1 2018-05-07 Luis Machado <luis.machado@linaro.org>
5 2018-05-07 Luis Machado <luis.machado@linaro.org>
7 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
8 <prefetch_dynamic_strides>: New const bool field.
9 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
10 prefetch_dynamic_strides.
11 (exynosm1_prefetch_tune): Likewise.
12 (thunderxt88_prefetch_tune): Likewise.
13 (thunderx_prefetch_tune): Likewise.
14 (thunderx2t99_prefetch_tune): Likewise.
15 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
16 (aarch64_override_options_internal): Update to set
17 PARAM_PREFETCH_DYNAMIC_STRIDES.
18 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
19 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
20 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
21 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
22 prefetch-dynamic-strides setting.
24 2018-05-07 Luis Machado <luis.machado@linaro.org>
26 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
27 <minimum_stride>: New const int field.
28 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
30 (exynosm1_prefetch_tune): Likewise.
31 (thunderxt88_prefetch_tune): Likewise.
32 (thunderx_prefetch_tune): Likewise.
33 (thunderx2t99_prefetch_tune): Likewise.
34 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
35 (aarch64_override_options_internal): Update to set
36 PARAM_PREFETCH_MINIMUM_STRIDE.
37 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
38 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
39 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
40 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
41 stride is constant and is below the minimum stride threshold.
43 2018-05-07 Luis Machado <luis.machado@linaro.org>
45 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
48 2018-05-07 Luis Machado <luis.machado@linaro.org>
50 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
51 <prefetch_dynamic_strides>: New const bool field.
52 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
53 prefetch_dynamic_strides.
54 (exynosm1_prefetch_tune): Likewise.
55 (thunderxt88_prefetch_tune): Likewise.
56 (thunderx_prefetch_tune): Likewise.
57 (thunderx2t99_prefetch_tune): Likewise.
58 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
59 (aarch64_override_options_internal): Update to set
60 PARAM_PREFETCH_DYNAMIC_STRIDES.
61 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
62 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
63 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
64 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
65 prefetch-dynamic-strides setting.
67 2018-05-07 Luis Machado <luis.machado@linaro.org>
69 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
70 <minimum_stride>: New const int field.
71 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
73 (exynosm1_prefetch_tune): Likewise.
74 (thunderxt88_prefetch_tune): Likewise.
75 (thunderx_prefetch_tune): Likewise.
76 (thunderx2t99_prefetch_tune): Likewise.
77 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
78 (aarch64_override_options_internal): Update to set
79 PARAM_PREFETCH_MINIMUM_STRIDE.
80 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
81 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
82 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
83 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
84 stride is constant and is below the minimum stride threshold.
86 2018-05-06 Jakub Jelinek <jakub@redhat.com>
89 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
90 the type is addressable. Don't force op into register if it has
93 2018-05-05 Roland McGrath <mcgrathr@google.com>
96 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
97 any section for which we don't know a specific type it should have,
98 regardless of name. Previously this was done only for the exact
99 names ".init_array", ".fini_array", and ".preinit_array".
100 (default_elf_asm_named_section): Add comment about
101 relationship with default_section_type_flags and SECTION_NOTYPE.
102 (get_section): Don't consider it a type conflict if one side has
103 SECTION_NOTYPE and the other doesn't, as long as neither has the
104 SECTION_BSS et al used in the default_section_type_flags logic.
106 2018-05-05 Tom de Vries <tom@codesourcery.com>
109 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
110 (workaround_barsyncs): New function.
111 (nvptx_reorg): Use workaround_barsyncs.
112 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
113 (define_expand "nvptx_membar_cta"): New define_expand.
114 (define_insn "*nvptx_membar_cta"): New insn.
116 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
118 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
119 To improve optimization opportunities.
120 * builtin-types.def: The new needed builtin types for the above.
122 2018-05-04 Richard Biener <rguenther@suse.de>
124 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
125 * gimple-ssa-store-merging.c
126 (imm_store_chain_info::output_merged_store): Remove redundant create,
127 release split_store vector contents on failure.
128 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
129 scalar stmt vector on cache hit.
131 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
133 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
135 * config.gcc (powerpc-xilinx-eabi*): Remove.
136 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
138 (fusion_addis_mem_combo_load): Ditto.
139 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
141 (rs6000_cpu_cpp_builtins): Ditto.
142 * config/rs6000/rs6000-linux.c
143 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
144 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
145 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
147 (rs6000_setup_reg_addr_masks): Ditto.
148 (rs6000_init_hard_regno_mode_ok): Ditto.
149 (rs6000_option_override_internal): Ditto.
150 (legitimate_lo_sum_address_p): Ditto.
151 (rs6000_legitimize_address): Ditto.
152 (rs6000_legitimize_reload_address): Ditto.
153 (rs6000_legitimate_address_p): Ditto.
154 (abi_v4_pass_in_fpr): Ditto.
155 (setup_incoming_varargs): Ditto.
156 (rs6000_gimplify_va_arg): Ditto.
157 (rs6000_split_multireg_move): Ditto.
158 (rs6000_savres_strategy): Ditto.
159 (rs6000_emit_prologue_components): Ditto.
160 (rs6000_emit_epilogue_components): Ditto.
161 (rs6000_emit_prologue): Ditto.
162 (rs6000_emit_epilogue): Ditto.
163 (rs6000_elf_file_end): Ditto.
164 (rs6000_function_value): Ditto.
165 (rs6000_libcall_value): Ditto.
166 * config/rs6000/rs6000.h: Ditto.
167 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
168 (TARGET_MINMAX): ... this. New.
169 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
170 * config/rs6000/rs6000.md: Remove Xilinx FP support.
171 (*movsi_internal1_single): Delete.
172 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
173 mfpu=, mxilinx-fpu): Delete.
174 * config/rs6000/singlefp.h: Delete.
175 * config/rs6000/sysv4.h: Remove Xilinx FP support.
176 * config/rs6000/t-rs6000: Ditto.
177 * config/rs6000/t-xilinx: Delete.
178 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
179 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
180 (VStype_simple): Delete.
181 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
182 * config/rs6000/xfpu.h: Delete.
183 * config/rs6000/xfpu.md: Delete.
184 * config/rs6000/xilinx.h: Delete.
185 * config/rs6000/xilinx.opt: Delete.
186 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
187 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
189 2018-05-04 Tom de Vries <tom@codesourcery.com>
192 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
195 2018-05-04 Richard Biener <rguenther@suse.de>
198 * tree-complex.c (update_complex_assignment): We are always in SSA form.
199 (expand_complex_div_wide): Likewise.
200 (expand_complex_operations_1): Likewise.
201 (expand_complex_libcall): Preserve EH info of the original stmt.
202 (tree_lower_complex): Handle removed blocks.
203 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
204 on complex multiplication and division libcall builtins.
206 2018-05-04 Richard Biener <rguenther@suse.de>
209 * fold-const.c (negate_expr_p): Restrict negation of operand
210 zero of a division to when we know that can happen without
212 (fold_negate_expr_1): Likewise.
214 2018-05-04 Jakub Jelinek <jakub@redhat.com>
217 * real.h (real_nextafter): Declare.
218 * real.c (real_nextafter): New function.
219 * fold-const-call.c (fold_const_nextafter): New function.
220 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
222 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
223 even when arg1_mode is different from arg0_mode.
225 2018-05-03 Nathan Sidwell <nathan@acm.org>
227 * doc/extend.texi (Deprecated Features): Remove
229 (Backwards Compatibility): Likewise.
230 * doc/invoke.texi (C++ Language Options): Likewise.
231 (C++ Dialect Options): Likewise.
233 2018-05-03 Jakub Jelinek <jakub@redhat.com>
236 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
237 _mm512_mask_mullox_epi64): New intrinsics.
239 2018-05-03 Tom de Vries <tom@codesourcery.com>
242 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
243 dump files): Add offload-tree.
245 2018-05-03 Richard Biener <rguenther@suse.de>
247 PR tree-optimization/85615
248 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
249 to loops not nested in BBs loop father to avoid creating multi-entry
252 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
254 PR tree-optimization/70291
255 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
256 arguments. Change return type to tree. Emit libcall as a new
257 statement rather than replacing existing one when inplace_p is true.
258 (expand_complex_multiplication_components): New function.
259 (expand_complex_multiplication): Expand floating-point complex
260 multiplication using the above.
261 (expand_complex_division): Rename inner_type parameter to type.
262 Update expand_complex_libcall call-site.
263 (expand_complex_operations_1): Update expand_complex_multiplication
264 and expand_complex_division call-sites.
266 2018-05-02 Jakub Jelinek <jakub@redhat.com>
269 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
270 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
271 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
272 the highest significant bit of the shift count mask is clear. In
273 check whether and[sq]i3 is needed verify that all significant bits
274 of the shift count other than the highest are set.
276 2018-05-02 Tom de Vries <tom@codesourcery.com>
279 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
280 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
281 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
282 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
283 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
284 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
285 __builtin_goacc_parlevel_size.
287 2018-05-02 Richard Biener <rguenther@suse.de>
289 PR tree-optimization/85597
290 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
291 do not use split vect_get_vec_defs call but call vect_get_slp_defs
294 2018-05-02 Tom de Vries <tom@codesourcery.com>
297 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
298 dump files): Add ltrans-tree.
300 2018-05-02 Tom de Vries <tom@codesourcery.com>
303 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
304 dump files): Add wpa-ipa.
306 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
308 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
309 powerpc*-*-linux*paired* target.
310 * config/rs6000/750cl.h: Delete.
311 * config/rs6000/paired.h: Delete.
312 * config/rs6000/paired.md: Delete.
313 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
315 * config/rs6000/rs6000-builtin.def: Remove paired float support.
316 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
317 comment. Remove paired float support.
318 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
319 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
321 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
322 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
324 * config/rs6000/rs6000.c: Remove paired float support.
325 (paired_expand_vector_init, paired_expand_vector_move,
326 paired_emit_vector_compare, paired_emit_vector_cond_expr,
327 (paired_expand_lv_builtin, paired_expand_stv_builtin,
328 paired_expand_builtin, paired_expand_predicate_builtin,
329 paired_init_builtins): Delete.
330 * config/rs6000/rs6000.h: Remove paired float support.
331 * config/rs6000/rs6000.md: Remove paired float support.
332 (move_from_CR_ov_bit): Delete.
333 * config/rs6000/rs6000.opt (mpaired): Delete.
334 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
335 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
337 2018-05-02 Richard Biener <rguenther@suse.de>
340 * gimplify.c (gimplify_save_expr): When in SSA form allow
341 SAVE_EXPRs to compute to SSA vars.
343 2018-05-02 Jakub Jelinek <jakub@redhat.com>
346 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
347 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
348 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
349 clobber operands[2], instead use a new pseudo. Formatting fixes.
351 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
353 PR tree-optimization/85586
354 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
355 exit early for statements in the same group if the accesses are
358 2018-05-02 Tom de Vries <tom@codesourcery.com>
361 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
364 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
366 PR tree-optimization/85143
367 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
369 2018-05-01 Tom de Vries <tom@codesourcery.com>
372 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
373 not found" error message.
375 2018-05-01 Tom de Vries <tom@codesourcery.com>
378 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
379 * vec.c (test_ordered_remove_if): New function.
380 (vec_c_tests): Call test_ordered_remove_if.
381 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
382 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
383 * tree-vect-patterns.c (vect_pattern_recog_1): Use
384 VEC_ORDERED_REMOVE_IF.
386 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
388 PR tree-optimization/82665
389 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
390 pointer subtraction where arguments come from a memchr call.
392 2018-05-01 Jakub Jelinek <jakub@redhat.com>
394 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
395 --push-state --as-needed and --pop-state instead of --as-needed and
396 --no-as-needed if ld supports it.
397 * configure: Regenerated.
400 * doc/install.texi2html: Replace _002d with - and _002a with * in
401 generated html files using sed.
403 2018-04-30 David Malcolm <dmalcolm@redhat.com>
406 * gcc-rich-location.c (blank_line_before_p): New function.
407 (use_new_line): New function.
408 (gcc_rich_location::add_fixit_insert_formatted): New function.
409 * gcc-rich-location.h
410 (gcc_rich_location::add_fixit_insert_formatted): New function.
412 2018-04-30 David Malcolm <dmalcolm@redhat.com>
414 * selftest.c (assert_streq): Rename "expected" and "actual" to
415 "val1" and "val2". Extend NULL-handling to cover both inputs
416 symmetrically, while still requiring both to be non-NULL for a pass.
417 * selftest.h (assert_streq): Rename "expected" and "actual" to
419 (ASSERT_EQ): Likewise.
420 (ASSERT_EQ_AT): Likewise.
421 (ASSERT_KNOWN_EQ): Likewise.
422 (ASSERT_KNOWN_EQ_AT): Likewise.
423 (ASSERT_NE): Likewise.
424 (ASSERT_MAYBE_NE): Likewise.
425 (ASSERT_MAYBE_NE_AT): Likewise.
426 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
427 the assertion to pass.
428 (ASSERT_STREQ_AT): Likewise.
430 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
432 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
433 interaction with -pie.
435 2018-04-30 David Malcolm <dmalcolm@redhat.com>
437 * selftest.h: Fix alphabetization of per-source-file selftest
440 2018-04-30 Jason Merrill <jason@redhat.com>
442 PR c++/61982 - dead stores to destroyed objects.
443 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
446 2018-04-30 Jason Merrill <jason@redhat.com>
448 * tree.c (build_clobber): New.
449 * tree.h: Declare it.
450 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
452 2018-04-30 David Malcolm <dmalcolm@redhat.com>
454 * diagnostic-show-locus.c (layout::layout): Update for
455 location_get_source_line returning a char_span.
456 (struct char_span): Move to input.h.
457 (struct correction): Update for fields in char_span becoming
459 (struct source_line): Update for location_get_source_line
460 returning a char_span.
461 (layout::print_line): Likewise.
462 * edit-context.c (edited_file::print_content): Likewise.
463 (edited_file::print_diff_hunk): Likewise.
464 (edited_file::print_run_of_changed_lines): Likewise.
465 (edited_file::get_num_lines): Likewise.
466 (edited_line::edited_line): Likewise.
467 * final.c (asm_show_source): Likewise.
468 * input.c (location_get_source_line): Convert return type
469 from const char * to char_span, losing the final "line_len"
471 (dump_location_info): Update for the above.
472 (get_substring_ranges_for_loc): Likewise. Use a char_span
473 when handling the literal within the line.
474 (test_reading_source_line): Update for location_get_source_line
475 returning a char_span.
476 * input.h (class char_span): Move here from
477 diagnostic-show-locus.c, converting from a struct to a class.
478 Make data members private.
479 (char_span::operator bool): New.
480 (char_span::length): New.
481 (char_span::get_buffer): New.
482 (char_span::operator[]): New.
483 (char_span::subspan): Make const.
484 (char_span::xstrdup): New.
485 (location_get_source_line): Convert return type from const char *
486 to char_span, losing the final "line_size" param.
488 2018-04-30 Jan Hubicka <jh@suse.cz>
490 * lto-wrapper.c (ltrans_priorities): New static var.
492 (run_gcc): Read priorities and if doing parallel build order
493 the Makefile by them.
495 2018-04-30 David Malcolm <dmalcolm@redhat.com>
497 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
499 2018-04-30 Richard Biener <rguenther@suse.de>
501 * tree-cfg.c (verify_address): Remove base argument, add
502 flag whether to check TREE_ADDRESSABLE and do that.
503 (verify_expr): Remove.
504 (verify_types_in_gimple_reference): Add pieces from verify_expr.
505 (verify_gimple_assign_single): Likewise.
506 (verify_gimple_switch): Likewise.
507 (verify_expr_location_1): Dereference tp once. Add (disabled)
508 piece from verify_expr.
509 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
511 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
513 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
515 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
517 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
518 (small_data_pattern): Likewise.
519 (arc_rewrite_small_data): Likewise.
520 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
521 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
522 (get_symbol_alignment): New function.
523 (legitimate_small_data_address_p): Likewise.
524 (legitimate_scaled_address): Update, call
525 legitimate_small_data_address_p.
526 (output_sdata): New static variable.
527 (arc_print_operand): Update how we handle small data operands.
528 (arc_print_operand_address): Likewise.
529 (arc_legitimate_address_p): Update, use
530 legitimate_small_data_address_p.
531 (arc_rewrite_small_data_p): Remove.
532 (arc_rewrite_small_data_1): Likewise.
533 (arc_rewrite_small_data): Likewise.
534 (small_data_pattern): Likewise.
535 (compact_sda_memory_operand): Update to use
536 legitimate_small_data_address_p and get_symbol_alignment.
537 (prepare_move_operands): Don't rewite sdata pattern.
538 (prepare_extend_operands): Remove.
539 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
541 (zero_extendqisi2): Likewise.
542 (zero_extendhisi2): Likewise.
543 (extendqihi2): Likewise.
544 (extendqisi2): Likewise.
545 (extendhisi2): Likewise.
549 * config/arc/constraints.md (Usd): Change it to memory constraint.
551 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
553 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
554 as source of std instructions.
555 * config/arc/arc.md (movsi_insn): Update pattern predicate to
556 allow 6-bit constants as source for store instructions.
557 (movdi_insn): Update instruction pattern to allow 6-bit constants
558 as source for store instructions.
560 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
562 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
564 2018-04-30 Nathan Sidwell <nathan@acm.org>
565 Sandra Loosemore <sandra@codesourcery.com>
567 * dumpfile.c (dump_open): Allow '-' for stdout.
568 * doc/invoke.texi (Developer Options): Document dump filename
569 determination early. Document stdin/stdout selection.
571 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
573 Microblaze Target: PIC data text relative
575 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
576 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
578 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
579 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
580 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
581 New addressing mode for data-text relative position indepenedent code.
582 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
583 'ADDRESS_SYMBOLIC_TXT_REL'.
584 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
585 (microblaze_legitimate_pic_operand): Exclude function calls from
586 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
587 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
589 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
590 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
591 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
592 for 'address + offset'.
593 (microblaze_expand_prologue): Add new function prologue call for
595 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
596 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
597 table in case of TARGET_PIC_DATA_TEXT_REL.
598 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
599 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
600 Add new macros 'UNSPEC_TEXT',
601 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
602 + exclude function calls from 'UNSPEC_PLT' in case of data text
604 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
605 new target hook for generating address diff vector tables in case of
607 * doc/tm.texi : Regenerate.
608 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
609 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
610 of addr diff vector generation.
611 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
612 target hook definition.
613 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
614 Add default function for generate_pic_addr_diff_vec -> flag_pic.
615 * doc/invoke.texi (Add new pic option): Add new microblaze pic
616 option for data text relative.
618 2018-04-30 Richard Biener <rguenther@suse.de>
620 * tree-chrec.h (evolution_function_is_constant_p): Remove
622 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
624 2018-04-30 Richard Biener <rguenther@suse.de>
627 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
629 2018-04-30 Richard Biener <rguenther@suse.de>
631 PR tree-optimization/28364
632 PR tree-optimization/85275
633 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
634 copying first exit test.
636 2018-04-28 Mark Wielaard <mark@klomp.org>
638 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
640 (dwarf_AT): Handle DW_AT_addr_base.
641 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
643 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
646 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
647 (*ashl<dwi>3_doubleword_mask_1): Ditto.
648 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
649 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
651 2018-04-28 Richard Biener <rguenther@suse.de>
653 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
654 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
655 to reflect use. Only add interesting stmts.
657 2018-04-27 Martin Jambor <mjambor@suse.cz>
660 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
661 the jump function allows for passing through aggregate values.
663 2018-04-27 David Malcolm <dmalcolm@redhat.com>
665 * input.h (in_system_header_at): Convert from macro to inline
667 (from_macro_expansion_at): Likewise.
668 (from_macro_definition_at): Likewise.
670 2018-04-27 Jeff Law <law@redhat.com>
672 * config.gcc: Mark tile* targets as deprecated/obsolete.
674 2018-04-27 Richard Biener <rguenther@suse.de>
676 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
679 2018-04-27 Richard Biener <rguenther@suse.de>
681 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
683 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
685 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
686 with Yd constraint. Set "preferred_for_speed" attribute from
687 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
689 (*movdi_internal): Ditto.
690 (movti_interunit splitters): Remove
691 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
692 (movdi_interunit splitters): Ditto.
693 * config/i386/constraints.md (Ye): Remove.
694 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
696 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
699 * config/aarch64/constraints.md (Usg): Limit to 31.
702 2018-04-27 Jakub Jelinek <jakub@redhat.com>
704 PR tree-optimization/85529
705 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
706 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
707 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
708 zero extension or masking of the MSB bit.
709 (optimize_range_tests): Add FIRST_BB argument, pass it through
710 to optimize_range_tests_var_bound.
711 (maybe_optimize_range_tests, reassociate_bb): Adjust
712 optimize_range_tests callers.
714 2018-04-26 Richard Biener <rguenther@suse.de>
715 Jakub Jelinek <jakub@redhat.com>
717 * cgraph.h (symbol_table): Just declare debug method here.
718 * symtab.c (symbol_table::debug): Define.
720 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
722 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
724 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
726 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
727 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
728 (*movdi_internal): Substitute Yi and Yj constraint with x
729 and Ym and Yn constraint with y constraint. Update "isa"
730 attribute and set "preferred_for_speed" attribute from
731 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
732 (*movsi_internal): Ditto.
733 (*movdf_internal): Ditto.
734 (*movsf_internal): Ditto.
735 (*zero_extendsidi2): Ditto.
736 * config/i386/sse.md (vec_set<mode>_0): Ditto.
737 (sse2_loadld): Ditto.
738 (*vec_extract<ssevecmodelower>_0): Ditto.
739 (*vec_extractv4si_0_zext_sse4): Ditto.
740 (vec_concatv2di): Ditto.
741 (*vec_dup<mode>): Ditto.
742 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
743 * config/i386/constraints.md (Yi): Remove.
748 2018-04-26 Nathan Sidwell <nathan@acm.org>
750 * dumpfile.c (dump_open): New.
751 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
752 (dump_finish): Detect stdio/stderr by value not name.
754 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
756 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
758 2018-04-26 Tom de Vries <tom@codesourcery.com>
761 * config/nvptx/nvptx.c (verify_neutering_jumps)
762 (verify_neutering_labels): New function
763 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
765 2018-04-26 Tom de Vries <tom@codesourcery.com>
768 * config/nvptx/nvptx.c (needs_neutering_p): New function.
769 (nvptx_single): Use needs_neutering_p to skip over insns that do not
772 2018-04-26 Richard Biener <rguenther@suse.de>
773 Tom de Vries <tom@codesourcery.com>
776 * lto-streamer-out.c (output_function): Fixup loops if required to match
777 discovery done in the reader.
779 2018-04-26 Richard Biener <rguenther@suse.de>
781 PR tree-optimization/85116
782 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
783 have a loop exit from the single latch predecessor. Remove
784 case of header with just condition.
785 (ch_base::copy_headers): Exclude infinite loops from any
787 (pass_ch::execute): Record exits.
789 2018-04-26 Richard Biener <rguenther@suse.de>
791 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
792 prologue cost vector and pass it to vect_get_load_cost.
793 (vect_get_peeling_costs_all_drs): Likewise.
794 (vect_peeling_hash_get_lowest_cost): Likewise.
795 (vect_enhance_data_refs_alignment): Likewise.
797 2018-04-26 Richard Biener <rguenther@suse.de>
800 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
801 checking of integer<->pointer conversions.
802 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
803 sign-/zero-extending pointer types.
804 (expand_omp_for_static_chunk): Likewise.
806 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
807 Jean Lee <xiaoyur347@gmail.com>
809 * config/mips/mips.c (mips_asan_shadow_offset): New function.
810 (TARGET_ASAN_SHADOW_OFFSET): Define.
811 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
812 true for -fsanitize=address.
814 2018-04-25 Mark Wielaard <mark@klomp.org>
816 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
819 2018-04-25 Jakub Jelinek <jakub@redhat.com>
821 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
822 than "alu", remove explicit "memory" and "imm_disp" attributes.
823 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
826 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
827 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
830 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
833 * config/i386/i386.c (ix86_expand_builtin): Change memory
834 operand to XI, extend p0 to Pmode.
835 * config/i386/i386.md: Change unspec volatile and operand
836 1 mode to XI, change operand 0 mode to P.
838 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
840 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
841 GET_MODE_MASK before any checking.
842 (nds32_can_use_bset_p): Likewise.
843 (nds32_can_use_btgl_p): Likewise.
845 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
847 * config/nds32/nds32-doubleword.md: New define_split pattern for
848 illegal register number.
850 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
852 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
854 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
856 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
858 2018-04-25 Richard Biener <rguenther@suse.de>
860 * lto-streamer.h (LTO_major_version): Bump to 8.
862 2018-04-25 Jakub Jelinek <jakub@redhat.com>
864 * BASE-VER: Set to 9.0.0.
866 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
868 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
869 in __abskf2 and __powikf2.
871 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
874 * config/aarch64/constraints.md (Usg, Usj): New constraints.
875 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
876 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
877 Use the above on operand 2. Reindent.
878 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
880 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
883 * common/config/i386/i386-common.c (ix86_handle_option): Don't
885 * config/i386/i386.opt (mcet): Removed.
886 * doc/install.texi: Remove -mcet documentation.
887 * doc/invoke.texi: Likewise.
889 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
892 * doc/install.texi: Remove -mcet from bootstrap-cet.
894 2018-04-24 Jakub Jelinek <jakub@redhat.com>
897 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
898 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
902 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
903 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
904 containing a CONST_VECTOR.
906 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
908 * doc/install.texi: Update newlib dependency for nvptx.
910 2018-04-24 Jakub Jelinek <jakub@redhat.com>
913 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
914 instead of INTVAL when shifting x left.
916 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
918 PR tree-optimization/85478
919 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
920 vect_grouped_store_supported for single element vectors.
922 2018-04-24 Richard Biener <rguenther@suse.de>
925 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
926 load cost increase to the case of non-constant step.
928 2018-04-24 Jakub Jelinek <jakub@redhat.com>
931 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
932 destination if any_malformed_asm.
934 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
937 * expr.c (store_field): In the bitfield case, if the value comes from
938 a function call and is returned in registers by means of a PARALLEL,
939 do not change the mode of the temporary unless BLKmode and VOIDmode.
941 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
943 PR rtl-optimization/85423
944 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
945 dependencies to debug insns when the previous insn is non-debug.
947 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
949 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
950 enums into a single definition.
951 (fls): Fix predicates and printing.
954 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
956 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
957 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
958 and short u6 immediate.
959 (check_if_valid_sleep_operand): Remove.
960 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
962 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
964 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
965 flag_always_save_lp condition.
966 * config/nds32/nds32.opt (malways-save-lp): New option.
968 2018-04-22 Shiva Chen <shiva0217@gmail.com>
970 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
971 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
972 * config/nds32/nds32.h
973 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
974 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
976 2018-04-22 Shiva Chen <shiva0217@gmail.com>
978 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
979 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
981 2018-04-22 Shiva Chen <shiva0217@gmail.com>
982 Chung-Ju Wu <jasonwucj@gmail.com>
984 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
986 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
987 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
989 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
991 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
993 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
995 * config/nds32/nds32-protos.h (nds32_data_alignment,
996 nds32_local_alignment): Declare.
997 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
998 nds32_local_alignment): New functions.
999 (TARGET_CONSTANT_ALIGNMENT): Define.
1000 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
1002 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1004 * config/nds32/nds32.c
1005 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
1006 (TARGET_MODES_TIEABLE_P): Likewise.
1008 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1010 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
1013 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
1014 Chung-Ju Wu <jasonwucj@gmail.com>
1016 * config/nds32/constants.md (unspec_volatile_element): Add enum values
1017 for unaligned access.
1018 * config/nds32/nds32-intrinsic.c: Implementation of expanding
1020 * config/nds32/nds32-intrinsic.md: Likewise.
1021 * config/nds32/nds32_intrinsic.h: Likewise.
1022 * config/nds32/nds32.h (nds32_builtins): Likewise.
1023 * config/nds32/nds32.opt (munaligned-access): New option.
1024 * config/nds32/nds32.c (nds32_asm_file_start): Display
1025 flag_unaligned_access status.
1027 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
1029 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
1030 -mno-relax is present.
1031 * config/riscv/linux.h (LINK_SPEC): Ditto.
1033 2018-04-20 Martin Sebor <msebor@redhat.com>
1036 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
1038 (gimple_fold_builtin_stxcpy_chk): Same.
1039 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
1041 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
1044 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
1045 __powikf2 when long double is IEEE 128-bit.
1047 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
1049 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
1050 step to make sure stack always aligned.
1052 2018-04-20 Carl Love <cel@us.ibm.com>
1055 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
1056 size check for arg0.
1058 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
1059 Tom de Vries <tom@codesourcery.com>
1062 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
1063 Emit insns for calls too.
1064 (nvptx_find_par): Always look for worker-level predecessor insn.
1065 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
1067 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
1068 (nvptx_process_pars): Propagate frames for calls.
1070 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
1073 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
1075 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
1076 (ix86_handle_option): Don't handle OPT_mibt.
1077 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
1079 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
1081 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
1082 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
1083 (ix86_target_macros): Define __CET__ with flag_cf_protection
1084 for -fcf-protection.
1085 * config/i386/i386.c (isa2_opts): Remove -mibt.
1086 * config/i386/i386.h (TARGET_IBT): Removed.
1087 (TARGET_IBT_P): Likewise.
1088 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
1089 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
1090 * config/i386/i386.opt (mcet): Update help message.
1093 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
1094 -mcet as an alias for -mshstk.
1096 2018-04-20 Richard Biener <rguenther@suse.de>
1099 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
1100 complexity by forcing a single use of the multiply operand.
1102 2018-04-20 Martin Jambor <mjambor@suse.cz>
1105 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
1106 recursion dependency to only apply to non-clones.
1108 2018-04-20 Martin Jambor <mjambor@suse.cz>
1111 * ipa-cp.c (create_specialized_node): Check that clones of
1112 self-recursive edges exist during IPA-CP.
1114 2018-04-19 Toon Moene <toon@moene.org>
1116 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
1119 2018-04-19 Jakub Jelinek <jakub@redhat.com>
1121 PR tree-optimization/85467
1122 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
1123 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
1124 VECTOR_CST element to type.
1126 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1129 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
1130 * config/i386/i386.md (builtin_setjmp_setup): Removed.
1131 (builtin_longjmp): Likewise.
1132 (save_stack_nonlocal): New pattern.
1133 (restore_stack_nonlocal): Likewise.
1135 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1138 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
1139 Replace ASM_OUTPUT_LABEL with fprintf.
1141 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1144 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
1145 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
1146 * config/i386/i386-c.c (ix86_target_macros_internal): Also
1147 define __IBT__ and __SHSTK__ for -fcf-protection.
1148 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
1150 (ix86_trampoline_init): Likewise.
1151 (x86_output_mi_thunk): Likewise.
1152 (ix86_notrack_prefixed_insn_p): Likewise.
1153 (ix86_option_override_internal): Don't disallow -fcf-protection.
1154 * config/i386/i386.md (rdssp<mode>): Also enable for
1156 (incssp<mode>): Likewise.
1157 (nop_endbr): Likewise.
1158 * config/i386/i386.opt (mcet): Change help message to built-in
1162 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
1163 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
1164 enable CET built-in functions.
1166 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
1168 * common/config/i386/i386-common.c
1169 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
1170 OPTION_MASK_ISA_MOVDIRI_UNSET,
1171 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
1172 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
1173 * config.gcc (movdirintrin.h): New header.
1174 * config/i386/cpuid.h (bit_MOVDIRI,
1175 bit_MOVDIR64B): New bits.
1176 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
1178 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
1179 (VOID, PVOID, PCVOID)): New function types.
1180 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
1181 __builtin_ia32_directstoreu_u64,
1182 __builtin_ia32_movdir64b): New builtins.
1183 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
1184 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
1186 (ix86_valid_target_attribute_inner_p): Ditto.
1187 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
1188 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
1189 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
1190 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
1191 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
1192 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
1193 (movdiri<mode>, movdir64b_<mode>): New.
1194 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
1195 * config/i386/immintrin.h: Include movdirintrin.h.
1196 * config/i386/movdirintrin.h: New file.
1197 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
1199 2018-04-19 Richard Biener <rguenther@suse.de>
1202 * cfg.c (clear_bb_flags): When loop state says we have
1203 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
1205 2018-04-19 Richard Biener <rguenther@suse.de>
1207 PR tree-optimization/84737
1208 * tree-vect-data-refs.c (vect_copy_ref_info): New function
1209 copying restrict info.
1210 (vect_setup_realignment): Use it.
1211 * tree-vectorizer.h (vect_copy_ref_info): Declare.
1212 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
1213 the first DR to all generated stores.
1214 (vectorizable_load): Likewise for loads.
1216 2018-04-19 Jakub Jelinek <jakub@redhat.com>
1218 PR tree-optimization/85446
1219 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
1220 the integral and pointer types to have the same precision.
1222 * doc/install.texi: Document --disable-cet being the default and
1225 2018-04-18 Martin Liska <mliska@suse.cz>
1227 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
1230 2018-04-18 Martin Liska <mliska@suse.cz>
1233 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
1236 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
1237 arguments if they are comparable.
1239 2018-04-18 Martin Liska <mliska@suse.cz>
1242 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
1245 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
1248 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
1251 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
1252 ENDBR after calling __morestack.
1254 2018-04-18 David Malcolm <dmalcolm@redhat.com>
1257 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
1258 by using gcc_base_ver to generate a gcc_driver_version, and use
1259 it when generating GCC_DRIVER_NAME.
1260 * configure: Regenerate.
1262 2018-04-18 Jakub Jelinek <jakub@redhat.com>
1265 * config.gcc: Obsolete powerpc*-*-*spe*.
1267 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1270 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
1271 (stabstr_D): Change type of unum from unsigned int to
1272 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
1275 2018-04-17 Jim Wilson <jimw@sifive.com>
1278 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
1279 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
1280 Set arg_pointer_offset after using pretend_args_size.
1282 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1284 PR rtl-optimization/85431
1285 * dse.c (record_store): Ignore zero width stores.
1288 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
1289 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
1290 __builtin_stack_restore rather than after it.
1291 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
1292 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
1293 argument instead of virtual_dynamic_stack_rtx.
1295 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
1297 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
1299 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
1300 Add note to error message to explain internal mapping of overloaded
1301 built-in function name to non-overloaded built-in function name.
1302 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
1305 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
1308 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
1309 where the inputs overlap with the output.
1311 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1313 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
1314 (=v, v) alternative and explicit "memory" attribute.
1315 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
1316 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
1318 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
1319 "sselog1" type instead of "sselog".
1320 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
1321 "sselog". Remove explicit "memory" attribute.
1322 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
1323 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
1325 (vec_extract_hi_v32hi): Merge all alternatives into one, use
1326 "sselog1" type instead of "sselog". Remove explicit "memory"
1328 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
1329 use "sselog1" type instead of "sselog". Remove explicit "memory"
1331 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
1332 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
1334 (vec_extract_hi_v64qi): Merge all alternatives into one, use
1335 "sselog1" type instead of "sselog". Remove explicit "memory"
1337 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
1338 use "sselog1" type instead of "sselog". Remove explicit "memory"
1342 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
1345 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
1348 2018-04-17 Martin Jambor <mjambor@suse.cz>
1351 * ipa-cp.c (create_specialized_node): Call
1352 expand_all_artificial_thunks if necessary.
1354 2018-04-17 Martin Liska <mliska@suse.cz>
1357 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
1358 in message, remote space in between '_G' and '('.
1360 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1363 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
1364 avx512f_vmcmp<mode>3<round_saeonly_name>,
1365 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
1366 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
1367 avx512f_rndscale<mode><round_saeonly_name>,
1368 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
1369 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
1370 Use %<iptr>2 instead of %2 for -masm=intel.
1371 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
1372 avx512f_vcvttss2usi<round_saeonly_name>,
1373 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
1375 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
1376 avx512f_vcvttsd2usi<round_saeonly_name>,
1377 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
1378 Use %q1 instead of %1 for -masm=intel.
1379 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
1380 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
1381 of %3 for -masm=intel.
1382 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
1384 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
1386 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
1388 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
1389 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
1391 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
1393 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
1394 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
1395 %g1 and one with %0 and %1.
1396 (avx512er_vmrcp28<mode><round_saeonly_name>,
1397 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
1399 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
1400 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
1401 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
1402 of %0 and %{%4%} for -masm=intel.
1403 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
1404 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
1405 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
1406 order of %0 and %{%5%}%{z%} for -masm=intel.
1408 2018-04-17 Jan Hubicka <jh@suse.cz>
1411 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
1413 2018-04-17 Martin Liska <mliska@suse.cz>
1416 * multiple_target.c (create_dispatcher_calls): Set apostrophes
1417 for target_clone error message. Make default implementation
1418 clone to be a local declaration.
1419 (separate_attrs): Add new argument and check for an empty
1421 (expand_target_clones): Handle it.
1422 (ipa_target_clone): Make redirection just for target_clones
1425 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
1426 Tom de Vries <tom@codesourcery.com>
1429 * omp-expand.c (expand_oacc_for): Add dummy false branch for
1430 tiled basic blocks without omp continue statements.
1432 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
1435 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
1436 vec_extract expression as having side effects to make sure it gets
1439 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
1442 * config/i386/i386.c (get_builtin_code_for_version): Check
1445 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
1448 * gcc/config.gcc: Support "skylake".
1449 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
1451 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
1452 (processor_target_table): Add "skylake".
1453 (ix86_option_override_internal): Add "skylake".
1454 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
1455 PROCESSOR_CANNONLAKE.
1456 (get_builtin_code_for_version): Fix priority for
1457 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
1458 PROCESSOR_SKYLAKE-AVX512.
1459 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
1460 (processor_type): Add PROCESSOR_SKYLAKE.
1462 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
1463 Jason Merrill <jason@redhat.com>
1466 * convert.c (convert_to_integer_1): Use direct recursion for
1467 enumeral types and types with a precision less than the number
1468 of bits in their mode.
1470 2018-04-16 Julia Koval <julia.koval@intel.com>
1473 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
1474 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
1476 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
1479 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
1480 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
1481 and -mno-direct-move.
1483 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
1486 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
1487 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
1488 Ensure negative shifts result in {0}.
1490 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
1492 PR rtl-optimization/79916
1493 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
1494 regs (if any) to define how to gnerate SD moves when LRA is in
1497 2018-04-13 Jakub Jelinek <jakub@redhat.com>
1499 PR rtl-optimization/85393
1500 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
1501 * except.c (expand_dw2_landing_pad_for_region): Make static.
1502 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
1503 a label and unconditional jump to old_bb, rather than
1504 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
1507 PR rtl-optimization/85376
1508 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
1509 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
1510 instead of a specific value.
1512 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
1513 Bin Cheng <bin.cheng@arm.com>
1515 PR tree-optimization/82965
1516 PR tree-optimization/83991
1517 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
1518 by_profile_only parameter.
1519 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
1520 information if the loop was predicted to iterate too many times.
1521 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
1523 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
1526 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
1529 2018-04-13 Martin Liska <mliska@suse.cz>
1530 Jakub Jelinek <jakub@redhat.com>
1533 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
1534 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
1535 * builtins.c (expand_builtin_memory_copy_args): Use
1536 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
1537 handle dest_addr == pc_rtx.
1539 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
1542 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
1543 asked to not generate direct moves.
1544 (fix_trunc<mode>si2_stfiwx): Similar.
1545 (fix_trunc<mode>si2_internal): Similar.
1547 2018-04-12 Jakub Jelinek <jakub@redhat.com>
1550 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
1551 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
1552 lookup if dest in some wider mode is known to be const0_rtx and
1553 if so, record permanent equivalence for it to be ZERO_EXTEND of
1554 the narrower mode destination.
1556 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
1558 * lto-streamer-out.c (output_function): Revert 259346.
1559 * omp-expand.c (expand_oacc_for): Likewise.
1561 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
1563 PR rtl-optimization/85354
1564 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
1565 * sel-sched.c (sel_global_init): ... here.
1567 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
1570 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
1571 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
1572 mode for PE-COFF targets.
1573 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
1574 (i386_pe_asm_lto_end): Likewise.
1575 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
1576 (TARGET_ASM_LTO_END): Likewise.
1577 * config/i386/winnt.c (saved_debug_info_level): New static variable.
1578 (i386_pe_asm_lto_start): New function.
1579 (i386_pe_asm_lto_end): Likewise.
1581 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
1582 Richard Biener <rguenther@suse.de>
1585 * lto-streamer-out.c (output_function): Fix CFG loop state before
1587 * omp-expand.c (expand_oacc_for): Handle calls to internal
1588 functions like regular functions.
1590 2018-04-12 Richard Biener <rguenther@suse.de>
1593 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
1594 for the early LTO debug to properly generate references to it
1595 during DIE emission. Do not re-use that for the skeleton for
1597 (dwarf2out_early_finish): Likewise.
1599 2018-04-12 Jakub Jelinek <jakub@redhat.com>
1602 * config/i386/sse.md
1603 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
1604 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
1605 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
1606 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
1607 and output is a reg, avoid creating invalid lowpart subreg, but
1608 instead split into a 512-bit move. Don't split if not AVX512VL,
1609 input is xmm16+ reg and output is a mem.
1610 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
1611 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
1612 xmm16+ reg and output is a mem.
1614 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1616 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
1617 also for flag_dwarf2_cfi_asm.
1619 2018-04-12 Jakub Jelinek <jakub@redhat.com>
1621 PR rtl-optimization/85342
1622 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
1623 a bool scalar var inside of the loop instead. Don't try to update
1624 recog_data.operand after failed apply_change_group.
1626 2018-04-12 Tom de Vries <tom@codesourcery.com>
1629 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
1630 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
1631 array with flexible array member as array without given dimension.
1632 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
1633 argument for undefined param to true.
1635 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
1638 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
1639 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
1640 from PowerPC section.
1641 * config/rs6000/sysv4.opt (mcall-): Improve help text.
1642 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
1643 help text that is too long.
1644 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
1645 help text that is too long.
1646 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
1647 help text that is too long.
1649 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
1651 * config/alpha/alpha.md (stack_probe_internal): Rename
1652 from "probe_stack". Update all callers.
1654 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1656 PR rtl-optimization/84566
1657 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
1658 sched_macro_fuse_insns.
1660 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1663 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
1664 (compute_block_dependences): ... from here.
1666 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1668 PR tree-optimization/85331
1669 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
1670 from int to HOST_WIDE_INT.
1672 2018-04-11 Martin Jambor <mjambor@suse.cz>
1675 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
1676 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
1677 not the same as the source val.
1678 (cgraph_edge_brings_value_p): New parameter.
1679 (gather_edges_for_value): Pass destination value to
1680 cgraph_edge_brings_value_p.
1681 (perhaps_add_new_callers): Likewise.
1682 (get_info_about_necessary_edges): Likewise and exclude values brought
1683 only by self-recursive edges.
1684 (create_specialized_node): Redirect only clones of self-calling edges.
1685 (+self_recursive_pass_through_p): New function.
1686 (find_more_scalar_values_for_callers_subset): Use it.
1687 (find_aggregate_values_for_callers_subset): Likewise.
1688 (known_aggs_to_agg_replacement_list): Removed.
1689 (decide_whether_version_node): Re-calculate known constants for all
1690 remaining context clones.
1692 2018-04-11 Richard Biener <rguenther@suse.de>
1695 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
1696 from early DWARF output.
1697 (dwarf2out_early_finish): Output line info unconditionally into
1698 early DWARF and add reference to it.
1700 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1703 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
1704 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
1705 other than V2DFmode using iptr mode attribute.
1706 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
1708 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1710 PR rtl-optimization/84659
1711 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
1713 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1716 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
1718 (output_loc_list): Pass address of a dummy size variable even in the
1719 locview handling loop.
1720 (index_location_lists): Add comment on why skip_loc_list_entry can't
1723 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
1726 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
1729 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
1732 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
1733 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
1734 and -mstring-compare-inline-limit.
1736 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1739 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
1740 for stack clash protection in a register whenever we need it to be in
1743 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1745 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
1746 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
1748 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1751 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
1753 (mlong-double-): Ditto.
1754 * config/rs6000/sysv4.opt (msdata=): Ditto.
1755 (mtls-size=): Ditto.
1757 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
1759 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
1760 erroneous entries for
1761 "vector int vec_ldl (int, long int *)", and
1762 "vector unsigned int vec_ldl (int, unsigned long int *)".
