv8-M Mainline.
1757         * config/arm/arm.c (arm_options_perform_arch_sanity_checks): Remove
1758         error for using -mcmse when targeting Armv8.1-M Mainline.
1760 2020-01-16  Mihail-Calin Ionescu  <mihail.ionescu@arm.com>
1761             Thomas Preud'homme  <thomas.preudhomme@arm.com>
1763         * config/arm/arm.md (nonsecure_call_internal): Do not force memory
1764         address in r4 when targeting Armv8.1-M Mainline.
1765         (nonsecure_call_value_internal): Likewise.
1766         * config/arm/thumb2.md (nonsecure_call_reg_thumb2): Make memory address
1767         a register match_operand again.  Emit BLXNS when targeting
1768         Armv8.1-M Mainline.
1769         (nonsecure_call_value_reg_thumb2): Likewise.
1771 2020-01-16  Mihail-Calin Ionescu  <mihail.ionescu@arm.com>
1772             Thomas Preud'homme  <thomas.preudhomme@arm.com>
1774         * config/arm/arm.c (arm_add_cfa_adjust_cfa_note): Declare early.
1775         (cmse_nonsecure_call_inline_register_clear): Define new lazy_fpclear
1776         variable as true when floating-point ABI is not hard.  Replace
1777         check against TARGET_HARD_FLOAT_ABI by checks against lazy_fpclear.
1778         Generate VLSTM and VLLDM instruction respectively before and
1779         after a function call to cmse_nonsecure_call function.
1780         * config/arm/unspecs.md (VUNSPEC_VLSTM): Define unspec.
1781         (VUNSPEC_VLLDM): Likewise.
1782         * config/arm/vfp.md (lazy_store_multiple_insn): New define_insn.
1783         (lazy_load_multiple_insn): Likewise.
1785 2020-01-16  Mihail-Calin Ionescu  <mihail.ionescu@arm.com>
1786             Thomas Preud'homme  <thomas.preudhomme@arm.com>
1788         * config/arm/arm.c (vfp_emit_fstmd): Declare early.
1789         (arm_emit_vfp_multi_reg_pop): Likewise.
1790         (cmse_nonsecure_call_inline_register_clear): Abstract number of VFP
1791         registers to clear in max_fp_regno.  Emit VPUSH and VPOP to save and
1792         restore callee-saved VFP registers.
1794 2020-01-16  Mihail-Calin Ionescu  <mihail.ionescu@arm.com>
1795             Thomas Preud'homme  <thomas.preudhomme@arm.com>
1797         * config/arm/arm.c (arm_emit_multi_reg_pop): Declare early.
1798         (cmse_nonsecure_call_clear_caller_saved): Rename into ...
1799         (cmse_nonsecure_call_inline_register_clear): This.  Save and clear
1800         callee-saved GPRs as well as clear ip register before doing a nonsecure
1801         call then restore callee-saved GPRs after it when targeting
1802         Armv8.1-M Mainline.
1803         (arm_reorg): Adapt to function rename.
1805 2020-01-16  Mihail-Calin Ionescu  <mihail.ionescu@arm.com>
1806             Thomas Preud'homme  <thomas.preudhomme@arm.com>
1808         * config/arm/arm-protos.h (clear_operation_p): Adapt prototype.
1809         * config/arm/arm.c (clear_operation_p): Extend to be able to check a
1810         clear_vfp_multiple pattern based on a new vfp parameter.
1811         (cmse_clear_registers): Generate VSCCLRM to clear VFP registers when
1812         targeting Armv8.1-M Mainline.
1813         (cmse_nonsecure_entry_clear_before_return): Clear VFP registers
1814         unconditionally when targeting Armv8.1-M Mainline architecture.  Check
1815         whether VFP registers are available before looking call_used_regs for a
1816         VFP register.
1817         * config/arm/predicates.md (clear_multiple_operation): Adapt to change
1818         of prototype of clear_operation_p.
1819         (clear_vfp_multiple_operation): New predicate.
1820         * config/arm/unspecs.md (VUNSPEC_VSCCLRM_VPR): New volatile unspec.
1821         * config/arm/vfp.md (clear_vfp_multiple): New define_insn.
1823 2020-01-16  Mihail-Calin Ionescu  <mihail.ionescu@arm.com>
1824             Thomas Preud'homme  <thomas.preudhomme@arm.com>
1826         * config/arm/arm-protos.h (clear_operation_p): Declare.
1827         * config/arm/arm.c (clear_operation_p): New function.
1828         (cmse_clear_registers): Generate clear_multiple instruction pattern if
1829         targeting Armv8.1-M Mainline or successor.
1830         (output_return_instruction): Only output APSR register clearing if
1831         Armv8.1-M Mainline instructions not available.
1832         (thumb_exit): Likewise.
1833         * config/arm/predicates.md (clear_multiple_operation): New predicate.
1834         * config/arm/thumb2.md (clear_apsr): New define_insn.
1835         (clear_multiple): Likewise.
1836         * config/arm/unspecs.md (VUNSPEC_CLRM_APSR): New volatile unspec.
1838 2020-01-16  Mihail-Calin Ionescu  <mihail.ionescu@arm.com>
1839             Thomas Preud'homme  <thomas.preudhomme@arm.com>
1841         * config/arm/arm.c (fp_sysreg_names): Declare and define.
1842         (use_return_insn): Also return false for Armv8.1-M Mainline.
1843         (output_return_instruction): Skip FPSCR clearing if Armv8.1-M
1844         Mainline instructions are available.
1845         (arm_compute_frame_layout): Allocate space in frame for FPCXTNS
1846         when targeting Armv8.1-M Mainline Security Extensions.
1847         (arm_expand_prologue): Save FPCXTNS if this is an Armv8.1-M
1848         Mainline entry function.
1849         (cmse_nonsecure_entry_clear_before_return): Clear IP and r4 if
1850         targeting Armv8.1-M Mainline or successor.
1851         (arm_expand_epilogue): Fix indentation of caller-saved register
1852         clearing.  Restore FPCXTNS if this is an Armv8.1-M Mainline
1853         entry function.
1854         * config/arm/arm.h (TARGET_HAVE_FP_CMSE): New macro.
1855         (FP_SYSREGS): Likewise.
1856         (enum vfp_sysregs_encoding): Define enum.
1857         (fp_sysreg_names): Declare.
1858         * config/arm/unspecs.md (VUNSPEC_VSTR_VLDR): New volatile unspec.
1859         * config/arm/vfp.md (push_fpsysreg_insn): New define_insn.
1860         (pop_fpsysreg_insn): Likewise.
1862 2020-01-16  Mihail-Calin Ionescu  <mihail.ionescu@arm.com>
1863             Thomas Preud'homme  <thomas.preudhomme@arm.com>
1865         * config/arm/arm-cpus.in (armv8_1m_main): New feature.
1866         (ARMv4, ARMv4t, ARMv5t, ARMv5te, ARMv5tej, ARMv6, ARMv6j, ARMv6k,
1867         ARMv6z, ARMv6kz, ARMv6zk, ARMv6t2, ARMv6m, ARMv7, ARMv7a, ARMv7ve,
1868         ARMv7r, ARMv7m, ARMv7em, ARMv8a, ARMv8_1a, ARMv8_2a, ARMv8_3a,
1869         ARMv8_4a, ARMv8_5a, ARMv8m_base, ARMv8m_main, ARMv8r): Reindent.
1870         (ARMv8_1m_main): New feature group.
1871         (armv8.1-m.main): New architecture.
1872         * config/arm/arm-tables.opt: Regenerate.
1873         * config/arm/arm.c (arm_arch8_1m_main): Define and default initialize.
1874         (arm_option_reconfigure_globals): Initialize arm_arch8_1m_main.
1875         (arm_options_perform_arch_sanity_checks): Error out when targeting
1876         Armv8.1-M Mainline Security Extensions.
1877         * config/arm/arm.h (arm_arch8_1m_main): Declare.
1879 2020-01-16  Stam Markianos-Wright  <stam.markianos-wright@arm.com>
1881         * config/aarch64/aarch64-simd-builtins.def (aarch64_bfdot,
1882         aarch64_bfdot_lane, aarch64_bfdot_laneq): New.
1883         * config/aarch64/aarch64-simd.md (aarch64_bfdot, aarch64_bfdot_lane,
1884         aarch64_bfdot_laneq): New.
1885         * config/aarch64/arm_bf16.h (vbfdot_f32, vbfdotq_f32,
1886         vbfdot_lane_f32, vbfdotq_lane_f32, vbfdot_laneq_f32,
1887         vbfdotq_laneq_f32): New.
1888         * config/aarch64/iterators.md (UNSPEC_BFDOT, Vbfdottype,
1889         VBFMLA_W, VBF): New.
1890         (isquadop): Add V4BF, V8BF.
1892 2020-01-16  Stam Markianos-Wright  <stam.markianos-wright@arm.com>
1894         * config/aarch64/aarch64-builtins.c: (enum aarch64_type_qualifiers):
1895         New qualifier_lane_quadtup_index, TYPES_TERNOP_SSUS,
1896         TYPES_QUADOPSSUS_LANE_QUADTUP, TYPES_QUADOPSSSU_LANE_QUADTUP.
1897         (aarch64_simd_expand_args): Add case SIMD_ARG_LANE_QUADTUP_INDEX.
1898         (aarch64_simd_expand_builtin): Add qualifier_lane_quadtup_index.
1899         * config/aarch64/aarch64-simd-builtins.def (usdot, usdot_lane,
1900         usdot_laneq, sudot_lane,sudot_laneq): New.
1901         * config/aarch64/aarch64-simd.md (aarch64_usdot): New.
1902         (aarch64_<sur>dot_lane): New.
1903         * config/aarch64/arm_neon.h (vusdot_s32): New.
1904         (vusdotq_s32): New.
1905         (vusdot_lane_s32): New.
1906         (vsudot_lane_s32): New.
1907         * config/aarch64/iterators.md (DOTPROD_I8MM): New iterator.
1908         (UNSPEC_USDOT, UNSPEC_SUDOT): New unspecs.
1910 2020-01-16  Martin Liska  <mliska@suse.cz>
1912         * value-prof.c (dump_histogram_value): Fix
1913         obvious spacing issue.
1915 2020-01-16  Andrew Pinski  <apinski@marvell.com>
1917         * tree-ssa-sccvn.c(vn_reference_lookup_3): Check lhs for
1918         !storage_order_barrier_p.
1920 2020-01-16  Andrew Pinski  <apinski@marvell.com>
1922         * sched-int.h (_dep): Add unused bit-field field for the padding.
1923         * sched-deps.c (init_dep_1): Init unused field.
1925 2020-01-16  Andrew Pinski  <apinski@marvell.com>
1927         * optabs.h (create_expand_operand): Initialize target field also.
1929 2020-01-16  Andre Vieira  <andre.simoesdiasvieira@arm.com>
1931         PR tree-optimization/92429
1932         * tree-ssa-loop-niter.h (simplify_replace_tree): Add parameter.
1933         * tree-ssa-loop-niter.c (simplify_replace_tree): Add parameter to
1934         control folding.
1935         * tree-vect-loop.c (update_epilogue_vinfo): Do not fold when replacing
1936         tree.
1938 2020-01-16  Richard Sandiford  <richard.sandiford@arm.com>
1940         * config/aarch64/aarch64.c (aarch64_split_sve_subreg_move): Apply
1941         aarch64_sve_int_mode to each mode.
1943 2020-01-15  David Malcolm  <dmalcolm@redhat.com>
1945         * doc/analyzer.texi (Overview): Add note about
1946         -fdump-ipa-analyzer.
1948 2020-01-15  Wilco Dijkstra  <wdijkstr@arm.com>
1950         PR tree-optimization/93231
1951         * tree-ssa-forwprop.c (optimize_count_trailing_zeroes): Check
1952         input_type is unsigned.  Use tree_to_shwi for shift constant.
1953         Check CST_STRING element size is CHAR_TYPE_SIZE bits.
1954         (simplify_count_trailing_zeroes): Add test to handle known non-zero
1955         inputs more efficiently.
1957 2020-01-15  Uroš Bizjak  <ubizjak@gmail.com>
1959         * config/i386/i386.md (*movsf_internal): Do not require
1960         SSE2 ISA for alternatives 14 and 15.
1962 2020-01-15  Richard Biener  <rguenther@suse.de>
1964         PR middle-end/93273
1965         * tree-eh.c (sink_clobbers): If we already visited the destination
1966         block do not defer insertion.
1967         (pass_lower_eh_dispatch::execute): Maintain BB_VISITED for
1968         the purpose of defered insertion.
1970 2020-01-15  Jakub Jelinek  <jakub@redhat.com>
1972         * BASE-VER: Bump to 10.0.1.
1974 2020-01-15  Richard Sandiford  <richard.sandiford@arm.com>
1976         PR tree-optimization/93247
1977         * tree-vect-loop.c (update_epilogue_loop_vinfo): Check the access
1978         type of the stmt that we're going to vectorize.
1980 2020-01-15  Richard Sandiford  <richard.sandiford@arm.com>
1982         * tree-vect-slp.c (vectorize_slp_instance_root_stmt): Use a
1983         VIEW_CONVERT_EXPR if the vectorized constructor has a diffeent
1984         type from the lhs.
1986 2020-01-15  Martin Liska  <mliska@suse.cz>
1988         * ipa-profile.c (ipa_profile_read_edge_summary): Do not allow
1989         2 calls of streamer_read_hwi in a function call.
1991 2020-01-15  Richard Biener  <rguenther@suse.de>
1993         * alias.c (record_alias_subset): Avoid redundant work when
1994         subset is already recorded.
1996 2020-01-14  David Malcolm  <dmalcolm@redhat.com>
1998         * doc/invoke.texi (-fdiagnostics-show-cwe): Add note that some of
1999         the analyzer options provide CWE identifiers.
