1 2018-05-07 Nathan Sidwell <nathan@acm.org>
3 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
4 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
5 (Backwards Compatibility): Likewise.
7 2018-05-07 Luis Machado <luis.machado@linaro.org>
11 2018-05-07 Luis Machado <luis.machado@linaro.org>
13 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
14 <prefetch_dynamic_strides>: New const bool field.
15 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
16 prefetch_dynamic_strides.
17 (exynosm1_prefetch_tune): Likewise.
18 (thunderxt88_prefetch_tune): Likewise.
19 (thunderx_prefetch_tune): Likewise.
20 (thunderx2t99_prefetch_tune): Likewise.
21 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
22 (aarch64_override_options_internal): Update to set
23 PARAM_PREFETCH_DYNAMIC_STRIDES.
24 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
25 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
26 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
27 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
28 prefetch-dynamic-strides setting.
30 2018-05-07 Luis Machado <luis.machado@linaro.org>
32 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
33 <minimum_stride>: New const int field.
34 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
36 (exynosm1_prefetch_tune): Likewise.
37 (thunderxt88_prefetch_tune): Likewise.
38 (thunderx_prefetch_tune): Likewise.
39 (thunderx2t99_prefetch_tune): Likewise.
40 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
41 (aarch64_override_options_internal): Update to set
42 PARAM_PREFETCH_MINIMUM_STRIDE.
43 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
44 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
45 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
46 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
47 stride is constant and is below the minimum stride threshold.
49 2018-05-07 Luis Machado <luis.machado@linaro.org>
51 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
54 2018-05-07 Luis Machado <luis.machado@linaro.org>
56 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
57 <prefetch_dynamic_strides>: New const bool field.
58 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
59 prefetch_dynamic_strides.
60 (exynosm1_prefetch_tune): Likewise.
61 (thunderxt88_prefetch_tune): Likewise.
62 (thunderx_prefetch_tune): Likewise.
63 (thunderx2t99_prefetch_tune): Likewise.
64 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
65 (aarch64_override_options_internal): Update to set
66 PARAM_PREFETCH_DYNAMIC_STRIDES.
67 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
68 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
69 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
70 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
71 prefetch-dynamic-strides setting.
73 2018-05-07 Luis Machado <luis.machado@linaro.org>
75 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
76 <minimum_stride>: New const int field.
77 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
79 (exynosm1_prefetch_tune): Likewise.
80 (thunderxt88_prefetch_tune): Likewise.
81 (thunderx_prefetch_tune): Likewise.
82 (thunderx2t99_prefetch_tune): Likewise.
83 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
84 (aarch64_override_options_internal): Update to set
85 PARAM_PREFETCH_MINIMUM_STRIDE.
86 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
87 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
88 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
89 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
90 stride is constant and is below the minimum stride threshold.
92 2018-05-06 Jakub Jelinek <jakub@redhat.com>
95 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
96 the type is addressable. Don't force op into register if it has
99 2018-05-05 Roland McGrath <mcgrathr@google.com>
102 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
103 any section for which we don't know a specific type it should have,
104 regardless of name. Previously this was done only for the exact
105 names ".init_array", ".fini_array", and ".preinit_array".
106 (default_elf_asm_named_section): Add comment about
107 relationship with default_section_type_flags and SECTION_NOTYPE.
108 (get_section): Don't consider it a type conflict if one side has
109 SECTION_NOTYPE and the other doesn't, as long as neither has the
110 SECTION_BSS et al used in the default_section_type_flags logic.
112 2018-05-05 Tom de Vries <tom@codesourcery.com>
115 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
116 (workaround_barsyncs): New function.
117 (nvptx_reorg): Use workaround_barsyncs.
118 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
119 (define_expand "nvptx_membar_cta"): New define_expand.
120 (define_insn "*nvptx_membar_cta"): New insn.
122 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
124 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
125 To improve optimization opportunities.
126 * builtin-types.def: The new needed builtin types for the above.