1763 Add comments and entries for
1764 "vector bool char vec_ldl (int, bool char *)",
1765 "vector bool short vec_ldl (int, bool short *)",
1766 "vector bool int vec_ldl (int, bool int *)",
1767 "vector bool long long vec_ldl (int, bool long long *)",
1768 "vector pixel vec_ldl (int, pixel *)",
1769 "vector long long vec_ldl (int, long long *)",
1770 "vector unsigned long long vec_ldl (int, unsigned long long *)".
1771 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
1772 type tree bool_long_long_type_node and correct definition of
1773 bool_V2DI_type_node to make reference to this new type tree.
1774 (rs6000_mangle_type): Replace erroneous reference to
1775 bool_long_type_node with bool_long_long_type_node.
1776 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
1777 comments to emphasize sign distinctions for char and int types and
1778 replace RS6000_BTI_bool_long constant with
1779 RS6000_BTI_bool_long_long constant. Also add comment to restrict
1780 use of RS6000_BTI_pixel.
1781 (bool_long_type_node): Remove this macro definition.
1782 (bool_long_long_type_node): New macro definition
1784 2018-04-10 Jakub Jelinek <jakub@redhat.com>
1786 PR rtl-optimization/85300
1787 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
1788 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
1789 simplify_unary_operation fails.
1791 2018-04-10 Martin Liska <mliska@suse.cz>
1793 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
1794 cgraph_edge and ipa_ref.
1796 2018-04-10 Jakub Jelinek <jakub@redhat.com>
1800 * config/i386/sse.md
1801 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
1802 computation of the VEC_MERGE selector from mask.
1803 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
1804 Fix decoding of the VEC_MERGE selector into mask.
1806 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
1808 PR tree-optimization/85286
1809 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
1811 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
1813 * final.c (final_1): Set insn_last_address as well as
1814 insn_current_address.
1816 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1819 * explow.c (emit_stack_probe): Call validize_mem on memory location
1820 before passing it to gen_probe_stack. Create address operand and
1821 legitimize it for the probe_stack_address case.
1823 2018-04-09 Jan Hubicka <jh@suse.cz>
1826 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
1827 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
1828 * tree.c (free_lang_data_in_type): Fix handling of binfos;
1830 (free_lang_data): Rebuild type inheritance graph.
1832 2018-04-09 Martin Sebor <msebor@redhat.com>
1834 * invoke.texi (-finline-small-functions): Mention other optimization
1836 (-findirect-inlining, -fpartial-inlining): Same.
1837 (-finline-functions-called-once): Same.
1838 (-freorder-blocks-and-partition): Same.
1840 2018-04-09 Jan Hubicka <jh@suse.cz>
1843 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
1844 jumps; choose last target that matches the criteria (i.e.
1845 no partition changes for non-crossing jumps).
1846 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
1847 support for redirecting crossing jumps to non-crossing.
1849 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
1851 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
1852 also for naked functions.
1854 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
1856 * config/arc/arc.md (add_shift): New pattern.
1857 (add_shift2): Likewise.
1858 (sub_shift): Likewise.
1859 (sub_shift_cmp0_noout): Likewise.
1860 (compare_si_ashiftsi): Likewise.
1861 (xbfu_cmp0_noout): New combine pattern.
1862 (xbfu_cmp0"): Likewise.
1863 (movsi_set_cc_insn): Place the predicable variant first.
1864 (commutative_binary_cmp0_noout): Remove clobber.
1865 (commutative_binary_cmp0): New pattern.
1866 (noncommutative_binary_cmp0): Likewise.
1867 (noncommutative_binary_cmp0_noout): Likewise.
1868 (noncommutative_binary_comparison_result_used): Removed.
1869 (rsub_cmp0): New pattern.
1870 (rsub_cmp0_noout): Likewise.
1871 (extzvsi): Changed, keep only meaningful variants.
1872 (SQH, SEZ): New iterators.
1873 (SQH_postfix): New mode attribute.
1874 (SEZ_prefix): New code attribute.
1875 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
1876 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
1877 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
1879 (noncommutative_operator): Check the availability of barrel
1882 2018-04-09 Richard Biener <rguenther@suse.de>
1884 PR tree-optimization/85284
1885 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
1886 Only use the niter constraining form of simple_iv when the exit
1889 2018-04-09 Tom de Vries <tom@codesourcery.com>
1892 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
1893 (define_expand "*memory_barrier"): New define_expand.
1894 (define_insn "memory_barrier"): New insn.
1896 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1898 PR rtl-optimization/80463
1899 PR rtl-optimization/83972
1900 PR rtl-optimization/83480
1902 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
1903 correct producer for the insn.
1904 (tidy_control_flow): Fixup seqnos in case of debug insns.
1906 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1908 PR rtl-optimization/83913
1910 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
1911 different sched-times when merging exprs.
1913 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1915 PR rtl-optimization/83962
1917 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
1918 tidy_fallthru_edge and tidy_control_flow.
1920 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1922 PR rtl-optimization/83530
1924 * sel-sched.c (force_next_insn): New global variable.
1925 (remove_insn_for_debug): When force_next_insn is true, also leave only
1926 next insn in the ready list.
1927 (sel_sched_region): When the region wasn't scheduled, make another pass
1928 over it with force_next_insn set to 1.
1930 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
1932 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
1934 * config/nds32/constants.md (unspec_volatile_element): Add enum values
1935 for interrupt control.
1936 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
1937 functions for interrupt control.
1938 * config/nds32/nds32-intrinsic.md: Likewise.
1939 * config/nds32/nds32_intrinsic.h: Likewise.
1940 * config/nds32/nds32.h (nds32_builtins): Likewise.
1942 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
1944 * config/nds32/nds32.c (nds32_init_machine_status,
1945 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
1946 strict_aligned_p field.
1947 (nds32_expand_to_rtl_hook): New function.
1948 (TARGET_EXPAND_TO_RTL_HOOK): Define.
1949 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
1951 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1952 Chung-Ju Wu <jasonwucj@gmail.com>
1954 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
1955 * config/nds32/nds32-n7.md: New file.
1956 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
1957 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
1959 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
1960 * config/nds32/nds32.md (pipeline_model): Add n7.
1961 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
1962 * config/nds32/pipelines.md: Include n7 settings.
1964 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1965 Chung-Ju Wu <jasonwucj@gmail.com>
1967 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
1968 * config/nds32/nds32-e8.md: New file.
1969 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
1970 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
1972 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
1973 * config/nds32/nds32.md (pipeline_model): Add e8.
1974 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
1975 * config/nds32/pipelines.md: Include e8 settings.
1977 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1978 Chung-Ju Wu <jasonwucj@gmail.com>
1980 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
1981 * config/nds32/nds32-n8.md: New file.
1982 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
1983 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
1985 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
1986 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
1987 * config/nds32/nds32.md (pipeline_model): Add n8.
1988 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
1989 * config/nds32/pipelines.md: Include n8 settings.
1991 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1992 Chung-Ju Wu <jasonwucj@gmail.com>
1994 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
1995 * config/nds32/nds32-n9-2r1w.md: New file.
1996 * config/nds32/nds32-n9-3r2w.md: New file.
1997 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
1998 nds32_register_ports): New or modify for cpu n9.
1999 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
2001 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
2002 * config/nds32/nds32-utils.c: New file.
2003 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
2004 TARGET_MUL_SLOW): Define.
2005 * config/nds32/nds32.md (pipeline_model): New attribute.
2006 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
2007 New options that support cpu n9.
2008 * config/nds32/pipelines.md: Include n9 settings.
2009 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
2011 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
2013 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
2014 information if necessary.
2015 (output_cond_branch_compare_zero): Likewise.
2016 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
2017 (nds32_target_alignment): Refine for alignment.
2018 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
2019 (FUNCTION_BOUNDARY): Modify.
2020 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
2022 * config/nds32/nds32.opt (malways-align, malign-functions): New.
2024 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
2026 * config/nds32/constants.md (unspec_volatile_element): Add values for
2027 TLB operation and data prefetch.
2028 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
2029 functions for TLB operation and data prefetch.
2030 * config/nds32/nds32-intrinsic.md: Likewise.
2031 * config/nds32/nds32_intrinsic.h: Likewise.
2032 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
2033 (nds32_print_operand): Likewise.
2034 * config/nds32/nds32.h (nds32_builtins): Likewise.
2036 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
2037 Andrew Pinski <pinsika@gcc.gnu.org>
2040 * match.pd: Use constant_boolean_node of correct type instead of
2041 boolean_true_node or boolean_false_node for simplifying
2042 pointer comparisons to zero.
2044 2018-04-07 Jakub Jelinek <jakub@redhat.com>
2046 PR tree-optimization/80021
2047 * tree.c (verify_type_variant): Make error call in verify_variant_match
2048 translatable and remove final full stop.
2050 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2052 * config/nds32/constants.md (unspec_volatile_element): Add
2053 UNSPEC_VOLATILE_EH_RETURN.
2054 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
2055 nds32_output_stack_pop): Support dwarf exception handling process.
2056 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
2057 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
2058 exception handling process.
2059 (nds32_compute_stack_frame): Likewise.
2060 (nds32_return_addr_rtx): Likewise.
2061 (nds32_initial_elimination_offset): Likewise.
2062 (nds32_expand_prologue): Likewise.
2063 (nds32_expand_epilogue): Likewise.
2064 (nds32_dynamic_chain_address): New function.
2065 * config/nds32/nds32.h (machine_function): Add fields for dwarf
2067 (DYNAMIC_CHAIN_ADDRESS): Define.
2068 (EH_RETURN_DATA_REGNO): Define.
2069 (EH_RETURN_STACKADJ_RTX): Define.
2070 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
2071 patterns for dwarf exception handling.
2073 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2075 * config/nds32/nds32.h: Clean up obsolete macros.
2077 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2079 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2080 Add enum values for particular instructions.
2081 * config/nds32/nds32-intrinsic.c: Implementation of expanding
2082 particular intrinsic functions.
2083 * config/nds32/nds32-intrinsic.md: Likewise.
2084 * config/nds32/nds32_intrinsic.h: Likewise.
2085 * config/nds32/nds32.h (nds32_builtins): Likewise.
2086 * config/nds32/nds32.md (type): Add pbsad and pbsada.
2087 (btst, ave): New patterns for particular instructions.
2089 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2091 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2092 Add enum values for atomic load/store and memory sync.
2093 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
2095 * config/nds32/nds32-intrinsic.md: Likewise.
2096 * config/nds32/nds32_intrinsic.h: Likewise.
2097 * config/nds32/nds32.h (nds32_builtins): Likewise.
2099 2018-04-07 Jakub Jelinek <jakub@redhat.com>
2101 PR tree-optimization/85257
2102 * fold-const.c (native_encode_vector): If not all elts could fit
2103 and off is -1, return 0 rather than offset.
2104 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
2105 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
2106 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
2107 adjust buffer in native_interpret_expr call.
2109 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2111 * config/nds32/constants.md (unspec_volatile_element): Add cache
2112 control enum values.
2113 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
2114 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
2115 * config/nds32/nds32.c (nds32_cctl_names): New.
2116 (nds32_print_operand): Handle cache control register names.
2117 * config/nds32/nds32.h (nds32_builtins): New enum values.
2118 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
2120 * config/nds32/nds32.md (type): Add mmu.
2121 * config/nds32/pipelines.md (simple_insn): Add mmu.
2123 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2125 * config/nds32/nds32.md (type): Remove call.
2126 * config/nds32/pipelines.md (simple_insn): Likewise.
2128 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2130 * config/nds32/constants.md (unspec_volatile_element): Add
2131 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
2132 UNSPEC_VOLATILE_FMFCFG.
2133 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
2134 description for fmfcfg and fmfcsr.
2135 (bdesc_1arg): Add fmtcsr.
2136 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
2137 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
2138 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
2139 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
2140 unspec_fmfcfg): New patterns.
2141 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
2142 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
2143 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
2144 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
2145 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
2146 __nds32__fmfcfg): Define.
2148 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2150 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
2151 intrinsic register names.
2152 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
2153 intrinsic register enum values and macros.
2155 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2157 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
2158 for load/store addressing form.
2159 (nds32_print_operand_address): Likewise.
2161 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
2164 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
2165 based on LABEL_REF. Remove useless assertion.
2166 (pic_address_needs_scratch): Fix formatting.
2167 (sparc_legitimize_pic_address): Minor tweaks.
2168 (sparc_delegitimize_address): Adjust assertion accordingly.
2169 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
2170 into symbolic_operand.
2171 (movsi_high_pic_label_ref): Likewise.
2172 (movsi_lo_sum_pic_label_ref): Likewise.
2173 (movdi_pic_label_ref): Likewise.
2174 (movdi_high_pic_label_ref): Likewise.
2175 (movdi_lo_sum_pic_label_ref): Likewise.
2177 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
2179 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
2180 custom LIB_SPEC setup.
2182 2018-04-06 Ruslan Bukin <br@bsdpad.com>
2183 Kito Cheng <kito.cheng@gmail.com>
2185 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
2186 * config/riscv/freebsd.h: New.
2188 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
2190 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
2191 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
2194 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
2195 Kito Cheng <kito.cheng@gmail.com>
2197 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
2198 nds32_output_call, nds32_symbol_binds_local_p): New functions.
2199 * config/nds32/nds32-protos.h (nds32_output_call,
2200 nds32_output_return): Declare.
2201 * config/nds32/nds32.md: Refine all the call and return patterns.
2203 2018-04-06 Jakub Jelinek <jakub@redhat.com>
2206 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
2207 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
2209 PR rtl-optimization/84872
2210 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
2211 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
2214 2018-04-06 Tamar Christina <tamar.christina@arm.com>
2216 * expr.c (copy_blkmode_to_reg): Revert 254862.
2217 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
2219 2018-04-06 Richard Biener <rguenther@suse.de>
2222 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
2223 after seeing a component reference with an adjacent field. Treat
2224 refs to arrays at struct end of external decls similar to
2225 refs to unconstrained commons.
2227 2018-04-06 Jakub Jelinek <jakub@redhat.com>
2230 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
2231 look through SAVE_EXPRs with non-side-effects argument. Adjust
2233 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
2236 2018-04-06 Richard Biener <rguenther@suse.de>
2239 * alias.c (find_base_term): New wrapper around find_base_term
2240 unwinding CSELIB_VAL_PTR changes.
2241 (find_base_term): Do not restore CSELIB_VAL_PTR during the
2244 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2246 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
2248 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
2249 constant definitions.
2250 ("nop"): lr 0,0 -> nopr r0
2251 ("nop_lr0", "nop_lr1"): New insn definitions.
2253 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
2255 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
2256 NDS32_V3PUSH_AVAILABLE_P macro.
2258 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
2259 Chung-Ju Wu <jasonwucj@gmail.com>
2261 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
2262 (nds32*-*-*): Add float and fpu_config into supported_defaults.
2263 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
2264 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
2265 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
2266 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
2267 * config/nds32/constraints.md: New constraints and checking for hard
2268 float configuration.
2269 * config/nds32/iterators.md: New mode iterator and attribute for hard
2270 float configuration.
2271 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
2273 * config/nds32/nds32-fpu.md: New file.
2274 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
2275 deal with hard float code generation.
2276 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
2278 (abi_type, float_reg_number): New enum type.
2279 * config/nds32/nds32-predicates.c: New predicates for hard float.
2280 * config/nds32/nds32-protos.h: Declare functions for hard float.
2281 * config/nds32/nds32.c: Implementation for hard float configuration.
2282 * config/nds32/nds32.h: Definitions for hard float configuration.
2283 * config/nds32/nds32.md: Include hard float machine description and
2284 modify patterns for hard float configuration.
2285 * config/nds32/nds32.opt: New options for hard float configuration.
2286 * config/nds32/predicates.md: New predicates for hard float
2289 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
2291 * common/config/nds32/nds32-common.c
2292 (nds32_option_optimization_table): Enable -mreleax-hint by default.
2294 2018-04-05 Jakub Jelinek <jakub@redhat.com>
2297 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
2298 CONSTRUCTOR_ELT (ctor, ...)->value.
2300 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
2303 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
2305 2018-04-05 Tom de Vries <tom@codesourcery.com>
2308 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
2311 2018-04-05 Shiva Chen <shiva0217@gmail.com>
2312 Kito Cheng <kito.cheng@gmail.com>
2314 * config/nds32/constraints.md (U33): Fine-tune checking condition.
2315 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
2316 * config/nds32/nds32.h (nds32_16bit_address_type): Add
2317 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
2319 2018-04-05 Shiva Chen <shiva0217@gmail.com>
2320 Kito Cheng <kito.cheng@gmail.com>
2322 * config/nds32/constraints.md (Ufe): New memory constraint.
2323 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
2324 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
2325 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
2327 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
2328 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
2330 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2332 * config/nds32/nds32.md: Use optimize_size in the condition for
2333 alu-shift instructions.
2335 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2337 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
2339 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2341 * config/nds32/nds32.md (negsi2): Refine pattern.
2343 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
2344 Chung-Ju Wu <jasonwucj@gmail.com>
2346 * config/nds32/iterators.md (shift_rotate): New code iterator.
2347 (shift): New code attribute.
2348 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
2349 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
2350 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
2351 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
2352 bit-wise operations.
2353 (andsi3, *andsi3): Ditto.
2354 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
2355 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
2356 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
2357 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
2358 nds32_ior_operand, nds32_xor_operand): New predicates.
2360 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2362 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
2363 (addsi3, subsi3): ... this.
2365 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2367 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
2369 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2371 * config/nds32/nds32.md: Adjust indention.
2373 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
2375 * config/nds32/nds32.md (feature): New attribute.
2377 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2379 * config/nds32/nds32.md (subtype): New attribute.
2381 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
2384 * config/arm/arm-builtins.c (arm_expand_builtin): Change
2385 expansion to perform a bitwise AND of the argument followed by a
2386 boolean negation of the result.
2388 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
2390 PR rtl-optimization/84878
2391 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
2392 the basic block. Assert the use reference is not artificial and that
2393 it has an associated insn.
2395 2018-04-04 Michael Matz <matz@suse.de>
2397 * builtins.c (compute_objsize): Pass correct operand
2398 to array_at_struct_end_p.
2400 2018-04-04 Richard Biener <rguenther@suse.de>
2403 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
2404 from contexts for DINFO_LEVEL_TERSE and below.
2406 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
2408 * config/nds32/nds32-doubleword.md (move_<mode>): Require
2409 resiter_operand condition.
2410 * config/nds32/nds32.md (*move<mode>): Ditto.
2412 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
2413 Monk Chiang <sh.chiang04@gmail.com>
2415 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
2417 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2419 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
2421 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2422 Kito Cheng <kito.cheng@gmail.com>
2424 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
2425 nds32_cond_code_str, output_cond_branch,
2426 output_cond_branch_compare_zero, nds32_expand_cbranch,
2427 nds32_expand_cstore, nds32_expand_movcc,
2428 nds32_output_cbranchsi4_equality_zero,
2429 nds32_output_cbranchsi4_equality_reg,
2430 nds32_output_cbranchsi4_equality_reg_or_const_int,
2431 nds32_output_cbranchsi4_greater_less_zero: New functions.
2432 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
2433 nds32_expand_cstore, nds32_expand_movcc,
2434 nds32_output_cbranchsi4_equality_zero,
2435 nds32_output_cbranchsi4_equality_reg,
2436 nds32_output_cbranchsi4_equality_reg_or_const_int,
2437 nds32_output_cbranchsi4_greater_less_zero): Declare.
2438 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
2439 nds32_rimm11s_operand): New predicates.
2440 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
2441 * config/nds32/nds32.md: Rewrite all the branch and conditional move
2444 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
2446 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
2447 * config/nds32/nds32.md: Ditto.
2448 * config/nds32/pipelines.md: Ditto.
2450 2018-04-04 Richard Biener <rguenther@suse.de>
2452 PR tree-optimization/85168
2453 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
2454 propagating abnormals.
2456 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2458 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
2460 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2461 Kito Cheng <kito.cheng@gmail.com>
2463 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
2464 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
2465 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
2466 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
2467 * config/nds32/nds32.md (sibcall_internal): New.
2468 (sibcall_register): Remove.
2469 (sibcall_immediate): Remove.
2470 (sibcall_value_internal): New.
2471 (sibcall_value_register): Remove.
2472 (sibcall_value_immediate): Remove.
2473 * config/nds32/predicates.md (nds32_general_register_operand): New.
2474 (nds32_call_address_operand): New.
2476 2018-04-03 Jakub Jelinek <jakub@redhat.com>
2478 PR rtl-optimization/85167
2479 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
2480 bb_defs if *split_p, instead preinitialize it to NULL.
2482 PR tree-optimization/85156
2483 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
2484 evaluating the argument multiple times.
2486 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
2488 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
2490 (_mm_cvtpd_ps): Likewise.
2491 (_mm_cvttpd_epi32): Likewise.
2492 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
2493 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
2494 vector, pixel, and bool following altivec.h include.
2496 2018-04-03 Martin Sebor <msebor@redhat.com>
2498 * doc/extend.texi (Common Function Attributes): Clarify.
2499 (const attribute): Likewise.
2500 (pure attribute): Likewise.
2502 2018-04-03 Jakub Jelinek <jakub@redhat.com>
2505 * config/i386/i386.c (ix86_expand_vector_set): Use
2506 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
2508 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
2510 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
2511 instructions when changing rounding bits to preserve precision bits
2512 in the x87 control word.
2514 2018-04-03 Martin Liska <mliska@suse.cz>
2516 PR tree-optimization/82491
2517 * rtl.h (strip_offset_and_add): Replace += suboffset with
2518 poly_uint64 () + suboffset.
2520 2018-03-29 Martin Liska <mliska@suse.cz>
2521 Martin Jambor <mjambor@suse.cz>
2524 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
2525 param_type is not an integral or pointer type.
2527 2018-04-03 Richard Biener <rguenther@suse.de>
2529 * sese.h (recompute_all_dominators): Remove.
2531 2018-04-02 Martin Sebor <msebor@redhat.com>
2533 * doc/invoke.texi (-Wrestrict): Fix typos.
2535 2018-04-02 Jim Wilson <jimw@sifive.com>
2537 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
2538 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
2539 (<optab>di3, <optab>si3_extend): Likewise.
2540 (<optab>si3_mask, <optab>si3_mask_1): New.
2541 (<optab>di3_mask, <optab>di3_mask_1): New.
2542 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
2543 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
2544 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
2546 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
2548 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
2551 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
2553 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
2554 (nds32_canonicalize_comparison): New function.
2556 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
2557 Kito Cheng <kito.cheng@gmail.com>
2558 Kuan-Lin Chen <kuanlinchentw@gmail.com>
2560 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
2561 * config/nds32/constants.md (unspec_volatile_element): Add
2562 UNSPEC_VOLATILE_RELAX_GROUP.
2563 * config/nds32/nds32-relax-opt.c: New file.
2564 * config/nds32/nds32-predicates.c
2565 (nds32_symbol_load_store_p): New function.
2566 * config/nds32/nds32-protos.h
2567 (nds32_symbol_load_store_p): Declare function.
2568 (make_pass_nds32_relax_opt): Declare new rtl pass function.
2569 * config/nds32/nds32.c
2570 (nds32_register_pass): New function to register pass.
2571 (nds32_register_passes): New function to register passes.
2572 * config/nds32/nds32.md (relax_group): New pattern.
2573 * config/nds32/nds32.opt (mrelax-hint): New option.
2574 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
2576 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
2578 * config/nds32/t-nds32: Modify files dependency.
2580 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
2582 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
2583 (PROFILE_HOOK): Define its implementation.
2585 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
2587 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
2588 type and 32-bit size.
2590 2018-04-01 Jakub Jelinek <jakub@redhat.com>
2593 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
2594 (V_128_256): New mode iterator.
2595 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
2596 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
2597 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
2599 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
2600 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
2602 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
2605 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
2606 NaN inputs correctly.
2608 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
2611 * config/rs6000/vsx.md (??r): New mode attribute.
2612 (*vsx_mov<mode>_64bit): Use it.
2613 (*vsx_mov<mode>_32bit): Likewise.
2615 2018-03-30 Martin Sebor <msebor@redhat.com>
2617 PR tree-optimization/84818
2618 * builtins.c (check_access): Use warning_n.
2620 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2623 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
2625 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
2628 2018-03-30 Julia Koval <julia.koval@intel.com>
2631 * x86-tune.def (movx, partial_reg_dependency): Enable for
2634 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
2637 * lra-constraints.c (process_alt_operands): Move setting
2638 this_alternative_matches below.
2640 2018-03-29 Martin Liska <mliska@suse.cz>
2643 * doc/invoke.texi: Document how LTO works with debug info.
2644 Describe auto-load support of binutils. Mention 'x86-64'
2645 as valid option value of -march option.
2647 2018-03-29 Jakub Jelinek <jakub@redhat.com>
2649 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
2652 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
2653 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
2654 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
2657 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
2660 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
2661 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
2662 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
2663 for RS6000_BTM_POWERPC64.
2664 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
2665 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
2666 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
2671 2018-03-28 Carl Love <cel@us.ibm.com>
2674 2017-09-27 Carl Love <cel@us.ibm.com>
2676 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
2677 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
2678 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
2681 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2683 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
2684 instead of __vector bool.
2685 (_mm_max_pu8): Likewise.
2686 (_mm_min_pi16): Likewise.
2688 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
2691 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
2692 (DIVWEUO): Likewise.
2694 (DIVDEUO): Likewise.
2695 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
2696 DIVWEUO and DIVDEUO.
2697 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
2698 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
2699 (div_extend): Likewise.
2700 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
2702 (__builtin_divweuo): Likewise.
2703 (__builtin_divdeo): Likewise.
2704 (__builtin_divdeuo): Likewise.
2706 2018-03-28 Jakub Jelinek <jakub@redhat.com>
2709 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
2710 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
2712 PR tree-optimization/82004
2713 * gimple-match-head.c (optimize_pow_to_exp): New function.
2714 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
2715 Don't fold to exp if optimize_pow_to_exp is false.
2717 2018-03-28 Martin Liska <mliska@suse.cz>
2720 * calls.c (initialize_argument_information): Fix trailing space.
2721 * common.opt: Fix typo and provide better explanation for
2722 -fsanitize-coverage option.
2723 * config/i386/i386.opt: Fix typo.
2725 2018-03-28 Jakub Jelinek <jakub@redhat.com>
2726 Martin Liska <mliska@suse.cz>
2729 * gimplify.c (asan_poison_variable): Don't do the check for
2730 gimplify_omp_ctxp here.
2731 (gimplify_decl_expr): Do it here.
2732 (gimplify_target_expr): Likewise.
2734 2018-03-28 Martin Liska <mliska@suse.cz>
2737 * config/i386/i386.c (ix86_function_arg_advance): Do not call
2738 chkp_type_bounds_count if MPX is not enabled.
2740 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
2742 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
2744 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
2747 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
2748 function to create the function decl for complex long double
2749 multiply and divide for -mabi=ieeelongdouble.
2750 (init_float128_ieee): Call it.
2752 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
2755 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
2756 -fcf-protection=branch -mibt.
2757 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
2759 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2762 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
2764 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
2767 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
2768 extern array declarations.
2770 2018-03-27 Richard Biener <rguenther@suse.de>
2773 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
2774 explicit single_use checks.
2776 2018-03-27 Richard Biener <rguenther@suse.de>
2778 PR tree-optimization/85082
2779 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
2782 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2784 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
2785 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
2786 Turn on fasynchronous-unwind-tables and funwind-tables.
2788 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
2791 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
2792 (*bmi_blsr_<mode>_ccz): Ditto.
2794 2018-03-26 Tom de Vries <tom@codesourcery.com>
2796 PR tree-optimization/85063
2797 * omp-general.c (offloading_function_p): New function. Factor out
2799 * omp-offload.c (pass_omp_target_link::gate): ... here.
2800 * omp-general.h (offloading_function_p): Declare.
2801 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
2802 with attribute omp declare target for offloading functions.
2804 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
2806 PR tree-optimization/84005
2807 * tree-data-ref.h (get_base_for_alignment): Declare.
2808 * tree-data-ref.c (get_base_for_alignment_1): New function.
2809 (get_base_for_alignment): Likewise.
2810 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
2811 get_base_for_alignment to find a suitable base object, instead
2812 of always using drb->base_address.
2814 2018-03-23 Jakub Jelinek <jakub@redhat.com>
2817 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
2818 known size by default.
2820 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
2823 * lra-constraints.c (process_alt_operands): Don't match BLKmode
2824 and non BLKmode operands.
2826 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2829 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
2830 Clean up attributes.
2832 2018-03-23 Richard Biener <rguenther@suse.de>
2835 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
2836 we are going to emit early debug for LTO.
2838 2018-03-23 Jakub Jelinek <jakub@redhat.com>
2841 * function.c (match_asm_constraints_1): Don't optimize if input
2842 doesn't satisfy general_operand predicate for output's mode.
2845 * alias.c (write_dependence_p): Don't require for x_canonicalized
2846 non-VOIDmode if x has VOIDmode.
2849 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
2850 just don't try to optimize it rather than assert it never happens.
2852 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
2854 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
2855 macro expansions for definition of ST_INTERNAL_<mode> and
2856 LD_INTERNAL_<mode> builtins.
2857 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
2859 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
2861 (altivec_expand_st_builtin): Likewise.
2862 (altivec_expand_builtin): Remove calls to deleted functions.
2863 (rs6000_address_for_altivec): Delete this function.
2864 * config/rs6000/vector.md: Remove expands for
2865 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
2867 2018-03-22 Sudakshina Das <sudi.das@arm.com>
2870 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
2871 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
2872 re-computing once computed.
2873 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
2874 (arm_init_machine_status): Initialize
2875 machine->static_chain_stack_bytes.
2877 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
2880 * doc/extend.texi: Add four new prototypes for vec_ld.
2881 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
2882 definitions for more logical presentation.
2883 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
2884 entries for V1TI variants of __builtin_altivec_ld builtin.
2885 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
2886 handling of V1TI variant of LVX icode pattern.
2887 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
2888 (rs6000_gimple_fold_builtin): Likewise.
2889 (altivec_init_builtins): Add code to define
2890 __builtin_altivec_lvx_v1ti function.
2892 2018-03-22 Jakub Jelinek <jakub@redhat.com>
2895 * function.c (match_asm_constraints_1): Don't do the optimization
2896 if input isn't a REG, SUBREG, MEM or constant.
2898 2018-03-22 Tom de Vries <tom@codesourcery.com>
2900 PR tree-optimization/84956
2901 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
2902 bb_has_abnormal_pred.
2904 2018-03-22 Jakub Jelinek <jakub@redhat.com>
2907 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
2908 DECL_INITIAL (decl) to decl at the end.
2909 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
2912 2018-03-21 Joseph Myers <joseph@codesourcery.com>
2914 * doc/extend.texi (__builtin_tgmath): Document when complex
2915 integer types are treated as _Complex _Float64.
2917 2018-03-21 Tom de Vries <tom@codesourcery.com>
2919 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
2921 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2923 PR tree-optimization/84960
2924 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
2925 if it is ENTRY block, move them into single succ of ENTRY in that case.
2927 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
2929 PR tree-optimization/84811
2930 * poly-int.h (poly_span_traits): Remove the T3 parameter and
2931 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
2932 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
2933 (known_subrange_p): Update accordingly. Cast each value involved
2934 in the size comparison, rather than casting the result of the
2937 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2939 PR tree-optimization/84982
2940 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
2941 by flipping the least significant bit rather than all bits from
2942 bitpos to bitpos + bitsize - 1.
2944 2018-03-21 Nathan Sidwell <nathan@acm.org>
2946 * doc/extend.texi (Deprecated Features): Remove mention of
2947 long-deleted deprecations.
2949 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2952 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
2953 * configure: Regenerate.
2955 2018-03-21 Tom de Vries <tom@codesourcery.com>
2957 PR tree-optimization/83126
2958 * tree-parloops.c (num_phis): New function.
2959 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
2961 2018-03-21 Nathan Sidwell <nathan@acm.org>
2963 * doc/extend.texi (Deprecated Features): Update deprecated flags,
2964 mention anon-struct/union members and trailing attributes.
2966 2018-03-21 Bin Cheng <bin.cheng@arm.com>
2968 PR tree-optimization/84969
2969 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
2970 builtin memset partitions if they set different rhs values.
2972 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2974 PR rtl-optimization/84989
2975 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
2976 VEC_DUPLICATE with scalar result mode.
2978 2018-03-21 Martin Liska <mliska@suse.cz>
2981 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
2982 not intended return statement.
2984 2018-03-21 Martin Liska <mliska@suse.cz>
2987 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
2988 (chkp_find_bound_slots_1): Limit number of iterations.
2990 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
2993 * Minor grammar fixes for x86 options.
2995 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2998 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
2999 holding REG_CFA_RESTORE notes, instead turn them into a USE.
3001 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
3004 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
3005 (altivec_lvx_<mode>_1op): Likewise.
3006 (altivec_stvx_<mode>_2op): Likewise.
3007 (altivec_stvx_<mode>_1op): Likewise.
3008 (altivec_lvx_<VM2:mode>): New define_expand.
3009 (altivec_stvx_<VM2:mode>): Likewise.
3010 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
3011 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
3012 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
3013 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
3014 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
3015 (rs6000_gen_lvx): Likewise.
3016 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
3017 (altivec_expand_stv_builtin): Likewise.
3018 (altivec_expand_builtin): Likewise.
3019 * config/rs6000/vector.md: Likewise.
3021 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3024 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
3027 2018-03-20 Richard Biener <rguenther@suse.de>
3030 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
3031 sign-conversions as zero, fall back to standard scalar_stmt
3034 2018-03-20 Martin Liska <mliska@suse.cz>
3037 * predict.c (rebuild_frequencies): Handle case when we have
3038 PROFILE_ABSENT, but flag_guess_branch_prob is false.
3040 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3043 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
3044 flag_section_anchors.
3045 * varasm.c (use_blocks_for_decl_p): Remove hack for
3046 dw2_force_const_mem.
3049 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
3051 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
3052 be created, use lowpart_subreg of operands[0] rather than operands[0]
3054 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
3055 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
3056 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
3057 and n constraint instead of aarch64_shift_imm_di and Usd.
3058 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
3059 (*aarch64_<optab>_reg_minus<mode>3): ... this.
3061 2018-03-20 Sudakshina Das <sudi.das@arm.com>
3064 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
3065 to favor GPR over NEON registers.
3066 (<shift>di3_neon): Likewise.
3068 2018-03-20 Tom de Vries <tom@codesourcery.com>
3071 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
3072 (nvptx_process_pars): Emit bar.sync asap and alap.
3074 2018-03-20 Tom de Vries <tom@codesourcery.com>
3077 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
3078 seen_label if seen_label is already set.
3080 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3083 * config/i386/i386.c (fold_builtin_cpu): For features above 31
3084 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
3085 Use 1U instead of 1. Formatting fixes.
3088 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
3089 instead of TREE_TYPE (s1) for the return value.
3091 2018-03-19 Jakub Jelinek <jakub@redhat.com>
3093 PR tree-optimization/84946
3094 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
3095 bitsize + bitsize in poly_uint64 rather than poly_int64.
3098 * dwarf2asm.c: Include fold-const.c.
3099 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
3100 of decl rather than decl itself.
3102 PR rtl-optimization/84643
3103 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
3105 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
3108 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
3109 calling assemble_variable.
3111 2018-03-19 Sudakshina Das <sudi.das@arm.com>
3114 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
3115 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
3117 2018-03-19 Jim Wilson <jimw@sifive.com>
3120 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
3121 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
3122 (riscv_first_stack_step): Likewise.
3123 (riscv_option_override): Use STACK_BOUNDARY instead of
3125 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
3127 (BIGGEST_ALIGNMENT): Set to 128.
3128 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
3129 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
3132 2018-03-19 Richard Biener <rguenther@suse.de>
3134 PR tree-optimization/84933
3135 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
3136 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
3138 2018-03-19 Richard Biener <rguenther@suse.de>
3140 PR tree-optimization/84859
3141 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
3142 (cond_if_else_store_replacement): Perform sinking operation on
3143 single-store BBs regardless of MAX_STORES_TO_SINK setting.
3144 Generalize what a BB with a single eligible store is.
3146 2018-03-19 Richard Biener <rguenther@suse.de>
3148 PR tree-optimization/84929
3149 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
3150 chrec_is_positive against non-chrec arg.
3152 2018-03-19 Tamar Christina <tamar.christina@arm.com>
3155 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
3157 2018-03-18 Martin Liska <mliska@suse.cz>
3159 PR rtl-optimization/84635
3160 * regrename.c (build_def_use): Use matches_mode only when
3163 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
3165 PR tree-optimization/84913
3166 * tree-vect-loop.c (vectorizable_reduction): Don't try to
3167 vectorize chains of COND_EXPRs.
3169 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3171 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
3173 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3175 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
3177 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3179 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
3181 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
3182 Kito Cheng <kito.cheng@gmail.com>
3184 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
3185 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
3186 (nds32_adjust_reg_alloc_order): New function.
3187 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
3189 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
3191 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
3192 nds32_print_operand, nds32_print_operand_address): Use
3193 HOST_WIDE_INT_PRINT_DEC instead.
3195 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
3197 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
3199 2018-03-17 Jakub Jelinek <jakub@redhat.com>
3202 * config/i386/i386.c (initial_ix86_tune_features,
3203 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
3205 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
3206 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
3207 rather than 1u << ix86_tune. Formatting fix.
3208 (ix86_option_override_internal): Change ix86_arch_mask from
3209 unsigned int to unsigned HOST_WIDE_INT, initialize to
3210 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
3211 (ix86_function_specific_restore): Likewise.
3213 2018-03-16 Jakub Jelinek <jakub@redhat.com>
3216 * postreload.c (reload_combine_recognize_pattern): Perform
3217 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
3218 truncate_int_for_mode the result for the destination's mode.
3221 * hsa-gen.c (mem_type_for_type): Fix comment typo.
3222 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
3224 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
3227 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
3230 * lra-assigns.c (lra_split_hard_reg_for): Don't use
3231 regno_allocno_class_array and sorted_pseudos.
3232 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
3233 insns where regno is used.
3235 2018-03-16 Martin Liska <mliska@suse.cz>
3238 * multiple_target.c (create_dispatcher_calls): Redirect
3239 reference in the symbol table.
3241 2018-03-16 Martin Liska <mliska@suse.cz>
3244 * multiple_target.c (create_dispatcher_calls): Redirect also
3247 2018-03-16 Jakub Jelinek <jakub@redhat.com>
3251 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
3252 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
3253 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
3255 2018-03-16 Julia Koval <julia.koval@intel.com>
3257 * doc/invoke.texi (Skylake Server): Add CLWB.
3258 Cannonlake): Remove CLWB.
3260 2018-03-16 Jakub Jelinek <jakub@redhat.com>
3262 PR tree-optimization/84841
3263 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
3265 (FLOAT_ONE_CONST_TYPE): Define.
3266 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
3267 (sort_by_operand_rank): Put entries with higher constant_type last
3268 rather than first to match comments.
3270 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
3272 * config/nios2/nios2.md (movsi_internal): Fix thinko in
3275 2018-03-15 Jakub Jelinek <jakub@redhat.com>
3278 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
3279 check and use address of target always.
3281 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
3284 * config/i386/i386.c (indirect_thunk_needed): Update comments.
3285 (indirect_thunk_bnd_needed): Likewise.
3286 (indirect_thunks_used): Likewise.
3287 (indirect_thunks_bnd_used): Likewise.
3288 (indirect_return_needed): New.
3289 (indirect_return_bnd_needed): Likewise.
3290 (output_indirect_thunk_function): Add a bool argument for
3292 (output_indirect_thunk_function): Don't generate alias for
3293 function return thunk.
3294 (ix86_code_end): Call output_indirect_thunk_function to generate
3295 function return thunks.
3296 (ix86_output_function_return): Set indirect_return_bnd_needed
3297 and indirect_return_needed instead of indirect_thunk_bnd_needed
3298 and indirect_thunk_needed.
3300 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
3302 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
3303 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
3304 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
3306 2018-03-15 David Malcolm <dmalcolm@redhat.com>
3307 Paul Hua <paul.hua.gm@gmail.com>
3310 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
3312 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
3314 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
3315 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
3318 2018-03-15 Tamar Christina <tamar.christina@arm.com>
3321 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
3322 instead of GET_MODE_SIZE when comparing Units.
3324 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
3327 * varasm.c (hash_section): Return an unchangeble hash value
3328 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
3329 Return !aarch64_can_use_per_function_literal_pools_p ().
3331 2018-03-15 Jakub Jelinek <jakub@redhat.com>
3334 * optabs.c (emit_conditional_move): Pass address of cmode's copy
3335 rather than address of cmode as last argument to prepare_cmp_insn.
3337 2018-03-15 Julia Koval <julia.koval@intel.com>
3339 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
3340 F_AVX512VNNI, F_AVX512BITALG): New.
3342 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
3345 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
3346 insn for floating-point loads and stores.
3348 2018-03-14 Carl Love <cel@us.ibm.com>
3350 * config/rs6000/rs6000-c.c: Add macro definitions for
3351 ALTIVEC_BUILTIN_VEC_PERMXOR.
3352 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
3353 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
3354 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
3355 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
3357 * config/doc/extend.texi: Add prototypes for vec_permxor.
3359 2018-03-14 David Malcolm <dmalcolm@redhat.com>
3362 * diagnostic-show-locus.c (class layout_point): Convert m_line
3363 from int to linenum_type.
3364 (line_span::comparator): Use linenum "compare" function when
3365 comparing line numbers.
3366 (test_line_span): New function.
3367 (layout_range::contains_point): Convert param "row" from int to
3369 (layout_range::intersects_line_p): Likewise.
3370 (layout::will_show_line_p): Likewise.
3371 (layout::print_source_line): Likewise.
3372 (layout::should_print_annotation_line_p): Likewise.
3373 (layout::print_annotation_line): Likewise.
3374 (layout::print_leading_fixits): Likewise.
3375 (layout::annotation_line_showed_range_p): Likewise.
3376 (struct line_corrections): Likewise for field m_row.
3377 (line_corrections::line_corrections): Likewise for param "row".
3378 (layout::print_trailing_fixits): Likewise.
3379 (layout::get_state_at_point): Likewise.
3380 (layout::get_x_bound_for_row): Likewise.
3381 (layout::print_line): Likewise.
3382 (diagnostic_show_locus): Likewise for locals "last_line" and
3384 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
3385 * input.c (selftest::test_linenum_comparisons): New function.
3386 (selftest::input_c_tests): Call it.
3387 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
3388 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
3389 * selftest.h (ASSERT_GT): New macro.
3390 (ASSERT_GT_AT): New macro.
3391 (ASSERT_LT): New macro.
3392 (ASSERT_LT_AT): New macro.
3394 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
3396 PR rtl-optimization/84780
3397 * combine.c (distribute_links): Don't make a link based on pc_rtx.
3399 2018-03-14 Martin Liska <mliska@suse.cz>
3401 * tree.c (record_node_allocation_statistics): Use
3402 get_stats_node_kind.
3403 (get_stats_node_kind): New function extracted from
3404 record_node_allocation_statistics.
3405 (free_node): Use get_stats_node_kind.
3407 2018-03-14 Richard Biener <rguenther@suse.de>
3409 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
3410 that the value-set of ANTIC_IN doesn't grow.
3413 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
3415 (BB_VISITED_WITH_VISITED_SUCCS): New define.
3416 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
3418 2018-03-14 Julia Koval <julia.koval@intel.com>
3420 * config.gcc (icelake-client, icelake-server): New.
3422 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
3423 (initial_ix86_arch_features): Ditto.
3424 (PTA_SKYLAKE): Add SGX.
3425 (PTA_ICELAKE): Remove.
3426 (PTA_ICELAKE_CLIENT): New.
3427 (PTA_ICELAKE_SERVER): New.
3428 (ix86_option_override_internal): Split up icelake on icelake client and
3430 (get_builtin_code_for_version): Ditto.
3431 (fold_builtin_cpu): Ditto.
3432 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
3433 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
3434 * config/i386/i386.h (processor_type): Ditto.
3435 * doc/invoke.texi: Ditto.
3437 2018-03-14 Jakub Jelinek <jakub@redhat.com>
3440 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
3441 INTEGER_CST offset, add it together with bitpos / 8 and
3442 sign extend based on POINTER_SIZE.