2001 2020-01-14  David Malcolm  <dmalcolm@redhat.com>
2003         * tree-diagnostic-path.cc (path_summary::event_range::print):
2004         When testing for UNKNOWN_LOCATION, look through ad-hoc wrappers
2005         using get_pure_location.
2007 2020-01-15  Jakub Jelinek  <jakub@redhat.com>
2009         PR tree-optimization/93262
2010         * tree-ssa-dse.c (maybe_trim_memstar_call): For *_chk builtins,
2011         perform head trimming only if the last argument is constant,
2012         either all ones, or larger or equal to head trim, in the latter
2013         case decrease the last argument by head_trim.
2015         PR tree-optimization/93249
2016         * tree-ssa-dse.c: Include builtins.h and gimple-fold.h.
2017         (maybe_trim_memstar_call): Move head_trim and tail_trim vars to
2018         function body scope, reindent.  For BUILTIN_IN_STRNCPY*, don't
2019         perform head trim unless we can prove there are no '\0' chars
2020         from the source among the first head_trim chars.
2022 2020-01-14  David Malcolm  <dmalcolm@redhat.com>
2024         * Makefile.in (ANALYZER_OBJS): Add analyzer/function-set.o.
2026 2020-01-15  Jakub Jelinek  <jakub@redhat.com>
2028         PR target/93009
2029         * config/i386/sse.md
2030         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1,
2031         *<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1,
2032         *<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1,
2033         *<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Use
2034         just a single alternative instead of two, make operands 1 and 2
2035         commutative.
2037 2020-01-14  Jan Hubicka  <hubicka@ucw.cz>
2039         PR lto/91576
2040         * ipa-devirt.c (odr_types_equivalent_p): Compare TREE_ADDRESSABLE and
2041         TYPE_MODE.
2043 2020-01-14  David Malcolm  <dmalcolm@redhat.com>
2045         * Makefile.in (lang_opt_files): Add analyzer.opt.
2046         (ANALYZER_OBJS): New.
2047         (OBJS): Add digraph.o, graphviz.o, ordered-hash-map-tests.o,
2048         tristate.o and ANALYZER_OBJS.
2049         (TEXI_GCCINT_FILES): Add analyzer.texi.
2050         * common.opt (-fanalyzer): New driver option.
2051         * config.in: Regenerate.
2052         * configure: Regenerate.
2053         * configure.ac (--disable-analyzer, ENABLE_ANALYZER): New option.
2054         (gccdepdir): Also create depdir for "analyzer" subdir.
2055         * digraph.cc: New file.
2056         * digraph.h: New file.
2057         * doc/analyzer.texi: New file.
2058         * doc/gccint.texi ("Static Analyzer") New menu item.
2059         (analyzer.texi): Include it.
2060         * doc/invoke.texi ("Static Analyzer Options"): New list and new section.
2061         ("Warning Options"): Add static analysis warnings to the list.
2062         (-Wno-analyzer-double-fclose): New option.
2063         (-Wno-analyzer-double-free): New option.
2064         (-Wno-analyzer-exposure-through-output-file): New option.
2065         (-Wno-analyzer-file-leak): New option.
2066         (-Wno-analyzer-free-of-non-heap): New option.
2067         (-Wno-analyzer-malloc-leak): New option.
2068         (-Wno-analyzer-possible-null-argument): New option.
2069         (-Wno-analyzer-possible-null-dereference): New option.
2070         (-Wno-analyzer-null-argument): New option.
2071         (-Wno-analyzer-null-dereference): New option.
2072         (-Wno-analyzer-stale-setjmp-buffer): New option.
2073         (-Wno-analyzer-tainted-array-index): New option.
2074         (-Wno-analyzer-use-after-free): New option.
2075         (-Wno-analyzer-use-of-pointer-in-stale-stack-frame): New option.
2076         (-Wno-analyzer-use-of-uninitialized-value): New option.
2077         (-Wanalyzer-too-complex): New option.
2078         (-fanalyzer-call-summaries): New warning.
2079         (-fanalyzer-checker=): New warning.
2080         (-fanalyzer-fine-grained): New warning.
2081         (-fno-analyzer-state-merge): New warning.
2082         (-fno-analyzer-state-purge): New warning.
2083         (-fanalyzer-transitivity): New warning.
2084         (-fanalyzer-verbose-edges): New warning.
2085         (-fanalyzer-verbose-state-changes): New warning.
2086         (-fanalyzer-verbosity=): New warning.
2087         (-fdump-analyzer): New warning.
2088         (-fdump-analyzer-callgraph): New warning.
2089         (-fdump-analyzer-exploded-graph): New warning.
2090         (-fdump-analyzer-exploded-nodes): New warning.
2091         (-fdump-analyzer-exploded-nodes-2): New warning.
2092         (-fdump-analyzer-exploded-nodes-3): New warning.
2093         (-fdump-analyzer-supergraph): New warning.
2094         * doc/sourcebuild.texi (dg-require-dot): New.
2095         (dg-check-dot): New.
2096         * gdbinit.in (break-on-saved-diagnostic): New command.
2097         * graphviz.cc: New file.
2098         * graphviz.h: New file.
2099         * ordered-hash-map-tests.cc: New file.
2100         * ordered-hash-map.h: New file.
2101         * passes.def (pass_analyzer): Add before
2102         pass_ipa_whole_program_visibility.
2103         * selftest-run-tests.c (selftest::run_tests): Call
2104         selftest::ordered_hash_map_tests_cc_tests.
2105         * selftest.h (selftest::ordered_hash_map_tests_cc_tests): New
2106         decl.
2107         * shortest-paths.h: New file.
2108         * timevar.def (TV_ANALYZER): New timevar.
2109         (TV_ANALYZER_SUPERGRAPH): Likewise.
2110         (TV_ANALYZER_STATE_PURGE): Likewise.
2111         (TV_ANALYZER_PLAN): Likewise.
2112         (TV_ANALYZER_SCC): Likewise.
2113         (TV_ANALYZER_WORKLIST): Likewise.
2114         (TV_ANALYZER_DUMP): Likewise.
2115         (TV_ANALYZER_DIAGNOSTICS): Likewise.
2116         (TV_ANALYZER_SHORTEST_PATHS): Likewise.
2117         * tree-pass.h (make_pass_analyzer): New decl.
2118         * tristate.cc: New file.
2119         * tristate.h: New file.
2121 2020-01-14  Uroš Bizjak  <ubizjak@gmail.com>
2123         PR target/93254
2124         * config/i386/i386.md (*movsf_internal): Require SSE2 ISA for
2125         alternatives 9 and 10.
2127 2020-01-14  David Malcolm  <dmalcolm@redhat.com>
2129         * attribs.c (excl_hash_traits::empty_zero_p): New static constant.
2130         * gcov.c (function_start_pair_hash::empty_zero_p): Likewise.
2131         * graphite.c (struct sese_scev_hash::empty_zero_p): Likewise.
2132         * hash-map-tests.c (selftest::test_nonzero_empty_key): New selftest.
2133         (selftest::hash_map_tests_c_tests): Call it.
2134         * hash-map-traits.h (simple_hashmap_traits::empty_zero_p):
2135         New static constant, using the value of = H::empty_zero_p.
2136         (unbounded_hashmap_traits::empty_zero_p): Likewise, using the value
2137         from default_hash_traits <Value>.
2138         * hash-map.h (hash_map::empty_zero_p): Likewise, using the value
2139         from Traits.
2140         * hash-set-tests.c (value_hash_traits::empty_zero_p): Likewise.
2141         * hash-table.h (hash_table::alloc_entries): Guard the loop of
2142         calls to mark_empty with !Descriptor::empty_zero_p.
2143         (hash_table::empty_slow): Conditionalize the memset call with a
2144         check that Descriptor::empty_zero_p; otherwise, loop through the
2145         entries calling mark_empty on them.
2146         * hash-traits.h (int_hash::empty_zero_p): New static constant.
2147         (pointer_hash::empty_zero_p): Likewise.
2148         (pair_hash::empty_zero_p): Likewise.
2149         * ipa-devirt.c (default_hash_traits <type_pair>::empty_zero_p):
2150         Likewise.
2151         * ipa-prop.c (ipa_bit_ggc_hash_traits::empty_zero_p): Likewise.
2152         (ipa_vr_ggc_hash_traits::empty_zero_p): Likewise.
2153         * profile.c (location_triplet_hash::empty_zero_p): Likewise.
2154         * sanopt.c (sanopt_tree_triplet_hash::empty_zero_p): Likewise.
2155         (sanopt_tree_couple_hash::empty_zero_p): Likewise.
2156         * tree-hasher.h (int_tree_hasher::empty_zero_p): Likewise.
2157         * tree-ssa-sccvn.c (vn_ssa_aux_hasher::empty_zero_p): Likewise.
2158         * tree-vect-slp.c (bst_traits::empty_zero_p): Likewise.
2159         * tree-vectorizer.h
2160         (default_hash_traits<scalar_cond_masked_key>::empty_zero_p):
2161         Likewise.
2163 2020-01-14  Kewen Lin  <linkw@gcc.gnu.org>
2165         * cfgloopanal.c (average_num_loop_insns): Free bbs when early return,
2166         fix typo on return value.
2168 2020-01-14  Xiong Hu Luo  <luoxhu@linux.ibm.com>
2170         PR ipa/69678
2171         * cgraph.c (symbol_table::create_edge): Init speculative_id and
2172         target_prob.
2173         (cgraph_edge::make_speculative): Add param for setting speculative_id
2174         and target_prob.
2175         (cgraph_edge::speculative_call_info): Update comments and find reference
2176         by speculative_id for multiple indirect targets.
2177         (cgraph_edge::resolve_speculation): Decrease the speculations
2178         for indirect edge, drop it's speculative if not direct target
2179         left. Update comments.
2180         (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
2181         (cgraph_node::dump): Print num_speculative_call_targets.
2182         (cgraph_node::verify_node): Don't report error if speculative
2183         edge not include statement.
2184         (cgraph_edge::num_speculative_call_targets_p): New function.
2185         * cgraph.h (int common_target_id): Remove.
2186         (int common_target_probability): Remove.
2187         (num_speculative_call_targets): New variable.
2188         (make_speculative): Add param for setting speculative_id.
2189         (cgraph_edge::num_speculative_call_targets_p): New declare.
2190         (target_prob): New variable.
2191         (speculative_id): New variable.
2192         * ipa-fnsummary.c (analyze_function_body): Create and duplicate
2193           call summaries for multiple speculative call targets.
2194         * cgraphclones.c (cgraph_node::create_clone): Clone speculative_id.
2195         * ipa-profile.c (struct speculative_call_target): New struct.
2196         (class speculative_call_summary): New class.
2197         (class speculative_call_summaries): New class.
2198         (call_sums): New variable.
2199         (ipa_profile_generate_summary): Generate indirect multiple targets summaries.
2200         (ipa_profile_write_edge_summary): New function.
2201         (ipa_profile_write_summary): Stream out indirect multiple targets summaries.
2202         (ipa_profile_dump_all_summaries): New function.
2203         (ipa_profile_read_edge_summary): New function.
2204         (ipa_profile_read_summary_section): New function.
2205         (ipa_profile_read_summary): Stream in indirect multiple targets summaries.
2206         (ipa_profile): Generate num_speculative_call_targets from
2207         profile summaries.
2208         * ipa-ref.h (speculative_id): New variable.
2209         * ipa-utils.c (ipa_merge_profiles): Update with target_prob.
2210         * lto-cgraph.c (lto_output_edge): Remove indirect common_target_id and
2211         common_target_probability.   Stream out speculative_id and
2212         num_speculative_call_targets.
2213         (input_edge): Likewise.
2214         * predict.c (dump_prediction): Remove edges count assert to be
2215         precise.
2216         * symtab.c (symtab_node::create_reference): Init speculative_id.
2217         (symtab_node::clone_references): Clone speculative_id.
2218         (symtab_node::clone_referring): Clone speculative_id.
2219         (symtab_node::clone_reference): Clone speculative_id.
2220         (symtab_node::clear_stmts_in_references): Clear speculative_id.
2221         * tree-inline.c (copy_bb): Duplicate all the speculative edges
2222         if indirect call contains multiple speculative targets.
2223         * value-prof.h  (check_ic_target): Remove.
2224         * value-prof.c  (gimple_value_profile_transformations):
2225         Use void function gimple_ic_transform.
2226         * value-prof.c  (gimple_ic_transform): Handle topn case.
2227         Fix comment typos.  Change it to a void function.
2229 2020-01-13  Andrew Pinski  <apinski@marvell.com>
2231         * config/aarch64/aarch64-cores.def (octeontx2): New define.
2232         (octeontx2t98): New define.
2233         (octeontx2t96): New define.
2234         (octeontx2t93): New define.
2235         (octeontx2f95): New define.
2236         (octeontx2f95n): New define.
2237         (octeontx2f95mm): New define.
2238         * config/aarch64/aarch64-tune.md: Regenerate.
2239         * doc/invoke.texi (-mcpu=): Document the new cpu types.
2241 2020-01-13  Jason Merrill  <jason@redhat.com>
2243         PR c++/33799 - destroy return value if local cleanup throws.
2244         * gimplify.c (gimplify_return_expr): Handle COMPOUND_EXPR.
2246 2020-01-13  Martin Liska  <mliska@suse.cz>
2248         * ipa-cp.c (get_max_overall_size): Use newly
2249         renamed param param_ipa_cp_unit_growth.
2250         * params.opt: Remove legacy param name.
2252 2020-01-13  Martin Sebor  <msebor@redhat.com>
2254         PR tree-optimization/93213
2255         * tree-ssa-strlen.c (handle_store): Only allow single-byte nul-over-nul
2256         stores to be eliminated.