128 2018-05-04 Richard Biener <rguenther@suse.de>
130 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
131 * gimple-ssa-store-merging.c
132 (imm_store_chain_info::output_merged_store): Remove redundant create,
133 release split_store vector contents on failure.
134 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
135 scalar stmt vector on cache hit.
137 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
139 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
141 * config.gcc (powerpc-xilinx-eabi*): Remove.
142 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
144 (fusion_addis_mem_combo_load): Ditto.
145 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
147 (rs6000_cpu_cpp_builtins): Ditto.
148 * config/rs6000/rs6000-linux.c
149 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
150 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
151 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
153 (rs6000_setup_reg_addr_masks): Ditto.
154 (rs6000_init_hard_regno_mode_ok): Ditto.
155 (rs6000_option_override_internal): Ditto.
156 (legitimate_lo_sum_address_p): Ditto.
157 (rs6000_legitimize_address): Ditto.
158 (rs6000_legitimize_reload_address): Ditto.
159 (rs6000_legitimate_address_p): Ditto.
160 (abi_v4_pass_in_fpr): Ditto.
161 (setup_incoming_varargs): Ditto.
162 (rs6000_gimplify_va_arg): Ditto.
163 (rs6000_split_multireg_move): Ditto.
164 (rs6000_savres_strategy): Ditto.
165 (rs6000_emit_prologue_components): Ditto.
166 (rs6000_emit_epilogue_components): Ditto.
167 (rs6000_emit_prologue): Ditto.
168 (rs6000_emit_epilogue): Ditto.
169 (rs6000_elf_file_end): Ditto.
170 (rs6000_function_value): Ditto.
171 (rs6000_libcall_value): Ditto.
172 * config/rs6000/rs6000.h: Ditto.
173 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
174 (TARGET_MINMAX): ... this. New.
175 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
176 * config/rs6000/rs6000.md: Remove Xilinx FP support.
177 (*movsi_internal1_single): Delete.
178 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
179 mfpu=, mxilinx-fpu): Delete.
180 * config/rs6000/singlefp.h: Delete.
181 * config/rs6000/sysv4.h: Remove Xilinx FP support.
182 * config/rs6000/t-rs6000: Ditto.
183 * config/rs6000/t-xilinx: Delete.
184 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
185 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
186 (VStype_simple): Delete.
187 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
188 * config/rs6000/xfpu.h: Delete.
189 * config/rs6000/xfpu.md: Delete.
190 * config/rs6000/xilinx.h: Delete.
191 * config/rs6000/xilinx.opt: Delete.
192 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
193 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
195 2018-05-04 Tom de Vries <tom@codesourcery.com>
198 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
201 2018-05-04 Richard Biener <rguenther@suse.de>
204 * tree-complex.c (update_complex_assignment): We are always in SSA form.
205 (expand_complex_div_wide): Likewise.
206 (expand_complex_operations_1): Likewise.
207 (expand_complex_libcall): Preserve EH info of the original stmt.
208 (tree_lower_complex): Handle removed blocks.
209 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
210 on complex multiplication and division libcall builtins.
212 2018-05-04 Richard Biener <rguenther@suse.de>
215 * fold-const.c (negate_expr_p): Restrict negation of operand
216 zero of a division to when we know that can happen without
218 (fold_negate_expr_1): Likewise.
220 2018-05-04 Jakub Jelinek <jakub@redhat.com>
223 * real.h (real_nextafter): Declare.
224 * real.c (real_nextafter): New function.
225 * fold-const-call.c (fold_const_nextafter): New function.
226 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
228 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
229 even when arg1_mode is different from arg0_mode.
231 2018-05-03 Nathan Sidwell <nathan@acm.org>
233 * doc/extend.texi (Deprecated Features): Remove
235 (Backwards Compatibility): Likewise.
236 * doc/invoke.texi (C++ Language Options): Likewise.
237 (C++ Dialect Options): Likewise.
239 2018-05-03 Jakub Jelinek <jakub@redhat.com>
242 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
243 _mm512_mask_mullox_epi64): New intrinsics.
245 2018-05-03 Tom de Vries <tom@codesourcery.com>
248 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
249 dump files): Add offload-tree.