3446 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
3449 * config/i386/constraints.md (Yc): New register constraint.
3450 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
3451 Use Yc constraint for alternative 2 of operand 0. Remove
3452 preferred_for_speed attribute.
3454 2018-03-14 Richard Biener <rguenther@suse.de>
3456 PR tree-optimization/84830
3457 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
3458 with the old one to avoid oscillations.
3460 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
3463 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
3465 (assign_by_spills): Return a flag of reload assignment failure.
3466 Do not process the reload assignment failures. Do not spill other
3467 reload pseudos if they has the same reg class. Update n if
3469 (lra_assign): Add a return arg. Set up from the result of
3470 assign_by_spills call.
3471 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
3472 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
3473 usage_insns if it is not NULL.
3474 (spill_hard_reg_in_range): New function.
3475 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
3476 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
3477 function prototypes.
3478 (lra_assign): Change prototype.
3479 * lra.c (lra): Add code to deal with fails by splitting hard reg
3482 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
3484 * config/riscv/riscv.opt (mrelax): New option.
3485 * config/riscv/riscv.c (riscv_file_start): Emit ".option
3486 "norelax" when riscv_mrelax is disabled.
3487 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
3489 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3492 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
3493 reassociation for int modes.
3495 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
3497 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
3498 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
3500 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
3501 * config/aarch64/aarch64-sve.md
3502 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
3503 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
3504 (*extend<mode><Vwide>2): Rename to...
3505 (aarch64_sve_extend<mode><Vwide>2): ...this.
3506 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
3507 renaming the old pattern to...
3508 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
3510 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
3511 define_expand, renaming the old pattern to...
3512 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
3513 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
3514 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
3515 account when deciding which SVE instruction the optab should use.
3516 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
3518 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
3520 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
3521 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
3522 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
3523 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
3524 (tlsdesc_small_<mode>): Turn a define_expand and use
3525 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
3526 (tlsdesc_small_advsimd_<mode>): ...this.
3527 (tlsdesc_small_sve_<mode>): New pattern.
3529 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
3531 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
3532 (UNSPEC_UMUL_HIGHPART): New constants.
3533 (MUL_HIGHPART): New int iteraor.
3534 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
3535 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
3537 (*<su>mul<mode>3_highpart): New define_insn.
3539 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
3542 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
3545 2018-03-13 Martin Liska <mliska@suse.cz>
3548 * (sem_item_optimizer::sem_item_optimizer): Initialize new
3550 (sem_item_optimizer::~sem_item_optimizer): Release it.
3551 (sem_item_optimizer::merge_classes): Register variable aliases.
3552 (sem_item_optimizer::fixup_pt_set): New function.
3553 (sem_item_optimizer::fixup_points_to_sets): Likewise.
3554 * ipa-icf.h: Declare new variables and functions.
3556 2018-03-13 Jakub Jelinek <jakub@redhat.com>
3559 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
3560 integer_pow2p@2 and test integer_pow2p in condition.
3561 (A < 0 ? C : 0): Similarly for @1.
3564 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
3565 characters starting at p contain '\0' character, don't look beyond
3569 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
3570 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
3573 * reg-stack.c (change_stack): Change update_end var from int to
3574 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
3575 also call set_block_for_insn on the newly added insns and rescan.
3578 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
3579 on the last operand.
3582 * tree.c (stabilize_reference_1): Return save_expr (e) for
3583 STATEMENT_LIST even if it doesn't have side-effects.
3585 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
3587 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
3589 2018-03-12 Renlin Li <renlin.li@arm.com>
3591 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
3592 aarch64_output_scalar_simd_mov_immediate.
3594 2018-03-12 Martin Sebor <msebor@redhat.com>
3596 PR tree-optimization/83456
3597 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
3598 for perfectly overlapping calls to memcpy.
3599 (gimple_fold_builtin_memory_chk): Same.
3600 (gimple_fold_builtin_strcpy): Handle no-warning.
3601 (gimple_fold_builtin_stxcpy_chk): Same.
3602 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
3604 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
3606 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
3607 parameter. Use it for SFmode.
3608 (rs6000_function_arg_advance_1): Adjust.
3609 (rs6000_function_arg): Adjust.
3610 (rs6000_gimplify_va_arg): Pass false for that new parameter.
3612 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
3614 PR rtl-optimization/84169
3615 PR rtl-optimization/84780
3616 * combine.c (can_combine_p): Check for a 2-insn combination whether
3617 the destination register is used between the two insns, too.
3619 2018-03-12 Richard Biener <rguenther@suse.de>
3621 PR tree-optimization/84803
3622 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
3623 for refs DR analysis didn't process.
3625 2018-03-12 Richard Biener <rguenther@suse.de>
3627 PR tree-optimization/84777
3628 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
3629 force-vectorize loops ignore whether we are optimizing for size.
3631 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
3633 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
3634 (TARGET_MD_ASM_ADJUST): Define.
3636 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
3637 Kito Cheng <kito.cheng@gmail.com>
3638 Chung-Ju Wu <jasonwucj@gmail.com>
3640 * config/nds32/nds32.c (nds32_compute_stack_frame,
3641 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
3642 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
3643 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
3644 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
3645 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
3646 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
3647 * config/nds32/nds32.md (prologue, epilogue): Use macro
3648 NDS32_V3PUSH_AVAILABLE_P to do checking.
3650 2018-03-11 Jakub Jelinek <jakub@redhat.com>
3653 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
3654 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
3655 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
3656 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
3657 addition of most attributes on !orig_type_die or the attribute not
3658 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
3660 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3661 Chung-Ju Wu <jasonwucj@gmail.com>
3663 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
3665 * config/nds32/nds32.opt (mvh): New option.
3667 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3668 Chung-Ju Wu <jasonwucj@gmail.com>
3670 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
3672 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
3673 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
3676 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3677 Chung-Ju Wu <jasonwucj@gmail.com>
3679 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
3681 * config/nds32/nds32-multiple.md (strlensi): New pattern.
3682 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
3684 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
3685 Kito Cheng <kito.cheng@gmail.com>
3686 Chung-Ju Wu <jasonwucj@gmail.com>
3688 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
3689 UNSPEC_FFMISM and UNSPEC_FLMISM.
3690 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
3691 for ffb, ffmism and flmism.
3692 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
3693 (unspec_ffmism): Ditto.
3694 (unspec_flmism): Ditto.
3695 (nds32_expand_builtin_impl): Check if string extension is available.
3696 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
3697 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
3699 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
3702 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
3705 * lra-assigns.c (assign_by_spills): Return a flag of reload
3706 assignment failure. Do not process the reload assignment
3707 failures. Do not spill other reload pseudos if they has the same
3709 (lra_assign): Add a return arg. Set up from the result of
3710 assign_by_spills call.
3711 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
3712 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
3713 usage_insns if it is not NULL.
3714 (spill_hard_reg_in_range): New function.
3715 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
3716 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
3717 function prototypes.
3718 (lra_assign): Change prototype.
3719 * lra.c (lra): Add code to deal with fails by splitting hard reg
3722 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
3725 * config/i386/i386.opt: Replace Enforcment with Enforcement.
3727 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
3730 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
3731 (dw_val_node): Add val_symbolic_view.
3732 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
3733 (symview_upper_bound): New.
3734 (new_line_info_table): Initialize symviews_since_reset.
3735 (dwarf2out_source_line): Count symviews_since_reset and set
3736 symview_upper_bound.
3737 (dw_val_equal_p): Handle symview.
3738 (add_AT_symview): New.
3739 (print_dw_val): Handle symview.
3740 (attr_checksum, attr_checksum_ordered): Likewise.
3741 (same_dw_val_p, size_of_die): Likewise.
3742 (value_format, output_die): Likewise.
3743 (add_high_low_attributes): Use add_AT_symview for entry_view.
3744 (dwarf2out_finish): Reset symview_upper_bound, clear
3747 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
3750 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
3751 Add strict argument and use it.
3752 (rs6000_split_multireg_move): Update for new strict argument.
3753 (mem_operand_gpr): Disallow all non-offsettable addresses.
3754 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
3756 2018-03-09 Jakub Jelinek <jakub@redhat.com>
3759 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
3760 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
3761 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
3764 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
3765 decl, use remap_type if we want to use the type.
3767 2018-03-09 Martin Sebor <msebor@redhat.com>
3769 PR tree-optimization/84526
3770 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
3772 (builtin_access::generic_overlap): Be prepared to handle non-array
3775 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
3777 PR rtl-optimization/84682
3778 * lra-constraints.c (process_address_1): Check is_address flag
3779 for address constraints.
3780 (process_alt_operands): Likewise.
3781 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
3782 preprocess_constraints.
3783 * recog.h (preprocess_constraints): Add oploc parameter.
3785 * recog.c (preprocess_constraints): Test address_operand for
3786 CT_ADDRESS constraints.
3788 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
3791 * lra-assigns.c (assign_by_spills): Return a flag of reload
3792 assignment failure. Do not process the reload assignment
3793 failures. Do not spill other reload pseudos if they has the same
3795 (lra_assign): Add a return arg. Set up from the result of
3796 assign_by_spills call.
3797 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
3798 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
3799 usage_insns if it is not NULL.
3800 (spill_hard_reg_in_range): New function.
3801 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
3802 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
3803 function prototypes.
3804 (lra_assign): Change prototype.
3805 * lra.c (lra): Add code to deal with fails by splitting hard reg
3808 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3811 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
3812 Accept complain bool parameter. Only emit errors if it is true.
3813 (arm_parse_cpu_option_name): Likewise.
3814 (arm_target_thumb_only): Adjust callers of the above.
3815 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
3816 prototype to take a default true bool parameter.
3817 (arm_parse_arch_option_name): Likewise.
3819 2018-03-09 David Malcolm <dmalcolm@redhat.com>
3820 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
3824 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
3825 * configure: Regenerate.
3826 * configure.ac ("linker --version-script option"): New.
3827 ("linker soname option"): New.
3829 2018-03-09 Richard Biener <rguenther@suse.de>
3831 PR tree-optimization/84775
3832 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
3833 immediate uses of predicate stmts and mark them modified.
3836 PR tree-optimization/84178
3837 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
3839 (version_loop_for_if_conversion): Delay update_ssa call.
3840 (tree_if_conversion): Delay update_ssa until after predicate
3843 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
3846 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
3847 when the function accesses prior frames.
3849 2018-03-08 Jakub Jelinek <jakub@redhat.com>
3852 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
3853 gen_llsym, otherwise call maybe_gen_llsym.
3856 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
3857 has ',' character inside of it.
3859 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3862 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
3863 as clobbering CC_REGNUM.
3865 2018-03-08 Richard Biener <rguenther@suse.de>
3868 * tree-scalar-evolution.c: Include tree-into-ssa.h.
3869 (follow_copies_to_constant): Do not follow SSA names registered
3872 2018-03-08 Richard Biener <rguenther@suse.de>
3874 PR tree-optimization/84178
3875 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
3877 (version_loop_for_if_conversion): Delay update_ssa call.
3878 (tree_if_conversion): Delay update_ssa until after predicate
3881 2018-03-08 David Malcolm <dmalcolm@redhat.com>
3883 PR tree-optimization/84178
3884 * tree-if-conv.c (release_bb_predicate): Remove the
3885 the assertion that the stmts have NULL use_ops.
3886 Discard the statements, asserting that they haven't
3887 yet been added to a BB.
3889 2018-03-08 Richard Biener <rguenther@suse.de>
3891 PR tree-optimization/84746
3892 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
3893 (phi_translate): Pass in destination ANTIC_OUT set.
3894 (phi_translate_1): Likewise. For a simplified result lookup
3895 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
3896 (phi_translate_set): Adjust.
3897 (do_pre_regular_insertion): Likewise.
3898 (do_pre_partial_partial_insertion): Likewise.
3900 2018-03-08 Martin Liska <mliska@suse.cz>
3902 PR gcov-profile/84735
3903 * doc/gcov.texi: Document usage of profile files.
3904 * gcov-io.h: Document changes in the format.
3906 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
3910 * dwarf2out.c (struct dw_line_info_table): Update comments for
3912 (FORCE_RESET_NEXT_VIEW): New.
3913 (FORCE_RESETTING_VIEW_P): New.
3914 (RESETTING_VIEW_P): Check for -1 too.
3915 (ZERO_VIEW_P): Likewise.
3916 (new_line_info_table): Force-reset next view.
3917 (dwarf2out_begin_function): Likewise.
3918 (dwarf2out_source_line): Simplify zero_view_p initialization.
3919 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
3920 view directly. Omit view when omitting .loc at line 0.
3922 2018-03-08 Jakub Jelinek <jakub@redhat.com>
3924 PR tree-optimization/84740
3925 * tree-switch-conversion.c (process_switch): Call build_constructors
3926 only if info.phi_count is non-zero.
3928 PR tree-optimization/84739
3929 * tree-tailcall.c (find_tail_calls): Check call arguments against
3930 DECL_ARGUMENTS (current_function_decl) rather than
3931 DECL_ARGUMENTS (func) when checking for tail recursion.
3933 2018-03-07 Jakub Jelinek <jakub@redhat.com>
3935 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
3936 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
3937 Volker Reichelt's entry and add entries for people that perform
3938 GCC fuzzy testing and report numerous bugs.
3940 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
3943 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
3944 readonly data in sdata, if that is disabled.
3945 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
3946 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
3947 -mreadonly-in-sdata option.
3949 2018-03-07 Martin Sebor <msebor@redhat.com>
3951 PR tree-optimization/84468
3952 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
3953 basic block when looking for nul assignment.
3955 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
3958 * except.h (output_function_exception_table): Adjust prototype.
3959 * except.c (output_function_exception_table): Remove FNNAME parameter
3960 and add SECTION parameter. Ouput one part of the table at a time.
3961 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
3962 the first part of the exception table and emit unwind directives.
3963 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
3964 (i386_pe_seh_cold_init): Likewise.
3965 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
3966 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
3967 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
3968 (ix86_output_call_insn): Emit a nop in one more case for SEH.
3969 * config/i386/winnt.c: Include except.h.
3970 (struct seh_frame_state): Add reg_offset, after_prologue and
3971 in_cold_section fields.
3972 (i386_pe_seh_end_prologue): Set seh->after_prologue.
3973 (i386_pe_seh_cold_init): New function.
3974 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
3975 to seh->in_cold_section.
3976 (seh_emit_push): Record the offset of the push.
3977 (seh_emit_save): Record the offet of the save.
3978 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
3979 Test seh->after_prologue to disregard the epilogue.
3980 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
3981 (i386_pe_end_cold_function): New function.
3983 2018-03-07 Jakub Jelinek <jakub@redhat.com>
3986 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
3987 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
3990 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
3992 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
3993 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
3996 * multiple_target.c: Include tree-inline.h and intl.h.
3997 (expand_target_clones): Diagnose and fail if node->definition and
3998 !tree_versionable_function_p (node->decl).
4000 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
4002 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
4004 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
4005 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
4006 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
4008 2018-03-06 Jakub Jelinek <jakub@redhat.com>
4011 * combine.c (try_combine): Use reg_or_subregno instead of handling
4012 just paradoxical SUBREGs and REGs.
4014 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
4016 * config/arc/arc.c (arc_finalize_pic): Remove function.
4017 (arc_must_save_register): We use single base PIC register, remove
4018 checks to save/restore the PIC register.
4019 (arc_expand_prologue): Likewise.
4020 * config/arc/arc-protos.h (arc_set_default_type_attributes):
4022 (arc_verify_short): Likewise.
4023 (arc_attr_type): Likewise.
4024 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
4025 (walk_stores): Likewise.
4026 (arc_address_cost): Make it static.
4027 (arc_verify_short): Likewise.
4028 (branch_dest): Likewise.
4029 (arc_attr_type): Likewise.
4030 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
4031 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
4032 (arc_final_prescan_insn): Remove inserting the nops due to
4033 hardware hazards. It is done in reorg step.
4034 (insn_length_variant_t): Remove.
4035 (insn_length_parameters_t): Likewise.
4036 (arc_insn_length_parameters): Likewise.
4037 (arc_get_insn_variants): Likewise.
4038 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
4040 2018-03-06 Jakub Jelinek <jakub@redhat.com>
4043 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
4046 PR tree-optimization/84687
4047 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
4049 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
4051 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4053 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
4054 Rename to ppc_speculation_barrier.
4055 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
4056 __builtin_ppc_speculation_barrier.
4058 2018-03-05 Jakub Jelinek <jakub@redhat.com>
4061 * combine.c (combine_simplify_rtx): Don't try to simplify if
4062 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
4065 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
4067 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
4068 to 32 bytes when compiling for POWER9.
4070 2018-03-05 Jakub Jelinek <jakub@redhat.com>
4073 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
4074 regparm >= 3 with no arg reg available also for calls with
4075 flag_force_indirect_call. Pass decl to ix86_function_regparm.
4078 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
4080 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
4082 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
4085 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
4087 2018-03-05 Richard Biener <rguenther@suse.de>
4089 PR tree-optimization/84486
4090 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
4091 When inserting a __builtin_assume_aligned call set the LHS
4092 SSA name alignment info accordingly.
4094 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
4096 PR tree-optimization/84114
4097 * config/aarch64/aarch64.c (aarch64_reassociation_width)
4098 Avoid reassociation of FLOAT_MODE addition.
4100 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
4102 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
4103 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
4104 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
4105 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
4106 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
4107 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
4108 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
4110 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
4111 __builtin_ia32_wbinvd): New builtins.
4112 (SPECIAL_ARGS2): New.
4113 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
4114 (SPECIAL_ARGS2): New.
4115 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
4116 (ix86_valid_target_attribute_inner_p): Ditto.
4117 (ix86_init_mmx_sse_builtins): Add special_args2.
4118 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
4119 TARGET_WBNOINVD_P): New.
4120 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
4121 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
4122 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
4123 * config/i386/immintrin.h (_wbinvd): New intrinsic.
4124 * config/i386/pconfigintrin.h: New file.
4125 * config/i386/wbnoinvdintrin.h: Ditto.
4126 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
4127 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
4129 2018-03-05 Richard Biener <rguenther@suse.de>
4131 PR tree-optimization/84670
4132 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4134 (BB_VISITED_WITH_VISITED_SUCCS): New define.
4135 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4136 (compute_antic_aux): Only assert the number of values in ANTIC_IN
4137 doesn't grow if all successors (recursively) were visited at least
4140 2018-03-05 Richard Biener <rguenther@suse.de>
4142 PR tree-optimization/84650
4143 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
4144 if executed in the loop pipeline.
4146 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
4148 * doc/configfiles.texi (Configuration Files): Move info about
4149 conditionalizing $target-protos.h to...
4150 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
4151 differs from $target-protos.h.
4153 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
4154 Chung-Ju Wu <jasonwucj@gmail.com>
4156 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
4157 * config/nds32/nds32-multiple.md (setmemsi): Define.
4158 * config/nds32/nds32-memory-manipulation.c
4159 (nds32_gen_dup_4_byte_to_word_value): New.
4160 (emit_setmem_word_loop): New.
4161 (emit_setmem_byte_loop): New.
4162 (nds32_expand_setmem_loop): New.
4163 (nds32_expand_setmem_loop_v3m): New.
4164 (nds32_expand_setmem_unroll): New.
4165 (nds32_expand_setmem): New.
4167 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4168 Chung-Ju Wu <jasonwucj@gmail.com>
4170 * config/nds32/nds32-memory-manipulation.c
4171 (nds32_emit_load_store): New.
4172 (nds32_emit_post_inc_load_store): New.
4173 (nds32_emit_mem_move): New.
4174 (nds32_emit_mem_move_block): New.
4175 (nds32_expand_movmemsi_loop_unknown_size): New.
4176 (nds32_expand_movmemsi_loop_known_size): New.
4177 (nds32_expand_movmemsi_loop): New.
4178 (nds32_expand_movmemsi_unroll): New.
4179 (nds32_expand_movmemqi): Rename ...
4180 (nds32_expand_movmemsi): ... to this.
4181 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
4182 (movmemsi): ... to this.
4183 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
4184 (nds32_expand_movmemsi): ... to this.
4186 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4187 Monk Chiang <sh.chiang04@gmail.com>
4188 Chung-Ju Wu <jasonwucj@gmail.com>
4190 * config/nds32/nds32-protos.h
4191 (nds32_expand_load_multiple): New arguments.
4192 (nds32_expand_store_multiple): Ditto.
4193 (nds32_valid_multiple_load_store): Rename ...
4194 (nds32_valid_multiple_load_store_p): ... to this.
4195 * config/nds32/nds32-memory-manipulation.c
4196 (nds32_expand_load_multiple): Refine implementation.
4197 (nds32_expand_store_multiple): Ditto.
4198 * config/nds32/nds32-multiple.md
4199 (load_multiple): Update nds32_expand_load_multiple interface.
4200 (store_multiple): Update nds32_expand_store_multiple interface.
4201 * config/nds32/nds32-predicates.c
4202 (nds32_valid_multiple_load_store): Rename ...
4203 (nds32_valid_multiple_load_store_p): ... to this and refine
4205 * config/nds32/predicates.md
4206 (nds32_load_multiple_and_update_address_operation): New predicate.
4207 (nds32_store_multiple_and_update_address_operation): New predicate.
4209 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4210 Chung-Ju Wu <jasonwucj@gmail.com>
4212 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
4213 (combo): New attribute.
4214 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
4216 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
4218 * config/nds32/nds32.opt: Change -mcmodel= default value.
4220 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
4221 Monk Chiang <sh.chiang04@gmail.com>
4222 Chung-Ju Wu <jasonwucj@gmail.com>
4224 * config/nds32/constants.md (unspec_element): New enum.
4225 * config/nds32/constraints.md (Umw): New constraint.
4226 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
4227 * config/nds32/nds32-intrinsic.md: Likewise.
4228 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
4229 (nds32_valid_smw_lwm_base_p): New.
4230 (nds32_output_smw_single_word): New.
4231 (nds32_output_lmw_single_word): New.
4232 (nds32_expand_unaligned_load): New.
4233 (nds32_expand_unaligned_store): New.
4234 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
4235 (nds32_output_smw_single_word): Declare.
4236 (nds32_output_lmw_single_word): Declare.
4237 (nds32_expand_unaligned_load): Declare.
4238 (nds32_expand_unaligned_store): Declare.
4239 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
4240 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
4241 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
4242 NDS32_BUILTIN_UASTORE_DW.
4243 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
4246 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
4247 Kito Cheng <kito.cheng@gmail.com>
4248 Chung-Ju Wu <jasonwucj@gmail.com>
4250 * config/nds32/nds32-intrinsic.c
4251 (nds32_expand_builtin_null_ftype_reg): Delete.
4252 (nds32_expand_builtin_reg_ftype_imm): Ditto.
4253 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
4254 (nds32_read_argument): New.
4255 (nds32_legitimize_target): Ditto.
4256 (nds32_legitimize_argument): Ditto.
4257 (nds32_check_constant_argument): Ditto.
4258 (nds32_expand_unop_builtin): Ditto.
4259 (nds32_expand_unopimm_builtin): Ditto.
4260 (nds32_expand_binop_builtin): Ditto.
4261 (nds32_builtin_decl_impl): Ditto.
4262 (builtin_description): Ditto.
4263 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
4264 (nds32_init_builtins_impl): Ditto.
4265 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
4266 (nds32_builtin_decl): New.
4267 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
4268 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
4270 2018-03-02 Jeff Law <law@redhat.com>
4272 * reorg.c (stop_search_p): Handle DEBUG_INSN.
4273 (redundant_insn, fill_simple_delay_slots): Likewise.
4274 (fill_slots_from_thread): Likewise.
4275 * resource.c (mark_referenced_resources): Likewise.
4276 (mark_set_resources, find_dead_or_set_registers): Likewise.
4278 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4280 * substring-locations.h (format_warning_va): Formatting fix for
4282 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
4284 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
4285 * substring-locations.c: Include intl.h.
4286 (format_warning_va): Turned into small wrapper around
4287 format_warning_n_va, renamed to ...
4288 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
4289 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
4291 (format_warning_at_substring_n): New function.
4292 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
4293 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
4294 format_warning_at_substring with just a shorter name instead of
4295 const function pointer.
4296 (fmtwarn_n): New function.
4297 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
4298 appropriate, get rid of all the fmtstr temporaries, move conditionals
4299 with G_() wrapped string literals directly into fmtwarn arguments,
4300 cast dir.len to (int), formatting fixes.
4302 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
4304 * doc/invoke.texi: Remove "Cilk Plus" references.
4306 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4307 Richard Biener <rguenther@suse.de>
4310 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
4311 for error or warning attributes if CALL_FROM_THUNK_P is set.
4314 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4317 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
4318 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
4321 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
4322 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
4324 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
4325 instead of -1U in last predictors element's probability member.
4327 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
4330 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
4331 arguments if they are comparable.
4333 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
4335 PR tree-optimization/84634
4336 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
4337 masks and masked_loop_p with a single loop_masks, making sure it's
4338 null for bb vectorization.
4340 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
4342 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
4343 (vect_analyze_data_ref_access): Use loop->safe_len rather than
4344 loop->force_vectorize to check whether there is no alias.
4346 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4349 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
4351 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
4353 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
4354 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
4355 instead of a loop around prev_real_insn.
4356 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
4360 * config/i386/i386.c (ix86_print_operand): Use conditional
4361 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
4364 2018-03-02 Richard Biener <rguenther@suse.de>
4366 PR tree-optimization/84427
4367 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
4368 (bitmap_set_subtract_values): Rewrite to handle multiple
4371 (prune_clobbered_mems): Likewise.
4372 (phi_translate): Take edge instead of pred/phiblock.
4373 (phi_translate_1): Likewise.
4374 (phi_translate_set): Likewise. Insert all translated
4375 exprs for a value into the set, keeping possibly multiple
4376 expressions per value.
4377 (compute_antic_aux): Adjust for phi_translate changes.
4378 When intersecting union the expressions and prune those
4379 not in the final value set, keeping possibly multiple
4380 expressions per value. Do not use value-insertion
4381 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
4382 all expressions. Add verification that the value-sets
4383 only shrink during iteration.
4384 (compute_partial_antic_aux): Adjust for the phi_translate changes.
4385 (do_pre_regular_insertion): Likewise.
4386 (do_pre_partial_partial_insertion): Likewise.
4388 2018-03-02 Richard Biener <rguenther@suse.de>
4391 * config/darwin.c (saved_debug_info_level): New static global.
4392 (darwin_asm_lto_start): Disable debug info generation for LTO out.
4393 (darwin_asm_lto_end): Restore debug info generation settings.
4395 2018-03-01 Martin Liska <mliska@suse.cz>
4398 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
4401 2018-03-01 Richard Biener <rguenther@suse.de>
4404 * dwarf2out.c (gen_variable_die): Properly handle late VLA
4405 type annotation with LTO when debug was disabled at compile-time.
4407 2018-03-01 Matthew Fortune <mfortune@gmail.com>
4409 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
4411 (mips_final_postscan_insn): Likewise.
4413 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
4415 PR rtl-optimization/84528
4416 * alias.c (init_alias_target): Add commentary.
4417 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
4418 a unique base value if the frame pointer is not eliminated
4419 to the stack pointer.
4421 2018-03-01 Tom de Vries <tom@codesourcery.com>
4423 PR rtl-optimization/83327
4424 * lra-int.h (hard_regs_spilled_into): Declare.
4425 * lra.c (hard_regs_spilled_into): Define.
4426 (init_reg_info): Init hard_regs_spilled_into.
4427 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
4428 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
4429 (process_bb_lives): Handle hard_regs_spilled_into.
4430 (lra_create_live_ranges_1): Before doing liveness propagation, clear
4431 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
4433 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
4435 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
4436 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
4437 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
4438 * config/rs6000/aix72.h: New file.
4440 2018-02-28 Jakub Jelinek <jakub@redhat.com>
4442 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
4443 instead of warning_at with conditional singular and plural messages
4447 * stor-layout.c (update_alignment_for_field): For
4448 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
4449 && !DECL_PACKED (field), do the alignment update, just use
4450 only desired_align instead of MAX (type_align, desired_align)
4452 (place_field): Don't do known_align < desired_align handling
4453 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
4454 is non-NULL, instead do it after rli->prev_field handling and
4455 only if not within a bitfield word. For DECL_PACKED (field)
4456 use type_align of BITS_PER_UNIT.
4458 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
4460 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
4461 superfluous parentheses and trailing spaces.
4463 2018-02-28 Richard Biener <rguenther@suse.de>
4465 PR tree-optimization/84584
4466 * graphite-scop-detection.c (scop_detection::add_scop): Discard
4467 SCoPs with fake exit edge.
4469 2018-02-28 Martin Liska <mliska@suse.cz>
4472 * timevar.c (timer::print): Fix format to properly print 100%
4475 2018-02-28 Richard Biener <rguenther@suse.de>
4478 * genmatch.c (capture_info::walk_match): Do not mark
4479 captured expressions without operands as expr_p given
4480 they act more like predicates and should be subject to
4481 "lost tail" side-effect preserving.
4483 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
4485 PR rtl-optimization/81611
4486 * auto-inc-dec.c (attempt_change): Move dead note from
4487 mem_insn if it's the next use of regno
4488 (find_address): Take address use of reg holding
4489 non-incremented value. Add parm to limit search to the named
4491 (merge_in_block): Attempt to use a mem insn that is the next
4492 use of the original regno.
4494 2018-02-27 Martin Sebor <msebor@redhat.com>
4497 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
4498 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
4500 2018-02-27 Martin Sebor <msebor@redhat.com>
4502 PR translation/84207
4503 * diagnostic-core.h (warning_n, error_n, inform_n): Change
4504 n argument to unsigned HOST_WIDE_INT.
4505 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
4506 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
4507 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
4508 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
4510 2018-02-27 Richard Biener <rguenther@suse.de>
4512 PR tree-optimization/84512
4513 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
4514 Do not use the estimate returned from record_stmt_cost for
4515 the scalar iteration cost but sum properly using add_stmt_cost.
4517 2018-02-27 Richard Biener <rguenther@suse.de>
4519 PR tree-optimization/84466
4520 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
4521 Adjust last change to less strictly validate use operands.
4523 2018-02-27 Martin Liska <mliska@suse.cz>
4525 PR gcov-profile/84548
4526 * gcov.c (process_file): Allow partial overlap and consider it
4527 also as group functions.
4528 (output_lines): Properly calculate range of lines for a group.
4530 2018-02-27 Martin Liska <mliska@suse.cz>
4532 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
4533 'ggc' suffixes. Change first column width.
4534 (timer::print): Fix formatting of the column.
4536 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
4538 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
4539 preserve inline entry blocks for the sake of debug inline
4540 entry point markers alone.
4541 (remove_unused_locals): Suggest in comments a better place to
4542 force the preservation of inline entry blocks that are
4543 otherwise unused, but do not preserve them.
4545 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
4547 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
4549 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
4552 * config/i386/constraints.md (Bs): Replace
4553 ix86_indirect_branch_register with
4554 TARGET_INDIRECT_BRANCH_REGISTER.
4556 * config/i386/i386.md (indirect_jump): Likewise.
4557 (tablejump): Likewise.
4558 (*sibcall_memory): Likewise.
4559 (*sibcall_value_memory): Likewise.
4560 Peepholes of indirect call and jump via memory: Likewise.
4561 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
4562 (*sibcall_value_GOT_32): Likewise.
4563 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4564 (GOT_memory_operand): Likewise.
4565 (call_insn_operand): Likewise.
4566 (sibcall_insn_operand): Likewise.
4567 (GOT32_symbol_operand): Likewise.
4568 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
4570 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
4572 PR rtl-optimization/83496
4573 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
4574 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
4575 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
4576 redundant insn, if any.
4577 (relax_delay_slots): Likewise.
4578 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
4580 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
4582 PR tree-optimization/83965
4583 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
4584 that grouped statements are part of a reduction chain. Return
4585 true if the statement is not marked as a reduction itself but
4587 (vect_recog_dot_prod_pattern): Don't check whether the statement
4588 is part of a group here.
4589 (vect_recog_sad_pattern): Likewise.
4590 (vect_recog_widen_sum_pattern): Likewise.
4592 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
4595 * final.c (rest_of_clean_state): Also look for calls inside sequences.
4597 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
4600 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
4602 (ix86_output_indirect_function_return): New prototype.
4603 (ix86_split_simple_return_pop_internal): Likewise.
4604 * config/i386/i386.c (indirect_return_via_cx): New.
4605 (indirect_return_via_cx_bnd): Likewise.
4606 (indirect_thunk_name): Handle return va CX_REG.
4607 (output_indirect_thunk_function): Create alias for
4608 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
4609 (ix86_output_indirect_jmp): Remove the bool argument.
4610 (ix86_output_indirect_function_return): New function.
4611 (ix86_split_simple_return_pop_internal): Likewise.
4612 * config/i386/i386.md (*indirect_jump): Don't pass false
4613 to ix86_output_indirect_jmp.
4614 (*tablejump_1): Likewise.
4615 (simple_return_pop_internal): Change it to define_insn_and_split.
4616 Call ix86_split_simple_return_pop_internal to split it for
4618 (simple_return_indirect_internal): Call
4619 ix86_output_indirect_function_return instead of
4620 ix86_output_indirect_jmp.
4622 2018-02-26 Jakub Jelinek <jakub@redhat.com>
4625 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
4626 memset and value initialization afterwards.
4628 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
4630 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
4632 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4635 * common/config/aarch64/aarch64-common.c
4636 (aarch_option_optimization_table[]): Switch
4637 off fomit-frame-pointer
4639 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
4640 Chung-Ju Wu <jasonwucj@gmail.com>
4642 * config/nds32/nds32-multiple.md (load_multiple): Disallow
4644 (store_multiple): Ditto.
4646 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
4648 * config.gcc: Add --with-cpu support for nds32 target.
4649 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
4650 * config/nds32/nds32.opt: Add -mcpu= option.
4652 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
4654 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
4655 isel=yes): Warn for these deprecated options.
4657 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
4659 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
4660 ISA_2_5_MASKS_EMBEDDED.
4662 2018-02-23 Jakub Jelinek <jakub@redhat.com>
4664 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
4665 p->max as pointers rather than using iterative_hash_expr.
4667 2018-02-23 Carl Love <cel@us.ibm.com>
4669 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
4670 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
4672 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
4673 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
4674 P8V_BUILTIN_VEC_VUNSIGNED2.
4676 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
4679 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
4680 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
4682 * lra-constraints.c (curr_insn_transform): Set up
4683 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
4685 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
4686 * lra-eliminations.c (spill_pseudos): Ditto.
4687 (process_insn_for_elimination): Ditto.
4688 * lra-lives.c (reg_early_clobber_p): Use the new macros.
4689 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
4690 LRA_NON_CLOBBERED_ALT.
4692 2018-02-22 Martin Sebor <msebor@redhat.com>
4694 PR tree-optimization/84480
4695 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
4696 to maybe_diag_stxncpy_trunc. Call it.
4697 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
4698 from gimple_fold_builtin_strcpy. Print inlining stack.
4699 (handle_builtin_stxncpy): Print inlining stack.
4700 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
4702 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
4705 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
4706 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
4707 and -fcheck-pointer-bounds are used together.
4708 (indirect_thunk_prefix): New enum.
4709 (indirect_thunk_need_prefix): New function.
4710 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
4711 "_nt" instead of "_bnd" for NOTRACK prefix.
4712 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
4713 (output_indirect_thunk_function): Likewise.
4715 (ix86_code_end): Update output_indirect_thunk_function calls.
4716 (ix86_output_indirect_branch_via_reg): Replace
4717 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
4718 (ix86_output_indirect_branch_via_push): Likewise.
4719 (ix86_output_function_return): Likewise.
4720 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
4721 incompatible with -fcf-protection=branch and
4722 -fcheck-pointer-bounds.
4724 2018-02-22 Steve Ellcey <sellcey@cavium.com>
4727 * config/aarch64/aarch64.c (aarch64_print_address_internal):
4728 Change gcc_assert call to output_operand_lossage.
4730 2018-02-22 Steve Ellcey <sellcey@cavium.com>
4732 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
4734 2018-02-22 DJ Delorie <dj@redhat.com>
4735 Sebastian Perta <sebastian.perta@renesas.com>
4736 Oleg Endo <olegendo@gcc.gnu.org>
4738 * config/rx/rx.c (rx_rtx_costs): New function.
4739 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
4741 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
4743 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
4745 2018-02-22 Martin Liska <mliska@suse.cz>
4748 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
4749 Add "native" as a possible value.
4751 2018-02-22 Martin Liska <mliska@suse.cz>
4754 * config/i386/i386.c (ix86_option_override_internal):
4755 Add "native" as a possible value for -march and -mtune.
4757 2018-02-22 Jakub Jelinek <jakub@redhat.com>
4760 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
4761 to all type variants.
4763 PR tree-optimization/84503
4764 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
4765 width as info->bitpos + info->bitsize - start.
4766 (merged_store_group::merge_overlapping): Simplify width computation.
4767 (check_no_overlap): New function.
4768 (imm_store_chain_info::try_coalesce_bswap): Compute expected
4769 start + width and last_order of the group, fail if check_no_overlap
4771 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
4772 to group if check_no_overlap fails.
4774 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
4776 * config/rs6000/altivec.md: Delete contraint arguments to
4777 define_expand, define_split, and define_peephole2, and in
4778 define_insn_and_split if always unused.
4779 * config/rs6000/darwin.md: Ditto.
4780 * config/rs6000/dfp.md: Ditto.
4781 * config/rs6000/rs6000.md: Ditto.
4782 * config/rs6000/sync.md: Ditto.
4783 * config/rs6000/vector.md: Ditto.
4784 * config/rs6000/vsx.md: Ditto.
4786 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
4788 * config/rs6000/altivec.md: Write output control strings as braced
4789 blocks instead of double-quoted strings.
4790 * config/rs6000/darwin.md: Ditto.
4791 * config/rs6000/rs6000.md: Ditto.
4792 * config/rs6000/vector.md: Ditto.
4793 * config/rs6000/vsx.md: Ditto.
4795 2018-02-21 Jason Merrill <jason@redhat.com>
4797 PR c++/84314 - ICE with templates and fastcall attribute.
4798 * attribs.c (build_type_attribute_qual_variant): Remove assert.
4800 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
4802 * ipa-cp.c (determine_versionability): Fix comment typos.
4804 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
4807 * ipa-cp.c (determine_versionability): Do not version functions caling
4810 2018-02-21 Martin Liska <mliska@suse.cz>
4813 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
4814 Add "native" as a possible value.
4815 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
4816 the macro when native cpu detection is available.
4818 2018-02-21 Martin Liska <mliska@suse.cz>
4821 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
4822 Add "native" as a possible value.
4823 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
4824 when native cpu detection is available.
4826 2018-02-21 Jakub Jelinek <jakub@redhat.com>
4827 Martin Sebor <msebor@redhat.com>
4829 PR tree-optimization/84478
4830 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
4832 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
4833 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
4834 support which is conservatively correct, for 2 only stay conservative
4835 for maxlen. Formatting and comment capitalization fixes. Add STRICT
4836 argument to the 2 argument get_range_strlen, adjust 6 arg
4837 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
4839 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
4840 (gimple_fold_builtin_strlen): Pass true as last argument to
4843 2018-02-20 Martin Sebor <msebor@redhat.com>
4846 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
4847 (builtin_memref::set_base_and_offset): Same. Handle inner references.
4848 (builtin_memref::builtin_memref): Factor out parts into
4849 set_base_and_offset and call it.
4851 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
4854 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
4855 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
4856 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
4857 search at the associated MODE_INT.
4859 2018-02-20 Jeff Law <law@redhat.com>
4862 PR tree-optimization/81592
4864 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
4865 for range data rather than using global data.
4866 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
4867 range data rather than using global data.
4868 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
4869 pass it to children as needed.
4870 (struct directive::fmtresult): Similarly.
4871 (struct directive::set_width): Similarly.
4872 (struct directive::set_precision): Similarly.
4873 (format_integer, format_directive, parse_directive): Similarly.
4874 (format_none): Accept unnamed vr_values parameter.
4875 (format_percent, format_floating, format_character): Similarly.
4876 (format_string, format_plain): Similarly.
4877 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
4878 the EVRP range analyzer for range data rather than using global data.
4879 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
4880 gimple-ssa-evrp-analyze.h
4881 (class sprintf_dom_walker): Add after_dom_children member function.
4882 Add evrp_range_analyzer member.
4883 (sprintf_dom_walker::before_dom_children): Call into the EVRP
4884 range analyzer as needed.
4885 (sprintf_dom_walker::after_dom_children): New member function.
4886 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
4888 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
4889 (evrp_range_analyzer::pop_to_marker): Likewise.
4891 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
4893 PR tree-optimization/84419
4894 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
4895 with the required type if its current type is compatible but
4898 2018-02-20 Jakub Jelinek <jakub@redhat.com>
4901 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
4902 after vectorization.
4904 2018-02-20 Martin Liska <mliska@suse.cz>
4907 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
4908 possible values if we don't have a hint.
4910 2018-02-20 Martin Liska <mliska@suse.cz>
4914 * final.c (shorten_branches): Build align_tab array with one
4916 * opts.c (finish_options): Add alignment option limit check.
4917 (MAX_CODE_ALIGN): Likewise.
4918 (MAX_CODE_ALIGN_VALUE): Likewise.
4919 * doc/invoke.texi: Document maximum allowed option value for
4920 all -falign-* options.
4922 2018-02-19 Jakub Jelinek <jakub@redhat.com>
4925 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
4926 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
4927 * var-tracking.c (emit_note_insn_var_location): Remove all references
4928 to NOTE_INSN_CALL_ARG_LOCATION.
4929 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
4930 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
4931 Use copy_rtx_if_shared.
4932 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
4933 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
4934 (dwarf2out_var_location): Remove handling of
4935 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
4937 * final.c (final_scan_insn): Remove all references to
4938 NOTE_INSN_CALL_ARG_LOCATION.
4939 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
4940 before dumping final insns.
4941 * except.c (emit_note_eh_region_end): Remove all references to
4942 NOTE_INSN_CALL_ARG_LOCATION.
4943 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
4944 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
4945 * config/arc/arc.c (hwloop_optimize): Likewise.
4946 * config/arm/arm.c (create_fix_barrier): Likewise.
4947 * config/s390/s390.c (s390_chunkify_start): Likewise.
4948 * config/sh/sh.c (find_barrier): Likewise.
4949 * config/i386/i386.c (rest_of_insert_endbranch,
4950 ix86_seh_fixup_eh_fallthru): Likewise.
4951 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
4952 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
4953 * config/frv/frv.c (frv_function_prologue): Likewise.
4954 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
4956 (note_outside_basic_block_p): Remove all references to
4957 NOTE_INSN_CALL_ARG_LOCATION.
4958 * gengtype.c (adjust_field_rtx_def): Likewise.
4959 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
4961 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
4962 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
4965 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
4968 PR tree-optimization/84452
4969 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
4970 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
4973 2018-02-19 Martin Liska <mliska@suse.cz>
4976 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
4978 2018-02-19 Martin Liska <mliska@suse.cz>
4979 Richard Sandiford <richard.sandiford@linaro.org>
4981 PR tree-optimization/82491
4982 * gimple-fold.c (get_base_constructor): Make earlier bail out
4985 2018-02-19 Carl Love <cel@us.ibm.com>
4987 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
4988 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
4990 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
4991 P8V_BUILTIN_VEC_NEG.
4993 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
4995 * config/rl78/rl78.md (movdf): New define expand.
4997 2018-02-19 Martin Liska <mliska@suse.cz>
5000 * doc/invoke.texi: Fix typo.
5001 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
5003 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
5005 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
5006 handle rs6000_single_float and rs6000_double_float specially for
5009 2018-02-16 Jeff Law <law@redhat.com>
5011 * config/rx/rx.c (add_pop_cfi_notes): New function.;
5014 2018-02-16 Jakub Jelinek <jakub@redhat.com>
5017 * ipa-inline.c (inline_small_functions): Fix a typo.
5019 2018-02-16 Nathan Sidwell <nathan@acm.org>
5021 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
5023 2018-02-16 Carl Love <cel@us.ibm.com>
5025 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
5026 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
5027 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
5028 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
5029 expansion to P8V_BUILTIN_VEC_FLOAT2.
5031 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
5033 PR rtl-optimization/70023
5034 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
5035 src_regno into account.
