1 2020-02-08 Jim Wilson <jimw@sifive.com>
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>
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.
21 2020-02-07 Richard Biener <rguenther@suse.de>
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
29 2020-02-07 H.J. Lu <hongjiu.lu@intel.com>
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
35 (ix86_function_arg): Pass arg.type to function_arg_ms_64.
37 2020-02-07 Jakub Jelinek <jakub@redhat.com>
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
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
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>
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
76 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
78 * config/aarch64/aarch64.md (aarch64_movk<mode>): Add a type
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>
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.
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>
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>
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.
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.
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>
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
193 (*ashl<mode>3_doubleword): Ditto.
194 (*<shift_insn><mode>3_doubleword): Ditto.
196 2020-02-05 Michael Meissner <meissner@linux.ibm.com>
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>
208 (Special Functions for Debugging the Analyzer): Update description
209 of __analyzer_dump_exploded_nodes.
211 2020-02-05 Jakub Jelinek <jakub@redhat.com>
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.
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
226 2020-02-05 Richard Biener <rguenther@suse.de>
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.
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
242 2020-02-05 Martin Liska <mliska@suse.cz>
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
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
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>
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>
336 * config/arm/arm-protos.h: New function arm_gen_far_branch prototype.
337 * config/arm/arm.c (arm_gen_far_branch): New function
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>
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
353 2020-02-02 Vladimir Makarov <vmakarov@redhat.com>
355 PR rtl-optimization/91333
356 * ira-color.c (struct allocno_color_data): Add member
358 (init_allocno_threads): Set the member up.
359 (bucket_allocno_compare_func): Add compare hard reg
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)
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):
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
437 (aarch64_full_sve_mode): Likewise.
438 * config/aarch64/iterators.md (SVE_STRUCT): Add VNx16BF, VNx24BF
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
445 * config/aarch64/aarch64-sve-builtins.h (TYPE_bfloat): New
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)
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.
502 (ternary_resize2_opt_n_base): Add TYPE_CLASS2 and TYPE_CLASS3
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):
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
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
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.
563 (*<code><mode>3): 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
603 (maskload<mode>di): Zero-initialize the destination.
606 2020-01-30 David Malcolm <dmalcolm@redhat.com>
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>
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.
620 * combine.c (simplify_comparison) <case ROTATE>: Punt on out of range
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>
641 * config/arm/arm.md (uaddvdi4): Actually emit what gen_uaddvsi4
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.
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>
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>
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>
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
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
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
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
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>
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
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
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
828 * profile-count.h: (profile_probability::adjusted): New.
829 * tree-inline.c (copy_bb): Update to new speculative call API; fix
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>
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
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
866 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
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>
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>
933 * config/aarch64/aarch64-simd.md (aarch64_get_half<mode>): New
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>
950 * digraph.cc (test_edge::test_edge): Specify template for base
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.
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
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.
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
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
991 (arc_isuncached_mem_p): Propagate uncached attribute to each
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.
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
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>
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>
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>
1051 * sanopt.c (sanitize_rewrite_addressable_params): Avoid crash on
1054 2020-01-26 Jason Merrill <jason@redhat.com>
1057 * tree.c (verify_type_variant): Only verify TYPE_NEEDS_CONSTRUCTING
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>
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.
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>
1080 * ipa-cp.c (get_info_about_necessary_edges): Remove value
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
1089 2020-01-24 Jakub Jelinek <jakub@redhat.com>
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>.
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
1108 2020-01-24 Jeff Law <law@redhat.com>
1111 * config/h8300/h8300.c (h8300_print_operand): Only call byte_reg
1112 for REGs. Call output_operand_lossage to get more reasonable
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>
1153 * config/h8300/h8300.h (FUNCTION_PROFILER): Fix emission of
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
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
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>
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>.