251 2018-05-03 Richard Biener <rguenther@suse.de>
253 PR tree-optimization/85615
254 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
255 to loops not nested in BBs loop father to avoid creating multi-entry
258 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
260 PR tree-optimization/70291
261 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
262 arguments. Change return type to tree. Emit libcall as a new
263 statement rather than replacing existing one when inplace_p is true.
264 (expand_complex_multiplication_components): New function.
265 (expand_complex_multiplication): Expand floating-point complex
266 multiplication using the above.
267 (expand_complex_division): Rename inner_type parameter to type.
268 Update expand_complex_libcall call-site.
269 (expand_complex_operations_1): Update expand_complex_multiplication
270 and expand_complex_division call-sites.
272 2018-05-02 Jakub Jelinek <jakub@redhat.com>
275 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
276 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
277 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
278 the highest significant bit of the shift count mask is clear. In
279 check whether and[sq]i3 is needed verify that all significant bits
280 of the shift count other than the highest are set.
282 2018-05-02 Tom de Vries <tom@codesourcery.com>
285 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
286 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
287 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
288 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
289 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
290 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
291 __builtin_goacc_parlevel_size.
293 2018-05-02 Richard Biener <rguenther@suse.de>
295 PR tree-optimization/85597
296 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
297 do not use split vect_get_vec_defs call but call vect_get_slp_defs
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 ltrans-tree.
306 2018-05-02 Tom de Vries <tom@codesourcery.com>
309 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
310 dump files): Add wpa-ipa.
312 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
314 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
315 powerpc*-*-linux*paired* target.
316 * config/rs6000/750cl.h: Delete.
317 * config/rs6000/paired.h: Delete.
318 * config/rs6000/paired.md: Delete.
319 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
321 * config/rs6000/rs6000-builtin.def: Remove paired float support.
322 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
323 comment. Remove paired float support.
324 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
325 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
327 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
328 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
330 * config/rs6000/rs6000.c: Remove paired float support.
331 (paired_expand_vector_init, paired_expand_vector_move,
332 paired_emit_vector_compare, paired_emit_vector_cond_expr,
333 (paired_expand_lv_builtin, paired_expand_stv_builtin,
334 paired_expand_builtin, paired_expand_predicate_builtin,
335 paired_init_builtins): Delete.
336 * config/rs6000/rs6000.h: Remove paired float support.
337 * config/rs6000/rs6000.md: Remove paired float support.
338 (move_from_CR_ov_bit): Delete.
339 * config/rs6000/rs6000.opt (mpaired): Delete.
340 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
341 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
343 2018-05-02 Richard Biener <rguenther@suse.de>
346 * gimplify.c (gimplify_save_expr): When in SSA form allow
347 SAVE_EXPRs to compute to SSA vars.
349 2018-05-02 Jakub Jelinek <jakub@redhat.com>
352 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
353 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
354 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
355 clobber operands[2], instead use a new pseudo. Formatting fixes.
357 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
359 PR tree-optimization/85586
360 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
361 exit early for statements in the same group if the accesses are
364 2018-05-02 Tom de Vries <tom@codesourcery.com>
367 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
370 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
372 PR tree-optimization/85143
373 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
375 2018-05-01 Tom de Vries <tom@codesourcery.com>
378 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
379 not found" error message.
381 2018-05-01 Tom de Vries <tom@codesourcery.com>
384 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
385 * vec.c (test_ordered_remove_if): New function.
386 (vec_c_tests): Call test_ordered_remove_if.
387 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
388 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
389 * tree-vect-patterns.c (vect_pattern_recog_1): Use
390 VEC_ORDERED_REMOVE_IF.
392 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
394 PR tree-optimization/82665
395 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
396 pointer subtraction where arguments come from a memchr call.
398 2018-05-01 Jakub Jelinek <jakub@redhat.com>
400 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
401 --push-state --as-needed and --pop-state instead of --as-needed and
402 --no-as-needed if ld supports it.
403 * configure: Regenerated.
406 * doc/install.texi2html: Replace _002d with - and _002a with * in
407 generated html files using sed.