5037 2018-02-16 Carl Love <cel@us.ibm.com>
5039 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
5040 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
5041 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
5042 * config/rs6000/rs6000.c: Remove case statements for
5043 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
5044 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
5045 and P9V_BUILTIN_VEC_VINSERT4B.
5046 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
5047 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
5048 * config/rs6000/vsx.md:
5049 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
5052 2018-02-16 Carl Love <cel@us.ibm.com>
5054 * config/rs6000/altivec.h: Add builtin names vec_extract4b
5056 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
5058 * config/rs6000/rs6000-c.c: Add the definitions for
5059 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
5060 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
5061 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
5062 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
5063 definition for insert4b and define insn *insert3b_internal.
5064 * doc/extend.texi: Add documentation for vec_extract4b.
5066 2018-02-16 Nathan Sidwell <nathan@acm.org>
5068 * doc/extend.texi (Backwards Compatibility): Mention friend
5069 injection. Note for-scope is deprecated.
5070 * doc/invoke.texi (-ffriend-injection): Deprecate.
5072 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
5074 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
5075 that moved to I2, also allow destinations that are a paradoxical
5076 subreg (instead of a normal reg).
5078 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
5081 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
5084 2018-02-16 Richard Biener <rguenther@suse.de>
5086 PR tree-optimization/84037
5087 PR tree-optimization/84016
5089 * config/i386/i386.c (ix86_builtin_vectorization_cost):
5090 Adjust vec_construct for the fact we need additional higher latency
5091 128bit inserts for AVX256 and AVX512 vector builds.
5092 (ix86_add_stmt_cost): Scale vector construction cost for
5095 2018-02-16 Richard Biener <rguenther@suse.de>
5097 PR tree-optimization/84417
5098 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
5099 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
5100 (non_rewritable_lvalue_p): Likewise, use poly-ints.
5102 2018-02-16 Martin Liska <mliska@suse.cz>
5105 * internal-fn.def (ASAN_CHECK): Set proper flags.
5106 (ASAN_MARK): Likewise.
5108 2018-02-16 Julia Koval <julia.koval@intel.com>
5110 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
5111 from PTA_CANNONLAKE.
5113 2018-02-16 Jakub Jelinek <jakub@redhat.com>
5116 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
5117 Use ++iter rather than iter++ for std::list iterators.
5118 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
5119 defer deleting them until all nodes in the forest are processed. Do
5120 free even leaf nodes. Change to_process into auto_vec.
5123 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
5124 * vec.h (vec_default_construct): Use memset instead of placement new
5125 if BROKEN_VALUE_INITIALIZATION is defined.
5126 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
5127 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
5130 PR rtl-optimization/83723
5131 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
5132 * lra.c (lra_substitute_pseudo): Likewise. If true, use
5133 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
5135 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
5137 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
5139 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
5141 PR rtl-optimization/81443
5142 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
5143 from inner REGs to paradoxical SUBREGs.
5145 2018-02-16 Richard Biener <rguenther@suse.de>
5147 PR tree-optimization/84399
5148 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
5149 For operands we can analyze at their definition make sure we can
5150 analyze them at each use as well.
5152 2018-02-16 Richard Biener <rguenther@suse.de>
5154 PR tree-optimization/84190
5155 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
5156 volatile accesses if the decl isn't volatile.
5158 2018-02-15 Jason Merrill <jason@redhat.com>
5160 PR c++/84314 - ICE with templates and fastcall attribute.
5161 * attribs.c (build_type_attribute_qual_variant): Don't clobber
5162 TYPE_CANONICAL on an existing type.
5164 2018-02-15 Jakub Jelinek <jakub@redhat.com>
5166 PR tree-optimization/84383
5167 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
5168 dstoff nor call operand_equal_p if dstbase is NULL.
5170 PR tree-optimization/84334
5171 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
5172 also a CONSTANT_CLASS_P, punt.
5174 2018-02-14 Jim Wilson <jimw@sifive.com>
5176 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
5177 first SMALL_OPERAND check. New local min_second_step. Move assert
5178 to where locals are set. Add TARGET_RVC support.
5179 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
5181 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
5183 * doc/invoke.texi: Correct -Wformat-overflow code sample.
5185 2018-02-14 Martin Sebor <msebor@redhat.com>
5187 PR tree-optimization/83698
5188 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
5189 arrays constrain the offset range to their bounds.
5190 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
5191 (builtin_access::overlap): Avoid setting the size of overlap if it's
5193 (maybe_diag_overlap): Also consider arrays when deciding what values
5194 of offsets to include in diagnostics.
5196 2018-02-14 Martin Sebor <msebor@redhat.com>
5199 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
5200 that correspond to the kind of a declaration.
5202 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
5205 * config/pa/pa.md: Load address of PIC label using the linkage table
5206 if the label is nonlocal.
5208 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
5210 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
5211 warning message if user requests -maltivec=be.
5212 * doc/invoke.texi: Document deprecation of -maltivec=be.
5214 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
5217 * config/rs6000/rs6000-c.c: Update definitions for
5218 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
5219 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
5221 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
5224 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
5225 add _get_ssp intrinsics. Remove argument from
5226 __builtin_ia32_rdssp[d|q].
5227 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
5228 * config/i386/i386-builtin.def: Remove argument from
5229 __builtin_ia32_rdssp[d|q].
5230 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
5231 ix86_expand_special_args_builtin for _rdssp[d|q].
5232 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
5233 Clear register before usage.
5234 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
5235 Add documentation for new _get_ssp and _inc_ssp intrinsics.
5237 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
5239 PR tree-optimization/84357
5240 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
5241 operand 1 of an ARRAY_REF too.
5243 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
5246 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
5247 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
5249 (set_of_reg): New struct.
5250 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
5251 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
5252 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
5254 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
5255 Split into bitclr, bitset, bitinvert patterns if appropriate.
5256 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
5257 use rx_fuse_in_memory_bitop.
5258 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
5259 to named insn, correct maximum insn length.
5261 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
5264 * machmode.def: Define a complex mode for PARTIAL_INT.
5265 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
5267 * doc/rtl.texi: Document CSPImode.
5268 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
5270 (msp430_hard_regno_nregs_with_padding): Likewise.
5272 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
5275 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
5277 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
5279 PR rtl-optimization/84169
5280 * combine.c (try_combine): New variable split_i2i3. Set it to true if
5281 we generated a parallel as new i3 and we split that to new i2 and i3
5282 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
5283 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
5284 those to i2, not i1. Partially rewrite this scan code.
5286 2018-02-13 Jakub Jelinek <jakub@redhat.com>
5289 * stor-layout.c (place_field): For variable length fields, adjust
5290 offset_align afterwards not just based on the field's alignment,
5291 but also on the size.
5294 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
5295 of exps and logs in the use_exp2 case.
5297 2018-02-13 Jeff Law <law@redhat.com>
5299 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
5302 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
5305 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
5309 * common.opt (gas-loc-support, gas-locview-support): New.
5310 (ginline-points, ginternal-reset-location-views): New.
5311 * doc/invoke.texi: Document them. Use @itemx where intended.
5312 (gvariable-location-views): Adjust.
5313 * target.def (reset_location_view): New.
5314 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
5315 (TARGET_RESET_LOCATION_VIEW): New.
5316 * doc/tm.texi: Rebuilt.
5317 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
5318 (dwarf2out_default_as_locview_support): New.
5319 (output_asm_line_debug_info): Use option variables.
5320 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
5321 (output_loc_list): Likewise.
5322 (add_high_low_attributes): Check option variables.
5323 Don't output entry view attribute in strict mode.
5324 (gen_inlined_subroutine_die): Check option variables.
5325 (dwarf2out_inline_entry): Likewise.
5326 (init_sections_and_labels): Likewise.
5327 (dwarf2out_early_finish): Likewise.
5328 (maybe_reset_location_view): New, from...
5329 (dwarf2out_var_location): ... here. Call it.
5330 * debug.h (dwarf2out_default_as_loc_support): Declare.
5331 (dwarf2out_default_as_locview_support): Declare.
5332 * hooks.c (hook_int_rtx_insn_0): New.
5333 * hooks.h (hook_int_rtx_insn_0): Declare.
5334 * toplev.c (process_options): Take -gas-loc-support and
5335 -gas-locview-support from dwarf2out. Enable
5336 -gvariable-location-views by default only with locview
5337 assembler support. Enable -ginternal-reset-location-views by
5338 default only if the target defines the corresponding hook.
5339 Enable -ginline-points by default if location views are
5340 enabled; force it disabled if statement frontiers are
5342 * tree-inline.c (expand_call_inline): Check option variables.
5343 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
5345 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
5347 PR tree-optimization/84321
5348 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
5349 handling. Also check whether the anti-range contains any values
5350 that satisfy the mask; switch to a VR_RANGE if not.
5352 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
5355 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
5357 2018-02-13 Martin Jambor <mjambor@suse.cz>
5360 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
5361 of call statements, also set location of a load to a temporary.
5363 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
5365 * config/rl78/rl78.c (add_vector_labels): New function.
5366 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
5367 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
5368 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
5369 which checks that no arguments are passed.
5370 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
5371 * doc/extend.texi: Documentation for the new attribute.
5373 2018-02-13 Andreas Schwab <schwab@suse.de>
5375 * config/riscv/linux.h (CPP_SPEC): Define.
5377 2018-02-13 Jakub Jelinek <jakub@redhat.com>
5380 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
5381 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
5382 OPTION_MASK_ISA_AES as first argument to def_builtin_const
5383 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
5384 instead of OPTION_MASK_ISA_PCLMUL as first argument to
5385 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
5386 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
5387 temporarily for AES and PCLMUL builtins.
5389 PR tree-optimization/84339
5390 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
5391 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
5395 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
5396 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
5397 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
5399 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
5401 * omp-simd-clone.h: New file.
5402 * omp-simd-clone.c: Include omp-simd-clone.h.
5403 (expand_simd_clones): No longer static.
5404 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
5405 cgraph.h and omp-simd-clone.h.
5406 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
5407 (vect_recog_widen_shift_pattern): Formatting fix.
5408 (vect_pattern_recog_1): Don't check optab for calls.
5411 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
5412 operands[2] into a REG before using gen_lowpart on it.
5414 2018-02-12 Jeff Law <law@redhat.com>
5417 * config/sh/sh.c (find_barrier): Consider a sibling call
5420 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
5421 successfully back substituting a reg.
5423 2018-02-12 Richard Biener <rguenther@suse.de>
5425 PR tree-optimization/84037
5426 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
5427 parameter, move visited init to caller.
5428 (vect_slp_analyze_operations): Separate cost from validity
5429 check, initialize visited once for all instances.
5430 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
5432 * tree-vect-stmts.c (vect_model_simple_cost): Make early
5434 (vect_model_promotion_demotion_cost): Likewise.
5435 (vectorizable_bswap): Guard cost modeling with !slp_node
5436 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
5438 (vectorizable_call): Likewise.
5439 (vectorizable_conversion): Likewise.
5440 (vectorizable_assignment): Likewise.
5441 (vectorizable_shift): Likewise.
5442 (vectorizable_operation): Likewise.
5443 (vectorizable_store): Likewise.
5444 (vectorizable_load): Likewise.
5445 (vectorizable_condition): Likewise.
5446 (vectorizable_comparison): Likewise.
5448 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
5451 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
5452 (ASAN_MARK): Fix fnspec to account for return value, change pointer
5453 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
5455 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
5458 * params.def (inline-min-speedup): Increase from 8 to 15.
5459 (max-inline-insns-auto): Decrease from 40 to 30.
5460 * ipa-split.c (consider_split): Add some buffer for function to
5461 be considered inlining candidate.
5462 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
5465 2018-02-12 Richard Biener <rguenther@suse.de>
5467 PR tree-optimization/84037
5468 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
5469 matched stmts if we cannot swap the non-matched ones.
5471 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
5473 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
5474 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
5475 _mm_maskz_scalef_round_ss): New intrinsics.
5476 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
5477 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
5478 __builtin_ia32_scalefss_round): Remove.
5479 (__builtin_ia32_scalefsd_mask_round,
5480 __builtin_ia32_scalefss_mask_round): New intrinsics.
5481 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
5482 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
5483 ((match_operand:VF_128 2 "<round_nimm_predicate>"
5484 "<round_constraint>")): Changed to ...
5485 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
5486 "<round_scalar_constraint>")): ... this.
5487 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
5488 %0, %1, %2<round_op3>}"): Changed to ...
5489 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
5490 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
5491 %2<round_scalar_mask_op3>}"): ... this.
5492 * config/i386/subst.md (round_scalar_nimm_predicate): New.
5494 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
5496 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
5497 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
5498 (_mm_maskz_sqrt_round_ss): New intrinsics.
5499 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
5500 (__builtin_ia32_sqrtsd_mask_round)
5501 (__builtin_ia32_sqrtss_mask_round): New builtins.
5502 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
5503 (__builtin_ia32_sqrtss_round): Remove.
5504 (__builtin_ia32_sqrtsd_mask_round)
5505 (__builtin_ia32_sqrtss_mask_round): New builtins.
5506 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
5507 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
5508 ((match_operand:VF_128 1 "vector_operand"
5509 "xBm,<round_constraint>")): Changed to ...
5510 ((match_operand:VF_128 1 "vector_operand"
5511 "xBm,<round_scalar_constraint>")): ... this.
5512 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
5513 %0, %2, %<iptr>1<round_op3>}): Changed to ...
5514 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
5515 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
5516 %<iptr>1<round_scalar_mask_op3>}): ... this.
5517 ((set_attr "prefix" "<round_prefix>")): Changed to ...
5518 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
5520 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
5523 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
5524 Cast vec_cmpeq result to correct type.
5525 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
5526 Cast vec_cmpgt result to correct type.
5528 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
5530 * final.c (final_scan_insn_1): Renamed from...
5531 (final_scan_insn): ... this. New wrapper, to recover
5532 seen from the outermost call in recursive ones.
5533 * config/sparc/sparc.c (output_return): Drop seen from call.
5534 (output_sibcall): Likewise.
5535 * config/visium/visium.c (output_branch): Likewise.
5537 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
5539 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
5542 2018-02-10 Alan Modra <amodra@gmail.com>
5545 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
5546 Specify LR as an input.
5548 2018-02-10 Jakub Jelinek <jakub@redhat.com>
5551 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
5552 remove_member_access_dummy_vars): New functions.
5553 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
5554 lower_omp_1, execute_lower_omp): Use them.
5556 PR rtl-optimization/84308
5557 * shrink-wrap.c (spread_components): Release todo vector.
5559 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
5561 PR rtl-optimization/57193
5562 * ira-color.c (struct allocno_color_data): Add member
5563 conflict_allocno_hard_prefs.
5564 (update_conflict_allocno_hard_prefs): New.
5565 (bucket_allocno_compare_func): Add a preference based on
5566 conflict_allocno_hard_prefs.
5567 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
5568 (color_allocnos): Remove a dead code. Initiate
5569 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
5571 2018-02-09 Jakub Jelinek <jakub@redhat.com>
5574 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
5575 constraint from =wa to wa. Avoid a subreg on the output operand,
5576 instead use a pseudo and subreg it in a move.
5577 (p9_xxbrd_<mode>): Changed to ...
5578 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
5579 (p9_xxbrd_v2df): New expander.
5580 (p9_xxbrw_<mode>): Changed to ...
5581 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
5582 (p9_xxbrw_v4sf): New expander.
5584 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
5586 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
5588 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
5591 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
5592 multiply in 32-bit mode.
5593 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
5594 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
5597 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
5599 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
5600 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
5601 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
5602 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
5604 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
5607 * dwarf2out.c (is_trivial_indirect_ref): New function.
5608 (dwarf2out_late_global_decl): Do not generate a location
5609 attribute for variables that have a non-trivial DECL_VALUE_EXPR
5610 and that are not defined in the current unit.
5612 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
5614 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
5615 instead of a libcall for UNORDERED.
5617 2018-02-09 Tamar Christina <tamar.christina@arm.com>
5620 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
5621 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
5623 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5626 * config/s390/s390.c (s390_set_current_function): Invoke
5627 s390_indirect_branch_settings also if fndecl didn't change.
5629 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
5631 * config/rs6000/rs6000.md (blockage): Set length to zero.
5633 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
5635 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
5637 2018-02-09 Jakub Jelinek <jakub@redhat.com>
5640 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
5641 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
5645 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
5646 PARALLEL incoming that failed vt_get_decl_and_offset check.
5649 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
5650 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
5652 (get_variable_section): For decls in named .bss* sections pass true as
5653 second argument to bss_initializer_p.
5655 2018-02-09 Marek Polacek <polacek@redhat.com>
5656 Jakub Jelinek <jakub@redhat.com>
5659 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
5660 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
5661 Sync some changes from cxx_fold_indirect_ref.
5663 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
5665 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
5667 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
5668 (BLOCK_INLINE_ENTRY_LABEL): New.
5669 (dwarf2out_var_location): Disregard inline entry markers.
5670 (inline_entry_data): New struct.
5671 (inline_entry_data_hasher): New hashtable type.
5672 (inline_entry_data_hasher::hash): New.
5673 (inline_entry_data_hasher::equal): New.
5674 (inline_entry_data_table): New variable.
5675 (add_high_low_attributes): Add DW_AT_entry_pc and
5676 DW_AT_GNU_entry_view attributes if a pending entry is found
5677 in inline_entry_data_table. Add old entry_pc attribute only
5678 if debug nonbinding markers are disabled.
5679 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
5680 markers are enabled.
5681 (block_within_block_p, dwarf2out_inline_entry): New.
5682 (dwarf2out_finish): Check that no entries remained in
5683 inline_entry_data_table.
5684 * final.c (reemit_insn_block_notes): Handle inline entry notes.
5685 (final_scan_insn, notice_source_line): Likewise.
5686 (rest_of_clean_state): Skip inline entry markers.
5687 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
5689 * gimple.c (gimple_build_debug_inline_entry): New.
5690 * gimple.h (enum gimple_debug_subcode): Add
5691 GIMPLE_DEBUG_INLINE_ENTRY.
5692 (gimple_build_debug_inline_entry): Declare.
5693 (gimple_debug_inline_entry_p): New.
5694 (gimple_debug_nonbind_marker_p): Adjust.
5695 * insn-notes.def (INLINE_ENTRY): New.
5696 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
5697 inline entry marker notes.
5698 (print_insn): Likewise.
5699 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
5700 (INSN_DEBUG_MARKER_KIND): Likewise.
5701 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
5702 * tree-inline.c (expand_call_inline): Build and insert
5703 debug_inline_entry stmt.
5704 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
5705 inline entry blocks early, if nonbind markers are enabled.
5706 (dump_scope_block): Dump fragment info.
5707 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
5708 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
5709 (gimple_build_debug_inline_entry): New.
5710 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
5711 Enable/disable inline entry points too.
5712 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
5713 (DEBUG_INSN): Describe inline entry markers.
5715 * common.opt (gvariable-location-views): New.
5716 (gvariable-location-views=incompat5): New.
5717 * config.in: Rebuilt.
5718 * configure: Rebuilt.
5719 * configure.ac: Test assembler for view support.
5720 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
5721 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
5722 * dwarf2out.c (var_loc_view): New typedef.
5723 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
5724 (dwarf2out_locviews_in_attribute): New.
5725 (dwarf2out_locviews_in_loclist): New.
5726 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
5727 (enum dw_line_info_opcode): Add LI_adv_address.
5728 (struct dw_line_info_table): Add view.
5729 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
5730 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
5731 (zero_view_p): New variable.
5732 (ZERO_VIEW_P): New macro.
5733 (output_asm_line_debug_info): New.
5734 (struct var_loc_node): Add view.
5735 (add_AT_view_list, AT_loc_list): New.
5736 (add_var_loc_to_decl): Add view param. Test it against last.
5737 (new_loc_list): Add view params. Record them.
5738 (AT_loc_list_ptr): Handle loc and view lists.
5739 (view_list_to_loc_list_val_node): New.
5740 (print_dw_val): Handle dw_val_class_view_list.
5741 (size_of_die): Likewise.
5742 (value_format): Likewise.
5743 (loc_list_has_views): New.
5744 (gen_llsym): Set vl_symbol too.
5745 (maybe_gen_llsym, skip_loc_list_entry): New.
5746 (dwarf2out_maybe_output_loclist_view_pair): New.
5747 (output_loc_list): Output view list or entries too.
5748 (output_view_list_offset): New.
5749 (output_die): Handle dw_val_class_view_list.
5750 (output_dwarf_version): New.
5751 (output_compilation_unit_header): Use it.
5752 (output_skeleton_debug_sections): Likewise.
5753 (output_rnglists, output_line_info): Likewise.
5754 (output_pubnames, output_aranges): Update version comments.
5755 (output_one_line_info_table): Output view numbers in asm comments.
5756 (dw_loc_list): Determine current endview, pass it to new_loc_list.
5757 Call maybe_gen_llsym.
5758 (loc_list_from_tree_1): Adjust.
5759 (add_AT_location_description): Create view list attribute if
5760 needed, check it's absent otherwise.
5761 (convert_cfa_to_fb_loc_list): Adjust.
5762 (maybe_emit_file): Call output_asm_line_debug_info for test.
5763 (dwarf2out_var_location): Reset views as needed. Precompute
5764 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
5765 attribute. Set view.
5766 (new_line_info_table): Reset next view.
5767 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
5768 (dwarf2out_source_line): Likewise. Output view resets and labels to
5769 the assembler, or select appropriate line info opcodes.
5770 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
5771 (optimize_string_length): Catch it. Adjust.
5772 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
5773 dw_val_class_view_list, and remove it if no longer needed.
5774 (hash_loc_list): Hash view numbers.
5775 (loc_list_hasher::equal): Compare them.
5776 (optimize_location_lists): Check whether a view list symbol is
5777 needed, and whether the locview attribute is present, and
5778 whether they match. Remove the locview attribute if no longer
5780 (index_location_lists): Call skip_loc_list_entry for test.
5781 (dwarf2out_finish): Call output_asm_line_debug_info for test.
5782 Use output_dwarf_version.
5783 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
5784 (struct dw_val_node): Add val_view_list.
5785 * final.c (SEEN_NEXT_VIEW): New.
5786 (set_next_view_needed): New.
5787 (clear_next_view_needed): New.
5788 (maybe_output_next_view): New.
5789 (final_start_function): Rename to...
5790 (final_start_function_1): ... this. Take pointer to FIRST,
5791 add SEEN parameter. Emit param bindings in the initial view.
5792 (final_start_function): Reintroduce SEEN-less interface.
5793 (final): Rename to...
5794 (final_1): ... this. Take SEEN parameter. Output final pending
5795 next view at the end.
5796 (final): Reintroduce seen-less interface.
5797 (final_scan_insn): Output pending next view before switching
5798 sections or ending a block. Mark the next view as needed when
5799 outputting variable locations. Notify debug backend of section
5800 changes, and of location view changes.
5801 (rest_of_handle_final): Adjust.
5802 * toplev.c (process_options): Autodetect value for debug variable
5803 location views option. Warn on incompat5 without -gdwarf-5.
5804 * doc/invoke.texi (gvariable-location-views): New.
5805 (gvariable-location-views=incompat5): New.
5806 (gno-variable-location-views): New.
5808 2018-02-08 David Malcolm <dmalcolm@redhat.com>
5810 PR tree-optimization/84136
5811 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
5812 that the result of find_edge is non-NULL.
5814 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
5817 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
5818 storing integer register in SImode. Fix cost of 256 and 512
5819 byte aligned SSE register store.
5821 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
5823 * config/i386/i386.c (ix86_multiplication_cost): Fix
5824 multiplication cost for TARGET_AVX512DQ.
5826 2018-02-08 Marek Polacek <polacek@redhat.com>
5828 PR tree-optimization/84238
5829 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
5832 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5834 PR tree-optimization/84265
5835 * tree-vect-stmts.c (vectorizable_store): Don't treat
5836 VMAT_CONTIGUOUS accesses as grouped.
5837 (vectorizable_load): Likewise.
5839 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5841 PR tree-optimization/81635
5842 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
5843 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
5844 (test_round_for_mask): New functions.
5845 (wide_int_cc_tests): Call test_round_for_mask.
5846 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
5847 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
5848 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
5849 range returned by get_range_info.
5851 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
5854 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
5855 * symtab.c: Include builtins.h
5856 (symtab_node::output_to_lto_symbol_table_p): Move here
5857 from lto-streamer-out.c:output_symbol_p.
5858 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
5859 (output_symbol_p): Move all logic to symtab.c
5860 (produce_symtab): Update.
5862 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5864 * config/s390/s390-opts.h (enum indirect_branch): Define.
5865 * config/s390/s390-protos.h (s390_return_addr_from_memory)
5866 (s390_indirect_branch_via_thunk)
5867 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
5868 (enum s390_indirect_branch_type): Define.
5869 * config/s390/s390.c (struct s390_frame_layout, struct
5870 machine_function): Remove.
5871 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
5872 (indirect_branch_table_label_no, indirect_branch_table_name):
5874 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
5875 (enum s390_indirect_branch_option): Define.
5876 (s390_return_addr_from_memory): New function.
5877 (s390_handle_string_attribute): New function.
5878 (s390_attribute_table): Add new attribute handler.
5879 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
5880 (s390_indirect_branch_via_thunk): New function.
5881 (s390_indirect_branch_via_inline_thunk): New function.
5882 (s390_function_ok_for_sibcall): When jumping via thunk disallow
5883 sibling call optimization for non z10 compiles.
5884 (s390_emit_call): Force indirect branch target to be a single
5885 register. Add r1 clobber for non-z10 compiles.
5886 (s390_emit_epilogue): Emit return jump via return_use expander.
5887 (s390_reorg): Handle JUMP_INSNs as execute targets.
5888 (s390_option_override_internal): Perform validity checks for the
5889 new command line options.
5890 (s390_indirect_branch_attrvalue): New function.
5891 (s390_indirect_branch_settings): New function.
5892 (s390_set_current_function): Invoke s390_indirect_branch_settings.
5893 (s390_output_indirect_thunk_function): New function.
5894 (s390_code_end): Implement target hook.
5895 (s390_case_values_threshold): Implement target hook.
5896 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
5898 * config/s390/s390.h (struct s390_frame_layout)
5899 (struct machine_function): Move here from s390.c.
5900 (TARGET_INDIRECT_BRANCH_NOBP_RET)
5901 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
5902 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
5903 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
5904 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
5905 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
5906 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
5907 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
5908 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
5909 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
5910 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
5911 (mnemonic attribute): Add values which aren't recognized
5913 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
5914 pattern for branch conversion. Fix mnemonic attribute.
5915 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
5916 indirect branch via thunk if requested.
5917 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
5918 ("*indirect_jump"): Disable for branch conversion using out of
5920 ("indirect_jump_via_thunk<mode>_z10")
5921 ("indirect_jump_via_thunk<mode>")
5922 ("indirect_jump_via_inlinethunk<mode>_z10")
5923 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
5924 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
5925 ("casesi_jump_via_inlinethunk<mode>_z10")
5926 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
5927 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
5928 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
5929 ("*indirect2_jump"): Disable for branch conversion.
5930 ("casesi_jump"): Turn into expander and expand patterns for branch
5932 ("return_use"): New expander.
5933 ("*return"): Emit return via thunk and rename it to ...
5934 ("*return<mode>"): ... this one.
5935 * config/s390/s390.opt: Add new options and and enum for the
5938 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5940 * lra-constraints.c (match_reload): Unconditionally use
5941 gen_lowpart_SUBREG, rather than selecting between that
5942 and equivalent gen_rtx_SUBREG code.
5944 2018-02-08 Richard Biener <rguenther@suse.de>
5946 PR tree-optimization/84233
5947 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
5948 changed flag instead of boguously re-using phi_inserted.
5950 2018-02-08 Martin Jambor <mjambor@suse.cz>
5952 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
5953 static local variables.
5955 2018-02-08 Richard Biener <rguenther@suse.de>
5957 PR tree-optimization/84278
5958 * tree-vect-stmts.c (vectorizable_store): When looking for
5959 smaller vector types to perform grouped strided loads/stores
5960 make sure the mode is supported by the target.
5961 (vectorizable_load): Likewise.
5963 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
5965 * config/aarch64/aarch64.c (aarch64_components_for_bb):
5966 Increase LDP/STP opportunities by adding adjacent callee-saves.
5968 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
5970 PR rtl-optimization/84068
5971 PR rtl-optimization/83459
5972 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
5974 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
5976 PR tree-optimization/84224
5977 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
5978 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
5981 2018-02-07 Iain Sandoe <iain@codesourcery.com>
5984 * config/rs6000/altivec.md (*restore_world): Remove LR use.
5985 * config/rs6000/predicates.md (restore_world_operation): Adjust op
5986 count, remove one USE.
5988 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
5990 * doc/install.texi (Configuration): Document the
5991 --with-long-double-format={ibm,ieee} PowerPC configuration
5995 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
5996 Convert from define_expand to be define_insn_and_split. Rework
5997 float/double/_Float128 conversions to QI/HI/SImode to work with
5998 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
5999 conversions to QI/HImode types did a store and then a load to
6000 truncate the value. For conversions to VSX registers, don't split
6001 the insn, instead emit the code directly. Use the code iterator
6002 any_fix to combine signed and unsigned conversions.
6003 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
6004 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
6005 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
6006 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
6007 (fix_<mode>di2_hw): Likewise.
6008 (fixuns_<mode>di2_hw): Likewise.
6009 (fix_<mode>si2_hw): Likewise.
6010 (fixuns_<mode>si2_hw): Likewise.
6011 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
6012 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
6013 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
6014 fix<uns>_trunc<SFDF:mode>si2_p8.
6015 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
6017 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
6018 (fix<uns>_<mode>_mem): Likewise.
6019 (fctiw<u>z_<mode>_mem): Likewise.
6020 (fix<uns>_<mode>_mem): Likewise.
6021 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
6022 the register allocator from doing a direct move to the GPRs to do
6023 a store, and instead use the ISA 3.0 store byte/half-word from
6024 vector register instruction. For IEEE 128-bit floating point,
6025 also optimize stores of 32-bit ints.
6026 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
6028 2018-02-07 Alan Hayward <alan.hayward@arm.com>
6030 * genextract.c (push_pathstr_operand): New function to support
6032 (walk_rtx): Call push_pathstr_operand.
6033 (print_path): Support [a-zA-Z].
6035 2018-02-07 Richard Biener <rguenther@suse.de>
6037 PR tree-optimization/84037
6038 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
6039 (cse_and_gimplify_to_preheader): Declare.
6040 (vect_get_place_in_interleaving_chain): Likewise.
6041 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
6043 (_loop_vec_info::~_loop_vec_info): Delete it.
6044 (cse_and_gimplify_to_preheader): New function.
6045 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
6046 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
6047 (vectorizable_load): Likewise. For grouped stores always base
6048 the IV on the first element.
6049 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
6050 condition before gimplifying.
6052 2018-02-07 Jakub Jelinek <jakub@redhat.com>
6054 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
6055 *DIV_EXPR and *MOD_EXPR.
6057 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
6060 * config/i386/i386.c (ix86_option_override_internal): Mask out
6061 the CF_SET bit when checking -fcf-protection.
6063 2018-02-07 Tom de Vries <tom@codesourcery.com>
6066 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
6069 2018-02-07 Richard Biener <rguenther@suse.de>
6071 PR tree-optimization/84204
6072 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
6075 PR tree-optimization/84205
6076 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
6077 special-case isl_ast_op_zdiv_r.
6079 PR tree-optimization/84223
6080 * graphite-scop-detection.c (gather_bbs::before_dom_children):
6081 Only add conditions from within the region.
6082 (gather_bbs::after_dom_children): Adjust.
6084 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
6087 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
6088 * config/avr/avr.md: Only post-reload split REG-REG moves if
6089 either register is GENERAL_REG_P.
6091 2018-02-07 Jakub Jelinek <jakub@redhat.com>
6093 PR tree-optimization/84235
6094 * tree-ssa-scopedtables.c
6095 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
6096 if the subtraction is performed in floating point type where NaNs are
6097 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
6098 build 1. Formatting fix.
6100 2018-02-06 Jakub Jelinek <jakub@redhat.com>
6103 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
6104 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
6105 and skip it regardless of bb boundaries. Use CALL_P macro,
6106 don't test INSN_P (insn) together with CALL_P or JUMP_P check
6107 unnecessarily, formatting fix.
6109 2018-02-06 Michael Collison <michael.collison@arm.com>
6111 * config/arm/thumb2.md:
6112 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
6113 (*thumb_mov_notscc): Ditto.
6115 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
6118 * config/rs6000/rs6000.md (su code attribute): Use "u" for
6119 unsigned_fix, not "s".
6121 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6123 * configure.ac (gcc_fn_eh_frame_ro): New function.
6124 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
6125 correct .eh_frame permissions.
6126 * configure: Regenerate.
6128 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
6130 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
6133 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6135 * config/rs6000/rs6000.c (rs6000_option_override_internal):
6136 Display warning message for -mno-speculate-indirect-jumps.
6138 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
6140 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
6142 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
6144 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
6146 PR tree-optimization/84225
6147 * tree-eh.c (find_trapping_overflow): Only call
6148 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
6150 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6153 * config/i386/i386.c: Reimplement the check of possible options
6154 -mibt/-mshstk conbination. Change error messages.
6155 * doc/invoke.texi: Fix a typo: remove extra '='.
6157 2018-02-06 Marek Polacek <polacek@redhat.com>
6159 PR tree-optimization/84228
6160 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
6162 2018-02-06 Tamar Christina <tamar.christina@arm.com>
6165 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
6166 emitted arch directives.
6167 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
6168 __ARM_FEATURE_COPROC before changing architectures.
6170 2018-02-06 Richard Biener <rguenther@suse.de>
6172 * config/i386/i386.c (print_reg): Fix typo.
6173 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
6175 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
6177 * configure: Regenerate.
6179 2018-02-05 Martin Sebor <msebor@redhat.com>
6181 PR tree-optimization/83369
6182 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
6185 2018-02-05 Martin Liska <mliska@suse.cz>
6187 * doc/invoke.texi: Cherry-pick upstream r323995.
6189 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
6191 * ira.c (ira_init_register_move_cost): Adjust comment.
6193 2018-02-05 Martin Liska <mliska@suse.cz>
6195 PR gcov-profile/84137
6196 * doc/gcov.texi: Fix typo in documentation.
6198 2018-02-05 Martin Liska <mliska@suse.cz>
6200 PR gcov-profile/83879
6201 * doc/gcov.texi: Document necessity of --dynamic-list-data when
6202 using dlopen functionality.
6204 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
6206 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
6207 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
6208 _mm_maskz_range_ss, _mm_mask_range_round_ss,
6209 _mm_maskz_range_round_ss): New intrinsics.
6210 (__builtin_ia32_rangesd128_round)
6211 (__builtin_ia32_rangess128_round): Remove.
6212 (__builtin_ia32_rangesd128_mask_round,
6213 __builtin_ia32_rangess128_mask_round): New builtins.
6214 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
6215 __builtin_ia32_rangess128_round): Remove.
6216 (__builtin_ia32_rangesd128_mask_round,
6217 __builtin_ia32_rangess128_mask_round): New builtins.
6218 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
6219 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
6220 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
6221 "<round_saeonly_constraint>")): Changed to ...
6222 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
6223 "<round_saeonly_scalar_constraint>")): ... this.
6224 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
6225 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
6226 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
6227 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
6228 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
6230 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
6232 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
6234 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
6235 Remove all values except native, 8540 and 8548.
6237 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
6239 * config/i386/i386.c (ix86_output_function_return): Pass
6240 INVALID_REGNUM, instead of -1, as invalid register number to
6241 indirect_thunk_name and output_indirect_thunk.
6243 2018-02-02 Julia Koval <julia.koval@intel.com>
6245 * config.gcc: Add -march=icelake.
6246 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
6247 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
6248 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
6249 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
6250 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
6251 (processor_target_table): Add icelake.
6252 (ix86_option_override_internal): Handle new PTAs.
6253 (get_builtin_code_for_version): Handle icelake.
6254 (M_INTEL_COREI7_ICELAKE): New.
6255 (fold_builtin_cpu): Handle icelake.
6256 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
6257 * doc/invoke.texi: Add -march=icelake.
6259 2018-02-02 Julia Koval <julia.koval@intel.com>
6261 * config/i386/i386.c (ix86_option_override_internal): Change flags type
6262 to wide_int_bitmask.
6263 * wide-int-bitmask.h: New.
6265 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6268 * config/i386/i386.md: Replace Pmode with word_mode in
6269 builtin_setjmp_setup and builtin_longjmp to support x32.
6271 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
6275 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
6277 (rs6000_supported_cpu_names): New static variable.
6278 (linux_cpu_translation_table): Likewise.
6279 (elf_platform) <cpu>: Define new static variable and use it.
6280 Translate kernel AT_PLATFORM name to canonical name if needed.
6281 Error if platform name is unknown.
6283 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
6286 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
6288 2018-02-01 Jeff Law <law@redhat.com>
6291 * config/i386/i386.c (release_scratch_register_on_entry): Add new
6292 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
6293 the scratch if RELEASE_VIA_POP is false.
6294 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
6295 If we have to save a temporary register, decrement SIZE appropriately.
6296 Pass new arguments to release_scratch_register_on_entry.
6297 (ix86_adjust_stack_and_probe): Likewise.
6298 (ix86_emit_probe_stack_range): Pass new arguments to
6299 release_scratch_register_on_entry.
6301 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
6303 PR rtl-optimization/84157
6304 * combine.c (change_zero_ext): Use REG_P predicate in
6305 front of HARD_REGISTER_P predicate.
6307 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
6309 * config/avr/avr.c (avr_option_override): Move disabling of
6310 -fdelete-null-pointer-checks to...
6311 * common/config/avr/avr-common.c (avr_option_optimization_table):
6314 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6316 PR tree-optimization/81635
6317 * tree-data-ref.c (split_constant_offset_1): For types that
6318 wrap on overflow, try to use range info to prove that wrapping
6321 2018-02-01 Renlin Li <renlin.li@arm.com>
6324 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
6326 (aarch64_register_move_cost): Likewise.
6327 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
6329 (REG_CLASS_NAMES): Likewise.
6330 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
6331 TAILCALL_ADDR_REGS. Remove IP registers.
6332 * config/aarch64/aarch64.md (Ucs): Update register constraint.
6334 2018-02-01 Richard Biener <rguenther@suse.de>
6336 * domwalk.h (dom_walker::dom_walker): Add additional constructor
6337 for specifying RPO order and allow NULL for that.
6338 * domwalk.c (dom_walker::dom_walker): Likewise.
6339 (dom_walker::walk): Handle NULL RPO order.
6340 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
6342 (rewrite_update_dom_walker): Likewise.
6343 (mark_def_dom_walker): Likewise.
6345 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6347 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
6348 (aarch64_maybe_expand_sve_subreg_move): Declare.
6349 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
6350 * config/aarch64/predicates.md (aarch64_any_register_operand): New
6352 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
6353 that are semantically a reverse operation.
6354 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
6355 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
6356 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
6358 (aarch64_can_change_mode_class): For big-endian, forbid changes
6359 between two SVE modes if they have different element sizes.
6361 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6363 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
6364 the TImode handling for big-endian targets.
6366 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6368 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
6369 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
6371 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
6372 Remove BSWAP handing for big-endian targets and use the form of
6373 LD1RQ appropariate for the mode.
6375 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6377 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
6378 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
6381 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6384 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
6385 check for operands that need to go through aarch64_sve_reload_be.
6387 2018-02-01 Jakub Jelinek <jakub@redhat.com>
6389 PR tree-optimization/81661
6390 PR tree-optimization/84117
6391 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
6392 * tree-eh.c: Include gimplify.h.
6393 (find_trapping_overflow, replace_trapping_overflow,
6394 rewrite_to_non_trapping_overflow): New functions.
6395 * tree-vect-loop.c: Include tree-eh.h.
6396 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
6397 * tree-data-ref.c: Include tree-eh.h.
6398 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
6400 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
6402 PR rtl-optimization/84123
6403 * combine.c (change_zero_ext): Check if hard register satisfies
6404 can_change_dest_mode before calling gen_lowpart_SUBREG.
6406 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
6409 * ira.c (ira_init_register_move_cost): Remove assert.
6411 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
6413 PR rtl-optimization/84071
6414 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
6415 * doc/tm.texi: Regenerate.
6417 2018-01-31 Richard Biener <rguenther@suse.de>
6419 PR tree-optimization/84132
6420 * tree-data-ref.c (analyze_miv_subscript): Properly
6421 check whether evolution_function_is_affine_multivariate_p
6422 before calling gcd_of_steps_may_divide_p.
6424 2018-01-31 Julia Koval <julia.koval@intel.com>
6427 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
6428 * config/i386/i386.md (rdpid_rex64) New.
6429 (rdpid): Make 32bit only.
6431 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
6434 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
6435 an IDENTIFIER_NODE for FUNCTION_TYPE's.
6437 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
6440 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
6442 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
6444 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
6446 PR rtl-optimization/84071
6447 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
6448 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
6450 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
6452 * config/arc/arc.c (arc_handle_aux_attribute): New function.
6453 (arc_attribute_table): Add 'aux' attribute.
6454 (arc_in_small_data_p): Consider aux like variables.
6455 (arc_is_aux_reg_p): New function.
6456 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
6457 (arc_get_aux_arg): New function.
6458 (prepare_move_operands): Handle aux-register access.
6459 (arc_handle_aux_attribute): New function.
6460 * doc/extend.texi (ARC Variable attributes): Add subsection.
6462 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
6464 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
6465 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
6466 (arc_attribute_table): Add 'uncached' attribute.
6467 (arc_print_operand): Print '.di' flag for uncached memory
6469 (arc_in_small_data_p): Do not consider for small data the uncached
6471 (arc_is_uncached_mem_p): New function.
6472 * config/arc/predicates.md (compact_store_memory_operand): Check
6473 for uncached memory accesses.
6474 (nonvol_nonimm_operand): Likewise.
6475 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
6477 2018-01-31 Jakub Jelinek <jakub@redhat.com>
6480 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
6481 falign-loops=): Add Optimization flag.
6483 2018-01-30 Jeff Law <law@redhat.com>
6486 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
6487 INT_REGISTERS_SAVED. Check it prior to calling
6488 get_scratch_register_on_entry.
6489 (ix86_adjust_stack_and_probe): Similarly.
6490 (ix86_emit_probe_stack_range): Similarly.
6491 (ix86_expand_prologue): Corresponding changes.
6493 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6496 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
6497 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
6499 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
6502 * lra-constraints.c (curr_insn_transform): Process AND in the
6505 2018-01-30 Jakub Jelinek <jakub@redhat.com>
6507 PR rtl-optimization/83986
6508 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
6509 dependence against last_pending_memory_flush in addition to
6512 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
6514 PR tree-optimization/81611
6515 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
6518 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6521 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
6524 2018-01-30 Richard Biener <rguenther@suse.de>
6525 Jakub Jelinek <jakub@redhat.com>
6527 PR tree-optimization/84111
6528 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
6529 inner loops added during recursion, as they don't have up-to-date
6532 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
6535 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
6536 (can_inline_edge_by_limits_p): ... here.
6537 (can_early_inline_edge_p, check_callers,
6538 update_caller_keys, update_callee_keys, recursive_inlining,
6539 add_new_edges_to_heap, speculation_useful_p,
6540 inline_small_functions,
6541 inline_small_functions, flatten_function,
6542 inline_to_all_callers_1): Update.
6544 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
6546 * profile-count.c (profile_count::combine_with_ipa_count): Handle
6549 2018-01-30 Richard Biener <rguenther@suse.de>
6551 PR tree-optimization/83008
6552 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
6553 invariant and constant vector uses in stmts when they need
6556 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6559 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
6560 * configure: Regenerate.
6562 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
6564 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
6566 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
6567 Use gen_rtx_REG rather than gen_lowpart.
6569 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
6571 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
6572 rather than 0 when creating partial subregs.
6574 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
6576 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
6579 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
6582 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
6583 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
6584 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
6585 flags. This restores the settings used before the 2017-07-24.
6586 Turning off pre increment/decrement/modify allows IVOPTS to
6587 optimize DF/SF loops where the index is an int.