2258 2020-01-13  Martin Liska  <mliska@suse.cz>
2260         * opts.c (print_help): Do not print CL_PARAM
2261         and CL_WARNING for CL_OPTIMIZATION.
2263 2020-01-13  Jonathan Wakely  <jwakely@redhat.com>
2265         PR driver/92757
2266         * doc/invoke.texi (Warning Options): Add caveat about some warnings
2267         depending on optimization settings.
2269 2020-01-13  Jakub Jelinek  <jakub@redhat.com>
2271         PR tree-optimization/90838
2272         * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
2273         SCALAR_INT_TYPE_MODE directly in CTZ_DEFINED_VALUE_AT_ZERO macro
2274         argument rather than to initialize temporary for targets that
2275         don't use the mode argument at all.  Initialize ctzval to avoid
2276         warning at -O0.
2278 2020-01-10  Thomas Schwinge  <thomas@codesourcery.com>
2280         * tree.h (OMP_CLAUSE_USE_DEVICE_PTR_IF_PRESENT): New definition.
2281         * tree-core.h: Document it.
2282         * gimplify.c (gimplify_omp_workshare): Set it.
2283         * omp-low.c (lower_omp_target): Use it.
2284         * tree-pretty-print.c (dump_omp_clause): Print it.
2286         * omp-low.c (lower_omp_target) <OMP_CLAUSE_USE_DEVICE_PTR etc.>:
2287         Assert that for OpenACC we always have 'GOMP_MAP_USE_DEVICE_PTR'.
2289 2020-01-10  David Malcolm  <dmalcolm@redhat.com>
2291         * Makefile.in (OBJS): Add tree-diagnostic-path.o.
2292         * common.opt (fdiagnostics-path-format=): New option.
2293         (diagnostic_path_format): New enum.
2294         (fdiagnostics-show-path-depths): New option.
2295         * coretypes.h (diagnostic_event_id_t): New forward decl.
2296         * diagnostic-color.c (color_dict): Add "path".
2297         * diagnostic-event-id.h: New file.
2298         * diagnostic-format-json.cc (json_from_expanded_location): Make
2299         non-static.
2300         (json_end_diagnostic): Call context->make_json_for_path if it
2301         exists and the diagnostic has a path.
2302         (diagnostic_output_format_init): Clear context->print_path.
2303         * diagnostic-path.h: New file.
2304         * diagnostic-show-locus.c (colorizer::set_range): Special-case
2305         when printing a run of events in a diagnostic_path so that they
2306         all get the same color.
2307         (layout::m_diagnostic_path_p): New field.
2308         (layout::layout): Initialize it.
2309         (layout::print_any_labels): Don't colorize the label text for an
2310         event in a diagnostic_path.
2311         (gcc_rich_location::add_location_if_nearby): Add
2312         "restrict_to_current_line_spans" and "label" params.  Pass the
2313         former to layout.maybe_add_location_range; pass the latter
2314         when calling add_range.
2315         * diagnostic.c: Include "diagnostic-path.h".
2316         (diagnostic_initialize): Initialize context->path_format and
2317         context->show_path_depths.
2318         (diagnostic_show_any_path): New function.
2319         (diagnostic_path::interprocedural_p): New function.
2320         (diagnostic_report_diagnostic): Call diagnostic_show_any_path.
2321         (simple_diagnostic_path::num_events): New function.
2322         (simple_diagnostic_path::get_event): New function.
2323         (simple_diagnostic_path::add_event): New function.
2324         (simple_diagnostic_event::simple_diagnostic_event): New ctor.
2325         (simple_diagnostic_event::~simple_diagnostic_event): New dtor.
2326         (debug): New overload taking a diagnostic_path *.
2327         * diagnostic.def (DK_DIAGNOSTIC_PATH): New.
2328         * diagnostic.h (enum diagnostic_path_format): New enum.
2329         (json::value): New forward decl.
2330         (diagnostic_context::path_format): New field.
2331         (diagnostic_context::show_path_depths): New field.
2332         (diagnostic_context::print_path): New callback field.
2333         (diagnostic_context::make_json_for_path): New callback field.
2334         (diagnostic_show_any_path): New decl.
2335         (json_from_expanded_location): New decl.
2336         * doc/invoke.texi (-fdiagnostics-path-format=): New option.
2337         (-fdiagnostics-show-path-depths): New option.
2338         (-fdiagnostics-color): Add "path" to description of default
2339         GCC_COLORS; describe it.
2340         (-fdiagnostics-format=json): Document how diagnostic paths are
2341         represented in the JSON output format.
2342         * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
2343         Add optional params "restrict_to_current_line_spans" and "label".
2344         * opts.c (common_handle_option): Handle
2345         OPT_fdiagnostics_path_format_ and
2346         OPT_fdiagnostics_show_path_depths.
2347         * pretty-print.c: Include "diagnostic-event-id.h".
2348         (pp_format): Implement "%@" format code for printing
2349         diagnostic_event_id_t *.
2350         (selftest::test_pp_format): Add tests for "%@".
2351         * selftest-run-tests.c (selftest::run_tests): Call
2352         selftest::tree_diagnostic_path_cc_tests.
2353         * selftest.h (selftest::tree_diagnostic_path_cc_tests): New decl.
2354         * toplev.c (general_init): Initialize global_dc->path_format and
2355         global_dc->show_path_depths.
2356         * tree-diagnostic-path.cc: New file.
2357         * tree-diagnostic.c (maybe_unwind_expanded_macro_loc): Make
2358         non-static.  Drop "diagnostic" param in favor of storing the
2359         original value of "where" and re-using it.
2360         (virt_loc_aware_diagnostic_finalizer): Update for dropped param of
2361         maybe_unwind_expanded_macro_loc.
2362         (tree_diagnostics_defaults): Initialize context->print_path and
2363         context->make_json_for_path.
2364         * tree-diagnostic.h (default_tree_diagnostic_path_printer): New
2365         decl.
2366         (default_tree_make_json_for_path): New decl.
2367         (maybe_unwind_expanded_macro_loc): New decl.
2369 2020-01-10  Jakub Jelinek  <jakub@redhat.com>
2371         PR tree-optimization/93210
2372         * fold-const.h (native_encode_initializer,
2373         can_native_interpret_type_p): Declare.
2374         * fold-const.c (native_encode_string): Fix up handling with off != -1,
2375         simplify.
2376         (native_encode_initializer): New function, moved from dwarf2out.c.
2377         Adjust to native_encode_expr compatible arguments, including dry-run
2378         and partial extraction modes.  Don't handle STRING_CST.
2379         (can_native_interpret_type_p): No longer static.
2380         * gimple-fold.c (fold_ctor_reference): For native_encode_expr, verify
2381         offset / BITS_PER_UNIT fits into int and don't call it if
2382         can_native_interpret_type_p fails.  If suboff is NULL and for
2383         CONSTRUCTOR fold_{,non}array_ctor_reference returns NULL, retry with
2384         native_encode_initializer.
2385         (fold_const_aggregate_ref_1): Formatting fix.
2386         * dwarf2out.c (native_encode_initializer): Moved to fold-const.c.
2387         (tree_add_const_value_attribute): Adjust caller.
2389         PR tree-optimization/90838
2390         * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
2391         SCALAR_INT_TYPE_MODE instead of TYPE_MODE as operand of
2392         CTZ_DEFINED_VALUE_AT_ZERO.
2394 2020-01-10  Vladimir Makarov  <vmakarov@redhat.com>
2396         PR inline-asm/93027
2397         * lra-constraints.c (match_reload): Permit input operands have the
2398         same mode as output while other input operands have a different
2399         mode.
2401 2020-01-10  Wilco Dijkstra  <wdijkstr@arm.com>
2403         PR tree-optimization/90838
2404         * tree-ssa-forwprop.c (check_ctz_array): Add new function.
2405         (check_ctz_string): Likewise.
2406         (optimize_count_trailing_zeroes): Likewise.
2407         (simplify_count_trailing_zeroes): Likewise.
2408         (pass_forwprop::execute): Try ctz simplification.
2409         * match.pd: Add matching for ctz idioms.
2411 2020-01-10  Stam Markianos-Wright  <stam.markianos-wright@arm.com>
2413         * config/aarch64/aarch64.c (aarch64_invalid_conversion): New function
2414         for target hook.
2415         (aarch64_invalid_unary_op): New function for target hook.
2416         (aarch64_invalid_binary_op): New function for target hook.
2418 2020-01-10  Stam Markianos-Wright  <stam.markianos-wright@arm.com>
2420         * config.gcc: Add arm_bf16.h.
2421         * config/aarch64/aarch64-builtins.c
2422         (aarch64_simd_builtin_std_type): Add BFmode.
2423         (aarch64_init_simd_builtin_types): Define element types for vector
2424         types.
2425         (aarch64_init_bf16_types): New function.
2426         (aarch64_general_init_builtins): Add arm_init_bf16_types function call.
2427         * config/aarch64/aarch64-modes.def: Add BFmode and V4BF, V8BF vector
2428         modes.
2429         * config/aarch64/aarch64-simd-builtin-types.def: Add BF SIMD types.
2430         * config/aarch64/aarch64-simd.md: Add BF vector types to NEON move
2431         patterns.
2432         * config/aarch64/aarch64.h (AARCH64_VALID_SIMD_DREG_MODE): Add V4BF.
2433         (AARCH64_VALID_SIMD_QREG_MODE): Add V8BF.
2434         * config/aarch64/aarch64.c
2435         (aarch64_classify_vector_mode): Add support for BF types.
2436         (aarch64_gimplify_va_arg_expr): Add support for BF types.
2437         (aarch64_vq_mode): Add support for BF types.
2438         (aarch64_simd_container_mode): Add support for BF types.
2439         (aarch64_mangle_type): Add support for BF scalar type.
2440         * config/aarch64/aarch64.md: Add BFmode to movhf pattern.
2441         * config/aarch64/arm_bf16.h: New file.
2442         * config/aarch64/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
2443         * config/aarch64/iterators.md: Add BF types to mode attributes.
2444         (HFBF, GPF_TF_F16_MOV, VDMOV, VQMOV, VQMOV_NO2Em VALL_F16MOV): New.
2446 2020-01-10  Jason Merrill  <jason@redhat.com>
2448         PR c++/93173 - incorrect tree sharing.
2449         * gimplify.c (copy_if_shared): No longer static.
2450         * gimplify.h: Declare it.
2452 2020-01-10  Richard Sandiford  <richard.sandiford@arm.com>
2454         * doc/invoke.texi (-msve-vector-bits=): Document that
2455         -msve-vector-bits=128 now generates VL-specific code for
2456         little-endian targets.
2457         * config/aarch64/aarch64-sve-builtins.cc (register_builtin_types): Use
2458         build_vector_type_for_mode to construct the data vector types.
2459         * config/aarch64/aarch64.c (aarch64_convert_sve_vector_bits): Generate
2460         VL-specific code for -msve-vector-bits=128 on little-endian targets.
2461         (aarch64_simd_container_mode): Always prefer Advanced SIMD modes
2462         for 128-bit vectors.
2464 2020-01-10  Richard Sandiford  <richard.sandiford@arm.com>
2466         * config/aarch64/aarch64.c (aarch64_evpc_sel): Fix gen_vcond_mask
2467         invocation.
2469 2020-01-10  Richard Sandiford  <richard.sandiford@arm.com>
2471         * config/aarch64/aarch64-builtins.c
2472         (aarch64_builtin_vectorized_function): Check for specific vector modes,
2473         rather than checking the number of elements and the element mode.
2475 2020-01-10  Richard Sandiford  <richard.sandiford@arm.com>
2477         * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
2478         get_related_vectype_for_scalar_type rather than build_vector_type
2479         to create the index type for a conditional reduction.
2481 2020-01-10  Richard Sandiford  <richard.sandiford@arm.com>
2483         * tree-vect-loop.c (update_epilogue_loop_vinfo): Update DR_REF
2484         for any type of gather or scatter, including strided accesses.
2486 2020-01-10  Andre Vieira  <andre.simoesdiasvieira@arm.com>
2488         * tree-vectorizer.h (get_dr_vinfo_offset): Add missing function
2489          comment.
2491 2020-01-10  Andre Vieira  <andre.simoesdiasvieira@arm.com>
2493         * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref): Use
2494         get_dr_vinfo_offset
2495         * tree-vect-loop.c (update_epilogue_loop_vinfo):  Remove orig_drs_init
2496         parameter and its use to reset DR_OFFSET's.
2497         (vect_transform_loop): Remove orig_drs_init argument.
2498         * tree-vect-loop-manip.c (vect_update_init_of_dr): Update the offset
2499         member of dr_vec_info rather than the offset of the associated
2500         data_reference's innermost_loop_behavior.
2501         (vect_update_init_of_dr): Pass dr_vec_info instead of data_reference.
2502         (vect_do_peeling): Remove orig_drs_init parameter and its construction.
2503         * tree-vect-stmts.c (check_scan_store): Replace use of DR_OFFSET with
2504         get_dr_vinfo_offset.
2505         (vectorizable_store): Likewise.
2506         (vectorizable_load): Likewise.
2508 2020-01-10  Richard Biener  <rguenther@suse.de>
2510         * gimple-ssa-store-merging
2511         (pass_store_merging::terminate_all_aliasing_chains): Cache alias info.
2513 2020-01-10  Martin Liska  <mliska@suse.cz>
2515         PR ipa/93217
2516         * ipa-inline-analysis.c (offline_size): Make proper parenthesis
2517         encapsulation that was there before r280040.
2519 2020-01-10  Richard Biener  <rguenther@suse.de>
2521         PR middle-end/93199
2522         * tree-eh.c (sink_clobbers): Move clobbers to out-of-IL
2523         sequences to avoid walking them again for secondary opportunities.