409 2018-04-30 David Malcolm <dmalcolm@redhat.com>
412 * gcc-rich-location.c (blank_line_before_p): New function.
413 (use_new_line): New function.
414 (gcc_rich_location::add_fixit_insert_formatted): New function.
415 * gcc-rich-location.h
416 (gcc_rich_location::add_fixit_insert_formatted): New function.
418 2018-04-30 David Malcolm <dmalcolm@redhat.com>
420 * selftest.c (assert_streq): Rename "expected" and "actual" to
421 "val1" and "val2". Extend NULL-handling to cover both inputs
422 symmetrically, while still requiring both to be non-NULL for a pass.
423 * selftest.h (assert_streq): Rename "expected" and "actual" to
425 (ASSERT_EQ): Likewise.
426 (ASSERT_EQ_AT): Likewise.
427 (ASSERT_KNOWN_EQ): Likewise.
428 (ASSERT_KNOWN_EQ_AT): Likewise.
429 (ASSERT_NE): Likewise.
430 (ASSERT_MAYBE_NE): Likewise.
431 (ASSERT_MAYBE_NE_AT): Likewise.
432 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
433 the assertion to pass.
434 (ASSERT_STREQ_AT): Likewise.
436 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
438 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
439 interaction with -pie.
441 2018-04-30 David Malcolm <dmalcolm@redhat.com>
443 * selftest.h: Fix alphabetization of per-source-file selftest
446 2018-04-30 Jason Merrill <jason@redhat.com>
448 PR c++/61982 - dead stores to destroyed objects.
449 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
452 2018-04-30 Jason Merrill <jason@redhat.com>
454 * tree.c (build_clobber): New.
455 * tree.h: Declare it.
456 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
458 2018-04-30 David Malcolm <dmalcolm@redhat.com>
460 * diagnostic-show-locus.c (layout::layout): Update for
461 location_get_source_line returning a char_span.
462 (struct char_span): Move to input.h.
463 (struct correction): Update for fields in char_span becoming
465 (struct source_line): Update for location_get_source_line
466 returning a char_span.
467 (layout::print_line): Likewise.
468 * edit-context.c (edited_file::print_content): Likewise.
469 (edited_file::print_diff_hunk): Likewise.
470 (edited_file::print_run_of_changed_lines): Likewise.
471 (edited_file::get_num_lines): Likewise.
472 (edited_line::edited_line): Likewise.
473 * final.c (asm_show_source): Likewise.
474 * input.c (location_get_source_line): Convert return type
475 from const char * to char_span, losing the final "line_len"
477 (dump_location_info): Update for the above.
478 (get_substring_ranges_for_loc): Likewise. Use a char_span
479 when handling the literal within the line.
480 (test_reading_source_line): Update for location_get_source_line
481 returning a char_span.
482 * input.h (class char_span): Move here from
483 diagnostic-show-locus.c, converting from a struct to a class.
484 Make data members private.
485 (char_span::operator bool): New.
486 (char_span::length): New.
487 (char_span::get_buffer): New.
488 (char_span::operator[]): New.
489 (char_span::subspan): Make const.
490 (char_span::xstrdup): New.
491 (location_get_source_line): Convert return type from const char *
492 to char_span, losing the final "line_size" param.
494 2018-04-30 Jan Hubicka <jh@suse.cz>
496 * lto-wrapper.c (ltrans_priorities): New static var.
498 (run_gcc): Read priorities and if doing parallel build order
499 the Makefile by them.
501 2018-04-30 David Malcolm <dmalcolm@redhat.com>
503 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
505 2018-04-30 Richard Biener <rguenther@suse.de>
507 * tree-cfg.c (verify_address): Remove base argument, add
508 flag whether to check TREE_ADDRESSABLE and do that.
509 (verify_expr): Remove.
510 (verify_types_in_gimple_reference): Add pieces from verify_expr.
511 (verify_gimple_assign_single): Likewise.
512 (verify_gimple_switch): Likewise.
513 (verify_expr_location_1): Dereference tp once. Add (disabled)
514 piece from verify_expr.