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>
1193 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
1196 * config/aarch64/aarch64.md (UNSPEC_SPECULATION_TRACKER_REV): New
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
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>
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>
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>
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
1268 2020-01-22 Jakub Jelinek <jakub@redhat.com>
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>
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
1296 2020-01-21 H.J. Lu <hongjiu.lu@intel.com>
1297 Uros Bizjak <ubizjak@gmail.com>
1300 * config/i386/i386.c (ix86_tls_module_base): Replace Pmode
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>
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>
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,
1342 2020-01-21 David Malcolm <dmalcolm@redhat.com>
1345 * ipa-profile.c (ipa_profile): Delete call_sums and set it to
1348 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
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
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
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>
1372 * tree-into-ssa.c (prepare_block_for_update_1): Split out
1374 (prepare_block_for_update): This. Use a worklist instead of
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>
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>
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>
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>
1410 * targhooks.c (default_print_patchable_function_entry): Align to
1413 2020-01-20 H.J. Lu <hongjiu.lu@intel.com>
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>
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
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>
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
1486 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
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
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.
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
1539 2020-01-18 Jakub Jelinek <jakub@redhat.com>
1541 * config/arm/vfp.md (*clear_vfp_multiple): Remove unused variable.
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.
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
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
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
1609 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): New
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>
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>
1658 * config/aarch64/aarch64.h (REVERSIBLE_CC_MODE): Return false
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
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
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
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>
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.
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
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
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
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
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
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.
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
1935 * tree-vect-loop.c (update_epilogue_vinfo): Do not fold when replacing
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>
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
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>
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
2037 2020-01-14 Jan Hubicka <hubicka@ucw.cz>
2040 * ipa-devirt.c (odr_types_equivalent_p): Compare TREE_ADDRESSABLE and
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
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>
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
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):
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.
2160 (default_hash_traits<scalar_cond_masked_key>::empty_zero_p):
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>
2171 * cgraph.c (symbol_table::create_edge): Init speculative_id and
2173 (cgraph_edge::make_speculative): Add param for setting speculative_id
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
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
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>
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
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
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
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,
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>
2397 * lra-constraints.c (match_reload): Permit input operands have the
2398 same mode as output while other input operands have a different
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
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
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
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
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
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
2491 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
2493 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref): Use
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>
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>
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
2527 2020-01-10 Richard Biener <rguenther@suse.de>
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
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
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
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
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
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
2656 (<SHRNT:r>shrnt<mode>): Generalize to...
2657 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWT:sve_int_op><mode>): ...this
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
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
2766 (lr): Handle the new unspecs.
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
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
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
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
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
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>):
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>):
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
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
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
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>
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
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>
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>
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.
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
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
3183 2020-01-08 Jakub Jelinek <jakub@redhat.com>
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.
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>
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>
3253 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
3255 2019-01-08 Richard Biener <rguenther@suse.de>
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
3278 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
3280 Add -nodevicespecs option for avr.
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.
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>
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>
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
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.
3353 (Db): New constraint.
3354 * config/gcn/gcn-protos.h (gcn_inline_constant64_p): Add second
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>
3435 * doc/extend.texi: Explain cloning for target_clone
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
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
3454 2020-01-07 Martin Liska <mliska@suse.cz>
3456 PR optimization/92860
3457 * params.opt: Mark param_min_crossjump_insns with Optimization
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
3474 * doc/md.texi (RS/6000 constraints): Update 'Q' cosntraint
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
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
3525 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
3527 * config/aarch64/t-aarch64 ($(srcdir)/config/aarch64/aarch64-tune.md):
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
3543 2020-01-06 Martin Liska <mliska@suse.cz>
3545 PR tree-optimization/92860
3546 * params.opt: Mark param_max_combine_insns with Optimization
3549 2020-01-05 Jakub Jelinek <jakub@redhat.com>
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
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>
3577 * ipa-cp.c (print_all_lattices): Skip functions without info.
3579 2020-01-03 Jakub Jelinek <jakub@redhat.com>
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.
3588 * config/i386/i386.opt (x_prefer_vector_width_type): Remove TargetSave
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.
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
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>
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>
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.
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
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
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.