6589 2018-01-29 Richard Biener <rguenther@suse.de>
6590 Kelvin Nilsen <kelvin@gcc.gnu.org>
6593 * tree-vect-stmts.c (vectorizable_call): Don't call
6594 targetm.vectorize_builtin_md_vectorized_function if callee is
6597 2018-01-22 Carl Love <cel@us.ibm.com>
6599 * doc/extend.tex: Fix typo in second arg in
6600 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
6602 2018-01-29 Richard Biener <rguenther@suse.de>
6604 PR tree-optimization/84086
6605 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
6606 (flush_ssaname_freelist): When SSA names were released reset
6607 the SCEV hash table.
6609 2018-01-29 Richard Biener <rguenther@suse.de>
6611 PR tree-optimization/84057
6612 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
6613 removed paths when removing edges.
6615 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
6617 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
6618 -mfunction-return=@var{choice}.
6620 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
6623 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
6625 (layout::print_source_line): Likewise.
6626 (test_get_line_width_without_trailing_whitespace): Add test cases.
6628 2018-01-27 Jakub Jelinek <jakub@redhat.com>
6631 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
6634 2018-01-26 Jim Wilson <jimw@sifive.com>
6636 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
6638 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
6641 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6643 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
6644 and CMP + SUB-immediate -> SUBS.
6646 2018-01-26 Martin Sebor <msebor@redhat.com>
6648 PR tree-optimization/83896
6649 * tree-ssa-strlen.c (get_string_len): Rename...
6650 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
6651 Avoid assuming length is constant.
6652 (handle_char_store): Use HOST_WIDE_INT for string length.
6654 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
6657 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
6658 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
6660 2018-01-26 Richard Biener <rguenther@suse.de>
6662 PR rtl-optimization/84003
6663 * dse.c (record_store): Only record redundant stores when
6664 the earlier store aliases at least all accesses the later one does.
6666 2018-01-26 Jakub Jelinek <jakub@redhat.com>
6668 PR rtl-optimization/83985
6669 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
6670 REG_CFA_RESTORE insns.
6671 (delete_unmarked_insns): Don't ignore separate shrink wrapping
6672 REG_CFA_RESTORE insns here.
6675 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
6676 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
6678 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6680 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
6681 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
6682 (arc_init): Likewise.
6683 (arc_override_options): Likewise.
6684 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
6686 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
6688 * config/arc/arc.h (TARGET_DBNZ): Define.
6689 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
6690 properly set the tune attribute.
6691 (dbnz): Use TARGET_DBNZ guard.
6692 * config/arc/arc.opt (mtune): Add core3 option.
6694 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6696 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
6697 recognize new pic like addresses.
6698 (arc_delegitimize_address): Clean up.
6700 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6702 * config/arc/arc-arches.def: Option mrf16 valid for all
6704 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
6705 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
6706 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
6707 * config/arc/arc-tables.opt: Regenerate.
6708 * config/arc/arc.c (arc_conditional_register_usage): Handle
6709 reduced register file case.
6710 (arc_file_start): Set must have build attributes.
6711 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
6713 * config/arc/arc.opt (mrf16): Add new option.
6714 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
6715 * config/arc/genmultilib.awk: Handle new mrf16 option.
6716 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
6717 * config/arc/t-multilib: Regenerate.
6718 * doc/invoke.texi (ARC Options): Document mrf16 option.
6720 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6722 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
6723 * config/arc/arc.c (arc_handle_secure_attribute): New function.
6724 (arc_attribute_table): Add 'secure_call' attribute.
6725 (arc_print_operand): Print secure call operand.
6726 (arc_function_ok_for_sibcall): Don't optimize tail calls when
6728 (arc_is_secure_call_p): New function. * config/arc/arc.md
6729 (call_i): Add support for sjli instruction.
6730 (call_value_i): Likewise.
6731 * config/arc/constraints.md (Csc): New constraint.
6733 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6734 John Eric Martin <John.Martin@emmicro-us.com>
6736 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
6737 * config/arc/arc.c (_arc_jli_section): New struct.
6738 (arc_jli_section): New type.
6739 (rc_jli_sections): New static variable.
6740 (arc_handle_jli_attribute): New function.
6741 (arc_attribute_table): Add jli_always and jli_fixed attribute.
6742 (arc_file_end): New function.
6743 (TARGET_ASM_FILE_END): Define.
6744 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
6745 (arc_add_jli_section): New function.
6746 (jli_call_scan): Likewise.
6747 (arc_reorg): Call jli_call_scan.
6748 (arc_output_addsi): Remove 'S' from printing asm operand.
6749 (arc_is_jli_call_p): New function.
6750 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
6752 (movhi_insn): Likewise.
6753 (movsi_insn): Likewise.
6754 (movsi_set_cc_insn): Likewise.
6755 (loadqi_update): Likewise.
6756 (load_zeroextendqisi_update): Likewise.
6757 (load_signextendqisi_update): Likewise.
6758 (loadhi_update): Likewise.
6759 (load_zeroextendhisi_update): Likewise.
6760 (load_signextendhisi_update): Likewise.
6761 (loadsi_update): Likewise.
6762 (loadsf_update): Likewise.
6763 (movsicc_insn): Likewise.
6764 (bset_insn): Likewise.
6765 (bxor_insn): Likewise.
6766 (bclr_insn): Likewise.
6767 (bmsk_insn): Likewise.
6768 (bicsi3_insn): Likewise.
6769 (cmpsi_cc_c_insn): Likewise.
6770 (movsi_ne): Likewise.
6771 (movsi_cond_exec): Likewise.
6772 (clrsbsi2): Likewise.
6783 (call_i): Remove 'S' asm letter, add jli instruction.
6784 (call_value_i): Likewise.
6785 * config/arc/arc.op (mjli-always): New option.
6786 * config/arc/constraints.md (Cji): New constraint.
6787 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
6789 (subsf3_fpx): Likewise.
6790 (mulsf3_fpx): Likewise.
6791 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
6793 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
6795 * doc/invoke.texi (ARC): Document mjli-always option.
6797 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
6799 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
6800 avoid addition with 0 and use incw and decw where possible.
6802 2018-01-26 Richard Biener <rguenther@suse.de>
6804 PR tree-optimization/81082
6805 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
6806 association if it requires casting to unsigned.
6807 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
6808 from fold_plusminus_mult_expr to catch important cases late when
6809 range info is available.
6811 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6813 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
6814 * configure.ac (hidden_linkonce): New test.
6815 * configure: Regenerate.
6816 * config.in: Regenerate.
6818 2018-01-26 Julia Koval <julia.koval@intel.com>
6820 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
6821 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
6822 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
6823 _mm_mask_bitshuffle_epi64_mask): Fix type.
6824 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
6825 USI_FTYPE_V4DI_V4DI_USI): Remove.
6826 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
6827 __builtin_ia32_vpshufbitqmb256_mask,
6828 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
6829 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
6830 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
6832 2018-01-26 Alan Modra <amodra@gmail.com>
6835 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
6836 UNSPEC_VBPERMQ. Sort other unspecs.
6838 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
6840 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
6842 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
6845 * predict.c (drop_profile): Do not push/pop cfun; update also
6847 (handle_missing_profiles): Fix logic looking for zero profiles.
6849 2018-01-25 Jakub Jelinek <jakub@redhat.com>
6852 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
6853 on functions with #pragma omp declare simd or functions with simd
6855 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
6856 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
6857 Remove trailing \n from warning_at calls.
6859 2018-01-25 Tom de Vries <tom@codesourcery.com>
6862 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
6863 for neutered workers.
6865 2018-01-24 Joseph Myers <joseph@codesourcery.com>
6868 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
6869 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
6871 2018-01-24 Jeff Law <law@redhat.com>
6874 * i386.c (get_probe_interval): Move to earlier point.
6875 (ix86_compute_frame_layout): If -fstack-clash-protection and
6876 the frame is larger than the probe interval, then use pushes
6877 to save registers rather than reg->mem moves.
6878 (ix86_expand_prologue): Remove conditional for int_registers_saved
6881 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
6884 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
6885 min/max for never referenced object.
6887 2018-01-24 Jakub Jelinek <jakub@redhat.com>
6890 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
6892 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
6893 attributes from DECL_ATTRIBUTES (decl) without affecting
6894 DECL_ATTRIBUTES (current_function_decl).
6895 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
6896 functions with non-NULL DECL_ABSTRACT_ORIGIN.
6898 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
6900 PR tree-optimization/83979
6901 * fold-const.c (fold_comparison): Use constant_boolean_node
6902 instead of boolean_{true,false}_node.
6904 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
6906 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
6909 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6911 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
6912 Simplify the clause that sets the length attribute.
6913 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
6914 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
6915 clause that sets the length attribute.
6916 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
6918 2018-01-24 Tom de Vries <tom@codesourcery.com>
6921 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
6922 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
6923 Add strict parameter.
6924 (prevent_branch_around_nothing): Insert dummy insn between branch to
6925 label and label with no ptx insn inbetween.
6926 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
6928 2018-01-24 Tom de Vries <tom@codesourcery.com>
6931 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
6932 for neutered threads in warp.
6933 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
6935 2018-01-24 Richard Biener <rguenther@suse.de>
6937 PR tree-optimization/83176
6938 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
6941 2018-01-24 Richard Biener <rguenther@suse.de>
6943 PR tree-optimization/82819
6944 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
6945 code generating pluses that are no-ops in the target precision.
6947 2018-01-24 Richard Biener <rguenther@suse.de>
6950 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
6952 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
6954 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
6955 to merge probabilities.
6956 * predict.c (probably_never_executed): Also mark as cold functions
6957 with global 0 profile and guessed local profile.
6958 * profile-count.c (profile_probability::combine_with_count): New
6960 * profile-count.h (profile_probability::operator*,
6961 profile_probability::operator*=, profile_probability::operator/,
6962 profile_probability::operator/=): Reduce precision to adjusted
6963 and set value to guessed on contradictory divisions.
6964 (profile_probability::combine_with_freq): Remove.
6965 (profile_probability::combine_wiht_count): Declare.
6966 (profile_count::force_nonzero):: Set to adjusted.
6967 (profile_count::probability_in):: Set quality to adjusted.
6968 * tree-ssa-tail-merge.c (replace_block_by): Use
6971 2018-01-23 Andrew Waterman <andrew@sifive.com>
6972 Jim Wilson <jimw@sifive.com>
6974 * config/riscv/riscv.c (riscv_stack_boundary): New.
6975 (riscv_option_override): Set riscv_stack_boundary. Handle
6976 riscv_preferred_stack_boundary_arg.
6977 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
6978 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
6979 (STACK_BOUNDARY): Set to riscv_stack_boundary.
6980 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
6981 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
6982 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
6984 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
6987 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
6988 of struct ix86_frame.
6989 (ix86_expand_epilogue): Likewise. Add a local variable for
6990 the reg_save_offset field in struct ix86_frame.
6992 2018-01-23 Bin Cheng <bin.cheng@arm.com>
6994 PR tree-optimization/82604
6995 * tree-loop-distribution.c (enum partition_kind): New enum item
6996 PKIND_PARTIAL_MEMSET.
6997 (partition_builtin_p): Support above new enum item.
6998 (generate_code_for_partition): Ditto.
6999 (compute_access_range): Differentiate cases that equality can be
7000 proven at all loops, the innermost loops or no loops.
7001 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
7002 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
7003 (finalize_partitions, distribute_loop): Don't fuse partition of
7004 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
7005 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
7008 2018-01-23 Martin Liska <mliska@suse.cz>
7010 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
7011 order to ignore the predictor.
7012 (PRED_POLYMORPHIC_CALL): Likewise.
7013 (PRED_RECURSIVE_CALL): Likewise.
7015 2018-01-23 Martin Liska <mliska@suse.cz>
7017 * tree-profile.c (tree_profiling): Print function header to
7018 aware reader which function we are working on.
7019 * value-prof.c (gimple_find_values_to_profile): Do not print
7020 not interesting value histograms.
7022 2018-01-23 Martin Liska <mliska@suse.cz>
7024 * profile-count.h (enum profile_quality): Add
7025 profile_uninitialized as the first value. Do not number values
7026 as they are zero based.
7027 (profile_count::verify): Update sanity check.
7028 (profile_probability::verify): Likewise.
7030 2018-01-23 Nathan Sidwell <nathan@acm.org>
7032 * doc/invoke.texi (ffor-scope): Deprecate.
7034 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7036 PR tree-optimization/83510
7037 * domwalk.c (set_all_edges_as_executable): New function.
7038 (dom_walker::dom_walker): Convert bool param
7039 "skip_unreachable_blocks" to enum reachability. Move setup of
7040 edge flags to set_all_edges_as_executable and only do it when
7041 reachability is REACHABLE_BLOCKS.
7042 * domwalk.h (enum dom_walker::reachability): New enum.
7043 (dom_walker::dom_walker): Convert bool param
7044 "skip_unreachable_blocks" to enum reachability.
7045 (set_all_edges_as_executable): New decl.
7046 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
7047 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
7049 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
7050 but converting true to REACHABLE_BLOCKS.
7051 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
7053 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
7054 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
7055 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
7057 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
7058 if check_all_array_refs will be called.
7060 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7062 * tree.c (selftest::test_location_wrappers): Add more test
7065 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7067 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
7068 (selftest::test_bit_in_range): Likewise.
7070 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
7073 * doc/sourcebuild.texi (vect_float): Say that the selector
7074 only describes the situation when -funsafe-math-optimizations is on.
7075 (vect_float_strict): Document.
7077 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
7079 PR tree-optimization/83965
7080 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
7081 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
7082 instead of checking only for a reduction.
7083 (vect_recog_widen_sum_pattern): Likewise.
7085 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7087 * predict.c (probably_never_executed): Only use precise profile info.
7088 (compute_function_frequency): Skip after inlining hack since we now
7089 have quality checking.
7091 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7093 * profile-count.h (profile_probability::very_unlikely,
7094 profile_probability::unlikely, profile_probability::even): Set
7095 precision to guessed.
7097 2018-01-23 Richard Biener <rguenther@suse.de>
7099 PR tree-optimization/83963
7100 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
7101 Properly terminate dominator walk when crossing the exit edge not
7102 when visiting its source block.
7104 2018-01-23 Jakub Jelinek <jakub@redhat.com>
7107 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
7108 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
7110 2018-01-22 Jakub Jelinek <jakub@redhat.com>
7112 PR tree-optimization/83957
7113 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
7114 semicolon after for body surrounded by braces.
7116 PR tree-optimization/83081
7117 * profile-count.h (profile_probability::split): New method.
7118 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
7119 Use profile_probability::split.
7120 (do_compare_rtx_and_jump): Fix adjustment of probabilities
7121 when splitting a single conditional jump into 2.
7123 2018-01-22 David Malcolm <dmalcolm@redhat.com>
7125 PR tree-optimization/69452
7126 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
7129 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7131 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
7132 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
7133 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
7135 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7137 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
7138 * config/rl78/rl78.md (movdi): New define_expand.
7139 * config/rl78/rl78.c (rl78_split_movdi): New function.
7141 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
7144 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
7146 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
7147 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
7148 128-bit to produce an UNSPEC move to get the double word with the
7149 signbit and then a shift directly to do signbit.
7150 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
7151 implementation with a new version that just does either a direct
7152 move or a regular move. Move memory interface to separate insns.
7153 Move insns so they are next to the expander.
7154 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
7155 with signbit move. Split big and little endian case.
7156 (signbit<mode>2_dm_mem_le): Likewise.
7157 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
7158 (signbit<mode>2_dm2): Likewise.
7160 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7162 * config/rl78/rl78.md (anddi3): New define_expand.
7164 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7166 * config/rl78/rl78.md (umindi3): New define_expand.
7168 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7170 * config/rl78/rl78.md (smindi3): New define_expand.
7172 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7174 * config/rl78/rl78.md (smaxdi3): New define_expand.
7176 2018-01-22 Carl Love <cel@us.ibm.com>
7178 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
7179 LVX_V1TI): Add macro expansion.
7180 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
7181 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
7182 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
7183 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
7184 Change check to determine if the instruction is a byte reversing
7185 entry. Fix typo in comment.
7186 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
7187 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
7188 Add def_builtin calls for new builtins.
7189 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
7190 Add define_insn expansion.
7192 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7194 * config/rl78/rl78.md (umaxdi3): New define_expand.
7196 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7198 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
7199 for non-QImode registers.
7201 2018-01-22 Richard Biener <rguenther@suse.de>
7203 PR tree-optimization/83963
7204 * graphite-scop-detection.c (scop_detection::get_sese): Delay
7205 including the loop exit block.
7206 (scop_detection::merge_sese): Likewise.
7207 (scop_detection::add_scop): Do it here instead.
7209 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7211 * doc/sourcebuild.texi (arm_softfloat): Document.
7213 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
7216 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
7217 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
7218 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
7220 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7221 David Edelsohn <dje.gcc@gmail.com>
7224 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7225 Change "crset eq" to "crset 2".
7226 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7227 (*call_indirect_aix<mode>_nospec): Likewise.
7228 (*call_value_indirect_aix<mode>_nospec): Likewise.
7229 (*call_indirect_elfv2<mode>_nospec): Likewise.
7230 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
7231 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
7232 change assembly output from . to $.
7233 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7234 (indirect_jump<mode>_nospec): Change assembly output from . to $.
7235 (*tablejump<mode>_internal1_nospec): Likewise.
7237 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
7240 * config/sh/sh_optimize_sett_clrt.cc:
7241 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
7243 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
7245 PR tree-optimization/83940
7246 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
7247 offset_dt to vect_constant_def rather than vect_unknown_def_type.
7248 (vect_check_load_store_mask): Add a mask_dt_out parameter and
7249 use it to pass back the definition type.
7250 (vect_check_store_rhs): Likewise rhs_dt_out.
7251 (vect_build_gather_load_calls): Add a mask_dt argument and use
7252 it instead of a call to vect_is_simple_use.
7253 (vectorizable_store): Update calls to vect_check_load_store_mask
7254 and vect_check_store_rhs. Use the dt returned by the latter instead
7255 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
7256 instead of calls to vect_is_simple_use. Pass the scalar rather
7257 than the vector operand to vect_is_simple_use when handling
7258 second and subsequent copies of an rhs value.
7259 (vectorizable_load): Update calls to vect_check_load_store_mask
7260 and vect_build_gather_load_calls. Use the cached mask_dt and
7261 gs_info.offset_dt instead of calls to vect_is_simple_use.
7263 2018-01-20 Jakub Jelinek <jakub@redhat.com>
7266 * tree-emutls.c: Include gimplify.h.
7267 (lower_emutls_2): New function.
7268 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
7269 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
7270 it before further processing.
7273 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
7274 UINTVAL (trueop1) instead of INTVAL (op1).
7276 2018-01-19 Jakub Jelinek <jakub@redhat.com>
7280 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
7281 INCOMING_FRAME_SP_OFFSET if not defined.
7282 (scan_trace): Add ENTRY argument. If true and
7283 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
7284 emit a note to adjust the CFA offset.
7285 (create_cfi_notes): Adjust scan_trace callers.
7286 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
7287 INCOMING_FRAME_SP_OFFSET in the CIE.
7288 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
7289 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
7291 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
7292 * doc/tm.texi: Regenerated.
7294 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7296 PR rtl-optimization/83147
7297 * lra-constraints.c (remove_inheritance_pseudos): Use
7298 lra_substitute_pseudo_within_insn.
7300 2018-01-19 Tom de Vries <tom@codesourcery.com>
7301 Cesar Philippidis <cesar@codesourcery.com>
7304 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
7306 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
7309 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
7310 spaces for function labels.
7312 2018-01-19 Martin Liska <mliska@suse.cz>
7314 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
7315 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
7316 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
7317 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
7318 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
7319 (PRED_CONST_RETURN): Change from 69 to 65.
7320 (PRED_NULL_RETURN): Change from 91 to 71.
7321 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
7322 (PRED_LOOP_GUARD): Change from 66 to 73.
7324 2018-01-19 Martin Liska <mliska@suse.cz>
7326 * predict.c (predict_insn_def): Add new assert.
7327 (struct branch_predictor): Change type to signed integer.
7328 (test_prediction_value_range): Amend test to cover
7330 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
7331 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
7332 (PRED_LOOP_ITERATIONS_MAX): Likewise.
7333 (PRED_LOOP_IV_COMPARE): Likewise.
7334 * predict.h (PROB_UNINITIALIZED): Define new constant.
7336 2018-01-19 Martin Liska <mliska@suse.cz>
7338 * predict.c (dump_prediction): Add new format for
7339 analyze_brprob.py script which is enabled with -details
7341 * profile-count.h (precise_p): New function.
7343 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
7345 PR tree-optimization/83922
7346 * tree-vect-loop.c (vect_verify_full_masking): Return false if
7347 there are no statements that need masking.
7348 (vect_active_double_reduction_p): New function.
7349 (vect_analyze_loop_operations): Use it when handling phis that
7350 are not in the loop header.
7352 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
7354 PR tree-optimization/83914
7355 * tree-vect-loop.c (vectorizable_induction): Don't convert
7356 init_expr or apply the peeling adjustment for inductions
7357 that are nested within the vectorized loop.
7359 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7361 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
7364 2018-01-18 Jakub Jelinek <jakub@redhat.com>
7368 * function.h (gimplify_parameters): Add gimple_seq * argument.
7369 * function.c: Include gimple.h and options.h.
7370 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
7371 for the added local temporaries if needed.
7372 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
7373 if there are any parameter cleanups, wrap whole body into a
7374 try/finally with the cleanups.
7376 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
7379 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
7380 Use GET_MODE_CLASS for scalar floating point.
7382 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
7386 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
7387 Fix call of call_cgraph_insertion_hooks.
7389 2018-01-18 Martin Sebor <msebor@redhat.com>
7391 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
7393 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
7396 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
7399 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
7402 * common.opt: (-ffile-prefix-map): New option.
7403 * opts.c (common_handle_option): Defer it.
7404 * opts-global.c (handle_common_deferred_options): Handle it.
7405 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
7406 * file-prefix-map.h: New file.
7407 (remap_debug_filename, add_debug_prefix_map): ...here.
7408 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
7409 * final.c (debug_prefix_map, add_debug_prefix_map
7410 remap_debug_filename): Move to...
7411 * file-prefix-map.c: New file.
7412 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
7413 generalize, get rid of alloca(), use strrchr() instead of strchr().
7414 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
7415 Implement in terms of add_prefix_map().
7416 (remap_macro_filename, remap_debug_filename): Implement in term of
7418 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
7419 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
7420 * dbxout.c: Include file-prefix-map.h.
7421 * varasm.c: Likewise.
7422 * vmsdbgout.c: Likewise.
7423 * xcoffout.c: Likewise.
7424 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
7425 * doc/cppopts.texi (-fmacro-prefix-map): Document.
7426 * doc/invoke.texi (-ffile-prefix-map): Document.
7427 (-fdebug-prefix-map): Update description.
7429 2018-01-18 Martin Liska <mliska@suse.cz>
7431 * config/i386/i386.c (indirect_thunk_name): Document that also
7433 (output_indirect_thunk): Document why both instructions
7434 (pause and lfence) are generated.
7436 2018-01-18 Richard Biener <rguenther@suse.de>
7438 PR tree-optimization/83887
7439 * graphite-scop-detection.c
7440 (scop_detection::get_nearest_dom_with_single_entry): Remove.
7441 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
7442 (scop_detection::merge_sese): Re-implement with a flood-fill
7443 algorithm that properly finds a SESE region if it exists.
7445 2018-01-18 Jakub Jelinek <jakub@redhat.com>
7448 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
7449 pointer_diff optimizations use view_convert instead of convert.
7451 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7453 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7454 Generate different code for -mno-speculate-indirect-jumps.
7455 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7456 (*call_indirect_aix<mode>): Disable for
7457 -mno-speculate-indirect-jumps.
7458 (*call_indirect_aix<mode>_nospec): New define_insn.
7459 (*call_value_indirect_aix<mode>): Disable for
7460 -mno-speculate-indirect-jumps.
7461 (*call_value_indirect_aix<mode>_nospec): New define_insn.
7462 (*sibcall_nonlocal_sysv<mode>): Generate different code for
7463 -mno-speculate-indirect-jumps.
7464 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7466 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
7468 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
7469 long double type, set the flags for noting the default long double
7470 type, even if we don't pass or return a long double type.
7472 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
7475 * ipa-inline.c (flatten_function): Do not overwrite final inlining
7478 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
7480 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
7481 support for merge[hl].
7482 (fold_mergehl_helper): New helper function.
7483 (tree-vector-builder.h): New #include for tree_vector_builder usage.
7484 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
7485 (altivec_vmrglw_direct): Add xxmrglw insn.
7487 2018-01-17 Andrew Waterman <andrew@sifive.com>
7489 * config/riscv/riscv.c (riscv_conditional_register_usage): If
7490 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
7492 2018-01-17 David Malcolm <dmalcolm@redhat.com>
7495 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
7496 call the lto_location_cache before reading the
7497 DECL_SOURCE_LOCATION of the types.
7499 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
7500 Richard Sandiford <richard.sandiford@linaro.org>
7502 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
7503 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
7504 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
7505 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
7506 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
7508 * config/aarch64/constraints.md (aarch64_movti_operand):
7510 * config/aarch64/predicates.md (Uti): Add new constraint.
7512 2018-01-17 Carl Love <cel@us.ibm.com>
7514 * config/rs6000/vsx.md (define_expand xl_len_r,
7515 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
7516 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
7518 (define_expand, define_insn): Move the shift left from the
7519 define_insn to the define_expand for lxvl and stxvl instructions.
7520 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
7521 and XL_LEN_R definitions to PURE.
7523 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
7525 * config/i386/i386.c (indirect_thunk_name): Declare regno
7526 as unsigned int. Compare regno with INVALID_REGNUM.
7527 (output_indirect_thunk): Ditto.
7528 (output_indirect_thunk_function): Ditto.
7529 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
7530 in the call to output_indirect_thunk_function.
7532 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
7535 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
7536 rather than the size of inner_type to determine the stack slot size
7537 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
7539 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
7542 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
7545 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
7547 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
7548 endian Linux systems to optionally enable multilibs for selecting
7549 the long double type if the user configured an explicit type.
7550 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
7551 have no long double multilibs if not defined.
7552 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
7553 warn if the user used -mabi={ieee,ibm}longdouble and we built
7554 multilibs for long double.
7555 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
7556 appropriate multilib option.
7557 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
7559 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
7560 for building long double multilibs.
7561 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
7563 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
7565 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
7568 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
7570 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
7573 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
7576 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
7579 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
7581 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
7582 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
7584 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
7586 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
7587 different rtl trees depending on TARGET_64BIT.
7588 (rs6000_gen_lvx): Likewise.
7590 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
7592 * config/visium/visium.md (nop): Tweak comment.
7593 (hazard_nop): Likewise.
7595 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7597 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
7598 -mspeculate-indirect-jumps.
7599 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
7600 for -mno-speculate-indirect-jumps.
7601 (*call_indirect_elfv2<mode>_nospec): New define_insn.
7602 (*call_value_indirect_elfv2<mode>): Disable for
7603 -mno-speculate-indirect-jumps.
7604 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
7605 (indirect_jump): Emit different RTL for
7606 -mno-speculate-indirect-jumps.
7607 (*indirect_jump<mode>): Disable for
7608 -mno-speculate-indirect-jumps.
7609 (*indirect_jump<mode>_nospec): New define_insn.
7610 (tablejump): Emit different RTL for
7611 -mno-speculate-indirect-jumps.
7612 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
7613 (tablejumpsi_nospec): New define_expand.
7614 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
7615 (tablejumpdi_nospec): New define_expand.
7616 (*tablejump<mode>_internal1): Disable for
7617 -mno-speculate-indirect-jumps.
7618 (*tablejump<mode>_internal1_nospec): New define_insn.
7619 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
7622 2018-01-16 Artyom Skrobov tyomitch@gmail.com
7624 * caller-save.c (insert_save): Drop unnecessary parameter. All
7627 2018-01-16 Jakub Jelinek <jakub@redhat.com>
7628 Richard Biener <rguenth@suse.de>
7631 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
7632 return early, inline manually is_gimple_sizepos. Make sure if we
7633 call gimplify_expr we don't end up with a gimple constant.
7634 * tree.c (variably_modified_type_p): Don't return true for
7635 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
7636 * gimplify.h (is_gimple_sizepos): Remove.
7638 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7640 PR tree-optimization/83857
7641 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
7642 vectorizable_live_operation for pure SLP statements.
7643 (vectorizable_live_operation): Handle PHIs.
7645 2018-01-16 Richard Biener <rguenther@suse.de>
7647 PR tree-optimization/83867
7648 * tree-vect-stmts.c (vect_transform_stmt): Precompute
7649 nested_in_vect_loop_p since the scalar stmt may get invalidated.
7651 2018-01-16 Jakub Jelinek <jakub@redhat.com>
7654 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
7655 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
7656 If off is not INTEGER_CST, issue a may not be aligned warning
7657 rather than isn't aligned. Use isn%'t rather than isn't.
7658 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
7660 <case MULT_EXPR>: Improve the case when bottom and one of the
7661 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
7662 operand, in that case check if the other operand is multiple of
7663 bottom divided by the INTEGER_CST operand.
7665 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7668 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
7669 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
7670 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
7671 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
7672 * config/pa/pa.c (pa_function_arg_advance): Likewise.
7673 (pa_function_arg, pa_arg_partial_bytes): Likewise.
7674 (pa_function_arg_size): New function.
7676 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7678 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
7679 in a separate statement.
7681 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7683 PR tree-optimization/83847
7684 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
7685 group gathers and scatters.
7687 2018-01-16 Jakub Jelinek <jakub@redhat.com>
7689 PR rtl-optimization/86620
7690 * params.def (max-sched-ready-insns): Bump minimum value to 1.
7692 PR rtl-optimization/83213
7693 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
7694 to last if both are JUMP_INSNs.
7696 PR tree-optimization/83843
7697 * gimple-ssa-store-merging.c
7698 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
7699 store_immediate_info for bswap/nop orig_stores.
7701 2018-01-15 Andrew Waterman <andrew@sifive.com>
7703 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
7705 <UDIV>: Increase cost if !TARGET_DIV.
7707 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
7709 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
7710 (define_attr "cr_logical_3op"): New.
7711 (cceq_ior_compare): Adjust.
7712 (cceq_ior_compare_complement): Adjust.
7713 (*cceq_rev_compare): Adjust.
7714 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
7715 (is_cracked_insn): Adjust.
7716 (insn_must_be_first_in_group): Adjust.
7717 * config/rs6000/40x.md: Adjust.
7718 * config/rs6000/440.md: Adjust.
7719 * config/rs6000/476.md: Adjust.
7720 * config/rs6000/601.md: Adjust.
7721 * config/rs6000/603.md: Adjust.
7722 * config/rs6000/6xx.md: Adjust.
7723 * config/rs6000/7450.md: Adjust.
7724 * config/rs6000/7xx.md: Adjust.
7725 * config/rs6000/8540.md: Adjust.
7726 * config/rs6000/cell.md: Adjust.
7727 * config/rs6000/e300c2c3.md: Adjust.
7728 * config/rs6000/e500mc.md: Adjust.
7729 * config/rs6000/e500mc64.md: Adjust.
7730 * config/rs6000/e5500.md: Adjust.
7731 * config/rs6000/e6500.md: Adjust.
7732 * config/rs6000/mpc.md: Adjust.
7733 * config/rs6000/power4.md: Adjust.
7734 * config/rs6000/power5.md: Adjust.
7735 * config/rs6000/power6.md: Adjust.
7736 * config/rs6000/power7.md: Adjust.
7737 * config/rs6000/power8.md: Adjust.
7738 * config/rs6000/power9.md: Adjust.
7739 * config/rs6000/rs64.md: Adjust.
7740 * config/rs6000/titan.md: Adjust.
7742 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7744 * config/i386/predicates.md (indirect_branch_operand): Rewrite
7745 ix86_indirect_branch_register logic.
7747 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7749 * config/i386/constraints.md (Bs): Update
7750 ix86_indirect_branch_register check. Don't check
7751 ix86_indirect_branch_register with GOT_memory_operand.
7753 * config/i386/predicates.md (GOT_memory_operand): Don't check
7754 ix86_indirect_branch_register here.
7755 (GOT32_symbol_operand): Likewise.
7757 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7759 * config/i386/predicates.md (constant_call_address_operand):
7760 Rewrite ix86_indirect_branch_register logic.
7761 (sibcall_insn_operand): Likewise.
7763 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7765 * config/i386/constraints.md (Bs): Replace
7766 ix86_indirect_branch_thunk_register with
7767 ix86_indirect_branch_register.
7769 * config/i386/i386.md (indirect_jump): Likewise.
7770 (tablejump): Likewise.
7771 (*sibcall_memory): Likewise.
7772 (*sibcall_value_memory): Likewise.
7773 Peepholes of indirect call and jump via memory: Likewise.
7774 * config/i386/i386.opt: Likewise.
7775 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7776 (GOT_memory_operand): Likewise.
7777 (call_insn_operand): Likewise.
7778 (sibcall_insn_operand): Likewise.
7779 (GOT32_symbol_operand): Likewise.
7781 2018-01-15 Jakub Jelinek <jakub@redhat.com>
7784 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
7785 type rather than type addr's type points to.
7786 (expand_omp_atomic_mutex): Likewise.
7787 (expand_omp_atomic): Likewise.
7789 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7792 * config/i386/i386.c (output_indirect_thunk_function): Use
7793 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
7794 for __x86_return_thunk.
7796 2018-01-15 Richard Biener <rguenther@suse.de>
7799 * expmed.c (extract_bit_field_1): Fix typo.
7801 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7804 * config/arm/iterators.md (VF): New mode iterator.
7805 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
7806 Remove integer-related logic from pattern.
7807 (neon_vabd<mode>_3): Likewise.
7809 2018-01-15 Jakub Jelinek <jakub@redhat.com>
7812 * common.opt (fstrict-overflow): No longer an alias.
7813 (fwrapv-pointer): New option.
7814 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
7815 also for pointer types based on flag_wrapv_pointer.
7816 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
7817 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
7818 opts->x_flag_wrapv got set.
7819 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
7820 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
7821 POINTER_TYPE_OVERFLOW_UNDEFINED.
7822 * match.pd: Likewise in address comparison pattern.
7823 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
7825 2018-01-15 Richard Biener <rguenther@suse.de>
7828 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
7829 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
7830 Reset type names to their identifier if their TYPE_DECL doesn't
7831 have linkage (and thus is used for ODR and devirt).
7832 (save_debug_info_for_decl): Remove.
7833 (save_debug_info_for_type): Likewise.
7834 (add_tree_to_fld_list): Adjust.
7835 * tree-pretty-print.c (dump_generic_node): Make dumping of
7836 type names more robust.
7838 2018-01-15 Richard Biener <rguenther@suse.de>
7840 * BASE-VER: Bump to 8.0.1.
7842 2018-01-14 Martin Sebor <msebor@redhat.com>
7845 * builtins.c (check_access): Avoid warning when the no-warning bit
7848 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
7850 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
7851 * ira-color (allocno_hard_regs_compare): Likewise.
7853 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
7856 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
7857 Use .pushsection/.popsection.
7859 2018-01-14 Martin Sebor <msebor@redhat.com>
7862 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
7864 2018-01-14 Jakub Jelinek <jakub@redhat.com>
7866 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
7867 entry from extra_headers.
7868 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
7869 extra_headers, make the list bitwise identical to the i?86-*-* one.
7871 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7873 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
7874 -mcmodel=large with -mindirect-branch=thunk,
7875 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
7876 -mfunction-return=thunk-extern.
7877 * doc/invoke.texi: Document -mcmodel=large is incompatible with
7878 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
7879 -mfunction-return=thunk and -mfunction-return=thunk-extern.
7881 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7883 * config/i386/i386.c (print_reg): Print the name of the full
7884 integer register without '%'.
7885 (ix86_print_operand): Handle 'V'.
7886 * doc/extend.texi: Document 'V' modifier.
7888 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7890 * config/i386/constraints.md (Bs): Disallow memory operand for
7891 -mindirect-branch-register.
7893 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7894 (GOT_memory_operand): Likewise.
7895 (call_insn_operand): Likewise.
7896 (sibcall_insn_operand): Likewise.
7897 (GOT32_symbol_operand): Likewise.
7898 * config/i386/i386.md (indirect_jump): Call convert_memory_address
7899 for -mindirect-branch-register.
7900 (tablejump): Likewise.
7901 (*sibcall_memory): Likewise.
7902 (*sibcall_value_memory): Likewise.
7903 Disallow peepholes of indirect call and jump via memory for
7904 -mindirect-branch-register.
7905 (*call_pop): Replace m with Bw.
7906 (*call_value_pop): Likewise.
7907 (*sibcall_pop_memory): Replace m with Bs.
7908 * config/i386/i386.opt (mindirect-branch-register): New option.
7909 * doc/invoke.texi: Document -mindirect-branch-register option.
7911 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7913 * config/i386/i386-protos.h (ix86_output_function_return): New.
7914 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
7915 set function_return_type.
7916 (indirect_thunk_name): Add ret_p to indicate thunk for function
7918 (output_indirect_thunk_function): Pass false to
7919 indirect_thunk_name.
7920 (ix86_output_indirect_branch_via_reg): Likewise.
7921 (ix86_output_indirect_branch_via_push): Likewise.
7922 (output_indirect_thunk_function): Create alias for function
7923 return thunk if regno < 0.
7924 (ix86_output_function_return): New function.
7925 (ix86_handle_fndecl_attribute): Handle function_return.
7926 (ix86_attribute_table): Add function_return.
7927 * config/i386/i386.h (machine_function): Add
7928 function_return_type.
7929 * config/i386/i386.md (simple_return_internal): Use
7930 ix86_output_function_return.
7931 (simple_return_internal_long): Likewise.
7932 * config/i386/i386.opt (mfunction-return=): New option.
7933 (indirect_branch): Mention -mfunction-return=.
7934 * doc/extend.texi: Document function_return function attribute.
7935 * doc/invoke.texi: Document -mfunction-return= option.
7937 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7939 * config/i386/i386-opts.h (indirect_branch): New.
7940 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
7941 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
7942 with local indirect jump when converting indirect call and jump.
7943 (ix86_set_indirect_branch_type): New.
7944 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
7945 (indirectlabelno): New.
7946 (indirect_thunk_needed): Likewise.
7947 (indirect_thunk_bnd_needed): Likewise.
7948 (indirect_thunks_used): Likewise.
7949 (indirect_thunks_bnd_used): Likewise.
7950 (INDIRECT_LABEL): Likewise.
7951 (indirect_thunk_name): Likewise.
7952 (output_indirect_thunk): Likewise.
7953 (output_indirect_thunk_function): Likewise.
7954 (ix86_output_indirect_branch_via_reg): Likewise.
7955 (ix86_output_indirect_branch_via_push): Likewise.
7956 (ix86_output_indirect_branch): Likewise.
7957 (ix86_output_indirect_jmp): Likewise.
7958 (ix86_code_end): Call output_indirect_thunk_function if needed.
7959 (ix86_output_call_insn): Call ix86_output_indirect_branch if
7961 (ix86_handle_fndecl_attribute): Handle indirect_branch.
7962 (ix86_attribute_table): Add indirect_branch.
7963 * config/i386/i386.h (machine_function): Add indirect_branch_type
7964 and has_local_indirect_jump.
7965 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
7967 (tablejump): Likewise.
7968 (*indirect_jump): Use ix86_output_indirect_jmp.
7969 (*tablejump_1): Likewise.
7970 (simple_return_indirect_internal): Likewise.
7971 * config/i386/i386.opt (mindirect-branch=): New option.
7972 (indirect_branch): New.
7975 (thunk-inline): Likewise.
7976 (thunk-extern): Likewise.
7977 * doc/extend.texi: Document indirect_branch function attribute.
7978 * doc/invoke.texi: Document -mindirect-branch= option.
7980 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
7983 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
7985 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
7987 * ipa-inline.c (want_inline_small_function_p): Return false if
7988 inlining has already failed with CIF_FINAL_ERROR.
7989 (update_caller_keys): Call want_inline_small_function_p before
7991 (update_callee_keys): Likewise.
7993 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
7995 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
7997 (rs6000_quadword_masked_address_p): Likewise.
7998 (quad_aligned_load_p): Likewise.
7999 (quad_aligned_store_p): Likewise.
8000 (const_load_sequence_p): Add comment to describe the outer-most loop.
8001 (mimic_memory_attributes_and_flags): New function.
8002 (rs6000_gen_stvx): Likewise.
8003 (replace_swapped_aligned_store): Likewise.
8004 (rs6000_gen_lvx): Likewise.
8005 (replace_swapped_aligned_load): Likewise.
8006 (replace_swapped_load_constant): Capitalize argument name in
8007 comment describing this function.
8008 (rs6000_analyze_swaps): Add a third pass to search for vector loads
8009 and stores that access quad-word aligned addresses and replace
8010 with stvx or lvx instructions when appropriate.
8011 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
8012 New function prototype.
8013 (rs6000_quadword_masked_address_p): Likewise.
8014 (rs6000_gen_lvx): Likewise.
8015 (rs6000_gen_stvx): Likewise.
8016 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
8017 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
8018 when memory address is aligned.
8019 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
8020 this split to select lvx instruction when memory address is aligned.
8021 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
8022 instruction when memory address is aligned.
8023 (*vsx_le_perm_load_v16qi): Likewise.
8024 (four unnamed splitters): Modify to select the stvx instruction
8025 when memory is aligned.
8027 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
8029 * predict.c (determine_unlikely_bbs): Handle correctly BBs
8030 which appears in the queue multiple times.
8032 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8033 Alan Hayward <alan.hayward@arm.com>
8034 David Sherwood <david.sherwood@arm.com>
8036 * tree-vectorizer.h (vec_lower_bound): New structure.
8037 (_loop_vec_info): Add check_nonzero and lower_bounds.
8038 (LOOP_VINFO_CHECK_NONZERO): New macro.
8039 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
8040 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
8041 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
8042 fields. Make seg_len the distance travelled, not including the
8044 (dr_direction_indicator): Declare.
8045 (dr_zero_step_indicator): Likewise.
8046 (dr_known_forward_stride_p): Likewise.
8047 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
8049 (runtime_alias_check_p): Allow runtime alias checks with
8051 (operator ==): Compare access_size and align.
8052 (prune_runtime_alias_test_list): Rework for new distinction between
8053 the access_size and seg_len.
8054 (create_intersect_range_checks_index): Likewise. Cope with polynomial
8056 (get_segment_min_max): New function.
8057 (create_intersect_range_checks): Use it.
8058 (dr_step_indicator): New function.
8059 (dr_direction_indicator): Likewise.
8060 (dr_zero_step_indicator): Likewise.
8061 (dr_known_forward_stride_p): Likewise.
8062 * tree-loop-distribution.c (data_ref_segment_size): Return
8063 DR_STEP * (niters - 1).
8064 (compute_alias_check_pairs): Update call to the dr_with_seg_len
8066 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
8067 (vect_preserves_scalar_order_p): New function, split out from...
8068 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
8069 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
8070 (vect_vfa_access_size): New function.
8071 (vect_vfa_align): Likewise.
8072 (vect_compile_time_alias): Take access_size_a and access_b arguments.
8073 (dump_lower_bound): New function.
8074 (vect_check_lower_bound): Likewise.
8075 (vect_small_gap_p): Likewise.
8076 (vectorizable_with_step_bound_p): Likewise.
8077 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
8078 depencies if the vectorization factor is 1. Convert the checks
8079 for nonzero steps into checks on the bounds of DR_STEP. Try using
8080 a bunds check for variable steps if the minimum required step is
8081 relatively small. Update calls to the dr_with_seg_len
8082 constructor and to vect_compile_time_alias.
8083 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
8085 (vect_loop_versioning): Call it.
8086 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
8088 (vect_estimate_min_profitable_iters): Account for any bounds checks.
8090 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8091 Alan Hayward <alan.hayward@arm.com>
8092 David Sherwood <david.sherwood@arm.com>
8094 * doc/sourcebuild.texi (vect_scatter_store): Document.
8095 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
8097 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
8099 * genopinit.c (main): Add supports_vec_scatter_store and
8100 supports_vec_scatter_store_cached to target_optabs.
8101 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
8102 IFN_MASK_SCATTER_STORE.
8103 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
8105 * internal-fn.h (internal_store_fn_p): Declare.
8106 (internal_fn_stored_value_index): Likewise.
8107 * internal-fn.c (scatter_store_direct): New macro.
8108 (expand_scatter_store_optab_fn): New function.
8109 (direct_scatter_store_optab_supported_p): New macro.