515 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
517 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
519 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
521 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
523 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
524 (small_data_pattern): Likewise.
525 (arc_rewrite_small_data): Likewise.
526 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
527 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
528 (get_symbol_alignment): New function.
529 (legitimate_small_data_address_p): Likewise.
530 (legitimate_scaled_address): Update, call
531 legitimate_small_data_address_p.
532 (output_sdata): New static variable.
533 (arc_print_operand): Update how we handle small data operands.
534 (arc_print_operand_address): Likewise.
535 (arc_legitimate_address_p): Update, use
536 legitimate_small_data_address_p.
537 (arc_rewrite_small_data_p): Remove.
538 (arc_rewrite_small_data_1): Likewise.
539 (arc_rewrite_small_data): Likewise.
540 (small_data_pattern): Likewise.
541 (compact_sda_memory_operand): Update to use
542 legitimate_small_data_address_p and get_symbol_alignment.
543 (prepare_move_operands): Don't rewite sdata pattern.
544 (prepare_extend_operands): Remove.
545 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
547 (zero_extendqisi2): Likewise.
548 (zero_extendhisi2): Likewise.
549 (extendqihi2): Likewise.
550 (extendqisi2): Likewise.
551 (extendhisi2): Likewise.
555 * config/arc/constraints.md (Usd): Change it to memory constraint.
557 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
559 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
560 as source of std instructions.
561 * config/arc/arc.md (movsi_insn): Update pattern predicate to
562 allow 6-bit constants as source for store instructions.
563 (movdi_insn): Update instruction pattern to allow 6-bit constants
564 as source for store instructions.
566 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
568 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
570 2018-04-30 Nathan Sidwell <nathan@acm.org>
571 Sandra Loosemore <sandra@codesourcery.com>
573 * dumpfile.c (dump_open): Allow '-' for stdout.
574 * doc/invoke.texi (Developer Options): Document dump filename
575 determination early. Document stdin/stdout selection.
577 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
579 Microblaze Target: PIC data text relative
581 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
582 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
584 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
585 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
586 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
587 New addressing mode for data-text relative position indepenedent code.
588 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
589 'ADDRESS_SYMBOLIC_TXT_REL'.
590 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
591 (microblaze_legitimate_pic_operand): Exclude function calls from
592 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
593 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
595 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
596 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
597 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
598 for 'address + offset'.
599 (microblaze_expand_prologue): Add new function prologue call for
601 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
602 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
603 table in case of TARGET_PIC_DATA_TEXT_REL.
604 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
605 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
606 Add new macros 'UNSPEC_TEXT',
607 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
608 + exclude function calls from 'UNSPEC_PLT' in case of data text
610 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
611 new target hook for generating address diff vector tables in case of
613 * doc/tm.texi : Regenerate.
614 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
615 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
616 of addr diff vector generation.
617 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
618 target hook definition.
619 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
620 Add default function for generate_pic_addr_diff_vec -> flag_pic.
621 * doc/invoke.texi (Add new pic option): Add new microblaze pic
622 option for data text relative.
624 2018-04-30 Richard Biener <rguenther@suse.de>
626 * tree-chrec.h (evolution_function_is_constant_p): Remove
628 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
630 2018-04-30 Richard Biener <rguenther@suse.de>
633 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
635 2018-04-30 Richard Biener <rguenther@suse.de>
637 PR tree-optimization/28364
638 PR tree-optimization/85275
639 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
640 copying first exit test.
642 2018-04-28 Mark Wielaard <mark@klomp.org>
644 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
646 (dwarf_AT): Handle DW_AT_addr_base.
647 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
649 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
652 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
653 (*ashl<dwi>3_doubleword_mask_1): Ditto.
654 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
655 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
657 2018-04-28 Richard Biener <rguenther@suse.de>
659 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
660 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
661 to reflect use. Only add interesting stmts.
663 2018-04-27 Martin Jambor <mjambor@suse.cz>
666 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
667 the jump function allows for passing through aggregate values.