2524         (pass_lower_eh_dispatch::execute): Instead actually insert
2525         them here.
2527 2020-01-10  Richard Biener  <rguenther@suse.de>
2529         PR middle-end/93199
2530         * tree-eh.c (redirect_eh_edge_1): Avoid some work if possible.
2531         (cleanup_all_empty_eh): Walk landing pads in reverse order to
2532         avoid quadraticness.
2534 2020-01-10  Martin Jambor  <mjambor@suse.cz>
2536         * params.opt (param_ipa_sra_max_replacements): Mark as Optimization.
2537         * ipa-sra.c (pull_accesses_from_callee): New parameter caller, use it
2538         to get param_ipa_sra_max_replacements.
2539         (param_splitting_across_edge): Pass the caller to
2540         pull_accesses_from_callee.
2542 2020-01-10  Martin Jambor  <mjambor@suse.cz>
2544         * params.opt (param_ipcp_unit_growth): Mark as Optimization.
2545         * ipa-cp.c (max_new_size): Removed.
2546         (orig_overall_size): New variable.
2547         (get_max_overall_size): New function.
2548         (estimate_local_effects): Use it.  Adjust dump.
2549         (decide_about_value): Likewise.
2550         (ipcp_propagate_stage): Do not calculate max_new_size, just store
2551         orig_overall_size.  Adjust dump.
2552         (ipa_cp_c_finalize): Clear orig_overall_size instead of max_new_size.
2554 2020-01-10  Martin Jambor  <mjambor@suse.cz>
2556         * params.opt (param_ipa_max_agg_items): Mark as Optimization
2557         * ipa-cp.c (merge_agg_lats_step): New parameter max_agg_items, use
2558         instead of param_ipa_max_agg_items.
2559         (merge_aggregate_lattices): Extract param_ipa_max_agg_items from
2560         optimization info for the callee.
2562 2020-01-09  Kwok Cheung Yeung  <kcy@codesourcery.com>
2564         * lto-streamer-in.c (input_function): Remove streamed-in inline debug
2565         markers if debug_inline_points is false.
2567 2020-01-09  Richard Sandiford  <richard.sandiford@arm.com>
2569         * config.gcc (aarch64*-*-*): Add aarch64-sve-builtins-sve2.o to
2570         extra_objs.
2571         * config/aarch64/t-aarch64 (aarch64-sve-builtins.o): Depend on
2572         aarch64-sve-builtins-base.def, aarch64-sve-builtins-sve2.def and
2573         aarch64-sve-builtins-sve2.h.
2574         (aarch64-sve-builtins-sve2.o): New rule.
2575         * config/aarch64/aarch64.h (AARCH64_ISA_SVE2_AES): New macro.
2576         (AARCH64_ISA_SVE2_BITPERM, AARCH64_ISA_SVE2_SHA3): Likewise.
2577         (AARCH64_ISA_SVE2_SM4, TARGET_SVE2_AES, TARGET_SVE2_BITPERM): Likewise.
2578         (TARGET_SVE2_SHA, TARGET_SVE2_SM4): Likewise.
2579         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
2580         TARGET_SVE2_AES, TARGET_SVE2_BITPERM, TARGET_SVE2_SHA3 and
2581         TARGET_SVE2_SM4.
2582         * config/aarch64/aarch64-sve.md: Update comments with SVE2
2583         instructions that are handled here.
2584         (@cond_asrd<mode>): Generalize to...
2585         (@cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>): ...this.
2586         (*cond_asrd<mode>_2): Generalize to...
2587         (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_2): ...this.
2588         (*cond_asrd<mode>_z): Generalize to...
2589         (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_z): ...this.
2590         * config/aarch64/aarch64.md (UNSPEC_LDNT1_GATHER): New unspec.
2591         (UNSPEC_STNT1_SCATTER, UNSPEC_WHILEGE, UNSPEC_WHILEGT): Likewise.
2592         (UNSPEC_WHILEHI, UNSPEC_WHILEHS): Likewise.
2593         * config/aarch64/aarch64-sve2.md (@aarch64_gather_ldnt<mode>): New
2594         pattern.
2595         (@aarch64_gather_ldnt_<ANY_EXTEND:optab><SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
2596         (@aarch64_scatter_stnt<mode>): Likewise.
2597         (@aarch64_scatter_stnt_<SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
2598         (@aarch64_mul_lane_<mode>): Likewise.
2599         (@aarch64_sve_suqadd<mode>_const): Likewise.
2600         (*<sur>h<addsub><mode>): Generalize to...
2601         (@aarch64_pred_<SVE2_COND_INT_BINARY_REV:sve_int_op><mode>): ...this
2602         new pattern.
2603         (@cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>): New expander.
2604         (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_2): New pattern.
2605         (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_3): Likewise.
2606         (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_any): Likewise.
2607         (*cond_<SVE2_COND_INT_BINARY_NOREV:sve_int_op><mode>_z): Likewise.
2608         (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op><mode>):: Likewise.
2609         (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op>_lane_<mode>): Likewise.
2610         (@aarch64_pred_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): Likewise.
2611         (@cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): New expander.
2612         (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_2): New pattern.
2613         (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_3): Likewise.
2614         (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_any): Likewise.
2615         (@aarch64_sve_<SVE2_INT_TERNARY:sve_int_op><mode>): Likewise.
2616         (@aarch64_sve_<SVE2_INT_TERNARY_LANE:sve_int_op>_lane_<mode>)
2617         (@aarch64_sve_add_mul_lane_<mode>): Likewise.
2618         (@aarch64_sve_sub_mul_lane_<mode>): Likewise.
2619         (@aarch64_sve2_xar<mode>): Likewise.
2620         (@aarch64_sve2_bcax<mode>): Likewise.
2621         (*aarch64_sve2_eor3<mode>): Rename to...
2622         (@aarch64_sve2_eor3<mode>): ...this.
2623         (@aarch64_sve2_bsl<mode>): New expander.
2624         (@aarch64_sve2_nbsl<mode>): Likewise.
2625         (@aarch64_sve2_bsl1n<mode>): Likewise.
2626         (@aarch64_sve2_bsl2n<mode>): Likewise.
2627         (@aarch64_sve_add_<SHIFTRT:sve_int_op><mode>): Likewise.
2628         (*aarch64_sve2_sra<mode>): Add MOVPRFX support.
2629         (@aarch64_sve_add_<VRSHR_N:sve_int_op><mode>): New pattern.
2630         (@aarch64_sve_<SVE2_INT_SHIFT_INSERT:sve_int_op><mode>): Likewise.
2631         (@aarch64_sve2_<USMAX:su>aba<mode>): New expander.
2632         (*aarch64_sve2_<USMAX:su>aba<mode>): New pattern.
2633         (@aarch64_sve_<SVE2_INT_BINARY_WIDE:sve_int_op><mode>): Likewise.
2634         (<su>mull<bt><Vwide>): Generalize to...
2635         (@aarch64_sve_<SVE2_INT_BINARY_LONG:sve_int_op><mode>): ...this new
2636         pattern.
2637         (@aarch64_sve_<SVE2_INT_BINARY_LONG_lANE:sve_int_op>_lane_<mode>)
2638         (@aarch64_sve_<SVE2_INT_SHIFT_IMM_LONG:sve_int_op><mode>)
2639         (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG:sve_int_op><mode>)
2640         (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
2641         (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG:sve_int_op><mode>)
2642         (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
2643         (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG:sve_int_op><mode>)
2644         (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
2645         (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG:sve_int_op><mode>)
2646         (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
2647         (@aarch64_sve_<SVE2_FP_TERNARY_LONG:sve_fp_op><mode>): New patterns.
2648         (@aarch64_<SVE2_FP_TERNARY_LONG_LANE:sve_fp_op>_lane_<mode>)
2649         (@aarch64_sve_<SVE2_INT_UNARY_NARROWB:sve_int_op><mode>): Likewise.
2650         (@aarch64_sve_<SVE2_INT_UNARY_NARROWT:sve_int_op><mode>): Likewise.
2651         (@aarch64_sve_<SVE2_INT_BINARY_NARROWB:sve_int_op><mode>): Likewise.
2652         (@aarch64_sve_<SVE2_INT_BINARY_NARROWT:sve_int_op><mode>): Likewise.
2653         (<SHRNB:r>shrnb<mode>): Generalize to...
2654         (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWB:sve_int_op><mode>): ...this
2655         new pattern.
2656         (<SHRNT:r>shrnt<mode>): Generalize to...
2657         (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWT:sve_int_op><mode>): ...this
2658         new pattern.
2659         (@aarch64_pred_<SVE2_INT_BINARY_PAIR:sve_int_op><mode>): New pattern.
2660         (@aarch64_pred_<SVE2_FP_BINARY_PAIR:sve_fp_op><mode>): Likewise.
2661         (@cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>): New expander.
2662         (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_2): New pattern.
2663         (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_z): Likewise.
2664         (@aarch64_sve_<SVE2_INT_CADD:optab><mode>): Likewise.
2665         (@aarch64_sve_<SVE2_INT_CMLA:optab><mode>): Likewise.
2666         (@aarch64_<SVE2_INT_CMLA:optab>_lane_<mode>): Likewise.
2667         (@aarch64_sve_<SVE2_INT_CDOT:optab><mode>): Likewise.
2668         (@aarch64_<SVE2_INT_CDOT:optab>_lane_<mode>): Likewise.
2669         (@aarch64_pred_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): Likewise.
2670         (@cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New expander.
2671         (*cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New pattern.
2672         (@aarch64_sve2_cvtnt<mode>): Likewise.
2673         (@aarch64_pred_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): Likewise.
2674         (@cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): New expander.
2675         (*cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>_any): New pattern.
2676         (@aarch64_sve2_cvtxnt<mode>): Likewise.
2677         (@aarch64_pred_<SVE2_U32_UNARY:sve_int_op><mode>): Likewise.
2678         (@cond_<SVE2_U32_UNARY:sve_int_op><mode>): New expander.
2679         (*cond_<SVE2_U32_UNARY:sve_int_op><mode>): New pattern.
2680         (@aarch64_pred_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): Likewise.
2681         (@cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New expander.
2682         (*cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New pattern.
2683         (@aarch64_sve2_pmul<mode>): Likewise.
2684         (@aarch64_sve_<SVE2_PMULL:optab><mode>): Likewise.
2685         (@aarch64_sve_<SVE2_PMULL_PAIR:optab><mode>): Likewise.
2686         (@aarch64_sve2_tbl2<mode>): Likewise.
2687         (@aarch64_sve2_tbx<mode>): Likewise.
2688         (@aarch64_sve_<SVE2_INT_BITPERM:sve_int_op><mode>): Likewise.
2689         (@aarch64_sve2_histcnt<mode>): Likewise.
2690         (@aarch64_sve2_histseg<mode>): Likewise.
2691         (@aarch64_pred_<SVE2_MATCH:sve_int_op><mode>): Likewise.
2692         (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_cc): Likewise.
2693         (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_ptest): Likewise.
2694         (aarch64_sve2_aes<CRYPTO_AES:aes_op>): Likewise.
2695         (aarch64_sve2_aes<CRYPTO_AESMC:aesmc_op>): Likewise.
2696         (*aarch64_sve2_aese_fused, *aarch64_sve2_aesd_fused): Likewise.
2697         (aarch64_sve2_rax1, aarch64_sve2_sm4e, aarch64_sve2_sm4ekey): Likewise.
2698         (<su>mulh<r>s<mode>3): Update after above pattern name changes.
2699         * config/aarch64/iterators.md (VNx16QI_ONLY, VNx4SF_ONLY)
2700         (SVE_STRUCT2, SVE_FULL_BHI, SVE_FULL_HSI, SVE_FULL_HDI)
2701         (SVE2_PMULL_PAIR_I): New mode iterators.