8110 (internal_store_fn_p): New function.
8111 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
8112 IFN_MASK_SCATTER_STORE.
8113 (internal_fn_mask_index): Likewise.
8114 (internal_fn_stored_value_index): New function.
8115 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
8117 * optabs-query.h (supports_vec_scatter_store_p): Declare.
8118 * optabs-query.c (supports_vec_scatter_store_p): New function.
8119 * tree-vectorizer.h (vect_get_store_rhs): Declare.
8120 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
8121 true for scatter stores.
8122 (vect_gather_scatter_fn_p): Handle scatter stores too.
8123 (vect_check_gather_scatter): Consider using scatter stores if
8124 supports_vec_scatter_store_p.
8125 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
8127 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
8128 internal_fn_stored_value_index.
8129 (check_load_store_masking): Handle scatter stores too.
8130 (vect_get_store_rhs): Make public.
8131 (vectorizable_call): Use internal_store_fn_p.
8132 (vectorizable_store): Handle scatter store internal functions.
8133 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
8134 when deciding whether the end of the group has been reached.
8135 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
8136 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
8137 (mask_scatter_store<mode>): New insns.
8139 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8140 Alan Hayward <alan.hayward@arm.com>
8141 David Sherwood <david.sherwood@arm.com>
8143 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
8144 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
8145 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
8147 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
8148 Use vect_truncate_gather_scatter_offset if we can't treat the
8149 operation as a normal gather load or scatter store.
8150 (get_group_load_store_type): Take the gather_scatter_info
8151 as argument. Try using a gather load or scatter store for
8152 single-element groups.
8153 (get_load_store_type): Update calls to get_group_load_store_type
8154 and vect_use_strided_gather_scatters_p.
8156 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8157 Alan Hayward <alan.hayward@arm.com>
8158 David Sherwood <david.sherwood@arm.com>
8160 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
8161 optional tree argument.
8162 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
8164 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
8165 but continue to use the current value as a fallback.
8166 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
8167 to compare the updates.
8168 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
8169 (get_load_store_type): Use it when handling a strided access.
8170 (vect_get_strided_load_store_ops): New function.
8171 (vect_get_data_ptr_increment): Likewise.
8172 (vectorizable_load): Handle strided gather loads. Always pass
8173 a step to vect_create_data_ref_ptr and bump_vector_ptr.
8175 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8176 Alan Hayward <alan.hayward@arm.com>
8177 David Sherwood <david.sherwood@arm.com>
8179 * doc/md.texi (gather_load@var{m}): Document.
8180 (mask_gather_load@var{m}): Likewise.
8181 * genopinit.c (main): Add supports_vec_gather_load and
8182 supports_vec_gather_load_cached to target_optabs.
8183 * optabs-tree.c (init_tree_optimization_optabs): Use
8184 ggc_cleared_alloc to allocate target_optabs.
8185 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
8186 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
8188 * internal-fn.h (internal_load_fn_p): Declare.
8189 (internal_gather_scatter_fn_p): Likewise.
8190 (internal_fn_mask_index): Likewise.
8191 (internal_gather_scatter_fn_supported_p): Likewise.
8192 * internal-fn.c (gather_load_direct): New macro.
8193 (expand_gather_load_optab_fn): New function.
8194 (direct_gather_load_optab_supported_p): New macro.
8195 (direct_internal_fn_optab): New function.
8196 (internal_load_fn_p): Likewise.
8197 (internal_gather_scatter_fn_p): Likewise.
8198 (internal_fn_mask_index): Likewise.
8199 (internal_gather_scatter_fn_supported_p): Likewise.
8200 * optabs-query.c (supports_at_least_one_mode_p): New function.
8201 (supports_vec_gather_load_p): Likewise.
8202 * optabs-query.h (supports_vec_gather_load_p): Declare.
8203 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
8204 and memory_type field.
8205 (NUM_PATTERNS): Bump to 15.
8206 * tree-vect-data-refs.c: Include internal-fn.h.
8207 (vect_gather_scatter_fn_p): New function.
8208 (vect_describe_gather_scatter_call): Likewise.
8209 (vect_check_gather_scatter): Try using internal functions for
8210 gather loads. Recognize existing calls to a gather load function.
8211 (vect_analyze_data_refs): Consider using gather loads if
8212 supports_vec_gather_load_p.
8213 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
8214 (vect_get_gather_scatter_offset_type): Likewise.
8215 (vect_convert_mask_for_vectype): Likewise.
8216 (vect_add_conversion_to_patterm): Likewise.
8217 (vect_try_gather_scatter_pattern): Likewise.
8218 (vect_recog_gather_scatter_pattern): New pattern recognizer.
8219 (vect_vect_recog_func_ptrs): Add it.
8220 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
8221 internal_fn_mask_index and internal_gather_scatter_fn_p.
8222 (check_load_store_masking): Take the gather_scatter_info as an
8223 argument and handle gather loads.
8224 (vect_get_gather_scatter_ops): New function.
8225 (vectorizable_call): Check internal_load_fn_p.
8226 (vectorizable_load): Likewise. Handle gather load internal
8228 (vectorizable_store): Update call to check_load_store_masking.
8229 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
8230 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
8231 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
8232 (aarch64_gather_scale_operand_d): New predicates.
8233 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
8234 (mask_gather_load<mode>): New insns.
8236 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8237 Alan Hayward <alan.hayward@arm.com>
8238 David Sherwood <david.sherwood@arm.com>
8240 * optabs.def (fold_left_plus_optab): New optab.
8241 * doc/md.texi (fold_left_plus_@var{m}): Document.
8242 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
8243 * internal-fn.c (fold_left_direct): Define.
8244 (expand_fold_left_optab_fn): Likewise.
8245 (direct_fold_left_optab_supported_p): Likewise.
8246 * fold-const-call.c (fold_const_fold_left): New function.
8247 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
8248 * tree-parloops.c (valid_reduction_p): New function.
8249 (gather_scalar_reductions): Use it.
8250 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
8251 (vect_finish_replace_stmt): Declare.
8252 * tree-vect-loop.c (fold_left_reduction_fn): New function.
8253 (needs_fold_left_reduction_p): New function, split out from...
8254 (vect_is_simple_reduction): ...here. Accept reductions that
8255 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
8256 (vect_force_simple_reduction): Also store the reduction type in
8257 the assignment's STMT_VINFO_REDUC_TYPE.
8258 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
8259 (merge_with_identity): New function.
8260 (vect_expand_fold_left): Likewise.
8261 (vectorize_fold_left_reduction): Likewise.
8262 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
8263 scalar phi in place for it. Check for target support and reject
8264 cases that would reassociate the operation. Defer the transform
8265 phase to vectorize_fold_left_reduction.
8266 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
8267 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
8268 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
8270 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8272 * tree-if-conv.c (predicate_mem_writes): Remove redundant
8273 call to ifc_temp_var.
8275 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8276 Alan Hayward <alan.hayward@arm.com>
8277 David Sherwood <david.sherwood@arm.com>
8279 * target.def (legitimize_address_displacement): Take the original
8280 offset as a poly_int.
8281 * targhooks.h (default_legitimize_address_displacement): Update
8283 * targhooks.c (default_legitimize_address_displacement): Likewise.
8284 * doc/tm.texi: Regenerate.
8285 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
8286 as an argument, moving assert of ad->disp == ad->disp_term to...
8287 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
8288 Try calling targetm.legitimize_address_displacement before expanding
8289 the address rather than afterwards, and adjust for the new interface.
8290 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
8291 Match the new hook interface. Handle SVE addresses.
8292 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
8295 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8297 * Makefile.in (OBJS): Add early-remat.o.
8298 * target.def (select_early_remat_modes): New hook.
8299 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
8300 * doc/tm.texi: Regenerate.
8301 * targhooks.h (default_select_early_remat_modes): Declare.
8302 * targhooks.c (default_select_early_remat_modes): New function.
8303 * timevar.def (TV_EARLY_REMAT): New timevar.
8304 * passes.def (pass_early_remat): New pass.
8305 * tree-pass.h (make_pass_early_remat): Declare.
8306 * early-remat.c: New file.
8307 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
8309 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
8311 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8312 Alan Hayward <alan.hayward@arm.com>
8313 David Sherwood <david.sherwood@arm.com>
8315 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
8316 vfm1 with a bound_epilog parameter.
8317 (vect_do_peeling): Update calls accordingly, and move the prologue
8318 call earlier in the function. Treat the base bound_epilog as 0 for
8319 fully-masked loops and retain vf - 1 for other loops. Add 1 to
8320 this base when peeling for gaps.
8321 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
8322 with fully-masked loops.
8323 (vect_estimate_min_profitable_iters): Handle the single peeled
8324 iteration in that case.
8326 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8327 Alan Hayward <alan.hayward@arm.com>
8328 David Sherwood <david.sherwood@arm.com>
8330 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
8331 single-element interleaving even if the size is not a power of 2.
8332 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
8333 accesses for single-element interleaving if the group size is
8336 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8337 Alan Hayward <alan.hayward@arm.com>
8338 David Sherwood <david.sherwood@arm.com>
8340 * doc/md.texi (fold_extract_last_@var{m}): Document.
8341 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
8342 * optabs.def (fold_extract_last_optab): New optab.
8343 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
8344 * internal-fn.c (fold_extract_direct): New macro.
8345 (expand_fold_extract_optab_fn): Likewise.
8346 (direct_fold_extract_optab_supported_p): Likewise.
8347 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
8348 * tree-vect-loop.c (vect_model_reduction_cost): Handle
8349 EXTRACT_LAST_REDUCTION.
8350 (get_initial_def_for_reduction): Do not create an initial vector
8351 for EXTRACT_LAST_REDUCTION reductions.
8352 (vectorizable_reduction): Leave the scalar phi in place for
8353 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
8354 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
8355 epilogue code for EXTRACT_LAST_REDUCTION and defer the
8356 transform phase to vectorizable_condition.
8357 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
8359 (vect_finish_stmt_generation): ...here.
8360 (vect_finish_replace_stmt): New function.
8361 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
8362 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
8364 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
8366 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8367 Alan Hayward <alan.hayward@arm.com>
8368 David Sherwood <david.sherwood@arm.com>
8370 * doc/md.texi (extract_last_@var{m}): Document.
8371 * optabs.def (extract_last_optab): New optab.
8372 * internal-fn.def (EXTRACT_LAST): New internal function.
8373 * internal-fn.c (cond_unary_direct): New macro.
8374 (expand_cond_unary_optab_fn): Likewise.
8375 (direct_cond_unary_optab_supported_p): Likewise.
8376 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
8377 loops using EXTRACT_LAST.
8378 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
8379 (extract_last_<mode>): ...this optab.
8380 (vec_extract<mode><Vel>): Update accordingly.
8382 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8383 Alan Hayward <alan.hayward@arm.com>
8384 David Sherwood <david.sherwood@arm.com>
8386 * target.def (empty_mask_is_expensive): New hook.
8387 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
8388 * doc/tm.texi: Regenerate.
8389 * targhooks.h (default_empty_mask_is_expensive): Declare.
8390 * targhooks.c (default_empty_mask_is_expensive): New function.
8391 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
8392 if the target says that empty masks are expensive.
8393 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
8395 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
8397 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8398 Alan Hayward <alan.hayward@arm.com>
8399 David Sherwood <david.sherwood@arm.com>
8401 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
8402 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
8403 (vect_use_loop_mask_for_alignment_p): New function.
8404 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
8405 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
8406 niters_skip argument. Make sure that the first niters_skip elements
8407 of the first iteration are inactive.
8408 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
8409 Update call to vect_set_loop_masks_directly.
8410 (get_misalign_in_elems): New function, split out from...
8411 (vect_gen_prolog_loop_niters): ...here.
8412 (vect_update_init_of_dr): Take a code argument that specifies whether
8413 the adjustment should be added or subtracted.
8414 (vect_update_init_of_drs): Likewise.
8415 (vect_prepare_for_masked_peels): New function.
8416 (vect_do_peeling): Skip prologue peeling if we're using a mask
8417 instead. Update call to vect_update_inits_of_drs.
8418 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
8420 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
8421 alignment. Do not include the number of peeled iterations in
8422 the minimum threshold in that case.
8423 (vectorizable_induction): Adjust the start value down by
8424 LOOP_VINFO_MASK_SKIP_NITERS iterations.
8425 (vect_transform_loop): Call vect_prepare_for_masked_peels.
8426 Take the number of skipped iterations into account when calculating
8428 * tree-vect-stmts.c (vect_gen_while_not): New function.
8430 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8431 Alan Hayward <alan.hayward@arm.com>
8432 David Sherwood <david.sherwood@arm.com>
8434 * doc/sourcebuild.texi (vect_fully_masked): Document.
8435 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
8437 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
8439 (vect_analyze_loop_2): ...here. Don't check the vectorization
8440 factor against the number of loop iterations if the loop is
8443 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8444 Alan Hayward <alan.hayward@arm.com>
8445 David Sherwood <david.sherwood@arm.com>
8447 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
8448 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
8449 (dump_groups): Update accordingly.
8450 (iv_use::mem_type): New member variable.
8451 (address_p): New function.
8452 (record_use): Add a mem_type argument and initialize the new
8454 (record_group_use): Add a mem_type argument. Use address_p.
8455 Remove obsolete null checks of base_object. Update call to record_use.
8456 (find_interesting_uses_op): Update call to record_group_use.
8457 (find_interesting_uses_cond): Likewise.
8458 (find_interesting_uses_address): Likewise.
8459 (get_mem_type_for_internal_fn): New function.
8460 (find_address_like_use): Likewise.
8461 (find_interesting_uses_stmt): Try find_address_like_use before
8462 calling find_interesting_uses_op.
8463 (addr_offset_valid_p): Use the iv mem_type field as the type
8464 of the addressed memory.
8465 (add_autoinc_candidates): Likewise.
8466 (get_address_cost): Likewise.
8467 (split_small_address_groups_p): Use address_p.
8468 (split_address_groups): Likewise.
8469 (add_iv_candidate_for_use): Likewise.
8470 (autoinc_possible_for_pair): Likewise.
8471 (rewrite_groups): Likewise.
8472 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
8473 (determine_group_iv_cost): Update after split of USE_ADDRESS.
8474 (get_alias_ptr_type_for_ptr_address): New function.
8475 (rewrite_use_address): Rewrite address uses in calls that were
8476 identified by find_address_like_use.
8478 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8479 Alan Hayward <alan.hayward@arm.com>
8480 David Sherwood <david.sherwood@arm.com>
8482 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
8484 * gimple-expr.h (is_gimple_addressable: Likewise.
8485 * gimple-expr.c (is_gimple_address): Likewise.
8486 * internal-fn.c (expand_call_mem_ref): New function.
8487 (expand_mask_load_optab_fn): Use it.
8488 (expand_mask_store_optab_fn): Likewise.
8490 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8491 Alan Hayward <alan.hayward@arm.com>
8492 David Sherwood <david.sherwood@arm.com>
8494 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
8495 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
8496 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
8497 (cond_umax@var{mode}): Document.
8498 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
8499 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
8500 (cond_umin_optab, cond_umax_optab): New optabs.
8501 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
8502 (COND_IOR, COND_XOR): New internal functions.
8503 * internal-fn.h (get_conditional_internal_fn): Declare.
8504 * internal-fn.c (cond_binary_direct): New macro.
8505 (expand_cond_binary_optab_fn): Likewise.
8506 (direct_cond_binary_optab_supported_p): Likewise.
8507 (get_conditional_internal_fn): New function.
8508 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
8509 Cope with reduction statements that are vectorized as calls rather
8511 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
8512 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
8513 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
8514 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
8515 (UNSPEC_COND_EOR): New unspecs.
8516 (optab): Add mappings for them.
8517 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
8518 (sve_int_op, sve_fp_op): New int attributes.
8520 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8521 Alan Hayward <alan.hayward@arm.com>
8522 David Sherwood <david.sherwood@arm.com>
8524 * optabs.def (while_ult_optab): New optab.
8525 * doc/md.texi (while_ult@var{m}@var{n}): Document.
8526 * internal-fn.def (WHILE_ULT): New internal function.
8527 * internal-fn.h (direct_internal_fn_supported_p): New override
8528 that takes two types as argument.
8529 * internal-fn.c (while_direct): New macro.
8530 (expand_while_optab_fn): New function.
8531 (convert_optab_supported_p): Likewise.
8532 (direct_while_optab_supported_p): New macro.
8533 * wide-int.h (wi::udiv_ceil): New function.
8534 * tree-vectorizer.h (rgroup_masks): New structure.
8535 (vec_loop_masks): New typedef.
8536 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
8538 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
8539 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
8540 (vect_max_vf): New function.
8541 (slpeel_make_loop_iterate_ntimes): Delete.
8542 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
8543 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
8544 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
8545 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
8546 internal-fn.h, stor-layout.h and optabs-query.h.
8547 (vect_set_loop_mask): New function.
8548 (add_preheader_seq): Likewise.
8549 (add_header_seq): Likewise.
8550 (interleave_supported_p): Likewise.
8551 (vect_maybe_permute_loop_masks): Likewise.
8552 (vect_set_loop_masks_directly): Likewise.
8553 (vect_set_loop_condition_masked): Likewise.
8554 (vect_set_loop_condition_unmasked): New function, split out from
8555 slpeel_make_loop_iterate_ntimes.
8556 (slpeel_make_loop_iterate_ntimes): Rename to..
8557 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
8558 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
8559 (vect_do_peeling): Update call accordingly.
8560 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
8562 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
8563 mask_compare_type, can_fully_mask_p and fully_masked_p.
8564 (release_vec_loop_masks): New function.
8565 (_loop_vec_info): Use it to free the loop masks.
8566 (can_produce_all_loop_masks_p): New function.
8567 (vect_get_max_nscalars_per_iter): Likewise.
8568 (vect_verify_full_masking): Likewise.
8569 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
8570 retries, and free the mask rgroups before retrying. Check loop-wide
8571 reasons for disallowing fully-masked loops. Make the final decision
8572 about whether use a fully-masked loop or not.
8573 (vect_estimate_min_profitable_iters): Do not assume that peeling
8574 for the number of iterations will be needed for fully-masked loops.
8575 (vectorizable_reduction): Disable fully-masked loops.
8576 (vectorizable_live_operation): Likewise.
8577 (vect_halve_mask_nunits): New function.
8578 (vect_double_mask_nunits): Likewise.
8579 (vect_record_loop_mask): Likewise.
8580 (vect_get_loop_mask): Likewise.
8581 (vect_transform_loop): Handle the case in which the final loop
8582 iteration might handle a partial vector. Call vect_set_loop_condition
8583 instead of slpeel_make_loop_iterate_ntimes.
8584 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
8585 (check_load_store_masking): New function.
8586 (prepare_load_store_mask): Likewise.
8587 (vectorizable_store): Handle fully-masked loops.
8588 (vectorizable_load): Likewise.
8589 (supportable_widening_operation): Use vect_halve_mask_nunits for
8591 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
8592 (vect_gen_while): New function.
8593 * config/aarch64/aarch64.md (umax<mode>3): New expander.
8594 (aarch64_uqdec<mode>): New insn.
8596 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8597 Alan Hayward <alan.hayward@arm.com>
8598 David Sherwood <david.sherwood@arm.com>
8600 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
8601 (reduc_xor_scal_optab): New optabs.
8602 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
8603 (reduc_xor_scal_@var{m}): Document.
8604 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
8605 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
8607 * fold-const-call.c (fold_const_call): Handle them.
8608 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
8609 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
8610 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
8611 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
8612 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
8613 (UNSPEC_XORV): New unspecs.
8614 (optab): Add entries for them.
8615 (BITWISEV): New int iterator.
8616 (bit_reduc_op): New int attributes.
8618 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8619 Alan Hayward <alan.hayward@arm.com>
8620 David Sherwood <david.sherwood@arm.com>
8622 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
8623 * internal-fn.def (VEC_SHL_INSERT): New internal function.
8624 * optabs.def (vec_shl_insert_optab): New optab.
8625 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
8626 (duplicate_and_interleave): Likewise.
8627 * tree-vect-loop.c: Include internal-fn.h.
8628 (neutral_op_for_slp_reduction): New function, split out from
8629 get_initial_defs_for_reduction.
8630 (get_initial_def_for_reduction): Handle option 2 for variable-length
8631 vectors by loading the neutral value into a vector and then shifting
8632 the initial value into element 0.
8633 (get_initial_defs_for_reduction): Replace the code argument with
8634 the neutral value calculated by neutral_op_for_slp_reduction.
8635 Use gimple_build_vector for constant-length vectors.
8636 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
8637 but the first group_size elements have a neutral value.
8638 Use duplicate_and_interleave otherwise.
8639 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
8640 Update call to get_initial_defs_for_reduction. Handle SLP
8641 reductions for variable-length vectors by creating one vector
8642 result for each scalar result, with the elements associated
8643 with other scalar results stubbed out with the neutral value.
8644 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
8645 Require IFN_VEC_SHL_INSERT for double reductions on
8646 variable-length vectors, or SLP reductions that have
8647 a neutral value. Require can_duplicate_and_interleave_p
8648 support for variable-length unchained SLP reductions if there
8649 is no neutral value, such as for MIN/MAX reductions. Also require
8650 the number of vector elements to be a multiple of the number of
8651 SLP statements when doing variable-length unchained SLP reductions.
8652 Update call to vect_create_epilog_for_reduction.
8653 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
8654 and remove initial values.
8655 (duplicate_and_interleave): Make public.
8656 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
8657 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
8659 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8660 Alan Hayward <alan.hayward@arm.com>
8661 David Sherwood <david.sherwood@arm.com>
8663 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
8664 (can_duplicate_and_interleave_p): New function.
8665 (vect_get_and_check_slp_defs): Take the vector of statements
8666 rather than just the current one. Remove excess parentheses.
8667 Restriction rejectinon of vect_constant_def and vect_external_def
8668 for variable-length vectors to boolean types, or types for which
8669 can_duplicate_and_interleave_p is false.
8670 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
8671 (duplicate_and_interleave): New function.
8672 (vect_get_constant_vectors): Use gimple_build_vector for
8673 constant-length vectors and suitable variable-length constant
8674 vectors. Use duplicate_and_interleave for other variable-length
8675 vectors. Don't defer the update when inserting new statements.
8677 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8678 Alan Hayward <alan.hayward@arm.com>
8679 David Sherwood <david.sherwood@arm.com>
8681 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
8682 min_profitable_iters doesn't go negative.
8684 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8685 Alan Hayward <alan.hayward@arm.com>
8686 David Sherwood <david.sherwood@arm.com>
8688 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
8689 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
8690 * optabs.def (vec_mask_load_lanes_optab): New optab.
8691 (vec_mask_store_lanes_optab): Likewise.
8692 * internal-fn.def (MASK_LOAD_LANES): New internal function.
8693 (MASK_STORE_LANES): Likewise.
8694 * internal-fn.c (mask_load_lanes_direct): New macro.
8695 (mask_store_lanes_direct): Likewise.
8696 (expand_mask_load_optab_fn): Handle masked operations.
8697 (expand_mask_load_lanes_optab_fn): New macro.
8698 (expand_mask_store_optab_fn): Handle masked operations.
8699 (expand_mask_store_lanes_optab_fn): New macro.
8700 (direct_mask_load_lanes_optab_supported_p): Likewise.
8701 (direct_mask_store_lanes_optab_supported_p): Likewise.
8702 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
8704 (vect_load_lanes_supported): Likewise.
8705 * tree-vect-data-refs.c (strip_conversion): New function.
8706 (can_group_stmts_p): Likewise.
8707 (vect_analyze_data_ref_accesses): Use it instead of checking
8708 for a pair of assignments.
8709 (vect_store_lanes_supported): Take a masked_p parameter.
8710 (vect_load_lanes_supported): Likewise.
8711 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
8712 vect_store_lanes_supported and vect_load_lanes_supported.
8713 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
8714 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
8715 parameter. Don't allow gaps for masked accesses.
8716 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
8717 and vect_load_lanes_supported.
8718 (get_load_store_type): Take a masked_p parameter and update
8719 call to get_group_load_store_type.
8720 (vectorizable_store): Update call to get_load_store_type.
8721 Handle IFN_MASK_STORE_LANES.
8722 (vectorizable_load): Update call to get_load_store_type.
8723 Handle IFN_MASK_LOAD_LANES.
8725 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8726 Alan Hayward <alan.hayward@arm.com>
8727 David Sherwood <david.sherwood@arm.com>
8729 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
8731 * config/aarch64/aarch64-protos.h
8732 (aarch64_sve_struct_memory_operand_p): Declare.
8733 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
8734 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
8735 (VPRED, vpred): Handle SVE structure modes.
8736 * config/aarch64/constraints.md (Utx): New constraint.
8737 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
8738 (aarch64_sve_struct_nonimmediate_operand): New predicates.
8739 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
8740 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
8741 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
8742 structure modes. Split into pieces after RA.
8743 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
8744 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
8746 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
8747 SVE structure modes.
8748 (aarch64_classify_address): Likewise.
8749 (sizetochar): Move earlier in file.
8750 (aarch64_print_operand): Handle SVE register lists.
8751 (aarch64_array_mode): New function.
8752 (aarch64_sve_struct_memory_operand_p): Likewise.
8753 (TARGET_ARRAY_MODE): Redefine.
8755 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8756 Alan Hayward <alan.hayward@arm.com>
8757 David Sherwood <david.sherwood@arm.com>
8759 * target.def (array_mode): New target hook.
8760 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
8761 * doc/tm.texi: Regenerate.
8762 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
8763 * hooks.c (hook_optmode_mode_uhwi_none): New function.
8764 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
8766 * stor-layout.c (mode_for_array): Likewise. Support polynomial
8769 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8770 Alan Hayward <alan.hayward@arm.com>
8771 David Sherwood <david.sherwood@arm.com>
8773 * fold-const.c (fold_binary_loc): Check the argument types
8774 rather than the result type when testing for a vector operation.
8776 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8778 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
8779 * doc/tm.texi: Regenerate.
8781 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8782 Alan Hayward <alan.hayward@arm.com>
8783 David Sherwood <david.sherwood@arm.com>
8785 * doc/invoke.texi (-msve-vector-bits=): Document new option.
8786 (sve): Document new AArch64 extension.
8787 * doc/md.texi (w): Extend the description of the AArch64
8788 constraint to include SVE vectors.
8789 (Upl, Upa): Document new AArch64 predicate constraints.
8790 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
8792 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
8793 (msve-vector-bits=): New option.
8794 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
8795 SVE when these are disabled.
8796 (sve): New extension.
8797 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
8798 modes. Adjust their number of units based on aarch64_sve_vg.
8799 (MAX_BITSIZE_MODE_ANY_MODE): Define.
8800 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
8801 aarch64_addr_query_type.
8802 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
8803 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
8804 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
8805 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
8806 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
8807 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
8808 (aarch64_simd_imm_zero_p): Delete.
8809 (aarch64_check_zero_based_sve_index_immediate): Declare.
8810 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
8811 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
8812 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
8813 (aarch64_sve_float_mul_immediate_p): Likewise.
8814 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
8816 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
8817 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
8818 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
8819 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
8820 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
8821 (aarch64_regmode_natural_size): Likewise.
8822 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
8823 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
8825 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
8826 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
8827 for VG and the SVE predicate registers.
8828 (V_ALIASES): Add a "z"-prefixed alias.
8829 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
8830 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
8831 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
8832 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
8833 (REG_CLASS_NAMES): Add entries for them.
8834 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
8835 and the predicate registers.
8836 (aarch64_sve_vg): Declare.
8837 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
8838 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
8839 (REGMODE_NATURAL_SIZE): Define.
8840 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
8842 * config/aarch64/aarch64.c: Include cfgrtl.h.
8843 (simd_immediate_info): Add a constructor for series vectors,
8844 and an associated step field.
8845 (aarch64_sve_vg): New variable.
8846 (aarch64_dbx_register_number): Handle VG and the predicate registers.
8847 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
8848 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
8849 (VEC_ANY_DATA, VEC_STRUCT): New constants.
8850 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
8851 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
8852 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
8853 (aarch64_get_mask_mode): New functions.
8854 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
8855 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
8856 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
8857 predicate modes and predicate registers. Explicitly restrict
8858 GPRs to modes of 16 bytes or smaller. Only allow FP registers
8859 to store a vector mode if it is recognized by
8860 aarch64_classify_vector_mode.
8861 (aarch64_regmode_natural_size): New function.
8862 (aarch64_hard_regno_caller_save_mode): Return the original mode
8864 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
8865 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
8866 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
8867 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
8869 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
8870 does not overlap dest if the function is frame-related. Handle
8872 (aarch64_split_add_offset): New function.
8873 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
8874 them aarch64_add_offset.
8875 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
8876 and update call to aarch64_sub_sp.
8877 (aarch64_add_cfa_expression): New function.
8878 (aarch64_expand_prologue): Pass extra temporary registers to the
8879 functions above. Handle the case in which we need to emit new
8880 DW_CFA_expressions for registers that were originally saved
8881 relative to the stack pointer, but now have to be expressed
8882 relative to the frame pointer.
8883 (aarch64_output_mi_thunk): Pass extra temporary registers to the
8885 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
8886 IP0 and IP1 values for SVE frames.
8887 (aarch64_expand_vec_series): New function.
8888 (aarch64_expand_sve_widened_duplicate): Likewise.
8889 (aarch64_expand_sve_const_vector): Likewise.
8890 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
8891 Handle SVE constants. Use emit_move_insn to move a force_const_mem
8892 into the register, rather than emitting a SET directly.
8893 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
8894 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
8895 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
8896 (offset_9bit_signed_scaled_p): New functions.
8897 (aarch64_replicate_bitmask_imm): New function.
8898 (aarch64_bitmask_imm): Use it.
8899 (aarch64_cannot_force_const_mem): Reject expressions involving
8900 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
8901 (aarch64_classify_index): Handle SVE indices, by requiring
8902 a plain register index with a scale that matches the element size.
8903 (aarch64_classify_address): Handle SVE addresses. Assert that
8904 the mode of the address is VOIDmode or an integer mode.
8905 Update call to aarch64_classify_symbol.
8906 (aarch64_classify_symbolic_expression): Update call to
8907 aarch64_classify_symbol.
8908 (aarch64_const_vec_all_in_range_p): New function.
8909 (aarch64_print_vector_float_operand): Likewise.
8910 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
8911 "vN" for FP registers with SVE modes. Handle (const ...) vectors
8912 and the FP immediates 1.0 and 0.5.
8913 (aarch64_print_address_internal): Handle SVE addresses.
8914 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
8915 (aarch64_regno_regclass): Handle predicate registers.
8916 (aarch64_secondary_reload): Handle big-endian reloads of SVE
8918 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
8919 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
8920 (aarch64_convert_sve_vector_bits): New function.
8921 (aarch64_override_options): Use it to handle -msve-vector-bits=.
8922 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
8924 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
8925 Handle SVE vector and predicate modes. Accept VL-based constants
8926 that need only one temporary register, and VL offsets that require
8927 no temporary registers.
8928 (aarch64_conditional_register_usage): Mark the predicate registers
8929 as fixed if SVE isn't available.
8930 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
8931 Return true for SVE vector and predicate modes.
8932 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
8933 rather than an unsigned int. Handle SVE modes.
8934 (aarch64_preferred_simd_mode): Update call accordingly. Handle
8936 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
8938 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
8939 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
8940 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
8941 (aarch64_sve_float_mul_immediate_p): New functions.
8942 (aarch64_sve_valid_immediate): New function.
8943 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
8944 Explicitly reject structure modes. Check for INDEX constants.
8945 Handle PTRUE and PFALSE constants.
8946 (aarch64_check_zero_based_sve_index_immediate): New function.
8947 (aarch64_simd_imm_zero_p): Delete.
8948 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
8949 vector modes. Accept constants in the range of CNT[BHWD].
8950 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
8951 ask for an Advanced SIMD mode.
8952 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
8953 (aarch64_simd_vector_alignment): Handle SVE predicates.
8954 (aarch64_vectorize_preferred_vector_alignment): New function.
8955 (aarch64_simd_vector_alignment_reachable): Use it instead of
8957 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
8958 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
8960 (MAX_VECT_LEN): Delete.
8961 (expand_vec_perm_d): Add a vec_flags field.
8962 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
8963 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
8964 (aarch64_evpc_ext): Don't apply a big-endian lane correction
8966 (aarch64_evpc_rev): Rename to...
8967 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
8968 (aarch64_evpc_rev_global): New function.
8969 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
8970 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
8972 (aarch64_evpc_sve_tbl): New function.
8973 (aarch64_expand_vec_perm_const_1): Update after rename of
8974 aarch64_evpc_rev. Handle SVE permutes too, trying
8975 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
8976 than aarch64_evpc_tbl.
8977 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
8978 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
8979 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
8980 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
8981 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
8982 (aarch64_expand_sve_vcond): New functions.
8983 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
8984 of aarch64_vector_mode_p.
8985 (aarch64_dwarf_poly_indeterminate_value): New function.
8986 (aarch64_compute_pressure_classes): Likewise.
8987 (aarch64_can_change_mode_class): Likewise.
8988 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
8989 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
8990 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
8991 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
8992 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
8993 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
8994 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
8995 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
8997 (Dn, Dl, Dr): Accept const as well as const_vector.
8998 (Dz): Likewise. Compare against CONST0_RTX.
8999 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
9000 of "vector" where appropriate.
9001 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
9002 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
9003 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
9004 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
9005 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
9006 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
9007 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
9008 (v_int_equiv): Extend to SVE modes.
9009 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
9011 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
9012 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
9013 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
9014 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
9015 (SVE_COND_FP_CMP): New int iterators.
9016 (perm_hilo): Handle the new unpack unspecs.
9017 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
9019 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
9020 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
9021 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
9022 (aarch64_equality_operator, aarch64_constant_vector_operand)
9023 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
9024 (aarch64_sve_nonimmediate_operand): Likewise.
9025 (aarch64_sve_general_operand): Likewise.
9026 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
9027 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
9028 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
9029 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
9030 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
9031 (aarch64_sve_float_arith_immediate): Likewise.
9032 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
9033 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
9034 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
9035 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
9036 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
9037 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
9038 (aarch64_sve_float_arith_operand): Likewise.
9039 (aarch64_sve_float_arith_with_sub_operand): Likewise.
9040 (aarch64_sve_float_mul_operand): Likewise.
9041 (aarch64_sve_vec_perm_operand): Likewise.
9042 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
9043 (aarch64_mov_operand): Accept const_poly_int and const_vector.
9044 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
9045 as well as const_vector.
9046 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
9047 in file. Use CONST0_RTX and CONSTM1_RTX.
9048 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
9049 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
9050 Use aarch64_simd_imm_zero.
9051 * config/aarch64/aarch64-sve.md: New file.
9052 * config/aarch64/aarch64.md: Include it.
9053 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
9054 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
9055 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
9056 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
9057 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
9058 (sve): New attribute.
9059 (enabled): Disable instructions with the sve attribute unless
9061 (movqi, movhi): Pass CONST_POLY_INT operaneds through
9062 aarch64_expand_mov_immediate.
9063 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
9064 CNT[BHSD] immediates.
9065 (movti): Split CONST_POLY_INT moves into two halves.
9066 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
9067 Split additions that need a temporary here if the destination
9068 is the stack pointer.
9069 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
9070 (*add<mode>3_poly_1): New instruction.
9071 (set_clobber_cc): New expander.
9073 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9075 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
9076 parameter and use it instead of GET_MODE_SIZE (innermode). Use
9077 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
9078 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
9079 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
9080 Change innermode from fixed_mode_size to machine_mode.
9081 (simplify_subreg): Update call accordingly. Handle a constant-sized
9082 subreg of a variable-length CONST_VECTOR.
9084 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9085 Alan Hayward <alan.hayward@arm.com>
9086 David Sherwood <david.sherwood@arm.com>
9088 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
9089 (add_offset_to_base): New function, split out from...
9090 (create_mem_ref): ...here. When handling a scale other than 1,
9091 check first whether the address is valid without the offset.
9092 Add it into the base if so, leaving the index and scale as-is.
9094 2018-01-12 Jakub Jelinek <jakub@redhat.com>
9097 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
9098 fold_for_warn before checking if arg2 is INTEGER_CST.
9100 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
9102 * config/rs6000/predicates.md (load_multiple_operation): Delete.
9103 (store_multiple_operation): Delete.
9104 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
9105 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
9106 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
9107 guarded by TARGET_STRING.
9108 (rs6000_output_load_multiple): Delete.
9109 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
9110 OPTION_MASK_STRING / TARGET_STRING handling.
9111 (print_operand) <'N', 'O'>: Add comment that these are unused now.
9112 (const rs6000_opt_masks) <"string">: Change mask to 0.
9113 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
9114 (MASK_STRING): Delete.
9115 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
9117 (load_multiple): Delete.
9124 (store_multiple): Delete.
9131 (movmemsi_8reg): Delete.
9132 (corresponding unnamed define_insn): Delete.
9133 (movmemsi_6reg): Delete.
9134 (corresponding unnamed define_insn): Delete.
9135 (movmemsi_4reg): Delete.
9136 (corresponding unnamed define_insn): Delete.
9137 (movmemsi_2reg): Delete.
9138 (corresponding unnamed define_insn): Delete.
9139 (movmemsi_1reg): Delete.
9140 (corresponding unnamed define_insn): Delete.
9141 * config/rs6000/rs6000.opt (mno-string): New.
9142 (mstring): Replace by deprecation warning stub.
9143 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
9145 2018-01-12 Jakub Jelinek <jakub@redhat.com>
9147 * regrename.c (regrename_do_replace): If replacing the same
9148 reg multiple times, try to reuse last created gen_raw_REG.
9151 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
9152 main to workaround a bug in GDB.
9154 2018-01-12 Tom de Vries <tom@codesourcery.com>
9157 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
9159 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
9161 PR rtl-optimization/80481
9162 * ira-color.c (get_cap_member): New function.
9163 (allocnos_conflict_by_live_ranges_p): Use it.
9164 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
9165 (setup_slot_coalesced_allocno_live_ranges): Ditto.
9167 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
9170 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
9171 (*saddl_se_1): Ditto.
9173 (*ssubl_se_1): Ditto.
9175 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9177 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
9178 rather than wi::to_widest for DR_INITs.
9179 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
9180 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
9181 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
9183 (vect_analyze_group_access_1): Note that here.
9185 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9187 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
9188 polynomial type sizes.
9190 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9192 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
9193 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
9194 (gimple_add_tmp_var): Likewise.
9196 2018-01-12 Martin Liska <mliska@suse.cz>
9198 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
9199 (gimple_alloc_sizes): Likewise.
9200 (dump_gimple_statistics): Use PRIu64 in printf format.
9201 * gimple.h: Change uint64_t to int.
9203 2018-01-12 Martin Liska <mliska@suse.cz>
9205 * tree-core.h: Use uint64_t instead of int.
9206 * tree.c (tree_node_counts): Likewise.
9207 (tree_node_sizes): Likewise.
9208 (dump_tree_statistics): Use PRIu64 in printf format.
9210 2018-01-12 Martin Liska <mliska@suse.cz>
9212 * Makefile.in: As qsort_chk is implemented in vec.c, add
9213 vec.o to linkage of gencfn-macros.
9214 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
9215 passing the info to record_node_allocation_statistics.
9216 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
9218 * ggc-common.c (struct ggc_usage): Add operator== and use
9219 it in operator< and compare function.
9220 * mem-stats.h (struct mem_usage): Likewise.
9221 * vec.c (struct vec_usage): Remove operator< and compare
9222 function. Can be simply inherited.
9224 2018-01-12 Martin Jambor <mjambor@suse.cz>
9227 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
9228 * tree-ssa-math-opts.c: Include domwalk.h.
9229 (convert_mult_to_fma_1): New function.
9230 (fma_transformation_info): New type.
9231 (fma_deferring_state): Likewise.
9232 (cancel_fma_deferring): New function.
9233 (result_of_phi): Likewise.
9234 (last_fma_candidate_feeds_initial_phi): Likewise.
9235 (convert_mult_to_fma): Added deferring logic, split actual
9236 transformation to convert_mult_to_fma_1.
9237 (math_opts_dom_walker): New type.
9238 (math_opts_dom_walker::after_dom_children): New method, body moved
9239 here from pass_optimize_widening_mul::execute, added deferring logic
9241 (pass_optimize_widening_mul::execute): Moved most of code to
9242 math_opts_dom_walker::after_dom_children.
9243 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
9244 * config/i386/i386.c (ix86_option_override_internal): Added
9245 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
9247 2018-01-12 Richard Biener <rguenther@suse.de>
9250 * dwarf2out.c (gen_variable_die): Do not reset old_die for
9251 inline instance vars.
9253 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
9256 * config/rx/rx.c (rx_is_restricted_memory_address):
9259 2018-01-12 Richard Biener <rguenther@suse.de>
9261 PR tree-optimization/80846
9262 * target.def (split_reduction): New target hook.
9263 * targhooks.c (default_split_reduction): New function.
9264 * targhooks.h (default_split_reduction): Declare.
9265 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
9266 target requests first reduce vectors by combining low and high
9268 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
9269 (get_vectype_for_scalar_type_and_size): Export.
9270 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
9271 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
9272 * doc/tm.texi: Regenerate.
9273 * config/i386/i386.c (ix86_split_reduction): Implement
9274 TARGET_VECTORIZE_SPLIT_REDUCTION.
9276 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
9279 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
9280 in PIC mode except for TARGET_VXWORKS_RTP.
9281 * config/sparc/sparc.c: Include cfgrtl.h.
9282 (TARGET_INIT_PIC_REG): Define.
9283 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
9284 (sparc_pic_register_p): New predicate.
9285 (sparc_legitimate_address_p): Use it.
9286 (sparc_legitimize_pic_address): Likewise.
9287 (sparc_delegitimize_address): Likewise.
9288 (sparc_mode_dependent_address_p): Likewise.
9289 (gen_load_pcrel_sym): Remove 4th parameter.
9290 (load_got_register): Adjust call to above. Remove obsolete stuff.
9291 (sparc_expand_prologue): Do not call load_got_register here.
9292 (sparc_flat_expand_prologue): Likewise.
9293 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
9294 (sparc_use_pseudo_pic_reg): New function.
9295 (sparc_init_pic_reg): Likewise.
9296 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
9297 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
9299 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
9301 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
9302 Add item for branch_cost.
9304 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
9306 PR rtl-optimization/83565
9307 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
9308 not extend the result to a larger mode for rotate operations.
9309 (num_sign_bit_copies1): Likewise.
9311 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9314 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
9316 Use values-Xc.o for -pedantic.
9317 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
9319 2018-01-12 Martin Liska <mliska@suse.cz>
9322 * ipa-devirt.c (final_warning_record::grow_type_warnings):
9324 (possible_polymorphic_call_targets): Use it.
9325 (ipa_devirt): Likewise.
9327 2018-01-12 Martin Liska <mliska@suse.cz>
9329 * profile-count.h (enum profile_quality): Use 0 as invalid
9330 enum value of profile_quality.
9332 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
9334 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
9335 -mext-string options.
9337 2018-01-12 Richard Biener <rguenther@suse.de>
9339 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
9340 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
9341 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
9343 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
9345 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
9347 * configure.ac (--with-long-double-format): Add support for the
9348 configuration option to change the default long double format on
9350 * config.gcc (powerpc*-linux*-*): Likewise.
9351 * configure: Regenerate.
9352 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
9353 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
9354 used without modification.
9356 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9358 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
9359 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
9360 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
9361 MISC_BUILTIN_SPEC_BARRIER.
9362 (rs6000_init_builtins): Likewise.
9363 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
9365 (speculation_barrier): New define_insn.
9366 * doc/extend.texi: Document __builtin_speculation_barrier.
9368 2018-01-11 Jakub Jelinek <jakub@redhat.com>
9371 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
9372 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
9373 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
9375 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
9376 integral modes instead of "ss" and "sd".
9377 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
9378 vectors with 32-bit and 64-bit elements.
9379 (vecdupssescalarmodesuffix): New mode attribute.
9380 (vec_dup<mode>): Use it.
9382 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
9385 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
9386 frame if argument is passed on stack.
9388 2018-01-11 Jakub Jelinek <jakub@redhat.com>
9391 * ree.c (combine_reaching_defs): Optimize also
9392 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
9393 reg2=any_extend(exp); reg1=reg2;, formatting fix.
9395 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
9398 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
9400 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
9403 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
9404 after they are computed.