669 2018-04-27 David Malcolm <dmalcolm@redhat.com>
671 * input.h (in_system_header_at): Convert from macro to inline
673 (from_macro_expansion_at): Likewise.
674 (from_macro_definition_at): Likewise.
676 2018-04-27 Jeff Law <law@redhat.com>
678 * config.gcc: Mark tile* targets as deprecated/obsolete.
680 2018-04-27 Richard Biener <rguenther@suse.de>
682 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
685 2018-04-27 Richard Biener <rguenther@suse.de>
687 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
689 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
691 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
692 with Yd constraint. Set "preferred_for_speed" attribute from
693 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
695 (*movdi_internal): Ditto.
696 (movti_interunit splitters): Remove
697 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
698 (movdi_interunit splitters): Ditto.
699 * config/i386/constraints.md (Ye): Remove.
700 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
702 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
705 * config/aarch64/constraints.md (Usg): Limit to 31.
708 2018-04-27 Jakub Jelinek <jakub@redhat.com>
710 PR tree-optimization/85529
711 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
712 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
713 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
714 zero extension or masking of the MSB bit.
715 (optimize_range_tests): Add FIRST_BB argument, pass it through
716 to optimize_range_tests_var_bound.
717 (maybe_optimize_range_tests, reassociate_bb): Adjust
718 optimize_range_tests callers.
720 2018-04-26 Richard Biener <rguenther@suse.de>
721 Jakub Jelinek <jakub@redhat.com>
723 * cgraph.h (symbol_table): Just declare debug method here.
724 * symtab.c (symbol_table::debug): Define.
726 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
728 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
730 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
732 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
733 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
734 (*movdi_internal): Substitute Yi and Yj constraint with x
735 and Ym and Yn constraint with y constraint. Update "isa"
736 attribute and set "preferred_for_speed" attribute from
737 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
738 (*movsi_internal): Ditto.
739 (*movdf_internal): Ditto.
740 (*movsf_internal): Ditto.
741 (*zero_extendsidi2): Ditto.
742 * config/i386/sse.md (vec_set<mode>_0): Ditto.
743 (sse2_loadld): Ditto.
744 (*vec_extract<ssevecmodelower>_0): Ditto.
745 (*vec_extractv4si_0_zext_sse4): Ditto.
746 (vec_concatv2di): Ditto.
747 (*vec_dup<mode>): Ditto.
748 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
749 * config/i386/constraints.md (Yi): Remove.
754 2018-04-26 Nathan Sidwell <nathan@acm.org>
756 * dumpfile.c (dump_open): New.
757 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
758 (dump_finish): Detect stdio/stderr by value not name.
760 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
762 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
764 2018-04-26 Tom de Vries <tom@codesourcery.com>
767 * config/nvptx/nvptx.c (verify_neutering_jumps)
768 (verify_neutering_labels): New function
769 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
771 2018-04-26 Tom de Vries <tom@codesourcery.com>
774 * config/nvptx/nvptx.c (needs_neutering_p): New function.
775 (nvptx_single): Use needs_neutering_p to skip over insns that do not
778 2018-04-26 Richard Biener <rguenther@suse.de>
779 Tom de Vries <tom@codesourcery.com>
782 * lto-streamer-out.c (output_function): Fixup loops if required to match
783 discovery done in the reader.
785 2018-04-26 Richard Biener <rguenther@suse.de>
787 PR tree-optimization/85116
788 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
789 have a loop exit from the single latch predecessor. Remove
790 case of header with just condition.
791 (ch_base::copy_headers): Exclude infinite loops from any
793 (pass_ch::execute): Record exits.
795 2018-04-26 Richard Biener <rguenther@suse.de>
797 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
798 prologue cost vector and pass it to vect_get_load_cost.
799 (vect_get_peeling_costs_all_drs): Likewise.
800 (vect_peeling_hash_get_lowest_cost): Likewise.
801 (vect_enhance_data_refs_alignment): Likewise.
803 2018-04-26 Richard Biener <rguenther@suse.de>
806 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
807 checking of integer<->pointer conversions.
808 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
809 sign-/zero-extending pointer types.