2702         (UNSPEC_ADCLB, UNSPEC_ADCLT, UNSPEC_ADDHNB, UNSPEC_ADDHNT, UNSPEC_BDEP)
2703         (UNSPEC_BEXT, UNSPEC_BGRP, UNSPEC_CADD90, UNSPEC_CADD270, UNSPEC_CDOT)
2704         (UNSPEC_CDOT90, UNSPEC_CDOT180, UNSPEC_CDOT270, UNSPEC_CMLA)
2705         (UNSPEC_CMLA90, UNSPEC_CMLA180, UNSPEC_CMLA270, UNSPEC_COND_FCVTLT)
2706         (UNSPEC_COND_FCVTNT, UNSPEC_COND_FCVTX, UNSPEC_COND_FCVTXNT)
2707         (UNSPEC_COND_FLOGB, UNSPEC_EORBT, UNSPEC_EORTB, UNSPEC_FADDP)
2708         (UNSPEC_FMAXP, UNSPEC_FMAXNMP, UNSPEC_FMLALB, UNSPEC_FMLALT)
2709         (UNSPEC_FMLSLB, UNSPEC_FMLSLT, UNSPEC_FMINP, UNSPEC_FMINNMP)
2710         (UNSPEC_HISTCNT, UNSPEC_HISTSEG, UNSPEC_MATCH, UNSPEC_NMATCH)
2711         (UNSPEC_PMULLB, UNSPEC_PMULLB_PAIR, UNSPEC_PMULLT, UNSPEC_PMULLT_PAIR)
2712         (UNSPEC_RADDHNB, UNSPEC_RADDHNT, UNSPEC_RSUBHNB, UNSPEC_RSUBHNT)
2713         (UNSPEC_SLI, UNSPEC_SRI, UNSPEC_SABDLB, UNSPEC_SABDLT, UNSPEC_SADDLB)
2714         (UNSPEC_SADDLBT, UNSPEC_SADDLT, UNSPEC_SADDWB, UNSPEC_SADDWT)
2715         (UNSPEC_SBCLB, UNSPEC_SBCLT, UNSPEC_SMAXP, UNSPEC_SMINP)
2716         (UNSPEC_SQCADD90, UNSPEC_SQCADD270, UNSPEC_SQDMULLB, UNSPEC_SQDMULLBT)
2717         (UNSPEC_SQDMULLT, UNSPEC_SQRDCMLAH, UNSPEC_SQRDCMLAH90)
2718         (UNSPEC_SQRDCMLAH180, UNSPEC_SQRDCMLAH270, UNSPEC_SQRSHRNB)
2719         (UNSPEC_SQRSHRNT, UNSPEC_SQRSHRUNB, UNSPEC_SQRSHRUNT, UNSPEC_SQSHRNB)
2720         (UNSPEC_SQSHRNT, UNSPEC_SQSHRUNB, UNSPEC_SQSHRUNT, UNSPEC_SQXTNB)
2721         (UNSPEC_SQXTNT, UNSPEC_SQXTUNB, UNSPEC_SQXTUNT, UNSPEC_SSHLLB)
2722         (UNSPEC_SSHLLT, UNSPEC_SSUBLB, UNSPEC_SSUBLBT, UNSPEC_SSUBLT)
2723         (UNSPEC_SSUBLTB, UNSPEC_SSUBWB, UNSPEC_SSUBWT, UNSPEC_SUBHNB)
2724         (UNSPEC_SUBHNT, UNSPEC_TBL2, UNSPEC_UABDLB, UNSPEC_UABDLT)
2725         (UNSPEC_UADDLB, UNSPEC_UADDLT, UNSPEC_UADDWB, UNSPEC_UADDWT)
2726         (UNSPEC_UMAXP, UNSPEC_UMINP, UNSPEC_UQRSHRNB, UNSPEC_UQRSHRNT)
2727         (UNSPEC_UQSHRNB, UNSPEC_UQSHRNT, UNSPEC_UQXTNB, UNSPEC_UQXTNT)
2728         (UNSPEC_USHLLB, UNSPEC_USHLLT, UNSPEC_USUBLB, UNSPEC_USUBLT)
2729         (UNSPEC_USUBWB, UNSPEC_USUBWT): New unspecs.
2730         (UNSPEC_SMULLB, UNSPEC_SMULLT, UNSPEC_UMULLB, UNSPEC_UMULLT)
2731         (UNSPEC_SMULHS, UNSPEC_SMULHRS, UNSPEC_UMULHS, UNSPEC_UMULHRS)
2732         (UNSPEC_RSHRNB, UNSPEC_RSHRNT, UNSPEC_SHRNB, UNSPEC_SHRNT): Move
2733         further down file.
2734         (VNARROW, Ventype): New mode attributes.
2735         (Vewtype): Handle VNx2DI.  Fix typo in comment.
2736         (VDOUBLE): New mode attribute.
2737         (sve_lane_con): Handle VNx8HI.
2738         (SVE_INT_UNARY): Include ss_abs and ss_neg for TARGET_SVE2.
2739         (SVE_INT_BINARY): Likewise ss_plus, us_plus, ss_minus and us_minus.
2740         (sve_int_op, sve_int_op_rev): Handle the above codes.
2741         (sve_pred_int_rhs2_operand): Likewise.
2742         (MULLBT, SHRNB, SHRNT): Delete.
2743         (SVE_INT_SHIFT_IMM): New int iterator.
2744         (SVE_WHILE): Add UNSPEC_WHILEGE, UNSPEC_WHILEGT, UNSPEC_WHILEHI
2745         and UNSPEC_WHILEHS for TARGET_SVE2.
2746         (SVE2_U32_UNARY, SVE2_INT_UNARY_NARROWB, SVE2_INT_UNARY_NARROWT)
2747         (SVE2_INT_BINARY, SVE2_INT_BINARY_LANE, SVE2_INT_BINARY_LONG)
2748         (SVE2_INT_BINARY_LONG_LANE, SVE2_INT_BINARY_NARROWB)
2749         (SVE2_INT_BINARY_NARROWT, SVE2_INT_BINARY_PAIR, SVE2_FP_BINARY_PAIR)
2750         (SVE2_INT_BINARY_PAIR_LONG, SVE2_INT_BINARY_WIDE): New int iterators.
2751         (SVE2_INT_SHIFT_IMM_LONG, SVE2_INT_SHIFT_IMM_NARROWB): Likewise.
2752         (SVE2_INT_SHIFT_IMM_NARROWT, SVE2_INT_SHIFT_INSERT, SVE2_INT_CADD)
2753         (SVE2_INT_BITPERM, SVE2_INT_TERNARY, SVE2_INT_TERNARY_LANE): Likewise.
2754         (SVE2_FP_TERNARY_LONG, SVE2_FP_TERNARY_LONG_LANE, SVE2_INT_CMLA)
2755         (SVE2_INT_CDOT, SVE2_INT_ADD_BINARY_LONG, SVE2_INT_QADD_BINARY_LONG)
2756         (SVE2_INT_SUB_BINARY_LONG, SVE2_INT_QSUB_BINARY_LONG): Likewise.
2757         (SVE2_INT_ADD_BINARY_LONG_LANE, SVE2_INT_QADD_BINARY_LONG_LANE)
2758         (SVE2_INT_SUB_BINARY_LONG_LANE, SVE2_INT_QSUB_BINARY_LONG_LANE)
2759         (SVE2_COND_INT_UNARY_FP, SVE2_COND_FP_UNARY_LONG): Likewise.
2760         (SVE2_COND_FP_UNARY_NARROWB, SVE2_COND_INT_BINARY): Likewise.
2761         (SVE2_COND_INT_BINARY_NOREV, SVE2_COND_INT_BINARY_REV): Likewise.
2762         (SVE2_COND_INT_SHIFT, SVE2_MATCH, SVE2_PMULL): Likewise.
2763         (optab): Handle the new unspecs.
2764         (su, r): Remove entries for UNSPEC_SHRNB, UNSPEC_SHRNT, UNSPEC_RSHRNB
2765         and UNSPEC_RSHRNT.
2766         (lr): Handle the new unspecs.
2767         (bt): Delete.
2768         (cmp_op, while_optab_cmp, sve_int_op): Handle the new unspecs.
2769         (sve_int_op_rev, sve_int_add_op, sve_int_qadd_op, sve_int_sub_op)
2770         (sve_int_qsub_op): New int attributes.
2771         (sve_fp_op, rot): Handle the new unspecs.
2772         * config/aarch64/aarch64-sve-builtins.h
2773         (function_resolver::require_matching_pointer_type): Declare.
2774         (function_resolver::resolve_unary): Add an optional boolean argument.
2775         (function_resolver::finish_opt_n_resolution): Add an optional
2776         type_suffix_index argument.
2777         (gimple_folder::redirect_call): Declare.
2778         (gimple_expander::prepare_gather_address_operands): Add an optional
2779         bool parameter.
2780         * config/aarch64/aarch64-sve-builtins.cc: Include
2781         aarch64-sve-builtins-sve2.h.
2782         (TYPES_b_unsigned, TYPES_b_integer, TYPES_bh_integer): New macros.
2783         (TYPES_bs_unsigned, TYPES_hs_signed, TYPES_hs_integer): Likewise.
2784         (TYPES_hd_unsigned, TYPES_hsd_signed): Likewise.
2785         (TYPES_hsd_integer): Use TYPES_hsd_signed.
2786         (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): New macros.
2787         (TYPES_s_unsigned): Likewise.
2788         (TYPES_s_integer): Use TYPES_s_unsigned.
2789         (TYPES_sd_signed, TYPES_sd_unsigned): New macros.
2790         (TYPES_sd_integer): Use them.
2791         (TYPES_d_unsigned): New macro.
2792         (TYPES_d_integer): Use it.
2793         (TYPES_d_data, TYPES_cvt_long, TYPES_cvt_narrow_s): New macros.
2794         (TYPES_cvt_narrow): Likewise.
2795         (DEF_SVE_TYPES_ARRAY): Include the new types macros above.
2796         (preds_mx): New variable.
2797         (function_builder::add_overloaded_function): Allow the new feature
2798         set to be more restrictive than the original one.
2799         (function_resolver::infer_pointer_type): Remove qualifiers from
2800         the pointer type before printing it.
2801         (function_resolver::require_matching_pointer_type): New function.
2802         (function_resolver::resolve_sv_displacement): Handle functions
2803         that don't support 32-bit vector indices or svint32_t vector offsets.
2804         (function_resolver::finish_opt_n_resolution): Take the inferred type
2805         as a separate argument.
2806         (function_resolver::resolve_unary): Optionally treat all forms in
2807         the same way as normal merging functions.
2808         (gimple_folder::redirect_call): New function.
2809         (function_expander::prepare_gather_address_operands): Add an argument
2810         that says whether scaled forms are available.  If they aren't,
2811         handle scaling of vector indices and don't add the extension and
2812         scaling operands.
2813         (function_expander::map_to_unspecs): If aarch64_sve isn't available,
2814         fall back to using cond_* instead.
2815         * config/aarch64/aarch64-sve-builtins-functions.h (rtx_code_function):
2816         Split out the member variables into...
2817         (rtx_code_function_base): ...this new base class.
2818         (rtx_code_function_rotated): Inherit rtx_code_function_base.
2819         (unspec_based_function): Split out the member variables into...
2820         (unspec_based_function_base): ...this new base class.
2821         (unspec_based_function_rotated): Inherit unspec_based_function_base.
2822         (unspec_based_function_exact_insn): New class.
2823         (unspec_based_add_function, unspec_based_add_lane_function)
2824         (unspec_based_lane_function, unspec_based_pred_function)
2825         (unspec_based_qadd_function, unspec_based_qadd_lane_function)
2826         (unspec_based_qsub_function, unspec_based_qsub_lane_function)
2827         (unspec_based_sub_function, unspec_based_sub_lane_function): New
2828         typedefs.
2829         (unspec_based_fused_function): New class.
2830         (unspec_based_mla_function, unspec_based_mls_function): New typedefs.
2831         (unspec_based_fused_lane_function): New class.
2832         (unspec_based_mla_lane_function, unspec_based_mls_lane_function): New
2833         typedefs.
2834         (CODE_FOR_MODE1): New macro.
2835         (fixed_insn_function): New class.
2836         (while_comparison): Likewise.
2837         * config/aarch64/aarch64-sve-builtins-shapes.h (binary_long_lane)
2838         (binary_long_opt_n, binary_narrowb_opt_n, binary_narrowt_opt_n)
2839         (binary_to_uint, binary_wide, binary_wide_opt_n, compare, compare_ptr)
2840         (load_ext_gather_index_restricted, load_ext_gather_offset_restricted)
2841         (load_gather_sv_restricted, shift_left_imm_long): Declare.
2842         (shift_left_imm_to_uint, shift_right_imm_narrowb): Likewise.
2843         (shift_right_imm_narrowt, shift_right_imm_narrowb_to_uint): Likewise.
2844         (shift_right_imm_narrowt_to_uint, store_scatter_index_restricted)
2845         (store_scatter_offset_restricted, tbl_tuple, ternary_long_lane)
2846         (ternary_long_opt_n, ternary_qq_lane_rotate, ternary_qq_rotate)
2847         (ternary_shift_left_imm, ternary_shift_right_imm, ternary_uint)
2848         (unary_convert_narrowt, unary_long, unary_narrowb, unary_narrowt)
2849         (unary_narrowb_to_uint, unary_narrowt_to_uint, unary_to_int): Likewise.
2850         * config/aarch64/aarch64-sve-builtins-shapes.cc (apply_predication):
2851         Also add an initial argument for unary_convert_narrowt, regardless
2852         of the predication type.
2853         (build_32_64): Allow loads and stores to specify MODE_none.
2854         (build_sv_index64, build_sv_uint_offset): New functions.
2855         (long_type_suffix): New function.
2856         (binary_imm_narrowb_base, binary_imm_narrowt_base): New classes.
2857         (binary_imm_long_base, load_gather_sv_base): Likewise.
2858         (shift_right_imm_narrow_wrapper, ternary_shift_imm_base): Likewise.
2859         (ternary_resize2_opt_n_base, ternary_resize2_lane_base): Likewise.
2860         (unary_narrowb_base, unary_narrowt_base): Likewise.
2861         (binary_long_lane_def, binary_long_lane): New shape.
2862         (binary_long_opt_n_def, binary_long_opt_n): Likewise.
2863         (binary_narrowb_opt_n_def, binary_narrowb_opt_n): Likewise.
2864         (binary_narrowt_opt_n_def, binary_narrowt_opt_n): Likewise.
2865         (binary_to_uint_def, binary_to_uint): Likewise.
2866         (binary_wide_def, binary_wide): Likewise.
2867         (binary_wide_opt_n_def, binary_wide_opt_n): Likewise.
2868         (compare_def, compare): Likewise.
2869         (compare_ptr_def, compare_ptr): Likewise.
2870         (load_ext_gather_index_restricted_def,
2871         load_ext_gather_index_restricted): Likewise.
2872         (load_ext_gather_offset_restricted_def,
2873         load_ext_gather_offset_restricted): Likewise.
2874         (load_gather_sv_def): Inherit from load_gather_sv_base.
2875         (load_gather_sv_restricted_def, load_gather_sv_restricted): New shape.
2876         (shift_left_imm_def, shift_left_imm): Likewise.
2877         (shift_left_imm_long_def, shift_left_imm_long): Likewise.
2878         (shift_left_imm_to_uint_def, shift_left_imm_to_uint): Likewise.
2879         (store_scatter_index_restricted_def,
2880         store_scatter_index_restricted): Likewise.