9406 2018-01-11 Bin Cheng <bin.cheng@arm.com>
9408 PR tree-optimization/83695
9409 * gimple-loop-linterchange.cc
9410 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
9411 reset cached scev information after interchange.
9412 (pass_linterchange::execute): Remove call to scev_reset_htab.
9414 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9416 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
9417 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
9418 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
9419 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
9420 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
9421 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
9422 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
9423 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
9424 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
9425 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
9426 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
9427 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
9428 (V_lane_reg): Likewise.
9429 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
9431 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
9432 (vfmal_lane_low<mode>_intrinsic,
9433 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
9434 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
9435 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
9436 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
9437 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
9438 vfmsl_lane_high<mode>_intrinsic): New define_insns.
9440 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9442 * config/arm/arm-cpus.in (fp16fml): New feature.
9443 (ALL_SIMD): Add fp16fml.
9444 (armv8.2-a): Add fp16fml as an option.
9445 (armv8.3-a): Likewise.
9446 (armv8.4-a): Add fp16fml as part of fp16.
9447 * config/arm/arm.h (TARGET_FP16FML): Define.
9448 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
9450 * config/arm/arm-modes.def (V2HF): Define.
9451 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
9452 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
9453 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
9454 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
9455 vfmsl_low, vfmsl_high): New set of builtins.
9456 * config/arm/iterators.md (PLUSMINUS): New code iterator.
9457 (vfml_op): New code attribute.
9458 (VFMLHALVES): New int iterator.
9459 (VFML, VFMLSEL): New mode attributes.
9460 (V_reg): Define mapping for V2HF.
9461 (V_hi, V_lo): New mode attributes.
9462 (VF_constraint): Likewise.
9463 (vfml_half, vfml_half_selector): New int attributes.
9464 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
9466 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
9467 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
9469 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
9470 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
9471 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
9472 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
9474 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
9475 Document new effective target and option set.
9477 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9479 * config/arm/arm-cpus.in (armv8_4): New feature.
9480 (ARMv8_4a): New fgroup.
9481 (armv8.4-a): New arch.
9482 * config/arm/arm-tables.opt: Regenerate.
9483 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
9484 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
9485 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
9486 Add matching rules for -march=armv8.4-a and extensions.
9487 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
9489 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
9492 * config/rx/rx.md (BW): New mode attribute.
9493 (sync_lock_test_and_setsi): Add mode suffix to insn output.
9495 2018-01-11 Richard Biener <rguenther@suse.de>
9497 PR tree-optimization/83435
9498 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
9499 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
9500 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
9502 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9503 Alan Hayward <alan.hayward@arm.com>
9504 David Sherwood <david.sherwood@arm.com>
9506 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
9508 (aarch64_classify_address): Initialize it. Track polynomial offsets.
9509 (aarch64_print_address_internal): Use it to check for a zero offset.
9511 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9512 Alan Hayward <alan.hayward@arm.com>
9513 David Sherwood <david.sherwood@arm.com>
9515 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
9516 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
9517 Return a poly_int64 rather than a HOST_WIDE_INT.
9518 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
9519 rather than a HOST_WIDE_INT.
9520 * config/aarch64/aarch64.h (aarch64_frame): Protect with
9521 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
9522 hard_fp_offset, frame_size, initial_adjust, callee_offset and
9523 final_offset from HOST_WIDE_INT to poly_int64.
9524 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
9525 to_constant when getting the number of units in an Advanced SIMD
9527 (aarch64_builtin_vectorized_function): Check for a constant number
9529 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
9531 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
9532 attribute instead of GET_MODE_NUNITS.
9533 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
9534 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
9535 GET_MODE_SIZE for fixed-size registers.
9536 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
9537 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
9538 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
9539 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
9540 (aarch64_print_operand, aarch64_print_address_internal)
9541 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
9542 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
9543 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
9544 Handle polynomial GET_MODE_SIZE.
9545 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
9546 wider than SImode without modification.
9547 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
9548 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
9549 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
9550 passing and returning SVE modes.
9551 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
9552 rather than GEN_INT.
9553 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
9554 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
9555 (aarch64_allocate_and_probe_stack_space): Likewise.
9556 (aarch64_layout_frame): Cope with polynomial offsets.
9557 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
9558 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
9560 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
9561 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
9562 poly_int64 rather than a HOST_WIDE_INT.
9563 (aarch64_get_separate_components, aarch64_process_components)
9564 (aarch64_expand_prologue, aarch64_expand_epilogue)
9565 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
9566 (aarch64_anchor_offset): New function, split out from...
9567 (aarch64_legitimize_address): ...here.
9568 (aarch64_builtin_vectorization_cost): Handle polynomial
9569 TYPE_VECTOR_SUBPARTS.
9570 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
9572 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
9573 number of elements from the PARALLEL rather than the mode.
9574 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
9575 rather than GET_MODE_BITSIZE.
9576 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
9577 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
9578 (aarch64_expand_vec_perm_const_1): Handle polynomial
9579 d->perm.length () and d->perm elements.
9580 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
9581 Apply to_constant to d->perm elements.
9582 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
9583 polynomial CONST_VECTOR_NUNITS.
9584 (aarch64_move_pointer): Take amount as a poly_int64 rather
9586 (aarch64_progress_pointer): Avoid temporary variable.
9587 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
9588 the mode attribute instead of GET_MODE.
9590 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9591 Alan Hayward <alan.hayward@arm.com>
9592 David Sherwood <david.sherwood@arm.com>
9594 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
9595 x exists before using it.
9596 (aarch64_add_constant_internal): Rename to...
9597 (aarch64_add_offset_1): ...this. Replace regnum with separate
9598 src and dest rtxes. Handle the case in which they're different,
9599 including when the offset is zero. Replace scratchreg with an rtx.
9600 Use 2 additions if there is no spare register into which we can
9601 move a 16-bit constant.
9602 (aarch64_add_constant): Delete.
9603 (aarch64_add_offset): Replace reg with separate src and dest
9604 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
9605 Use aarch64_add_offset_1.
9606 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
9607 an rtx rather than an int. Take the delta as a poly_int64
9608 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
9609 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
9610 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
9611 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
9612 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
9614 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
9615 aarch64_add_constant.
9617 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9619 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
9620 Use scalar_float_mode.
9622 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9624 * config/aarch64/aarch64-simd.md
9625 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
9626 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
9627 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
9628 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
9629 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
9630 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
9631 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
9632 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
9633 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
9634 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
9636 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9639 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
9640 targ_options->x_arm_arch_string is non NULL.
9642 2018-01-11 Tamar Christina <tamar.christina@arm.com>
9644 * config/aarch64/aarch64.h
9645 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
9647 2018-01-11 Sudakshina Das <sudi.das@arm.com>
9650 * expmed.c (emit_store_flag_force): Swap if const op0
9651 and change VOIDmode to mode of op0.
9653 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9655 PR rtl-optimization/83761
9656 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
9657 than bytes to mode_for_size.
9659 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9662 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
9663 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
9666 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9669 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
9670 when in layout mode.
9671 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
9672 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
9675 2018-01-10 Michael Collison <michael.collison@arm.com>
9677 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
9678 * config/aarch64/aarch64-option-extension.def: Add
9679 AARCH64_OPT_EXTENSION of 'fp16fml'.
9680 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9681 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
9682 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
9683 * config/aarch64/constraints.md (Ui7): New constraint.
9684 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
9685 (VFMLA_SEL_W): Ditto.
9688 (VFMLA16_LOW): New int iterator.
9689 (VFMLA16_HIGH): Ditto.
9690 (UNSPEC_FMLAL): New unspec.
9691 (UNSPEC_FMLSL): Ditto.
9692 (UNSPEC_FMLAL2): Ditto.
9693 (UNSPEC_FMLSL2): Ditto.
9694 (f16mac): New code attribute.
9695 * config/aarch64/aarch64-simd-builtins.def
9696 (aarch64_fmlal_lowv2sf): Ditto.
9697 (aarch64_fmlsl_lowv2sf): Ditto.
9698 (aarch64_fmlalq_lowv4sf): Ditto.
9699 (aarch64_fmlslq_lowv4sf): Ditto.
9700 (aarch64_fmlal_highv2sf): Ditto.
9701 (aarch64_fmlsl_highv2sf): Ditto.
9702 (aarch64_fmlalq_highv4sf): Ditto.
9703 (aarch64_fmlslq_highv4sf): Ditto.
9704 (aarch64_fmlal_lane_lowv2sf): Ditto.
9705 (aarch64_fmlsl_lane_lowv2sf): Ditto.
9706 (aarch64_fmlal_laneq_lowv2sf): Ditto.
9707 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
9708 (aarch64_fmlalq_lane_lowv4sf): Ditto.
9709 (aarch64_fmlsl_lane_lowv4sf): Ditto.
9710 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
9711 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
9712 (aarch64_fmlal_lane_highv2sf): Ditto.
9713 (aarch64_fmlsl_lane_highv2sf): Ditto.
9714 (aarch64_fmlal_laneq_highv2sf): Ditto.
9715 (aarch64_fmlsl_laneq_highv2sf): Ditto.
9716 (aarch64_fmlalq_lane_highv4sf): Ditto.
9717 (aarch64_fmlsl_lane_highv4sf): Ditto.
9718 (aarch64_fmlalq_laneq_highv4sf): Ditto.
9719 (aarch64_fmlsl_laneq_highv4sf): Ditto.
9720 * config/aarch64/aarch64-simd.md:
9721 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
9722 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
9723 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
9724 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
9725 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
9726 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
9727 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
9728 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
9729 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
9730 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
9731 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
9732 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
9733 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
9734 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
9735 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
9736 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
9737 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
9738 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
9739 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
9740 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
9741 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
9742 (vfmlsl_low_u32): Ditto.
9743 (vfmlalq_low_u32): Ditto.
9744 (vfmlslq_low_u32): Ditto.
9745 (vfmlal_high_u32): Ditto.
9746 (vfmlsl_high_u32): Ditto.
9747 (vfmlalq_high_u32): Ditto.
9748 (vfmlslq_high_u32): Ditto.
9749 (vfmlal_lane_low_u32): Ditto.
9750 (vfmlsl_lane_low_u32): Ditto.
9751 (vfmlal_laneq_low_u32): Ditto.
9752 (vfmlsl_laneq_low_u32): Ditto.
9753 (vfmlalq_lane_low_u32): Ditto.
9754 (vfmlslq_lane_low_u32): Ditto.
9755 (vfmlalq_laneq_low_u32): Ditto.
9756 (vfmlslq_laneq_low_u32): Ditto.
9757 (vfmlal_lane_high_u32): Ditto.
9758 (vfmlsl_lane_high_u32): Ditto.
9759 (vfmlal_laneq_high_u32): Ditto.
9760 (vfmlsl_laneq_high_u32): Ditto.
9761 (vfmlalq_lane_high_u32): Ditto.
9762 (vfmlslq_lane_high_u32): Ditto.
9763 (vfmlalq_laneq_high_u32): Ditto.
9764 (vfmlslq_laneq_high_u32): Ditto.
9765 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
9766 (AARCH64_FL_FOR_ARCH8_4): New.
9767 (AARCH64_ISA_F16FML): New ISA flag.
9768 (TARGET_F16FML): New feature flag for fp16fml.
9769 (doc/invoke.texi): Document new fp16fml option.
9771 2018-01-10 Michael Collison <michael.collison@arm.com>
9773 * config/aarch64/aarch64-builtins.c:
9774 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
9775 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9776 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
9777 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
9778 (AARCH64_ISA_SHA3): New ISA flag.
9779 (TARGET_SHA3): New feature flag for sha3.
9780 * config/aarch64/iterators.md (sha512_op): New int attribute.
9781 (CRYPTO_SHA512): New int iterator.
9782 (UNSPEC_SHA512H): New unspec.
9783 (UNSPEC_SHA512H2): Ditto.
9784 (UNSPEC_SHA512SU0): Ditto.
9785 (UNSPEC_SHA512SU1): Ditto.
9786 * config/aarch64/aarch64-simd-builtins.def
9787 (aarch64_crypto_sha512hqv2di): New builtin.
9788 (aarch64_crypto_sha512h2qv2di): Ditto.
9789 (aarch64_crypto_sha512su0qv2di): Ditto.
9790 (aarch64_crypto_sha512su1qv2di): Ditto.
9791 (aarch64_eor3qv8hi): Ditto.
9792 (aarch64_rax1qv2di): Ditto.
9793 (aarch64_xarqv2di): Ditto.
9794 (aarch64_bcaxqv8hi): Ditto.
9795 * config/aarch64/aarch64-simd.md:
9796 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
9797 (aarch64_crypto_sha512su0qv2di): Ditto.
9798 (aarch64_crypto_sha512su1qv2di): Ditto.
9799 (aarch64_eor3qv8hi): Ditto.
9800 (aarch64_rax1qv2di): Ditto.
9801 (aarch64_xarqv2di): Ditto.
9802 (aarch64_bcaxqv8hi): Ditto.
9803 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
9804 (vsha512h2q_u64): Ditto.
9805 (vsha512su0q_u64): Ditto.
9806 (vsha512su1q_u64): Ditto.
9807 (veor3q_u16): Ditto.
9808 (vrax1q_u64): Ditto.
9810 (vbcaxq_u16): Ditto.
9811 * config/arm/types.md (crypto_sha512): New type attribute.
9812 (crypto_sha3): Ditto.
9813 (doc/invoke.texi): Document new sha3 option.
9815 2018-01-10 Michael Collison <michael.collison@arm.com>
9817 * config/aarch64/aarch64-builtins.c:
9818 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
9819 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9820 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
9821 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
9822 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
9823 (AARCH64_ISA_SM4): New ISA flag.
9824 (TARGET_SM4): New feature flag for sm4.
9825 * config/aarch64/aarch64-simd-builtins.def
9826 (aarch64_sm3ss1qv4si): Ditto.
9827 (aarch64_sm3tt1aq4si): Ditto.
9828 (aarch64_sm3tt1bq4si): Ditto.
9829 (aarch64_sm3tt2aq4si): Ditto.
9830 (aarch64_sm3tt2bq4si): Ditto.
9831 (aarch64_sm3partw1qv4si): Ditto.
9832 (aarch64_sm3partw2qv4si): Ditto.
9833 (aarch64_sm4eqv4si): Ditto.
9834 (aarch64_sm4ekeyqv4si): Ditto.
9835 * config/aarch64/aarch64-simd.md:
9836 (aarch64_sm3ss1qv4si): Ditto.
9837 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
9838 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
9839 (aarch64_sm4eqv4si): Ditto.
9840 (aarch64_sm4ekeyqv4si): Ditto.
9841 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
9842 (sm3part_op): Ditto.
9843 (CRYPTO_SM3TT): Ditto.
9844 (CRYPTO_SM3PART): Ditto.
9845 (UNSPEC_SM3SS1): New unspec.
9846 (UNSPEC_SM3TT1A): Ditto.
9847 (UNSPEC_SM3TT1B): Ditto.
9848 (UNSPEC_SM3TT2A): Ditto.
9849 (UNSPEC_SM3TT2B): Ditto.
9850 (UNSPEC_SM3PARTW1): Ditto.
9851 (UNSPEC_SM3PARTW2): Ditto.
9852 (UNSPEC_SM4E): Ditto.
9853 (UNSPEC_SM4EKEY): Ditto.
9854 * config/aarch64/constraints.md (Ui2): New constraint.
9855 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
9856 * config/arm/types.md (crypto_sm3): New type attribute.
9857 (crypto_sm4): Ditto.
9858 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
9859 (vsm3tt1aq_u32): Ditto.
9860 (vsm3tt1bq_u32): Ditto.
9861 (vsm3tt2aq_u32): Ditto.
9862 (vsm3tt2bq_u32): Ditto.
9863 (vsm3partw1q_u32): Ditto.
9864 (vsm3partw2q_u32): Ditto.
9865 (vsm4eq_u32): Ditto.
9866 (vsm4ekeyq_u32): Ditto.
9867 (doc/invoke.texi): Document new sm4 option.
9869 2018-01-10 Michael Collison <michael.collison@arm.com>
9871 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
9872 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
9873 (AARCH64_FL_FOR_ARCH8_4): New.
9874 (AARCH64_FL_V8_4): New flag.
9875 (doc/invoke.texi): Document new armv8.4-a option.
9877 2018-01-10 Michael Collison <michael.collison@arm.com>
9879 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9880 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
9881 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
9882 * config/aarch64/aarch64-option-extension.def: Add
9883 AARCH64_OPT_EXTENSION of 'sha2'.
9884 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
9885 (crypto): Disable sha2 and aes if crypto disabled.
9886 (crypto): Enable aes and sha2 if enabled.
9887 (simd): Disable sha2 and aes if simd disabled.
9888 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
9890 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
9891 (TARGET_SHA2): New feature flag for sha2.
9892 (TARGET_AES): New feature flag for aes.
9893 * config/aarch64/aarch64-simd.md:
9894 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
9895 conditional on TARGET_AES.
9896 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
9897 (aarch64_crypto_sha1hsi): Make pattern conditional
9899 (aarch64_crypto_sha1hv4si): Ditto.
9900 (aarch64_be_crypto_sha1hv4si): Ditto.
9901 (aarch64_crypto_sha1su1v4si): Ditto.
9902 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
9903 (aarch64_crypto_sha1su0v4si): Ditto.
9904 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
9905 (aarch64_crypto_sha256su0v4si): Ditto.
9906 (aarch64_crypto_sha256su1v4si): Ditto.
9907 (doc/invoke.texi): Document new aes and sha2 options.
9909 2018-01-10 Martin Sebor <msebor@redhat.com>
9911 PR tree-optimization/83781
9912 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
9915 2018-01-11 Martin Sebor <msebor@gmail.com>
9916 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9918 PR tree-optimization/83501
9919 PR tree-optimization/81703
9921 * tree-ssa-strlen.c (get_string_cst): Rename...
9922 (get_string_len): ...to this. Handle global constants.
9923 (handle_char_store): Adjust.
9925 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
9926 Jim Wilson <jimw@sifive.com>
9928 * config/riscv/riscv-protos.h (riscv_output_return): New.
9929 * config/riscv/riscv.c (struct machine_function): New naked_p field.
9930 (riscv_attribute_table, riscv_output_return),
9931 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
9932 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
9933 (riscv_compute_frame_info): Only compute frame->mask if not a naked
9935 (riscv_expand_prologue): Add early return for naked function.
9936 (riscv_expand_epilogue): Likewise.
9937 (riscv_function_ok_for_sibcall): Return false for naked function.
9938 (riscv_set_current_function): New.
9939 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
9940 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
9941 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
9942 * doc/extend.texi (RISC-V Function Attributes): New.
9944 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
9946 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
9947 check for 128-bit long double before checking TCmode.
9948 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
9949 128-bit long doubles before checking TFmode or TCmode.
9950 (FLOAT128_IBM_P): Likewise.
9952 2018-01-10 Martin Sebor <msebor@redhat.com>
9954 PR tree-optimization/83671
9955 * builtins.c (c_strlen): Unconditionally return zero for the empty
9957 Use -Warray-bounds for warnings.
9958 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
9959 for non-constant array indices with COMPONENT_REF, arrays of
9960 arrays, and pointers to arrays.
9961 (gimple_fold_builtin_strlen): Determine and set length range for
9962 non-constant character arrays.
9964 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
9967 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
9970 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
9972 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
9974 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
9977 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
9978 VECTOR_MEM_ALTIVEC_OR_VSX_P.
9979 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
9980 indexed_or_indirect_operand predicate.
9981 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
9982 (*vsx_le_perm_load_v8hi): Likewise.
9983 (*vsx_le_perm_load_v16qi): Likewise.
9984 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
9985 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
9986 (*vsx_le_perm_store_v8hi): Likewise.
9987 (*vsx_le_perm_store_v16qi): Likewise.
9988 (eight unnamed splitters): Likewise.
9990 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
9992 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
9993 * config/rs6000/emmintrin.h: Likewise.
9994 * config/rs6000/mmintrin.h: Likewise.
9995 * config/rs6000/xmmintrin.h: Likewise.
9997 2018-01-10 David Malcolm <dmalcolm@redhat.com>
10000 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
10002 * tree.c (tree_nop_conversion): Return true for location wrapper
10004 (maybe_wrap_with_location): New function.
10005 (selftest::check_strip_nops): New function.
10006 (selftest::test_location_wrappers): New function.
10007 (selftest::tree_c_tests): Call it.
10008 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
10009 (maybe_wrap_with_location): New decl.
10010 (EXPR_LOCATION_WRAPPER_P): New macro.
10011 (location_wrapper_p): New inline function.
10012 (tree_strip_any_location_wrapper): New inline function.
10014 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
10017 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
10018 stack_realign_offset for the largest alignment of stack slot
10020 (ix86_find_max_used_stack_alignment): New function.
10021 (ix86_finalize_stack_frame_flags): Use it. Set
10022 max_used_stack_alignment if we don't realign stack.
10023 * config/i386/i386.h (machine_function): Add
10024 max_used_stack_alignment.
10026 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
10028 * config/arm/arm.opt (-mbranch-cost): New option.
10029 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
10032 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
10035 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
10036 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
10038 2018-01-10 Richard Biener <rguenther@suse.de>
10041 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
10042 early out so it also covers the case where we have a non-NULL
10045 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
10047 PR tree-optimization/83753
10048 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
10049 for non-strided grouped accesses if the number of elements is 1.
10051 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10054 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
10055 * i386.h (TARGET_USE_GATHER): Define.
10056 * x86-tune.def (X86_TUNE_USE_GATHER): New.
10058 2018-01-10 Martin Liska <mliska@suse.cz>
10061 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
10062 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
10064 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
10065 CLEANUP_NO_PARTITIONING is not set.
10067 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
10069 * doc/rtl.texi: Remove documentation of (const ...) wrappers
10070 for vectors, as a partial revert of r254296.
10071 * rtl.h (const_vec_p): Delete.
10072 (const_vec_duplicate_p): Don't test for vector CONSTs.
10073 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
10074 * expmed.c (make_tree): Likewise.
10077 * common.md (E, F): Use CONSTANT_P instead of checking for
10079 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
10080 checking for CONST_VECTOR.
10082 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10084 PR middle-end/83575
10085 * predict.c (force_edge_cold): Handle in more sane way edges
10086 with no prediction.
10088 2018-01-09 Carl Love <cel@us.ibm.com>
10090 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
10092 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
10093 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
10094 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
10095 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
10096 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
10097 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
10098 * config/rs6000/rs6000-protos.h: Add extern defition for
10099 rs6000_generate_float2_double_code.
10100 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
10102 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
10103 (float2_v2df): Add define_expand.
10105 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
10108 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
10109 op_mode in the force_to_mode call.
10111 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10113 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
10114 instead of checking each element individually.
10115 (aarch64_evpc_uzp): Likewise.
10116 (aarch64_evpc_zip): Likewise.
10117 (aarch64_evpc_ext): Likewise.
10118 (aarch64_evpc_rev): Likewise.
10119 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
10120 instead of checking each element individually. Return true without
10122 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
10123 whether all selected elements come from the same input, instead of
10124 checking each element individually. Remove calls to gen_rtx_REG,
10125 start_sequence and end_sequence and instead assert that no rtl is
10128 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10130 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
10131 order of HIGH and CONST checks.
10133 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10135 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
10136 if the destination isn't an SSA_NAME.
10138 2018-01-09 Richard Biener <rguenther@suse.de>
10140 PR tree-optimization/83668
10141 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
10143 (canonicalize_loop_form): ... here, renamed from ...
10144 (canonicalize_loop_closed_ssa_form): ... this and amended to
10145 swap successor edges for loop exit blocks to make us use
10146 the RPO order we need for initial schedule generation.
10148 2018-01-09 Joseph Myers <joseph@codesourcery.com>
10150 PR tree-optimization/64811
10151 * match.pd: When optimizing comparisons with Inf, avoid
10152 introducing or losing exceptions from comparisons with NaN.
10154 2018-01-09 Martin Liska <mliska@suse.cz>
10157 * asan.c (shadow_mem_size): Add gcc_assert.
10159 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
10161 Don't save registers in main().
10164 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
10165 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
10166 * config/avr/avr.c (avr_set_current_function): Don't error if
10167 naked, OS_task or OS_main are specified at the same time.
10168 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
10170 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
10172 * common/config/avr/avr-common.c (avr_option_optimization_table):
10173 Switch on -mmain-is-OS_task for optimizing compilations.
10175 2018-01-09 Richard Biener <rguenther@suse.de>
10177 PR tree-optimization/83572
10178 * graphite.c: Include cfganal.h.
10179 (graphite_transform_loops): Connect infinite loops to exit
10180 and remove fake edges at the end.
10182 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10184 * ipa-inline.c (edge_badness): Revert accidental checkin.
10186 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10189 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
10190 symbols; not inline clones.
10192 2018-01-09 Jakub Jelinek <jakub@redhat.com>
10195 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
10196 hard registers. Formatting fixes.
10198 PR preprocessor/83722
10199 * gcc.c (try_generate_repro): Pass
10200 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
10201 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
10204 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
10205 Kito Cheng <kito.cheng@gmail.com>
10207 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
10208 (riscv_leaf_function_p): Delete.
10209 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
10211 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10213 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
10215 (do_ifelse): New function.
10216 (do_isel): New function.
10217 (do_sub3): New function.
10218 (do_add3): New function.
10219 (do_load_mask_compare): New function.
10220 (do_overlap_load_compare): New function.
10221 (expand_compare_loop): New function.
10222 (expand_block_compare): Call expand_compare_loop() when appropriate.
10223 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
10224 option description.
10225 (-mblock-compare-inline-loop-limit): New option.
10227 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10230 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
10231 Reverse order of second and third operands in first alternative.
10232 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
10233 of first and second elements in UNSPEC_VPERMR vector.
10234 (altivec_expand_vec_perm_le): Likewise.
10236 2018-01-08 Jeff Law <law@redhat.com>
10238 PR rtl-optimizatin/81308
10239 * tree-switch-conversion.c (cfg_altered): New file scoped static.
10240 (process_switch): If group_case_labels makes a change, then set
10242 (pass_convert_switch::execute): If a switch is converted, then
10243 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
10245 PR rtl-optimization/81308
10246 * recog.c (split_all_insns): Conditionally cleanup the CFG after
10249 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
10251 PR target/83663 - Revert r255946
10252 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
10253 generation for cases where splatting a value is not useful.
10254 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
10255 across a vec_duplicate and a paradoxical subreg forming a vector
10256 mode to a vec_concat.
10258 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10260 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
10261 -march=armv8.3-a variants.
10262 * config/arm/t-multilib: Likewise.
10263 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
10265 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10267 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
10269 (cceq_ior_compare_complement): Give it a name so I can use it, and
10270 change boolean_or_operator predicate to boolean_operator so it can
10271 be used to generate a crand.
10272 (eqne): New code iterator.
10273 (bd/bd_neg): New code_attrs.
10274 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
10275 a single define_insn.
10276 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
10277 decrement (bdnzt/bdnzf/bdzt/bdzf).
10278 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
10279 with the new names of the branch decrement patterns, and added the
10280 names of the branch decrement conditional patterns.
10282 2018-01-08 Richard Biener <rguenther@suse.de>
10284 PR tree-optimization/83563
10285 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
10288 2018-01-08 Richard Biener <rguenther@suse.de>
10290 PR middle-end/83713
10291 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
10293 2018-01-08 Richard Biener <rguenther@suse.de>
10295 PR tree-optimization/83685
10296 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
10297 references to abnormals.
10299 2018-01-08 Richard Biener <rguenther@suse.de>
10302 * dwarf2out.c (output_indirect_strings): Handle empty
10303 skeleton_debug_str_hash.
10304 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
10306 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
10308 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
10309 (emit_store_direct): Likewise.
10310 (arc_trampoline_adjust_address): Likewise.
10311 (arc_asm_trampoline_template): New function.
10312 (arc_initialize_trampoline): Use asm_trampoline_template.
10313 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
10314 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
10315 * config/arc/arc.md (flush_icache): Delete pattern.
10317 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
10319 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
10320 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
10323 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
10326 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
10327 by not USED_FOR_TARGET.
10328 (make_pass_resolve_sw_modes): Likewise.
10330 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
10332 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
10335 2018-01-08 Richard Biener <rguenther@suse.de>
10337 PR middle-end/83580
10338 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
10340 2018-01-08 Richard Biener <rguenther@suse.de>
10342 PR middle-end/83517
10343 * match.pd ((t * 2) / 2) -> t): Add missing :c.
10345 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
10347 PR middle-end/81897
10348 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
10349 basic blocks with a small number of successors.
10350 (convert_control_dep_chain_into_preds): Improve handling of
10352 (dump_predicates): Split apart into...
10353 (dump_pred_chain): ...here...
10354 (dump_pred_info): ...and here.
10355 (can_one_predicate_be_invalidated_p): Add debugging printfs.
10356 (can_chain_union_be_invalidated_p): Improve check for invalidation
10358 (uninit_uses_cannot_happen): Avoid unnecessary if
10359 convert_control_dep_chain_into_preds yielded nothing.
10361 2018-01-06 Martin Sebor <msebor@redhat.com>
10363 PR tree-optimization/83640
10364 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
10365 subtracting negative offset from size.
10366 (builtin_access::overlap): Adjust offset bounds of the access to fall
10367 within the size of the object if possible.
10369 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
10371 PR rtl-optimization/83699
10372 * expmed.c (extract_bit_field_1): Restrict the vector usage of
10373 extract_bit_field_as_subreg to cases in which the extracted
10374 value is also a vector.
10376 * lra-constraints.c (process_alt_operands): Test for the equivalence
10377 substitutions when detecting a possible reload cycle.
10379 2018-01-06 Jakub Jelinek <jakub@redhat.com>
10382 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
10383 by default if flag_selective_schedling{,2}. Formatting fixes.
10385 PR rtl-optimization/83682
10386 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
10387 if it has non-VECTOR_MODE element mode.
10388 (vec_duplicate_p): Likewise.
10390 PR middle-end/83694
10391 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
10392 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
10394 2018-01-05 Jakub Jelinek <jakub@redhat.com>
10397 * config/i386/i386-builtin.def
10398 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
10399 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
10400 Require also OPTION_MASK_ISA_AVX512F in addition to
10401 OPTION_MASK_ISA_GFNI.
10402 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
10403 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
10404 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
10405 to OPTION_MASK_ISA_GFNI.
10406 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
10407 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
10408 OPTION_MASK_ISA_AVX512BW.
10409 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
10410 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
10411 addition to OPTION_MASK_ISA_GFNI.
10412 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
10413 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
10414 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
10415 to OPTION_MASK_ISA_GFNI.
10416 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
10417 a requirement for all ISAs rather than any of them with a few
10419 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
10421 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
10422 bitmasks to be enabled with 3 exceptions, instead of requiring any
10423 enabled ISA with lots of exceptions.
10424 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
10425 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
10426 Change avx512bw in isa attribute to avx512f.
10427 * config/i386/sgxintrin.h: Add license boilerplate.
10428 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
10429 to __AVX512F__ and __AVX512VL to __AVX512VL__.
10430 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
10431 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
10433 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
10434 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
10435 temporarily sse2 rather than sse if not enabled already.
10438 * config/i386/sse.md (VI248_VLBW): Rename to ...
10439 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
10440 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
10441 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
10442 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
10443 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
10444 mode iterator instead of VI248_VLBW.
10446 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
10448 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
10449 (record_modified): Skip clobbers; add debug output.
10450 (param_change_prob): Use sreal frequencies.
10452 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
10454 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
10455 punt for user-aligned variables.
10457 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
10459 * tree-chrec.c (chrec_contains_symbols): Return true for
10462 2018-01-05 Sudakshina Das <sudi.das@arm.com>
10465 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
10466 of (x|y) == x for BICS pattern.
10468 2018-01-05 Jakub Jelinek <jakub@redhat.com>
10470 PR tree-optimization/83605
10471 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
10472 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
10475 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
10477 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
10478 * config/epiphany/rtems.h: New file.
10480 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10481 Uros Bizjak <ubizjak@gmail.com>
10484 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
10485 QIreg_operand instead of register_operand predicate.
10486 * config/i386/i386.c (ix86_rop_should_change_byte_p,
10487 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
10488 comments instead of -fmitigate[-_]rop.
10490 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10493 * cgraphunit.c (symbol_table::compile): Switch to text_section
10494 before calling assembly_start debug hook.
10495 * run-rtl-passes.c (run_rtl_passes): Likewise.
10498 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10500 * tree-vrp.c (extract_range_from_binary_expr_1): Check
10501 range_int_cst_p rather than !symbolic_range_p before calling
10502 extract_range_from_multiplicative_op_1.
10504 2018-01-04 Jeff Law <law@redhat.com>
10506 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
10507 redundant test in assertion.
10509 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10511 * doc/rtl.texi: Document machine_mode wrapper classes.
10513 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10515 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
10516 using tree_to_uhwi.
10518 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10520 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
10521 the VEC_PERM_EXPR fold to fail.
10523 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10526 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
10527 to switched_sections.
10529 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10532 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
10533 test for d.testing.
10535 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
10538 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
10539 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
10541 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10544 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
10545 is BLKmode and bitpos not zero or mode change is needed.
10547 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10550 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
10553 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
10556 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
10557 instead of MULT rtx. Update all corresponding splitters.
10558 (*saddl_se): Ditto.
10559 (*ssub<modesuffix>): Ditto.
10560 (*ssubl_se): Ditto.
10561 (*cmp_sadd_di): Update split patterns.
10562 (*cmp_sadd_si): Ditto.
10563 (*cmp_sadd_sidi): Ditto.
10564 (*cmp_ssub_di): Ditto.
10565 (*cmp_ssub_si): Ditto.
10566 (*cmp_ssub_sidi): Ditto.
10567 * config/alpha/predicates.md (const23_operand): New predicate.
10568 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
10569 Look for ASHIFT, not MULT inner operand.
10570 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
10572 2018-01-04 Martin Liska <mliska@suse.cz>
10574 PR gcov-profile/83669
10575 * gcov.c (output_intermediate_file): Add version to intermediate
10577 * doc/gcov.texi: Document new field 'version' in intermediate
10578 file format. Fix location of '-k' option of gcov command.
10580 2018-01-04 Martin Liska <mliska@suse.cz>
10583 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
10585 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10587 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
10589 2018-01-03 Martin Sebor <msebor@redhat.com>
10591 PR tree-optimization/83655
10592 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
10593 checking calls with invalid arguments.
10595 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10597 * tree-vect-stmts.c (vect_get_store_rhs): New function.
10598 (vectorizable_mask_load_store): Delete.
10599 (vectorizable_call): Return false for masked loads and stores.
10600 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
10601 instead of gimple_assign_rhs1.
10602 (vectorizable_load): Handle IFN_MASK_LOAD.
10603 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
10605 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10607 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
10609 (vectorizable_mask_load_store): ...here.
10610 (vectorizable_load): ...and here.
10612 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10614 * tree-vect-stmts.c (vect_build_all_ones_mask)
10615 (vect_build_zero_merge_argument): New functions, split out from...
10616 (vectorizable_load): ...here.
10618 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10620 * tree-vect-stmts.c (vect_check_store_rhs): New function,
10622 (vectorizable_mask_load_store): ...here.
10623 (vectorizable_store): ...and here.
10625 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10627 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
10629 (vectorizable_mask_load_store): ...here.
10631 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10633 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
10634 (vect_model_store_cost): Take a vec_load_store_type instead of a
10636 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
10637 (vect_model_store_cost): Take a vec_load_store_type instead of a
10639 (vectorizable_mask_load_store): Update accordingly.
10640 (vectorizable_store): Likewise.
10641 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
10643 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10645 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
10646 IFN_MASK_LOAD calls here rather than...
10647 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
10649 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10650 Alan Hayward <alan.hayward@arm.com>
10651 David Sherwood <david.sherwood@arm.com>
10653 * expmed.c (extract_bit_field_1): For vector extracts,
10654 fall back to extract_bit_field_as_subreg if vec_extract
10657 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10658 Alan Hayward <alan.hayward@arm.com>
10659 David Sherwood <david.sherwood@arm.com>
10661 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
10662 they are variable or constant sized.
10663 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
10664 slots for constant-sized data.
10666 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10667 Alan Hayward <alan.hayward@arm.com>
10668 David Sherwood <david.sherwood@arm.com>
10670 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
10671 handling COND_EXPRs with boolean comparisons, try to find a better
10672 basis for the mask type than the boolean itself.
10674 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10676 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
10677 is calculated and how it can be overridden.
10678 * genmodes.c (max_bitsize_mode_any_mode): New variable.
10679 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
10681 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
10684 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10685 Alan Hayward <alan.hayward@arm.com>
10686 David Sherwood <david.sherwood@arm.com>
10688 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
10689 Remove the mode argument.
10690 (aarch64_simd_valid_immediate): Remove the mode and inverse
10692 * config/aarch64/iterators.md (bitsize): New iterator.
10693 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
10694 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
10695 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
10696 aarch64_simd_valid_immediate.
10697 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
10698 (aarch64_reg_or_bic_imm): Likewise.
10699 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
10700 with an insn_type enum and msl with a modifier_type enum.
10701 Replace element_width with a scalar_mode. Change the shift
10702 to unsigned int. Add constructors for scalar_float_mode and
10703 scalar_int_mode elements.
10704 (aarch64_vect_float_const_representable_p): Delete.
10705 (aarch64_can_const_movi_rtx_p)
10706 (aarch64_simd_scalar_immediate_valid_for_move)
10707 (aarch64_simd_make_constant): Update call to
10708 aarch64_simd_valid_immediate.
10709 (aarch64_advsimd_valid_immediate_hs): New function.
10710 (aarch64_advsimd_valid_immediate): Likewise.
10711 (aarch64_simd_valid_immediate): Remove mode and inverse
10712 arguments. Rewrite to use the above. Use const_vec_duplicate_p
10713 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
10714 and aarch64_float_const_representable_p on the result.
10715 (aarch64_output_simd_mov_immediate): Remove mode argument.
10716 Update call to aarch64_simd_valid_immediate and use of
10717 simd_immediate_info.
10718 (aarch64_output_scalar_simd_mov_immediate): Update call
10721 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10722 Alan Hayward <alan.hayward@arm.com>
10723 David Sherwood <david.sherwood@arm.com>
10725 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
10726 (mode_nunits): Likewise CONST_MODE_NUNITS.
10727 * machmode.def (ADJUST_NUNITS): Document.
10728 * genmodes.c (mode_data::need_nunits_adj): New field.
10729 (blank_mode): Update accordingly.
10730 (adj_nunits): New variable.
10731 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
10733 (emit_mode_size_inline): Set need_bytesize_adj for all modes
10734 listed in adj_nunits.
10735 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
10736 listed in adj_nunits. Don't emit case statements for such modes.
10737 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
10738 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
10739 nothing if adj_nunits is nonnull.
10740 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
10741 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
10742 (emit_mode_fbit): Update use of print_maybe_const_decl.
10743 (emit_move_size): Likewise. Treat the array as non-const
10745 (emit_mode_adjustments): Handle adj_nunits.
10747 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10749 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
10750 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
10751 (VECTOR_MODES): Use it.
10752 (make_vector_modes): Take the prefix as an argument.
10754 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10755 Alan Hayward <alan.hayward@arm.com>
10756 David Sherwood <david.sherwood@arm.com>
10758 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
10759 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
10760 for MODE_VECTOR_BOOL.
10761 * machmode.def (VECTOR_BOOL_MODE): Document.
10762 * genmodes.c (VECTOR_BOOL_MODE): New macro.
10763 (make_vector_bool_mode): New function.
10764 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
10766 * lto-streamer-in.c (lto_input_mode_table): Likewise.
10767 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
10769 * stor-layout.c (int_mode_for_mode): Likewise.
10770 * tree.c (build_vector_type_for_mode): Likewise.
10771 * varasm.c (output_constant_pool_2): Likewise.
10772 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
10773 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
10774 for MODE_VECTOR_BOOL.
10775 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
10776 of mode class checks.
10777 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
10778 instead of a list of mode class checks.
10779 (expand_vector_scalar_condition): Likewise.
10780 (type_for_widest_vector_mode): Handle BImode as an inner mode.
10782 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10783 Alan Hayward <alan.hayward@arm.com>
10784 David Sherwood <david.sherwood@arm.com>
10786 * machmode.h (mode_size): Change from unsigned short to
10788 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
10789 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
10790 or if measurement_type is not polynomial.
10791 (fixed_size_mode::includes_p): Check for constant-sized modes.
10792 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
10793 return a poly_uint16 rather than an unsigned short.
10794 (emit_mode_size): Change the type of mode_size from unsigned short
10795 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
10796 (emit_mode_adjustments): Cope with polynomial vector sizes.
10797 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10799 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10801 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
10802 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
10803 * caller-save.c (setup_save_areas): Likewise.
10804 (replace_reg_with_saved_mem): Likewise.
10805 * calls.c (emit_library_call_value_1): Likewise.
10806 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
10807 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
10808 (gen_lowpart_for_combine): Likewise.
10809 * convert.c (convert_to_integer_1): Likewise.
10810 * cse.c (equiv_constant, cse_insn): Likewise.
10811 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
10812 (cselib_subst_to_values): Likewise.
10813 * dce.c (word_dce_process_block): Likewise.
10814 * df-problems.c (df_word_lr_mark_ref): Likewise.
10815 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
10816 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
10817 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
10818 (rtl_for_decl_location): Likewise.
10819 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
10820 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
10821 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
10822 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
10823 (expand_expr_real_1): Likewise.
10824 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
10825 (pad_below): Likewise.
10826 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
10827 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
10828 * ira.c (get_subreg_tracking_sizes): Likewise.
10829 * ira-build.c (ira_create_allocno_objects): Likewise.
10830 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
10831 (ira_sort_regnos_for_alter_reg): Likewise.
10832 * ira-costs.c (record_operand_costs): Likewise.
10833 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
10834 (resolve_simple_move): Likewise.
10835 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
10836 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
10837 (lra_constraints): Likewise.
10838 (CONST_POOL_OK_P): Reject variable-sized modes.
10839 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
10840 (add_pseudo_to_slot, lra_spill): Likewise.
10841 * omp-low.c (omp_clause_aligned_alignment): Likewise.
10842 * optabs-query.c (get_best_extraction_insn): Likewise.
10843 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
10844 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
10845 (expand_mult_highpart, valid_multiword_target_p): Likewise.
10846 * recog.c (offsettable_address_addr_space_p): Likewise.
10847 * regcprop.c (maybe_mode_change): Likewise.
10848 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
10849 * regrename.c (build_def_use): Likewise.
10850 * regstat.c (dump_reg_info): Likewise.
10851 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
10852 (find_reloads, find_reloads_subreg_address): Likewise.
10853 * reload1.c (eliminate_regs_1): Likewise.
10854 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
10855 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
10856 (simplify_binary_operation_1, simplify_subreg): Likewise.
10857 * targhooks.c (default_function_arg_padding): Likewise.
10858 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
10859 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
10860 (verify_gimple_assign_ternary): Likewise.
10861 * tree-inline.c (estimate_move_cost): Likewise.
10862 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10863 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
10864 (get_address_cost_ainc): Likewise.
10865 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
10866 (vect_supportable_dr_alignment): Likewise.
10867 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
10868 (vectorizable_reduction): Likewise.
10869 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
10870 (vectorizable_operation, vectorizable_load): Likewise.
10871 * tree.c (build_same_sized_truth_vector_type): Likewise.
10872 * valtrack.c (cleanup_auto_inc_dec): Likewise.
10873 * var-tracking.c (emit_note_insn_var_location): Likewise.
10874 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
10875 (ADDR_VEC_ALIGN): Likewise.
10877 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10878 Alan Hayward <alan.hayward@arm.com>
10879 David Sherwood <david.sherwood@arm.com>
10881 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
10883 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
10884 or if measurement_type is polynomial.
10885 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
10886 * combine.c (make_extraction): Likewise.
10887 * dse.c (find_shift_sequence): Likewise.
10888 * dwarf2out.c (mem_loc_descriptor): Likewise.
10889 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
10890 (extract_bit_field, extract_low_bits): Likewise.
10891 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
10892 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
10893 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
10894 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
10895 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
10896 * reload.c (find_reloads): Likewise.
10897 * reload1.c (alter_reg): Likewise.
10898 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
10899 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
10900 * tree-if-conv.c (predicate_mem_writes): Likewise.