810 (expand_omp_for_static_chunk): Likewise.
812 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
813 Jean Lee <xiaoyur347@gmail.com>
815 * config/mips/mips.c (mips_asan_shadow_offset): New function.
816 (TARGET_ASAN_SHADOW_OFFSET): Define.
817 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
818 true for -fsanitize=address.
820 2018-04-25 Mark Wielaard <mark@klomp.org>
822 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
825 2018-04-25 Jakub Jelinek <jakub@redhat.com>
827 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
828 than "alu", remove explicit "memory" and "imm_disp" attributes.
829 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
832 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
833 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
836 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
839 * config/i386/i386.c (ix86_expand_builtin): Change memory
840 operand to XI, extend p0 to Pmode.
841 * config/i386/i386.md: Change unspec volatile and operand
842 1 mode to XI, change operand 0 mode to P.
844 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
846 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
847 GET_MODE_MASK before any checking.
848 (nds32_can_use_bset_p): Likewise.
849 (nds32_can_use_btgl_p): Likewise.
851 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
853 * config/nds32/nds32-doubleword.md: New define_split pattern for
854 illegal register number.
856 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
858 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
860 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
862 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
864 2018-04-25 Richard Biener <rguenther@suse.de>
866 * lto-streamer.h (LTO_major_version): Bump to 8.
868 2018-04-25 Jakub Jelinek <jakub@redhat.com>
870 * BASE-VER: Set to 9.0.0.
872 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
874 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
875 in __abskf2 and __powikf2.
877 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
880 * config/aarch64/constraints.md (Usg, Usj): New constraints.
881 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
882 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
883 Use the above on operand 2. Reindent.
884 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
886 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
889 * common/config/i386/i386-common.c (ix86_handle_option): Don't
891 * config/i386/i386.opt (mcet): Removed.
892 * doc/install.texi: Remove -mcet documentation.
893 * doc/invoke.texi: Likewise.
895 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
898 * doc/install.texi: Remove -mcet from bootstrap-cet.
900 2018-04-24 Jakub Jelinek <jakub@redhat.com>
903 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
904 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
908 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
909 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
910 containing a CONST_VECTOR.
912 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
914 * doc/install.texi: Update newlib dependency for nvptx.
916 2018-04-24 Jakub Jelinek <jakub@redhat.com>
919 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
920 instead of INTVAL when shifting x left.
922 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
924 PR tree-optimization/85478
925 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
926 vect_grouped_store_supported for single element vectors.
928 2018-04-24 Richard Biener <rguenther@suse.de>
931 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
932 load cost increase to the case of non-constant step.
934 2018-04-24 Jakub Jelinek <jakub@redhat.com>
937 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
938 destination if any_malformed_asm.
940 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
943 * expr.c (store_field): In the bitfield case, if the value comes from
944 a function call and is returned in registers by means of a PARALLEL,
945 do not change the mode of the temporary unless BLKmode and VOIDmode.
947 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
949 PR rtl-optimization/85423
950 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
951 dependencies to debug insns when the previous insn is non-debug.
953 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
955 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
956 enums into a single definition.
957 (fls): Fix predicates and printing.
960 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
962 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
963 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
964 and short u6 immediate.
965 (check_if_valid_sleep_operand): Remove.
966 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
968 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
970 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
971 flag_always_save_lp condition.
972 * config/nds32/nds32.opt (malways-save-lp): New option.
974 2018-04-22 Shiva Chen <shiva0217@gmail.com>
976 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
977 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
978 * config/nds32/nds32.h
979 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
980 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
982 2018-04-22 Shiva Chen <shiva0217@gmail.com>
984 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
985 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
987 2018-04-22 Shiva Chen <shiva0217@gmail.com>
988 Chung-Ju Wu <jasonwucj@gmail.com>
990 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
992 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
993 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
995 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
997 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
999 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1001 * config/nds32/nds32-protos.h (nds32_data_alignment,
1002 nds32_local_alignment): Declare.
1003 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
1004 nds32_local_alignment): New functions.