2881         (store_scatter_offset_restricted_def,
2882         store_scatter_offset_restricted): Likewise.
2883         (tbl_tuple_def, tbl_tuple): Likewise.
2884         (ternary_long_lane_def, ternary_long_lane): Likewise.
2885         (ternary_long_opt_n_def, ternary_long_opt_n): Likewise.
2886         (ternary_qq_lane_def): Inherit from ternary_resize2_lane_base.
2887         (ternary_qq_lane_rotate_def, ternary_qq_lane_rotate): New shape
2888         (ternary_qq_opt_n_def): Inherit from ternary_resize2_opt_n_base.
2889         (ternary_qq_rotate_def, ternary_qq_rotate): New shape.
2890         (ternary_shift_left_imm_def, ternary_shift_left_imm): Likewise.
2891         (ternary_shift_right_imm_def, ternary_shift_right_imm): Likewise.
2892         (ternary_uint_def, ternary_uint): Likewise.
2893         (unary_convert): Fix typo in comment.
2894         (unary_convert_narrowt_def, unary_convert_narrowt): New shape.
2895         (unary_long_def, unary_long): Likewise.
2896         (unary_narrowb_def, unary_narrowb): Likewise.
2897         (unary_narrowt_def, unary_narrowt): Likewise.
2898         (unary_narrowb_to_uint_def, unary_narrowb_to_uint): Likewise.
2899         (unary_narrowt_to_uint_def, unary_narrowt_to_uint): Likewise.
2900         (unary_to_int_def, unary_to_int): Likewise.
2901         * config/aarch64/aarch64-sve-builtins-base.cc (unspec_cmla)
2902         (unspec_fcmla, unspec_cond_fcmla, expand_mla_mls_lane): New functions.
2903         (svasrd_impl): Delete.
2904         (svcadd_impl::expand): Handle integer operations too.
2905         (svcmla_impl::expand, svcmla_lane::expand): Likewise, using the
2906         new functions to derive the unspec numbers.
2907         (svmla_svmls_lane_impl): Replace with...
2908         (svmla_lane_impl, svmls_lane_impl): ...these new classes.  Handle
2909         integer operations too.
2910         (svwhile_impl): Rename to...
2911         (svwhilelx_impl): ...this and inherit from while_comparison.
2912         (svasrd): Use unspec_based_function.
2913         (svmla_lane): Use svmla_lane_impl.
2914         (svmls_lane): Use svmls_lane_impl.
2915         (svrecpe, svrsqrte): Handle unsigned integer operations too.
2916         (svwhilele, svwhilelt): Use svwhilelx_impl.
2917         * config/aarch64/aarch64-sve-builtins-sve2.h: New file.
2918         * config/aarch64/aarch64-sve-builtins-sve2.cc: Likewise.
2919         * config/aarch64/aarch64-sve-builtins-sve2.def: Likewise.
2920         * config/aarch64/aarch64-sve-builtins.def: Include
2921         aarch64-sve-builtins-sve2.def.
2923 2020-01-09  Richard Sandiford  <richard.sandiford@arm.com>
2925         * config/aarch64/aarch64-protos.h (aarch64_sve_arith_immediate_p)
2926         (aarch64_sve_sqadd_sqsub_immediate_p): Add a machine_mode argument.
2927         * config/aarch64/aarch64.c (aarch64_sve_arith_immediate_p)
2928         (aarch64_sve_sqadd_sqsub_immediate_p): Likewise.  Handle scalar
2929         immediates as well as vector ones.
2930         * config/aarch64/predicates.md (aarch64_sve_arith_immediate)
2931         (aarch64_sve_sub_arith_immediate, aarch64_sve_qadd_immediate)
2932         (aarch64_sve_qsub_immediate): Update calls accordingly.
2934 2020-01-09  Richard Sandiford  <richard.sandiford@arm.com>
2936         * config/aarch64/aarch64-sve2.md: Add banner comments.
2937         (<su>mulh<r>s<mode>3): Move further up file.
2938         (<su>mull<bt><Vwide>, <r>shrnb<mode>, <r>shrnt<mode>)
2939         (*aarch64_sve2_sra<mode>): Move further down file.
2940         * config/aarch64/t-aarch64 (s-check-sve-md): Check aarch64-sve2.md too.
2942 2020-01-09  Richard Sandiford  <richard.sandiford@arm.com>
2944         * config/aarch64/iterators.md (SVE_WHILE): Add UNSPEC_WHILERW
2945         and UNSPEC_WHILEWR.
2946         (while_optab_cmp): Handle them.
2947         * config/aarch64/aarch64-sve.md
2948         (*while_<while_optab_cmp><GPI:mode><PRED_ALL:mode>_ptest): Make public
2949         and add a "@" marker.
2950         * config/aarch64/aarch64-sve2.md (check_<raw_war>_ptrs<mode>): Use it
2951         instead of gen_aarch64_sve2_while_ptest.
2952         (@aarch64_sve2_while<cmp_op><GPI:mode><PRED_ALL:mode>_ptest): Delete.
2954 2020-01-09  Richard Sandiford  <richard.sandiford@arm.com>
2956         * config/aarch64/aarch64.md (UNSPEC_WHILE_LE): Rename to...
2957         (UNSPEC_WHILELE): ...this.
2958         (UNSPEC_WHILE_LO): Rename to...
2959         (UNSPEC_WHILELO): ...this.
2960         (UNSPEC_WHILE_LS): Rename to...
2961         (UNSPEC_WHILELS): ...this.
2962         (UNSPEC_WHILE_LT): Rename to...
2963         (UNSPEC_WHILELT): ...this.
2964         * config/aarch64/iterators.md (SVE_WHILE): Update accordingly.
2965         (cmp_op, while_optab_cmp): Likewise.
2966         * config/aarch64/aarch64.c (aarch64_sve_move_pred_via_while): Likewise.
2967         * config/aarch64/aarch64-sve-builtins-base.cc (svwhilele): Likewise.
2968         (svwhilelt): Likewise.
2970 2020-01-09  Richard Sandiford  <richard.sandiford@arm.com>
2972         * config/aarch64/aarch64-sve-builtins-shapes.h (unary_count): Delete.
2973         (unary_to_uint): Define.
2974         * config/aarch64/aarch64-sve-builtins-shapes.cc (unary_count_def)
2975         (unary_count): Rename to...
2976         (unary_to_uint_def, unary_to_uint): ...this.
2977         * config/aarch64/aarch64-sve-builtins-base.def: Update accordingly.
2979 2020-01-09  Richard Sandiford  <richard.sandiford@arm.com>
2981         * config/aarch64/aarch64-sve-builtins-functions.h
2982         (code_for_mode_function): New class.
2983         (CODE_FOR_MODE0, QUIET_CODE_FOR_MODE0): New macros.
2984         * config/aarch64/aarch64-sve-builtins-base.cc (svcompact_impl)
2985         (svext_impl, svmul_lane_impl, svsplice_impl, svtmad_impl): Delete.
2986         (svcompact, svext, svsplice): Use QUIET_CODE_FOR_MODE0.
2987         (svmul_lane, svtmad): Use CODE_FOR_MODE0.
2989 2020-01-09  Richard Sandiford  <richard.sandiford@arm.com>
2991         * config/aarch64/iterators.md (addsub): New code attribute.
2992         * config/aarch64/aarch64-simd.md (aarch64_<su_optab><optab><mode>):
2993         Re-express as...
2994         (aarch64_<su_optab>q<addsub><mode>): ...this, making the same change
2995         in the asm string and attributes.  Fix indentation.
2996         * config/aarch64/aarch64-sve.md (@aarch64_<su_optab><optab><mode>):
2997         Re-express as...
2998         (@aarch64_sve_<optab><mode>): ...this.
2999         * config/aarch64/aarch64-sve-builtins.h
3000         (function_expander::expand_signed_unpred_op): Delete.
3001         * config/aarch64/aarch64-sve-builtins.cc
3002         (function_expander::expand_signed_unpred_op): Likewise.
3003         (function_expander::map_to_rtx_codes): If the optab isn't defined,
3004         try using code_for_aarch64_sve instead.
3005         * config/aarch64/aarch64-sve-builtins-base.cc (svqadd_impl): Delete.
3006         (svqsub_impl): Likewise.
3007         (svqadd, svqsub): Use rtx_code_function instead.
3009 2020-01-09  Richard Sandiford  <richard.sandiford@arm.com>
3011         * config/aarch64/iterators.md (SRHSUB, URHSUB): Delete.
3012         (HADDSUB, sur, addsub): Remove them.
3014 2020-01-09  Richard Sandiford  <richard.sandiford@arm.com>
3016         * tree-nrv.c (pass_return_slot::execute): Handle all internal
3017         functions the same way, rather than singling out those that
3018         aren't mapped directly to optabs.
3020 2020-01-09  Richard Sandiford  <richard.sandiford@arm.com>
3022         * target.def (compatible_vector_types_p): New target hook.
3023         * hooks.h (hook_bool_const_tree_const_tree_true): Declare.
3024         * hooks.c (hook_bool_const_tree_const_tree_true): New function.
3025         * doc/tm.texi.in (TARGET_COMPATIBLE_VECTOR_TYPES_P): New hook.
3026         * doc/tm.texi: Regenerate.
3027         * gimple-expr.c: Include target.h.
3028         (useless_type_conversion_p): Use targetm.compatible_vector_types_p.
3029         * config/aarch64/aarch64.c (aarch64_compatible_vector_types_p): New
3030         function.
3031         (TARGET_COMPATIBLE_VECTOR_TYPES_P): Define.
3032         * config/aarch64/aarch64-sve-builtins.cc (gimple_folder::convert_pred):
3033         Use the original predicate if it already has a suitable type.
3035 2020-01-09  Martin Jambor  <mjambor@suse.cz>
3037         * cgraph.h (cgraph_edge): Make remove, set_call_stmt, make_direct,
3038         resolve_speculation and redirect_call_stmt_to_callee static.  Change
3039         return type of set_call_stmt to cgraph_edge *.
3040         * auto-profile.c (afdo_indirect_call): Adjust call to
3041         redirect_call_stmt_to_callee.
3042         * cgraph.c (cgraph_edge::set_call_stmt): Make return cgraph-edge *,
3043         make the this pointer explicit, adjust self-recursive calls and the
3044         call top make_direct.  Return the resulting edge.
3045         (cgraph_edge::remove): Make this pointer explicit.
3046         (cgraph_edge::resolve_speculation): Likewise, adjust call to remove.
3047         (cgraph_edge::make_direct): Likewise, adjust call to
3048         resolve_speculation.
3049         (cgraph_edge::redirect_call_stmt_to_callee): Likewise, also adjust
3050         call to set_call_stmt.
3051         (cgraph_update_edges_for_call_stmt_node): Update call to
3052         set_call_stmt and remove.
3053         * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
3054         Renamed edge to master_edge.  Adjusted calls to set_call_stmt.
3055         (cgraph_node::create_edge_including_clones): Moved "first" definition
3056         of edge to the block where it was used.  Adjusted calls to
3057         set_call_stmt.
3058         (cgraph_node::remove_symbol_and_inline_clones): Adjust call to
3059         cgraph_edge::remove.
3060         * cgraphunit.c (walk_polymorphic_call_targets): Adjusted calls to
3061         make_direct and redirect_call_stmt_to_callee.
3062         * ipa-fnsummary.c (redirect_to_unreachable): Adjust calls to
3063         resolve_speculation and make_direct.
3064         * ipa-inline-transform.c (inline_transform): Adjust call to
3065         redirect_call_stmt_to_callee.
3066         (check_speculations_1):: Adjust call to resolve_speculation.
3067         * ipa-inline.c (resolve_noninline_speculation): Adjust call to
3068         resolve-speculation.
3069         (inline_small_functions): Adjust call to resolve_speculation.
3070         (ipa_inline): Likewise.
3071         * ipa-prop.c (ipa_make_edge_direct_to_target): Adjust call to
3072         make_direct.
3073         * ipa-visibility.c (function_and_variable_visibility): Make iteration
3074         safe with regards to edge removal, adjust calls to
3075         redirect_call_stmt_to_callee.
3076         * ipa.c (walk_polymorphic_call_targets): Adjust calls to make_direct
3077         and redirect_call_stmt_to_callee.
3078         * multiple_target.c (create_dispatcher_calls): Adjust call to
3079         redirect_call_stmt_to_callee
3080         (redirect_to_specific_clone): Likewise.
3081         * tree-cfgcleanup.c (delete_unreachable_blocks_update_callgraph):
3082         Adjust calls to cgraph_edge::remove.
3083         * tree-inline.c (copy_bb): Adjust call to set_call_stmt.
3084         (redirect_all_calls): Adjust call to redirect_call_stmt_to_callee.
3085         (expand_call_inline): Adjust call to cgraph_edge::remove.
3087 2020-01-09  Martin Liska  <mliska@suse.cz>
3089         * params.opt: Set Optimization for
3090         param_max_speculative_devirt_maydefs.
3092 2020-01-09  Martin Sebor  <msebor@redhat.com>
3094         PR middle-end/93200
3095         PR fortran/92956
3096         * builtins.c (compute_objsize): Avoid handling MEM_REFs of vector type.
3098 2020-01-09  Martin Liska  <mliska@suse.cz>
3100         * auto-profile.c (auto_profile): Use opt_for_fn
3101         for a parameter.
3102         * ipa-cp.c (ipcp_lattice::add_value): Likewise.
3103         (propagate_vals_across_arith_jfunc): Likewise.
3104         (hint_time_bonus): Likewise.
3105         (incorporate_penalties): Likewise.
3106         (good_cloning_opportunity_p): Likewise.
3107         (perform_estimation_of_a_value): Likewise.
3108         (estimate_local_effects): Likewise.
3109         (ipcp_propagate_stage): Likewise.