10901 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
10902 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
10903 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
10904 * valtrack.c (dead_debug_insert_temp): Likewise.
10905 * varasm.c (mergeable_constant_section): Likewise.
10906 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
10908 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10909 Alan Hayward <alan.hayward@arm.com>
10910 David Sherwood <david.sherwood@arm.com>
10912 * expr.c (expand_assignment): Cope with polynomial mode sizes
10913 when assigning to a CONCAT.
10915 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10916 Alan Hayward <alan.hayward@arm.com>
10917 David Sherwood <david.sherwood@arm.com>
10919 * machmode.h (mode_precision): Change from unsigned short to
10921 (mode_to_precision): Return a poly_uint16 rather than an unsigned
10923 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
10924 or if measurement_type is not polynomial.
10925 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
10926 in which the mode is already known to be a scalar_int_mode.
10927 * genmodes.c (emit_mode_precision): Change the type of mode_precision
10928 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
10930 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10931 for GET_MODE_PRECISION.
10932 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10933 for GET_MODE_PRECISION.
10934 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
10936 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
10937 (expand_field_assignment, make_extraction): Likewise.
10938 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
10939 (get_last_value): Likewise.
10940 * convert.c (convert_to_integer_1): Likewise.
10941 * cse.c (cse_insn): Likewise.
10942 * expr.c (expand_expr_real_1): Likewise.
10943 * lra-constraints.c (simplify_operand_subreg): Likewise.
10944 * optabs-query.c (can_atomic_load_p): Likewise.
10945 * optabs.c (expand_atomic_load): Likewise.
10946 (expand_atomic_store): Likewise.
10947 * ree.c (combine_reaching_defs): Likewise.
10948 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
10949 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
10950 * tree.h (type_has_mode_precision_p): Likewise.
10951 * ubsan.c (instrument_si_overflow): Likewise.
10953 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10954 Alan Hayward <alan.hayward@arm.com>
10955 David Sherwood <david.sherwood@arm.com>
10957 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
10958 polynomial numbers of units.
10959 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
10960 (valid_vector_subparts_p): New function.
10961 (build_vector_type): Remove temporary shim and take the number
10962 of units as a poly_uint64 rather than an int.
10963 (build_opaque_vector_type): Take the number of units as a
10964 poly_uint64 rather than an int.
10965 * tree.c (build_vector_from_ctor): Handle polynomial
10966 TYPE_VECTOR_SUBPARTS.
10967 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
10968 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
10969 (build_vector_from_val): If the number of units is variable,
10970 use build_vec_duplicate_cst for constant operands and
10971 VEC_DUPLICATE_EXPR otherwise.
10972 (make_vector_type): Remove temporary is_constant ().
10973 (build_vector_type, build_opaque_vector_type): Take the number of
10974 units as a poly_uint64 rather than an int.
10975 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
10977 * cfgexpand.c (expand_debug_expr): Likewise.
10978 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
10979 (store_constructor, expand_expr_real_1): Likewise.
10980 (const_scalar_mask_from_tree): Likewise.
10981 * fold-const-call.c (fold_const_reduction): Likewise.
10982 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
10983 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
10984 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
10985 (fold_relational_const): Likewise.
10986 (native_interpret_vector): Likewise. Change the size from an
10987 int to an unsigned int.
10988 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
10989 TYPE_VECTOR_SUBPARTS.
10990 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
10991 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
10992 duplicating a non-constant operand into a variable-length vector.
10993 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
10994 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
10995 * ipa-icf.c (sem_variable::equals): Likewise.
10996 * match.pd: Likewise.
10997 * omp-simd-clone.c (simd_clone_subparts): Likewise.
10998 * print-tree.c (print_node): Likewise.
10999 * stor-layout.c (layout_type): Likewise.
11000 * targhooks.c (default_builtin_vectorization_cost): Likewise.
11001 * tree-cfg.c (verify_gimple_comparison): Likewise.
11002 (verify_gimple_assign_binary): Likewise.
11003 (verify_gimple_assign_ternary): Likewise.
11004 (verify_gimple_assign_single): Likewise.
11005 * tree-pretty-print.c (dump_generic_node): Likewise.
11006 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11007 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
11008 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
11009 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
11010 (vect_shift_permute_load_chain): Likewise.
11011 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
11012 (expand_vector_condition, optimize_vector_constructor): Likewise.
11013 (lower_vec_perm, get_compute_type): Likewise.
11014 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11015 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
11016 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
11017 (vect_recog_mask_conversion_pattern): Likewise.
11018 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
11019 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
11020 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11021 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
11022 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
11023 (vectorizable_shift, vectorizable_operation, vectorizable_store)
11024 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
11025 (supportable_widening_operation): Likewise.
11026 (supportable_narrowing_operation): Likewise.
11027 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
11029 * varasm.c (output_constant): Likewise.
11031 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11032 Alan Hayward <alan.hayward@arm.com>
11033 David Sherwood <david.sherwood@arm.com>
11035 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
11036 so that both the length == 3 and length != 3 cases set up their
11037 own permute vectors. Add comments explaining why we know the
11038 number of elements is constant.
11039 (vect_permute_load_chain): Likewise.
11041 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11042 Alan Hayward <alan.hayward@arm.com>
11043 David Sherwood <david.sherwood@arm.com>
11045 * machmode.h (mode_nunits): Change from unsigned char to
11047 (ONLY_FIXED_SIZE_MODES): New macro.
11048 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
11049 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
11050 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
11052 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
11053 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
11054 or if measurement_type is not polynomial.
11055 * genmodes.c (ZERO_COEFFS): New macro.
11056 (emit_mode_nunits_inline): Make mode_nunits_inline return a
11058 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
11059 Use ZERO_COEFFS when emitting initializers.
11060 * data-streamer.h (bp_pack_poly_value): New function.
11061 (bp_unpack_poly_value): Likewise.
11062 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11063 for GET_MODE_NUNITS.
11064 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11065 for GET_MODE_NUNITS.
11066 * tree.c (make_vector_type): Remove temporary shim and make
11067 the real function take the number of units as a poly_uint64
11068 rather than an int.
11069 (build_vector_type_for_mode): Handle polynomial nunits.
11070 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
11071 * emit-rtl.c (const_vec_series_p_1): Likewise.
11072 (gen_rtx_CONST_VECTOR): Likewise.
11073 * fold-const.c (test_vec_duplicate_folding): Likewise.
11074 * genrecog.c (validate_pattern): Likewise.
11075 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
11076 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
11077 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
11078 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
11079 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
11080 * rtlanal.c (subreg_get_info): Likewise.
11081 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11082 (vect_grouped_load_supported): Likewise.
11083 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
11084 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
11085 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
11086 (simplify_const_unary_operation, simplify_binary_operation_1)
11087 (simplify_const_binary_operation, simplify_ternary_operation)
11088 (test_vector_ops_duplicate, test_vector_ops): Likewise.
11089 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
11090 instead of CONST_VECTOR_NUNITS.
11091 * varasm.c (output_constant_pool_2): Likewise.
11092 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
11093 explicit-encoded elements in the XVEC for variable-length vectors.
11095 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11097 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
11099 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11100 Alan Hayward <alan.hayward@arm.com>
11101 David Sherwood <david.sherwood@arm.com>
11103 * coretypes.h (fixed_size_mode): Declare.
11104 (fixed_size_mode_pod): New typedef.
11105 * builtins.h (target_builtins::x_apply_args_mode)
11106 (target_builtins::x_apply_result_mode): Change type to
11107 fixed_size_mode_pod.
11108 * builtins.c (apply_args_size, apply_result_size, result_vector)
11109 (expand_builtin_apply_args_1, expand_builtin_apply)
11110 (expand_builtin_return): Update accordingly.
11112 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11114 * cse.c (hash_rtx_cb): Hash only the encoded elements.
11115 * cselib.c (cselib_hash_rtx): Likewise.
11116 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
11117 CONST_VECTOR encoding.
11119 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11120 Jeff Law <law@redhat.com>
11123 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
11124 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
11125 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
11126 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
11129 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
11130 explicitly probe *sp in a noreturn function if there were any callee
11131 register saves or frame pointer is needed.
11133 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11136 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
11137 BLKmode for ternary, binary or unary expressions.
11140 * var-tracking.c (delete_vta_debug_insn): New inline function.
11141 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
11142 insns from get_insns () to NULL instead of each bb separately.
11143 Use delete_vta_debug_insn. No longer static.
11144 (vt_debug_insns_local, variable_tracking_main_1): Adjust
11145 delete_vta_debug_insns callers.
11146 * rtl.h (delete_vta_debug_insns): Declare.
11147 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
11148 instead of variable_tracking_main.
11150 2018-01-03 Martin Sebor <msebor@redhat.com>
11152 PR tree-optimization/83603
11153 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
11154 arguments past the endof the argument list in functions declared
11155 without a prototype.
11156 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
11157 Avoid checking when arguments are null.
11159 2018-01-03 Martin Sebor <msebor@redhat.com>
11162 * doc/extend.texi (attribute const): Fix a typo.
11163 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
11164 issuing -Wsuggest-attribute for void functions.
11166 2018-01-03 Martin Sebor <msebor@redhat.com>
11168 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
11169 offset_int::from instead of wide_int::to_shwi.
11170 (maybe_diag_overlap): Remove assertion.
11171 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
11172 * gimple-ssa-sprintf.c (format_directive): Same.
11173 (parse_directive): Same.
11174 (sprintf_dom_walker::compute_format_length): Same.
11175 (try_substitute_return_value): Same.
11177 2018-01-03 Jeff Law <law@redhat.com>
11179 PR middle-end/83654
11180 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
11181 non-constant residual for zero at runtime and avoid probing in
11182 that case. Reorganize code for trailing problem to mirror handling
11185 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11187 PR tree-optimization/83501
11188 * tree-ssa-strlen.c (get_string_cst): New.
11189 (handle_char_store): Call get_string_cst.
11191 2018-01-03 Martin Liska <mliska@suse.cz>
11193 PR tree-optimization/83593
11194 * tree-ssa-strlen.c: Include tree-cfg.h.
11195 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
11196 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
11197 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
11199 (strlen_dom_walker::before_dom_children): Call
11200 gimple_purge_dead_eh_edges. Dump tranformation with details
11202 (strlen_dom_walker::before_dom_children): Update call by adding
11203 new argument cleanup_eh.
11204 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
11206 2018-01-03 Martin Liska <mliska@suse.cz>
11209 * cif-code.def (VARIADIC_THUNK): New enum value.
11210 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
11213 2018-01-03 Jan Beulich <jbeulich@suse.com>
11215 * sse.md (mov<mode>_internal): Tighten condition for when to use
11216 vmovdqu<ssescalarsize> for TI and OI modes.
11218 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11220 Update copyright years.
11222 2018-01-03 Martin Liska <mliska@suse.cz>
11225 * ipa-visibility.c (function_and_variable_visibility): Skip
11226 functions with noipa attribure.
11228 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11230 * gcc.c (process_command): Update copyright notice dates.
11231 * gcov-dump.c (print_version): Ditto.
11232 * gcov.c (print_version): Ditto.
11233 * gcov-tool.c (print_version): Ditto.
11234 * gengtype.c (create_file): Ditto.
11235 * doc/cpp.texi: Bump @copying's copyright year.
11236 * doc/cppinternals.texi: Ditto.
11237 * doc/gcc.texi: Ditto.
11238 * doc/gccint.texi: Ditto.
11239 * doc/gcov.texi: Ditto.
11240 * doc/install.texi: Ditto.
11241 * doc/invoke.texi: Ditto.
11243 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11245 * vector-builder.h (vector_builder::m_full_nelts): Change from
11246 unsigned int to poly_uint64.
11247 (vector_builder::full_nelts): Update prototype accordingly.
11248 (vector_builder::new_vector): Likewise.
11249 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
11250 (vector_builder::operator ==): Likewise.
11251 (vector_builder::finalize): Likewise.
11252 * int-vector-builder.h (int_vector_builder::int_vector_builder):
11253 Take the number of elements as a poly_uint64 rather than an
11255 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
11256 from unsigned int to poly_uint64.
11257 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
11258 (vec_perm_indices::new_vector): Likewise.
11259 (vec_perm_indices::length): Likewise.
11260 (vec_perm_indices::nelts_per_input): Likewise.
11261 (vec_perm_indices::input_nelts): Likewise.
11262 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
11263 number of elements per input as a poly_uint64 rather than an
11264 unsigned int. Use the original encoding for variable-length
11265 vectors, rather than clamping each individual element.
11266 For the second and subsequent elements in each pattern,
11267 clamp the step and base before clamping their sum.
11268 (vec_perm_indices::series_p): Handle polynomial element counts.
11269 (vec_perm_indices::all_in_range_p): Likewise.
11270 (vec_perm_indices_to_tree): Likewise.
11271 (vec_perm_indices_to_rtx): Likewise.
11272 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
11273 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
11274 (tree_vector_builder::new_binary_operation): Handle polynomial
11275 element counts. Return false if we need to know the number
11276 of elements at compile time.
11277 * fold-const.c (fold_vec_perm): Punt if the number of elements
11278 isn't known at compile time.
11280 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11282 * vec-perm-indices.h (vec_perm_builder): Change element type
11283 from HOST_WIDE_INT to poly_int64.
11284 (vec_perm_indices::element_type): Update accordingly.
11285 (vec_perm_indices::clamp): Handle polynomial element_types.
11286 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
11287 (vec_perm_indices::all_in_range_p): Likewise.
11288 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
11290 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
11291 polynomial vec_perm_indices element types.
11292 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
11293 * fold-const.c (fold_vec_perm): Likewise.
11294 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
11295 * tree-vect-generic.c (lower_vec_perm): Likewise.
11296 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
11297 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
11298 element type to HOST_WIDE_INT.
11300 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11301 Alan Hayward <alan.hayward@arm.com>
11302 David Sherwood <david.sherwood@arm.com>
11304 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
11305 rather than an int. Use plus_constant.
11306 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
11307 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
11309 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11310 Alan Hayward <alan.hayward@arm.com>
11311 David Sherwood <david.sherwood@arm.com>
11313 * calls.c (emit_call_1, expand_call): Change struct_value_size from
11314 a HOST_WIDE_INT to a poly_int64.
11316 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11317 Alan Hayward <alan.hayward@arm.com>
11318 David Sherwood <david.sherwood@arm.com>
11320 * calls.c (load_register_parameters): Cope with polynomial
11321 mode sizes. Require a constant size for BLKmode parameters
11322 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
11323 forces a parameter to be padded at the lsb end in order to
11324 fill a complete number of words, require the parameter size
11325 to be ordered wrt UNITS_PER_WORD.
11327 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11328 Alan Hayward <alan.hayward@arm.com>
11329 David Sherwood <david.sherwood@arm.com>
11331 * reload1.c (spill_stack_slot_width): Change element type
11332 from unsigned int to poly_uint64_pod.
11333 (alter_reg): Treat mode sizes as polynomial.
11335 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11336 Alan Hayward <alan.hayward@arm.com>
11337 David Sherwood <david.sherwood@arm.com>
11339 * reload.c (complex_word_subreg_p): New function.
11340 (reload_inner_reg_of_subreg, push_reload): Use it.
11342 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11343 Alan Hayward <alan.hayward@arm.com>
11344 David Sherwood <david.sherwood@arm.com>
11346 * lra-constraints.c (process_alt_operands): Reject matched
11347 operands whose sizes aren't ordered.
11348 (match_reload): Refer to this check here.
11350 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11351 Alan Hayward <alan.hayward@arm.com>
11352 David Sherwood <david.sherwood@arm.com>
11354 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
11355 that the mode size is in the set {1, 2, 4, 8, 16}.
11357 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11358 Alan Hayward <alan.hayward@arm.com>
11359 David Sherwood <david.sherwood@arm.com>
11361 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
11362 Use plus_constant instead of gen_rtx_PLUS.
11364 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11365 Alan Hayward <alan.hayward@arm.com>
11366 David Sherwood <david.sherwood@arm.com>
11368 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
11369 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
11370 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
11371 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
11372 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
11373 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
11374 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
11375 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
11376 * config/i386/i386.c (ix86_push_rounding): ...this new function.
11377 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
11379 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
11380 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
11381 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
11382 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
11383 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
11384 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
11385 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
11386 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
11387 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
11388 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
11390 * expr.c (emit_move_resolve_push): Treat the input and result
11391 of PUSH_ROUNDING as a poly_int64.
11392 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
11393 (emit_push_insn): Likewise.
11394 * lra-eliminations.c (mark_not_eliminable): Likewise.
11395 * recog.c (push_operand): Likewise.
11396 * reload1.c (elimination_effects): Likewise.
11397 * rtlanal.c (nonzero_bits1): Likewise.
11398 * calls.c (store_one_arg): Likewise. Require the padding to be
11399 known at compile time.
11401 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11402 Alan Hayward <alan.hayward@arm.com>
11403 David Sherwood <david.sherwood@arm.com>
11405 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
11406 Use plus_constant instead of gen_rtx_PLUS.
11408 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11409 Alan Hayward <alan.hayward@arm.com>
11410 David Sherwood <david.sherwood@arm.com>
11412 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
11413 rather than an int.
11415 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11416 Alan Hayward <alan.hayward@arm.com>
11417 David Sherwood <david.sherwood@arm.com>
11419 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
11420 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
11421 via stack temporaries. Treat the mode size as polynomial too.
11423 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11424 Alan Hayward <alan.hayward@arm.com>
11425 David Sherwood <david.sherwood@arm.com>
11427 * expr.c (expand_expr_real_2): When handling conversions involving
11428 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
11429 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
11430 as a poly_uint64 too.
11432 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11433 Alan Hayward <alan.hayward@arm.com>
11434 David Sherwood <david.sherwood@arm.com>
11436 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
11438 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11439 Alan Hayward <alan.hayward@arm.com>
11440 David Sherwood <david.sherwood@arm.com>
11442 * combine.c (can_change_dest_mode): Handle polynomial
11443 REGMODE_NATURAL_SIZE.
11444 * expmed.c (store_bit_field_1): Likewise.
11445 * expr.c (store_constructor): Likewise.
11446 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
11447 and polynomial REGMODE_NATURAL_SIZE.
11448 (gen_lowpart_common): Likewise.
11449 * reginfo.c (record_subregs_of_mode): Likewise.
11450 * rtlanal.c (read_modify_subreg_p): Likewise.
11452 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11453 Alan Hayward <alan.hayward@arm.com>
11454 David Sherwood <david.sherwood@arm.com>
11456 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
11457 numbers of elements.
11459 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11460 Alan Hayward <alan.hayward@arm.com>
11461 David Sherwood <david.sherwood@arm.com>
11463 * match.pd: Cope with polynomial numbers of vector elements.
11465 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11466 Alan Hayward <alan.hayward@arm.com>
11467 David Sherwood <david.sherwood@arm.com>
11469 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
11470 in a POINTER_PLUS_EXPR.
11472 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11473 Alan Hayward <alan.hayward@arm.com>
11474 David Sherwood <david.sherwood@arm.com>
11476 * omp-simd-clone.c (simd_clone_subparts): New function.
11477 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
11478 (ipa_simd_modify_function_body): Likewise.
11480 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11481 Alan Hayward <alan.hayward@arm.com>
11482 David Sherwood <david.sherwood@arm.com>
11484 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
11485 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
11486 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
11487 (expand_vector_condition, vector_element): Likewise.
11488 (subparts_gt): New function.
11489 (get_compute_type): Use subparts_gt.
11490 (count_type_subparts): Delete.
11491 (expand_vector_operations_1): Use subparts_gt instead of
11492 count_type_subparts.
11494 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11495 Alan Hayward <alan.hayward@arm.com>
11496 David Sherwood <david.sherwood@arm.com>
11498 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
11499 (vect_compile_time_alias): ...this new function. Do the calculation
11500 on poly_ints rather than trees.
11501 (vect_prune_runtime_alias_test_list): Update call accordingly.
11503 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11504 Alan Hayward <alan.hayward@arm.com>
11505 David Sherwood <david.sherwood@arm.com>
11507 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
11509 (vect_schedule_slp_instance): Likewise.
11511 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11512 Alan Hayward <alan.hayward@arm.com>
11513 David Sherwood <david.sherwood@arm.com>
11515 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
11516 constant and extern definitions for variable-length vectors.
11517 (vect_get_constant_vectors): Note that the number of units
11518 is known to be constant.
11520 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11521 Alan Hayward <alan.hayward@arm.com>
11522 David Sherwood <david.sherwood@arm.com>
11524 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
11525 of units as polynomial. Choose between WIDE and NARROW based
11528 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11529 Alan Hayward <alan.hayward@arm.com>
11530 David Sherwood <david.sherwood@arm.com>
11532 * tree-vect-stmts.c (simd_clone_subparts): New function.
11533 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
11535 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11536 Alan Hayward <alan.hayward@arm.com>
11537 David Sherwood <david.sherwood@arm.com>
11539 * tree-vect-stmts.c (vectorizable_call): Treat the number of
11540 vectors as polynomial. Use build_index_vector for
11541 IFN_GOMP_SIMD_LANE.
11543 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11544 Alan Hayward <alan.hayward@arm.com>
11545 David Sherwood <david.sherwood@arm.com>
11547 * tree-vect-stmts.c (get_load_store_type): Treat the number of
11548 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
11549 for variable-length vectors.
11550 (vectorizable_mask_load_store): Treat the number of units as
11551 polynomial, asserting that it is constant if the condition has
11552 already been enforced.
11553 (vectorizable_store, vectorizable_load): Likewise.
11555 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11556 Alan Hayward <alan.hayward@arm.com>
11557 David Sherwood <david.sherwood@arm.com>
11559 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
11560 of units as polynomial. Punt if we can't tell at compile time
11561 which vector contains the final result.
11563 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11564 Alan Hayward <alan.hayward@arm.com>
11565 David Sherwood <david.sherwood@arm.com>
11567 * tree-vect-loop.c (vectorizable_induction): Treat the number
11568 of units as polynomial. Punt on SLP inductions. Use an integer
11569 VEC_SERIES_EXPR for variable-length integer reductions. Use a
11570 cast of such a series for variable-length floating-point
11573 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11574 Alan Hayward <alan.hayward@arm.com>
11575 David Sherwood <david.sherwood@arm.com>
11577 * tree.h (build_index_vector): Declare.
11578 * tree.c (build_index_vector): New function.
11579 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
11580 of units as polynomial, forcibly converting it to a constant if
11581 vectorizable_reduction has already enforced the condition.
11582 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
11583 to create a {1,2,3,...} vector.
11584 (vectorizable_reduction): Treat the number of units as polynomial.
11585 Choose vectype_in based on the largest scalar element size rather
11586 than the smallest number of units. Enforce the restrictions
11589 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11590 Alan Hayward <alan.hayward@arm.com>
11591 David Sherwood <david.sherwood@arm.com>
11593 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
11594 number of units as polynomial.
11596 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11597 Alan Hayward <alan.hayward@arm.com>
11598 David Sherwood <david.sherwood@arm.com>
11600 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
11601 * target.def (autovectorize_vector_sizes): Return the vector sizes
11602 by pointer, using vector_sizes rather than a bitmask.
11603 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
11604 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
11605 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
11607 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
11608 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
11609 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
11610 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
11611 * omp-general.c (omp_max_vf): Likewise.
11612 * omp-low.c (omp_clause_aligned_alignment): Likewise.
11613 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
11614 * tree-vect-loop.c (vect_analyze_loop): Likewise.
11615 * tree-vect-slp.c (vect_slp_bb): Likewise.
11616 * doc/tm.texi: Regenerate.
11617 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
11619 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
11620 the vector size as a poly_uint64 rather than an unsigned int.
11621 (current_vector_size): Change from an unsigned int to a poly_uint64.
11622 (get_vectype_for_scalar_type): Update accordingly.
11623 * tree.h (build_truth_vector_type): Take the size and number of
11624 units as a poly_uint64 rather than an unsigned int.
11625 (build_vector_type): Add a temporary overload that takes
11626 the number of units as a poly_uint64 rather than an unsigned int.
11627 * tree.c (make_vector_type): Likewise.
11628 (build_truth_vector_type): Take the number of units as a poly_uint64
11629 rather than an unsigned int.
11631 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11632 Alan Hayward <alan.hayward@arm.com>
11633 David Sherwood <david.sherwood@arm.com>
11635 * target.def (get_mask_mode): Take the number of units and length
11636 as poly_uint64s rather than unsigned ints.
11637 * targhooks.h (default_get_mask_mode): Update accordingly.
11638 * targhooks.c (default_get_mask_mode): Likewise.
11639 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
11640 * doc/tm.texi: Regenerate.
11642 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11643 Alan Hayward <alan.hayward@arm.com>
11644 David Sherwood <david.sherwood@arm.com>
11646 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
11647 * omp-general.c (omp_max_vf): Likewise.
11648 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
11649 (expand_omp_simd): Handle polynomial safelen.
11650 * omp-low.c (omplow_simd_context): Add a default constructor.
11651 (omplow_simd_context::max_vf): Change from int to poly_uint64.
11652 (lower_rec_simd_input_clauses): Update accordingly.
11653 (lower_rec_input_clauses): Likewise.
11655 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11656 Alan Hayward <alan.hayward@arm.com>
11657 David Sherwood <david.sherwood@arm.com>
11659 * tree-vectorizer.h (vect_nunits_for_cost): New function.
11660 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
11661 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
11662 (vect_analyze_slp_cost): Likewise.
11663 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
11664 (vect_model_load_cost): Likewise.
11666 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11667 Alan Hayward <alan.hayward@arm.com>
11668 David Sherwood <david.sherwood@arm.com>
11670 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
11671 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
11672 from an unsigned int * to a poly_uint64_pod *.
11673 (calculate_unrolling_factor): New function.
11674 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
11676 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11677 Alan Hayward <alan.hayward@arm.com>
11678 David Sherwood <david.sherwood@arm.com>
11680 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
11681 from an unsigned int to a poly_uint64.
11682 (_loop_vec_info::slp_unrolling_factor): Likewise.
11683 (_loop_vec_info::vectorization_factor): Change from an int
11685 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
11686 (vect_get_num_vectors): New function.
11687 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
11688 (vect_get_num_copies): Use vect_get_num_vectors.
11689 (vect_analyze_data_ref_dependences): Change max_vf from an int *
11690 to an unsigned int *.
11691 (vect_analyze_data_refs): Change min_vf from an int * to a
11693 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
11694 than an unsigned HOST_WIDE_INT.
11695 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
11696 (vect_analyze_data_ref_dependence): Change max_vf from an int *
11697 to an unsigned int *.
11698 (vect_analyze_data_ref_dependences): Likewise.
11699 (vect_compute_data_ref_alignment): Handle polynomial vf.
11700 (vect_enhance_data_refs_alignment): Likewise.
11701 (vect_prune_runtime_alias_test_list): Likewise.
11702 (vect_shift_permute_load_chain): Likewise.
11703 (vect_supportable_dr_alignment): Likewise.
11704 (dependence_distance_ge_vf): Take the vectorization factor as a
11705 poly_uint64 rather than an unsigned HOST_WIDE_INT.
11706 (vect_analyze_data_refs): Change min_vf from an int * to a
11708 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
11709 vfm1 as a poly_uint64 rather than an int. Make the same change
11710 for the returned bound_scalar.
11711 (vect_gen_vector_loop_niters): Handle polynomial vf.
11712 (vect_do_peeling): Likewise. Update call to
11713 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
11714 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
11716 * tree-vect-loop.c (vect_determine_vectorization_factor)
11717 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
11718 (vect_get_known_peeling_cost): Likewise.
11719 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
11720 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
11721 (vect_transform_loop): Likewise. Use the lowest possible VF when
11722 updating the upper bounds of the loop.
11723 (vect_min_worthwhile_factor): Make static. Return an unsigned int
11724 rather than an int.
11725 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
11726 polynomial unroll factors.
11727 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
11728 (vect_make_slp_decision): Likewise.
11729 (vect_supported_load_permutation_p): Likewise, and polynomial
11731 (vect_analyze_slp_cost): Handle polynomial vf.
11732 (vect_slp_analyze_node_operations): Likewise.
11733 (vect_slp_analyze_bb_1): Likewise.
11734 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
11735 than an unsigned HOST_WIDE_INT.
11736 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
11737 (vectorizable_load): Handle polynomial vf.
11738 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
11740 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
11742 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11743 Alan Hayward <alan.hayward@arm.com>
11744 David Sherwood <david.sherwood@arm.com>
11746 * match.pd: Handle bit operations involving three constants
11747 and try to fold one pair.
11749 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11751 * tree-vect-loop-manip.c: Include gimple-fold.h.
11752 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
11753 niters_maybe_zero parameters. Handle other cases besides a step of 1.
11754 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
11755 Add a path that uses a step of VF instead of 1, but disable it
11757 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
11758 and niters_no_overflow parameters. Update calls to
11759 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
11760 Create a new SSA name if the latter choses to use a ste other
11761 than zero, and return it via niters_vector_mult_vf_var.
11762 * tree-vect-loop.c (vect_transform_loop): Update calls to
11763 vect_do_peeling, vect_gen_vector_loop_niters and
11764 slpeel_make_loop_iterate_ntimes.
11765 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
11766 (vect_gen_vector_loop_niters): Update declarations after above changes.
11768 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
11770 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
11771 128-bit round to integer instructions.
11772 (ceil<mode>2): Likewise.
11773 (btrunc<mode>2): Likewise.
11774 (round<mode>2): Likewise.
11776 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11778 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
11779 unaligned VSX load/store on P8/P9.
11780 (expand_block_clear): Allow the use of unaligned VSX
11781 load/store on P8/P9.
11783 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11785 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
11787 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
11788 swap associated with both a load and a store.
11790 2018-01-02 Andrew Waterman <andrew@sifive.com>
11792 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
11793 * config/riscv/riscv.md (clear_cache): Use it.
11795 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
11797 * web.c: Remove out-of-date comment.
11799 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11801 * expr.c (fixup_args_size_notes): Check that any existing
11802 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
11803 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
11804 (emit_single_push_insn): ...here.
11806 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11808 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
11809 (const_vector_encoded_nelts): New function.
11810 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
11811 (const_vector_int_elt, const_vector_elt): Declare.
11812 * emit-rtl.c (const_vector_int_elt_1): New function.
11813 (const_vector_elt): Likewise.
11814 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
11815 of CONST_VECTOR_ELT.
11817 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11819 * expr.c: Include rtx-vector-builder.h.
11820 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
11821 directly on the tree encoding.
11822 (const_vector_from_tree): Likewise.
11823 * optabs.c: Include rtx-vector-builder.h.
11824 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
11825 sequence of "u" values.
11826 * vec-perm-indices.c: Include rtx-vector-builder.h.
11827 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
11828 directly on the vec_perm_indices encoding.
11830 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11832 * doc/rtl.texi (const_vector): Describe new encoding scheme.
11833 * Makefile.in (OBJS): Add rtx-vector-builder.o.
11834 * rtx-vector-builder.h: New file.
11835 * rtx-vector-builder.c: Likewise.
11836 * rtl.h (rtx_def::u2): Add a const_vector field.
11837 (CONST_VECTOR_NPATTERNS): New macro.
11838 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
11839 (CONST_VECTOR_DUPLICATE_P): Likewise.
11840 (CONST_VECTOR_STEPPED_P): Likewise.
11841 (CONST_VECTOR_ENCODED_ELT): Likewise.
11842 (const_vec_duplicate_p): Check for a duplicated vector encoding.
11843 (unwrap_const_vec_duplicate): Likewise.
11844 (const_vec_series_p): Check for a non-duplicated vector encoding.
11845 Say that the function only returns true for integer vectors.
11846 * emit-rtl.c: Include rtx-vector-builder.h.
11847 (gen_const_vec_duplicate_1): Delete.
11848 (gen_const_vector): Call gen_const_vec_duplicate instead of
11849 gen_const_vec_duplicate_1.
11850 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
11851 (gen_const_vec_duplicate): Use rtx_vector_builder.
11852 (gen_const_vec_series): Likewise.
11853 (gen_rtx_CONST_VECTOR): Likewise.
11854 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
11855 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
11856 Build a new vector rather than modifying a CONST_VECTOR in-place.
11857 (handle_special_swappables): Update call accordingly.
11858 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
11859 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
11860 Build a new vector rather than modifying a CONST_VECTOR in-place.
11861 (handle_special_swappables): Update call accordingly.
11863 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11865 * simplify-rtx.c (simplify_const_binary_operation): Use
11866 CONST_VECTOR_ELT instead of XVECEXP.
11868 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11870 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
11871 the selector elements to be different from the data elements
11872 if the selector is a VECTOR_CST.
11873 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
11874 ssizetype for the selector.
11876 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11878 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
11879 before testing each element individually.
11880 * tree-vect-generic.c (lower_vec_perm): Likewise.
11882 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11884 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
11885 * selftest-run-tests.c (selftest::run_tests): Call it.
11886 * vector-builder.h (vector_builder::operator ==): New function.
11887 (vector_builder::operator !=): Likewise.
11888 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
11889 (vec_perm_indices::all_from_input_p): New function.
11890 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
11891 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
11892 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
11893 instead of reading the VECTOR_CST directly. Detect whether both
11894 vector inputs are the same before constructing the vec_perm_indices,
11895 and update the number of inputs argument accordingly. Use the
11896 utility functions added above. Only construct sel2 if we need to.
11898 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11900 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
11901 the broadcast of the low byte.
11902 (expand_mult_highpart): Use an explicit encoding for the permutes.
11903 * optabs-query.c (can_mult_highpart_p): Likewise.
11904 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
11905 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11906 (vectorizable_bswap): Likewise.
11907 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
11908 explicit encoding for the power-of-2 permutes.
11909 (vect_permute_store_chain): Likewise.
11910 (vect_grouped_load_supported): Likewise.
11911 (vect_permute_load_chain): Likewise.
11913 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11915 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
11916 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
11917 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
11918 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
11919 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
11920 (vect_gen_perm_mask_any): Likewise.
11922 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11924 * int-vector-builder.h: New file.
11925 * vec-perm-indices.h: Include int-vector-builder.h.
11926 (vec_perm_indices): Redefine as an int_vector_builder.
11927 (auto_vec_perm_indices): Delete.
11928 (vec_perm_builder): Redefine as a stand-alone class.
11929 (vec_perm_indices::vec_perm_indices): New function.
11930 (vec_perm_indices::clamp): Likewise.
11931 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
11932 (vec_perm_indices::new_vector): New function.
11933 (vec_perm_indices::new_expanded_vector): Update for new
11934 vec_perm_indices class.
11935 (vec_perm_indices::rotate_inputs): New function.
11936 (vec_perm_indices::all_in_range_p): Operate directly on the
11937 encoded form, without computing elided elements.
11938 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
11939 encoding. Update for new vec_perm_indices class.
11940 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
11941 the given vec_perm_builder.
11942 (expand_vec_perm_var): Update vec_perm_builder constructor.
11943 (expand_mult_highpart): Use vec_perm_builder instead of
11944 auto_vec_perm_indices.
11945 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
11946 vec_perm_indices instead of auto_vec_perm_indices. Use a single
11947 or double series encoding as appropriate.
11948 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
11949 vec_perm_indices instead of auto_vec_perm_indices.
11950 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11951 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11952 (vect_permute_store_chain): Likewise.
11953 (vect_grouped_load_supported): Likewise.
11954 (vect_permute_load_chain): Likewise.
11955 (vect_shift_permute_load_chain): Likewise.
11956 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
11957 (vect_transform_slp_perm_load): Likewise.
11958 (vect_schedule_slp_instance): Likewise.
11959 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11960 (vectorizable_mask_load_store): Likewise.
11961 (vectorizable_bswap): Likewise.
11962 (vectorizable_store): Likewise.
11963 (vectorizable_load): Likewise.
11964 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
11965 vec_perm_indices instead of auto_vec_perm_indices. Use
11966 tree_to_vec_perm_builder to read the vector from a tree.
11967 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
11968 vec_perm_builder instead of a vec_perm_indices.
11969 (have_whole_vector_shift): Use vec_perm_builder and
11970 vec_perm_indices instead of auto_vec_perm_indices. Leave the
11971 truncation to calc_vec_perm_mask_for_shift.
11972 (vect_create_epilog_for_reduction): Likewise.
11973 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
11974 from auto_vec_perm_indices to vec_perm_indices.
11975 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
11976 instead of changing individual elements.
11977 (aarch64_vectorize_vec_perm_const): Use new_vector to install
11978 the vector in d.perm.
11979 * config/arm/arm.c (expand_vec_perm_d::perm): Change
11980 from auto_vec_perm_indices to vec_perm_indices.
11981 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
11982 instead of changing individual elements.
11983 (arm_vectorize_vec_perm_const): Use new_vector to install
11984 the vector in d.perm.
11985 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
11986 Update vec_perm_builder constructor.
11987 (rs6000_expand_interleave): Likewise.
11988 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
11989 (rs6000_expand_interleave): Likewise.
11991 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11993 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
11994 to qimode could truncate the indices.
11995 * optabs.c (expand_vec_perm_var): Likewise.
11997 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11999 * Makefile.in (OBJS): Add vec-perm-indices.o.
12000 * vec-perm-indices.h: New file.
12001 * vec-perm-indices.c: Likewise.
12002 * target.h (vec_perm_indices): Replace with a forward class
12004 (auto_vec_perm_indices): Move to vec-perm-indices.h.
12005 * optabs.h: Include vec-perm-indices.h.
12006 (expand_vec_perm): Delete.
12007 (selector_fits_mode_p, expand_vec_perm_var): Declare.
12008 (expand_vec_perm_const): Declare.
12009 * target.def (vec_perm_const_ok): Replace with...
12010 (vec_perm_const): ...this new hook.
12011 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
12012 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
12013 * doc/tm.texi: Regenerate.
12014 * optabs.def (vec_perm_const): Delete.
12015 * doc/md.texi (vec_perm_const): Likewise.
12016 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
12017 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
12018 expand_vec_perm for constant permutation vectors. Assert that
12019 the mode of variable permutation vectors is the integer equivalent
12020 of the mode that is being permuted.
12021 * optabs-query.h (selector_fits_mode_p): Declare.
12022 * optabs-query.c: Include vec-perm-indices.h.
12023 (selector_fits_mode_p): New function.
12024 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
12025 is defined, instead of checking whether the vec_perm_const_optab
12026 exists. Use targetm.vectorize.vec_perm_const instead of
12027 targetm.vectorize.vec_perm_const_ok. Check whether the indices
12028 fit in the vector mode before using a variable permute.
12029 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
12030 vec_perm_indices instead of an rtx.
12031 (expand_vec_perm): Replace with...
12032 (expand_vec_perm_const): ...this new function. Take the selector
12033 as a vec_perm_indices rather than an rtx. Also take the mode of
12034 the selector. Update call to shift_amt_for_vec_perm_mask.
12035 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
12036 Use vec_perm_indices::new_expanded_vector to expand the original
12037 selector into bytes. Check whether the indices fit in the vector
12038 mode before using a variable permute.
12039 (expand_vec_perm_var): Make global.
12040 (expand_mult_highpart): Use expand_vec_perm_const.
12041 * fold-const.c: Includes vec-perm-indices.h.
12042 * tree-ssa-forwprop.c: Likewise.
12043 * tree-vect-data-refs.c: Likewise.
12044 * tree-vect-generic.c: Likewise.
12045 * tree-vect-loop.c: Likewise.
12046 * tree-vect-slp.c: Likewise.
12047 * tree-vect-stmts.c: Likewise.
12048 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
12050 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
12051 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
12052 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
12053 (aarch64_vectorize_vec_perm_const): ...this new function.
12054 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12055 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12056 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
12057 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
12058 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12059 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12060 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
12062 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
12063 check for NEON modes.
12064 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
12065 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
12066 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
12067 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
12069 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
12070 the old VEC_PERM_CONST conditions.
12071 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
12072 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
12073 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
12074 (ia64_vectorize_vec_perm_const_ok): Merge into...
12075 (ia64_vectorize_vec_perm_const): ...this new function.
12076 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
12077 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
12078 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
12079 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
12080 * config/mips/mips.c (mips_expand_vec_perm_const)
12081 (mips_vectorize_vec_perm_const_ok): Merge into...
12082 (mips_vectorize_vec_perm_const): ...this new function.
12083 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
12084 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
12085 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
12086 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
12087 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
12088 (rs6000_expand_vec_perm_const): Delete.
12089 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
12091 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12092 (altivec_expand_vec_perm_const_le): Take each operand individually.
12093 Operate on constant selectors rather than rtxes.
12094 (altivec_expand_vec_perm_const): Likewise. Update call to
12095 altivec_expand_vec_perm_const_le.
12096 (rs6000_expand_vec_perm_const): Delete.
12097 (rs6000_vectorize_vec_perm_const_ok): Delete.
12098 (rs6000_vectorize_vec_perm_const): New function.
12099 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
12100 an element count and rtx array.
12101 (rs6000_expand_extract_even): Update call accordingly.
12102 (rs6000_expand_interleave): Likewise.
12103 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
12104 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
12105 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
12106 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
12107 (rs6000_expand_vec_perm_const): Delete.
12108 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12109 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12110 (altivec_expand_vec_perm_const_le): Take each operand individually.
12111 Operate on constant selectors rather than rtxes.
12112 (altivec_expand_vec_perm_const): Likewise. Update call to
12113 altivec_expand_vec_perm_const_le.
12114 (rs6000_expand_vec_perm_const): Delete.
12115 (rs6000_vectorize_vec_perm_const_ok): Delete.
12116 (rs6000_vectorize_vec_perm_const): New function. Remove stray
12117 reference to the SPE evmerge intructions.
12118 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
12119 an element count and rtx array.
12120 (rs6000_expand_extract_even): Update call accordingly.
12121 (rs6000_expand_interleave): Likewise.
12122 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
12123 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
12125 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12127 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12129 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
12130 vector mode and that that mode matches the mode of the data
12132 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
12133 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
12134 directly using expand_vec_perm_1 when forcing selectors into
12136 (expand_vec_perm_var): New function, split out from expand_vec_perm.
12138 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12140 * optabs-query.h (can_vec_perm_p): Delete.
12141 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
12142 * optabs-query.c (can_vec_perm_p): Split into...
12143 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
12144 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
12145 particular selector is valid.
12146 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12147 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12148 (vect_grouped_load_supported): Likewise.
12149 (vect_shift_permute_load_chain): Likewise.
12150 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
12151 (vect_transform_slp_perm_load): Likewise.
12152 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12153 (vectorizable_bswap): Likewise.
12154 (vect_gen_perm_mask_checked): Likewise.
12155 * fold-const.c (fold_ternary_loc): Likewise. Don't take
12156 implementations of variable permutation vectors into account
12157 when deciding which selector to use.
12158 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
12159 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
12160 with a false third argument.
12161 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
12162 to test whether the constant selector is valid and can_vec_perm_var_p
12163 to test whether a variable selector is valid.
12165 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12167 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
12168 * optabs-query.c (can_vec_perm_p): Likewise.
12169 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
12170 instead of vec_perm_indices.
12171 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
12172 (vect_gen_perm_mask_checked): Likewise,
12173 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
12174 (vect_gen_perm_mask_checked): Likewise,
12176 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12178 * optabs-query.h (qimode_for_vec_perm): Declare.
12179 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
12180 (qimode_for_vec_perm): ...this new function.
12181 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
12183 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12185 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
12186 does not have a conditional at the top.
12188 2018-01-02 Richard Biener <rguenther@suse.de>
12190 * ipa-inline.c (big_speedup_p): Fix expression.
12192 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
12195 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
12198 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
12202 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
12203 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
12204 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
12205 cond_taken_branch_cost 3->4.
12207 2018-01-01 Jakub Jelinek <jakub@redhat.com>
12209 PR tree-optimization/83581
12210 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
12211 TODO_cleanup_cfg if any changes have been made.
12213 PR middle-end/83608
12214 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
12215 convert_modes if target mode has the right side, but different mode
12218 PR middle-end/83609
12219 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
12220 last argument when extracting from CONCAT. If either from_real or
12221 from_imag is NULL, use expansion through memory. If result is not
12222 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
12223 the parts directly to inner mode, if even that fails, use expansion
12226 PR middle-end/83623
12227 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
12228 check for bswap in mode rather than HImode and use that in expand_unop
12231 Copyright (C) 2018 Free Software Foundation, Inc.
12233 Copying and distribution of this file, with or without modification,
12234 are permitted in any medium without royalty provided the copyright
12235 notice and this notice are preserved.