3110         * ipa-fnsummary.c (decompose_param_expr): Likewise.
3111         (set_switch_stmt_execution_predicate): Likewise.
3112         (analyze_function_body): Likewise.
3113         * ipa-inline-analysis.c (offline_size): Likewise.
3114         * ipa-inline.c (early_inliner): Likewise.
3115         * ipa-prop.c (ipa_analyze_node): Likewise.
3116         (ipcp_transform_function): Likewise.
3117         * ipa-sra.c (process_scan_results): Likewise.
3118         (ipa_sra_summarize_function): Likewise.
3119         * params.opt: Rename ipcp-unit-growth to
3120         ipa-cp-unit-growth.  Add Optimization for various
3121         IPA-related parameters.
3123 2020-01-09  Richard Biener  <rguenther@suse.de>
3125         PR middle-end/93054
3126         * gimplify.c (gimplify_expr): Deal with NOP definitions.
3128 2020-01-09  Richard Biener  <rguenther@suse.de>
3130         PR tree-optimization/93040
3131         * gimple-ssa-store-merging.c (find_bswap_or_nop): Raise search limit.
3133 2020-01-09  Georg-Johann Lay  <avr@gjlay.de>
3135         * common/config/avr/avr-common.c (avr_option_optimization_table)
3136         [OPT_LEVELS_1_PLUS]: Set -fsplit-wide-types-early.
3138 2020-01-09  Martin Liska  <mliska@suse.cz>
3140         * cgraphclones.c (symbol_table::materialize_all_clones):
3141         Use cgraph_node::dump_name.
3143 2020-01-09  Jakub Jelinek  <jakub@redhat.com>
3145         PR inline-asm/93202
3146         * config/riscv/riscv.c (riscv_print_operand_reloc): Use
3147         output_operand_lossage instead of gcc_unreachable.
3148         * doc/md.texi (riscv f constraint): Fix typo.
3150         PR target/93141
3151         * config/i386/i386.md (subv<mode>4): Use SWIDWI iterator instead of
3152         SWI.  Use <general_hilo_operand> instead of <general_operand>.  Use
3153         CONST_SCALAR_INT_P instead of CONST_INT_P.
3154         (*subv<mode>4_1): Rename to ...
3155         (subv<mode>4_1): ... this.
3156         (*subv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
3157         define_insn_and_split patterns.
3158         (*subv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
3159         patterns.
3161 2020-01-08  David Malcolm  <dmalcolm@redhat.com>
3163         * vec.c (class selftest::count_dtor): New class.
3164         (selftest::test_auto_delete_vec): New test.
3165         (selftest::vec_c_tests): Call it.
3166         * vec.h (class auto_delete_vec): New class template.
3167         (auto_delete_vec<T>::~auto_delete_vec): New dtor.
3169 2020-01-08  David Malcolm  <dmalcolm@redhat.com>
3171         * sbitmap.h (auto_sbitmap): Add operator const_sbitmap.
3173 2020-01-08  Jim Wilson  <jimw@sifive.com>
3175         * config/riscv/riscv.c (riscv_legitimize_tls_address): Ifdef out
3176         use of TLS_MODEL_LOCAL_EXEC when not pic.
3178 2020-01-08  David Malcolm  <dmalcolm@redhat.com>
3180         * hash-map-tests.c (selftest::test_map_of_strings_to_int): Fix
3181         memory leak.
3183 2020-01-08  Jakub Jelinek  <jakub@redhat.com>
3185         PR target/93187
3186         * config/i386/i386.md (*stack_protect_set_2_<mode> peephole2,
3187         *stack_protect_set_3 peephole2): Also check that the second
3188         insns source is general_operand.
3190         PR target/93174
3191         * config/i386/i386.md (addcarry<mode>_0): Use nonimmediate_operand
3192         predicate for output operand instead of register_operand.
3193         (addcarry<mode>, addcarry<mode>_1): Likewise.  Add alternative with
3194         memory destination and non-memory operands[2].
3196 2020-01-08  Martin Liska  <mliska@suse.cz>
3198         * cgraph.c (cgraph_node::dump): Use ::dump_name or
3199         ::dump_asm_name instead of (::name or ::asm_name).
3200         * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
3201         * cgraphunit.c (walk_polymorphic_call_targets): Likewise.
3202         (analyze_functions): Likewise.
3203         (expand_all_functions): Likewise.
3204         * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
3205         (propagate_bits_across_jump_function): Likewise.
3206         (dump_profile_updates): Likewise.
3207         (ipcp_store_bits_results): Likewise.
3208         (ipcp_store_vr_results): Likewise.
3209         * ipa-devirt.c (dump_targets): Likewise.
3210         * ipa-fnsummary.c (analyze_function_body): Likewise.
3211         * ipa-hsa.c (check_warn_node_versionable): Likewise.
3212         (process_hsa_functions): Likewise.
3213         * ipa-icf.c (sem_item_optimizer::merge_classes): Likewise.
3214         (set_alias_uids): Likewise.
3215         * ipa-inline-transform.c (save_inline_function_body): Likewise.
3216         * ipa-inline.c (recursive_inlining): Likewise.
3217         (inline_to_all_callers_1): Likewise.
3218         (ipa_inline): Likewise.
3219         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
3220         (ipa_propagate_frequency): Likewise.
3221         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
3222         (remove_described_reference): Likewise.
3223         * ipa-pure-const.c (worse_state): Likewise.
3224         (check_retval_uses): Likewise.
3225         (analyze_function): Likewise.
3226         (propagate_pure_const): Likewise.
3227         (propagate_nothrow): Likewise.
3228         (dump_malloc_lattice): Likewise.
3229         (propagate_malloc): Likewise.
3230         (pass_local_pure_const::execute): Likewise.
3231         * ipa-visibility.c (optimize_weakref): Likewise.
3232         (function_and_variable_visibility): Likewise.
3233         * ipa.c (symbol_table::remove_unreachable_nodes): Likewise.
3234         (ipa_discover_variable_flags): Likewise.
3235         * lto-streamer-out.c (output_function): Likewise.
3236         (output_constructor): Likewise.
3237         * tree-inline.c (copy_bb): Likewise.
3238         * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
3239         * varpool.c (symbol_table::remove_unreferenced_decls): Likewise.
3241 2020-01-08  Richard Biener  <rguenther@suse.de>
3243         PR middle-end/93199
3244         * tree-eh.c (sink_clobbers): Update virtual operands for
3245         the first and last stmt only.  Add a dry-run capability.
3246         (pass_lower_eh_dispatch::execute): Perform clobber sinking
3247         after CFG manipulations and in RPO order to catch all
3248         secondary opportunities reliably.
3250 2020-01-08  Georg-Johann Lay  <avr@gjlay.de>
3252         PR target/93182
3253         * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
3255 2019-01-08  Richard Biener  <rguenther@suse.de>
3257         PR middle-end/93199
3258         * gimple-fold.c (rewrite_to_defined_overflow): Mark stmt modified.
3259         * tree-ssa-loop-im.c (move_computations_worker): Properly adjust
3260         virtual operand, also updating SSA use.
3261         * gimple-loop-interchange.cc (loop_cand::undo_simple_reduction):
3262         Update stmt after resetting virtual operand.
3263         (tree_loop_interchange::move_code_to_inner_loop): Likewise.
3264         * gimple-iterator.c (gsi_remove): When not removing the stmt
3265         permanently do not delink immediate uses or mark the stmt modified.
3267 2020-01-08  Martin Liska  <mliska@suse.cz>
3269         * ipa-fnsummary.c (dump_ipa_call_summary): Use symtab_node::dump_name.
3270         (ipa_call_context::estimate_size_and_time): Likewise.
3271         (inline_analyze_function): Likewise.
3273 2020-01-08  Martin Liska  <mliska@suse.cz>
3275         * cgraph.c (cgraph_node::dump): Use systematically
3276         dump_asm_name.
3278 2020-01-08  Georg-Johann Lay  <avr@gjlay.de>
3280         Add -nodevicespecs option for avr.
3282         PR target/93182
3283         * config/avr/avr.opt (-nodevicespecs): New driver option.
3284         * config/avr/driver-avr.c (avr_devicespecs_file): Only issue
3285         "-specs=device-specs/..." if that option is not set.
3286         * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
3288 2020-01-08  Georg-Johann Lay  <avr@gjlay.de>
3290         Implement 64-bit double functions for avr.
3292         PR target/92055
3293         * config.gcc (tm_defines) [target=avr]: Support --with-libf7,
3294         --with-double-comparison.
3295         * doc/install.texi: Document them.
3296         * config/avr/avr-c.c (avr_cpu_cpp_builtins)
3297         <WITH_LIBF7_LIBGCC, WITH_LIBF7_MATH, WITH_LIBF7_MATH_SYMBOLS>
3298         <WITH_DOUBLE_COMPARISON>: New built-in defines.
3299         * doc/invoke.texi (AVR Built-in Macros): Document them.
3300         * config/avr/avr-protos.h (avr_float_lib_compare_returns_bool): New.
3301         * config/avr/avr.c (avr_float_lib_compare_returns_bool): New function.
3302         * config/avr/avr.h (FLOAT_LIB_COMPARE_RETURNS_BOOL): New macro.
3304 2020-01-08  Richard Earnshaw  <rearnsha@arm.com>
3306         PR target/93188
3307         * config/arm/t-multilib (MULTILIB_MATCHES): Add rules to match
3308         armv7-a{+mp,+sec,+mp+sec} to appropriate armv7 multilib variants
3309         when only building rm-profile multilibs.
3311 2020-01-08  Feng Xue  <fxue@os.amperecomputing.com>
3313         PR ipa/93084
3314         * ipa-cp.c (self_recursively_generated_p): Find matched aggregate
3315         lattice for a value to check.
3316         (propagate_vals_across_arith_jfunc): Add an assertion to ensure
3317         finite propagation in self-recursive scc.
3319 2020-01-08  Luo Xiong Hu  <luoxhu@linux.ibm.com>
3321         * ipa-inline.c (caller_growth_limits): Restore the AND.
3323 2020-01-07  Andrew Stubbs  <ams@codesourcery.com>
3325         * config/gcn/gcn-valu.md (VEC_1REG_INT_ALT): Delete iterator.
3326         (VEC_ALLREG_ALT): New iterator.
3327         (VEC_ALLREG_INT_MODE): New iterator.
3328         (VCMP_MODE): New iterator.
3329         (VCMP_MODE_INT): New iterator.
3330         (vec_cmpu<mode>di): Use VCMP_MODE_INT.
3331         (vec_cmp<u>v64qidi): New define_expand.
3332         (vec_cmp<mode>di_exec): Use VCMP_MODE.
3333         (vec_cmpu<mode>di_exec): New define_expand.
3334         (vec_cmp<u>v64qidi_exec): New define_expand.
3335         (vec_cmp<mode>di_dup): Use VCMP_MODE.
3336         (vec_cmp<mode>di_dup_exec): Use VCMP_MODE.
3337         (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>): Rename ...
3338         (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): ... to this.
3339         (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>_exec): Rename ...
3340         (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): ... to this.
3341         (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>): Rename ...
3342         (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): ... to this.
3343         (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>_exec): Rename ...
3344         (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): ... to
3345         this.
3346         * config/gcn/gcn.c (print_operand): Fix 8 and 16 bit suffixes.
3347         * config/gcn/gcn.md (expander): Add sign_extend and zero_extend.
3349 2020-01-07  Andrew Stubbs  <ams@codesourcery.com>
3351         * config/gcn/constraints.md (DA): Update description and match.
3352         (DB): Likewise.
3353         (Db): New constraint.
3354         * config/gcn/gcn-protos.h (gcn_inline_constant64_p): Add second
3355         parameter.
3356         * config/gcn/gcn.c (gcn_inline_constant64_p): Add 'mixed' parameter.
3357         Implement 'Db' mixed immediate type.
3358         * config/gcn/gcn-valu.md (addcv64si3<exec_vcc>): Rework constraints.
3359         (addcv64si3_dup<exec_vcc>): Delete.
3360         (subcv64si3<exec_vcc>): Rework constraints.
3361         (addv64di3): Rework constraints.
3362         (addv64di3_exec): Rework constraints.
3363         (subv64di3): Rework constraints.
3364         (addv64di3_dup): Delete.
3365         (addv64di3_dup_exec): Delete.
3366         (addv64di3_zext): Rework constraints.
3367         (addv64di3_zext_exec): Rework constraints.
3368         (addv64di3_zext_dup): Rework constraints.
3369         (addv64di3_zext_dup_exec): Rework constraints.
3370         (addv64di3_zext_dup2): Rework constraints.
3371         (addv64di3_zext_dup2_exec): Rework constraints.
3372         (addv64di3_sext_dup2): Rework constraints.
3373         (addv64di3_sext_dup2_exec): Rework constraints.
3375 2020-01-07  Andre Vieira  <andre.simoesdiasvieira@arm.com>
3377         * doc/sourcebuild.texi (arm_little_endian, arm_nothumb): Documented
3378         existing target checks.
3380 2020-01-07  Richard Biener  <rguenther@suse.de>
3382         * doc/install.texi: Bump minimal supported MPC version.
3384 2020-01-07  Richard Sandiford  <richard.sandiford@arm.com>
3386         * langhooks-def.h (lhd_simulate_enum_decl): Declare.
3387         (LANG_HOOKS_SIMULATE_ENUM_DECL): Use it.
3388         * langhooks.c: Include stor-layout.h.
3389         (lhd_simulate_enum_decl): New function.
3390         * config/aarch64/aarch64-sve-builtins.cc (init_builtins): Call
3391         handle_arm_sve_h for the LTO frontend.
3392         (register_vector_type): Cope with null returns from pushdecl.
3394 2020-01-07  Richard Sandiford  <richard.sandiford@arm.com>
3396         * config/aarch64/aarch64-protos.h (aarch64_sve::svbool_type_p)
3397         (aarch64_sve::nvectors_if_data_type): Replace with...
3398         (aarch64_sve::builtin_type_p): ...this.
3399         * config/aarch64/aarch64-sve-builtins.cc: Include attribs.h.
3400         (find_vector_type): Delete.
3401         (add_sve_type_attribute): New function.
3402         (lookup_sve_type_attribute): Likewise.
3403         (register_builtin_types): Add an "SVE type" attribute to each type.
3404         (register_tuple_type): Likewise.
3405         (svbool_type_p, nvectors_if_data_type): Delete.
3406         (mangle_builtin_type): Use lookup_sve_type_attribute.
3407         (builtin_type_p): Likewise.  Add an overload that returns the
3408         number of constituent vector and predicate registers.
3409         * config/aarch64/aarch64.c (aarch64_sve_argument_p): Delete.
3410         (aarch64_returns_value_in_sve_regs_p): Use aarch64_sve::builtin_type_p
3411         instead of aarch64_sve_argument_p.
3412         (aarch64_takes_arguments_in_sve_regs_p): Likewise.
3413         (aarch64_pass_by_reference): Likewise.
3414         (aarch64_function_value_1): Likewise.
3415         (aarch64_return_in_memory): Likewise.
3416         (aarch64_layout_arg): Likewise.
3418 2020-01-07  Jakub Jelinek  <jakub@redhat.com>
3420         PR tree-optimization/93156
3421         * tree-ssa-ccp.c (bit_value_binop): For x * x note that the second
3422         least significant bit is always clear.
3424         PR tree-optimization/93118
3425         * match.pd ((x >> c) << c -> x & (-1<<c)): Add nop_convert?.  Add new
3426         simplifier with two intermediate conversions.
3428 2020-01-07  Martin Liska  <mliska@suse.cz>
3430         * params.opt: Add Optimization for various parameters.
3432 2020-01-07  Martin Liska  <mliska@suse.cz>
3434         PR ipa/83411
3435         * doc/extend.texi: Explain cloning for target_clone
3436         attribute.
3438 2020-01-07  Martin Liska  <mliska@suse.cz>
3440         PR tree-optimization/92860
3441         * common.opt: Make in Optimization option
3442         as it is affected by -O0, which is an Optimization
3443         option.
3444         * tree-inline.c (tree_inlinable_function_p):
3445         Use opt_for_fn for warn_inline.
3446         (expand_call_inline): Likewise.
3448 2020-01-07  Martin Liska  <mliska@suse.cz>
3450         PR tree-optimization/92860
3451         * common.opt: Make flag_ree as optimization
3452         attribute. 
3454 2020-01-07  Martin Liska  <mliska@suse.cz>
3456         PR optimization/92860
3457         * params.opt: Mark param_min_crossjump_insns with Optimization
3458         keyword.
3460 2020-01-07  Luo Xiong Hu  <luoxhu@linux.ibm.com>
3462         * ipa-inline-analysis.c (estimate_growth): Fix typo.
3463         * ipa-inline.c (caller_growth_limits): Use OR instead of AND.
3465 2020-01-06  Michael Meissner  <meissner@linux.ibm.com>
3467         * config/rs6000/rs6000.c (hard_reg_and_mode_to_addr_mask): New
3468         helper function to return the valid addressing formats for a given
3469         hard register and mode.
3470         (rs6000_adjust_vec_address): Call hard_reg_and_mode_to_addr_mask.
3472         * config/rs6000/constraints.md (Q constraint): Update
3473         documentation.
3474         * doc/md.texi (RS/6000 constraints): Update 'Q' cosntraint
3475         documentation.
3477         * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
3478         Use 'Q' for doing vector extract from memory.
3479         (vsx_extract_v4sf_var): Use 'Q' for doing vector extract from
3480         memory.
3481         (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Use 'Q' for
3482         doing vector extract from memory.
3483         (vsx_extract_<mode>_<VS_scalar>mode_var): Use 'Q' for doing vector
3484         extract from memory.
3486         * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add support
3487         for the offset being 34-bits when -mcpu=future is used.
3489 2020-01-06  John David Anglin  <danglin@gcc.gnu.org>
3491         * config/pa/pa.md: Revert change to use ordered_comparison_operator
3492         instead of cmpib_comparison_operator in cmpib patterns.
3493         * config/pa/predicates.md (cmpib_comparison_operator): Revert removal
3494         of cmpib_comparison_operator.  Revise comment.
3496 2020-01-06  Richard Sandiford  <richard.sandiford@arm.com>
3498         * tree-vect-slp.c (vect_build_slp_tree_1): Require all shifts
3499         in an IFN_DIV_POW2 node to be equal.
3501 2020-01-06  Richard Sandiford  <richard.sandiford@arm.com>
3503         * tree-vect-stmts.c (vect_check_load_store_mask): Rename to...
3504         (vect_check_scalar_mask): ...this.
3505         (vectorizable_store, vectorizable_load): Update call accordingly.
3506         (vectorizable_call): Use vect_check_scalar_mask to check the mask
3507         argument in calls to conditional internal functions.
3509 2020-01-06  Andrew Stubbs  <ams@codesourcery.com>
3511         * config/gcn/gcn-valu.md (subv64di3): Use separate alternatives for
3512         '0' matching inputs.
3513         (subv64di3_exec): Likewise.
3515 2020-01-06  Bryan Stenson  <bryan@siliconvortex.com>
3517         * config/mips/mips.c (vr4130_align_insns): Fix typo.
3518         * doc/md.texi (movstr): Likewise.
3520 2020-01-06  Andrew Stubbs  <ams@codesourcery.com>
3522         * config/gcn/gcn-valu.md (vec_extract<mode><scalar_mode>): Add early
3523         clobber.
3525 2020-01-06  Richard Sandiford  <richard.sandiford@arm.com>
3527         * config/aarch64/t-aarch64 ($(srcdir)/config/aarch64/aarch64-tune.md):
3528         Depend on...
3529         (s-aarch64-tune-md): ...this new stamp file.  Pipe the new contents
3530         to a temporary file and use move-if-change to update the real
3531         file where necessary.
3533 2020-01-06  Richard Sandiford  <richard.sandiford@arm.com>
3535         * config/aarch64/aarch64-sve.md (@aarch64_sel_dup<mode>): Use Upl
3536         rather than Upa for CPY /M.
3538 2020-01-06  Andrew Stubbs  <ams@codesourcery.com>
3540         * config/gcn/gcn.c (gcn_inline_constant_p): Allow 64 as an inline
3541         immediate.
3543 2020-01-06  Martin Liska  <mliska@suse.cz>
3545     PR tree-optimization/92860
3546     * params.opt: Mark param_max_combine_insns with Optimization
3547     keyword. 
3549 2020-01-05  Jakub Jelinek  <jakub@redhat.com>
3551         PR target/93141
3552         * config/i386/i386.md (SWIDWI): New mode iterator.
3553         (DWI, dwi): Add TImode variants.
3554         (addv<mode>4): Use SWIDWI iterator instead of SWI.  Use
3555         <general_hilo_operand> instead of <general_operand>.  Use
3556         CONST_SCALAR_INT_P instead of CONST_INT_P.
3557         (*addv<mode>4_1): Rename to ...
3558         (addv<mode>4_1): ... this.
3559         (QWI): New mode attribute.
3560         (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
3561         define_insn_and_split patterns.
3562         (*addv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
3563         patterns.
3564         (uaddv<mode>4): Use SWIDWI iterator instead of SWI.  Use
3565         <general_hilo_operand> instead of <general_operand>.
3566         (*addcarry<mode>_1): New define_insn.
3567         (*add<dwi>3_doubleword_cc_overflow_1): New define_insn_and_split.
3569 2020-01-03  Konstantin Kharlamov  <Hi-Angel@yandex.ru>
3571         * gdbinit.in (pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, pdd, pbs, pbm):
3572         Use "call" instead of "set".
3574 2020-01-03  Martin Jambor  <mjambor@suse.cz>
3576         PR ipa/92917
3577         * ipa-cp.c (print_all_lattices): Skip functions without info.
3579 2020-01-03  Jakub Jelinek  <jakub@redhat.com>
3581         PR target/93089
3582         * config/i386/i386-options.c (ix86_simd_clone_adjust): If
3583         TARGET_PREFER_AVX128, use prefer-vector-width=256 for 'c' and 'd'
3584         simd clones.  If TARGET_PREFER_AVX256, use prefer-vector-width=512
3585         for 'e' simd clones.
3587         PR target/93089
3588         * config/i386/i386.opt (x_prefer_vector_width_type): Remove TargetSave
3589         entry.
3590         (mprefer-vector-width=): Add Save.
3591         * config/i386/i386-options.c (ix86_target_string): Add PVW argument, print
3592         -mprefer-vector-width= if non-zero.  Fix up -mfpmath= comment.
3593         (ix86_debug_options, ix86_function_specific_print): Adjust
3594         ix86_target_string callers.
3595         (ix86_valid_target_attribute_inner_p): Handle prefer-vector-width=.
3596         (ix86_valid_target_attribute_tree): Likewise.
3597         * config/i386/i386-options.h (ix86_target_string): Add PVW argument.
3598         * config/i386/i386-expand.c (ix86_expand_builtin): Adjust
3599         ix86_target_string caller.
3601         PR target/93110
3602         * config/i386/i386.md (abs<mode>2): Use expand_simple_binop instead of
3603         emitting ASHIFTRT, XOR and MINUS by hand.  Use gen_int_mode with QImode
3604         instead of gen_int_shift_amount + convert_modes.
3606         PR rtl-optimization/93088
3607         * loop-iv.c (find_single_def_src): Punt after looking through
3608         128 reg copies for regs with single definitions.  Move definitions
3609         to first uses.
3611 2020-01-02  Dennis Zhang  <dennis.zhang@arm.com>
3613         * config/arm/arm-c.c (arm_cpu_builtins): Define
3614         __ARM_FEATURE_MATMUL_INT8, __ARM_FEATURE_BF16_VECTOR_ARITHMETIC,
3615         __ARM_FEATURE_BF16_SCALAR_ARITHMETIC, and
3616         __ARM_BF16_FORMAT_ALTERNATIVE when enabled.
3617         * config/arm/arm-cpus.in (armv8_6, i8mm, bf16): New features.
3618         * config/arm/arm-tables.opt: Regenerated.
3619         * config/arm/arm.c (arm_option_reconfigure_globals): Initialize
3620         arm_arch_i8mm and arm_arch_bf16 when enabled.
3621         * config/arm/arm.h (TARGET_I8MM): New macro.
3622         (TARGET_BF16_FP, TARGET_BF16_SIMD): Likewise.
3623         * config/arm/t-aprofile: Add matching rules for -march=armv8.6-a.
3624         * config/arm/t-arm-elf (all_v8_archs): Add armv8.6-a.
3625         * config/arm/t-multilib: Add matching rules for -march=armv8.6-a.
3626         (v8_6_a_simd_variants): New.
3627         (v8_*_a_simd_variants): Add i8mm and bf16.
3628         * doc/invoke.texi (armv8.6-a, i8mm, bf16): Document new options.
3630 2020-01-02  Jakub Jelinek  <jakub@redhat.com>
3632         PR ipa/93087
3633         * predict.c (compute_function_frequency): Don't call
3634         warn_function_cold on functions that already have cold attribute.
3636 2020-01-01  John David Anglin  <danglin@gcc.gnu.org>
3638         PR target/67834
3639         * config/pa/pa.c (pa_elf_select_rtx_section): New.  Put references to
3640         COMDAT group function labels in .data.rel.ro.local section.
3641         * config/pa/pa32-linux.h (TARGET_ASM_SELECT_RTX_SECTION): Define.
3643         PR target/93111
3644         * config/pa/pa.md (scc): Use ordered_comparison_operator instead of
3645         comparison_operator in B and S integer comparisons.  Likewise, use
3646         ordered_comparison_operator instead of cmpib_comparison_operator in
3647         cmpib patterns.
3648         * config/pa/predicates.md (cmpib_comparison_operator): Remove.
3650 2020-01-01  Jakub Jelinek  <jakub@redhat.com>
3652         Update copyright years.
3654         * gcc.c (process_command): Update copyright notice dates.
3655         * gcov-dump.c (print_version): Ditto.
3656         * gcov.c (print_version): Ditto.
3657         * gcov-tool.c (print_version): Ditto.
3658         * gengtype.c (create_file): Ditto.
3659         * doc/cpp.texi: Bump @copying's copyright year.
3660         * doc/cppinternals.texi: Ditto.
3661         * doc/gcc.texi: Ditto.
3662         * doc/gccint.texi: Ditto.
3663         * doc/gcov.texi: Ditto.
3664         * doc/install.texi: Ditto.
3665         * doc/invoke.texi: Ditto.
3667 2020-01-01  Jan Hubicka  <hubicka@ucw.cz>
3669         * ipa.c (walk_polymorphic_call_targets): Fix updating of overall
3670         summary.
3672 2020-01-01  Jakub Jelinek  <jakub@redhat.com>
3674         PR tree-optimization/93098
3675         * match.pd (popcount): For shift amounts, use integer_onep
3676         or wi::to_widest () == cst instead of tree_to_uhwi () == cst
3677         tests.  Make sure that precision is power of two larger than or equal
3678         to 16.  Ensure shift is never negative.  Use HOST_WIDE_INT_UC macro
3679         instead of ULL suffixed constants.  Formatting fixes.
3681 Copyright (C) 2020 Free Software Foundation, Inc.
3683 Copying and distribution of this file, with or without modification,
3684 are permitted in any medium without royalty provided the copyright
3685 notice and this notice are preserved.