1005 (TARGET_CONSTANT_ALIGNMENT): Define.
1006 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
1008 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1010 * config/nds32/nds32.c
1011 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
1012 (TARGET_MODES_TIEABLE_P): Likewise.
1014 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1016 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
1019 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
1020 Chung-Ju Wu <jasonwucj@gmail.com>
1022 * config/nds32/constants.md (unspec_volatile_element): Add enum values
1023 for unaligned access.
1024 * config/nds32/nds32-intrinsic.c: Implementation of expanding
1026 * config/nds32/nds32-intrinsic.md: Likewise.
1027 * config/nds32/nds32_intrinsic.h: Likewise.
1028 * config/nds32/nds32.h (nds32_builtins): Likewise.
1029 * config/nds32/nds32.opt (munaligned-access): New option.
1030 * config/nds32/nds32.c (nds32_asm_file_start): Display
1031 flag_unaligned_access status.
1033 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
1035 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
1036 -mno-relax is present.
1037 * config/riscv/linux.h (LINK_SPEC): Ditto.
1039 2018-04-20 Martin Sebor <msebor@redhat.com>
1042 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
1044 (gimple_fold_builtin_stxcpy_chk): Same.
1045 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
1047 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
1050 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
1051 __powikf2 when long double is IEEE 128-bit.
1053 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
1055 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
1056 step to make sure stack always aligned.
1058 2018-04-20 Carl Love <cel@us.ibm.com>
1061 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
1062 size check for arg0.
1064 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
1065 Tom de Vries <tom@codesourcery.com>
1068 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
1069 Emit insns for calls too.
1070 (nvptx_find_par): Always look for worker-level predecessor insn.
1071 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
1073 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
1074 (nvptx_process_pars): Propagate frames for calls.
1076 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
1079 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
1081 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
1082 (ix86_handle_option): Don't handle OPT_mibt.
1083 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
1085 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
1087 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
1088 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
1089 (ix86_target_macros): Define __CET__ with flag_cf_protection
1090 for -fcf-protection.
1091 * config/i386/i386.c (isa2_opts): Remove -mibt.
1092 * config/i386/i386.h (TARGET_IBT): Removed.
1093 (TARGET_IBT_P): Likewise.
1094 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
1095 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
1096 * config/i386/i386.opt (mcet): Update help message.
1099 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
1100 -mcet as an alias for -mshstk.
1102 2018-04-20 Richard Biener <rguenther@suse.de>
1105 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
1106 complexity by forcing a single use of the multiply operand.
1108 2018-04-20 Martin Jambor <mjambor@suse.cz>
1111 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
1112 recursion dependency to only apply to non-clones.
1114 2018-04-20 Martin Jambor <mjambor@suse.cz>
1117 * ipa-cp.c (create_specialized_node): Check that clones of
1118 self-recursive edges exist during IPA-CP.
1120 2018-04-19 Toon Moene <toon@moene.org>
1122 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
1125 2018-04-19 Jakub Jelinek <jakub@redhat.com>
1127 PR tree-optimization/85467
1128 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
1129 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
1130 VECTOR_CST element to type.
1132 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1135 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
1136 * config/i386/i386.md (builtin_setjmp_setup): Removed.
1137 (builtin_longjmp): Likewise.
1138 (save_stack_nonlocal): New pattern.
1139 (restore_stack_nonlocal): Likewise.
1141 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1144 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
1145 Replace ASM_OUTPUT_LABEL with fprintf.
1147 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1150 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
1151 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
1152 * config/i386/i386-c.c (ix86_target_macros_internal): Also
1153 define __IBT__ and __SHSTK__ for -fcf-protection.
1154 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
1156 (ix86_trampoline_init): Likewise.
1157 (x86_output_mi_thunk): Likewise.
1158 (ix86_notrack_prefixed_insn_p): Likewise.
1159 (ix86_option_override_internal): Don't disallow -fcf-protection.
1160 * config/i386/i386.md (rdssp<mode>): Also enable for
1162 (incssp<mode>): Likewise.
1163 (nop_endbr): Likewise.
1164 * config/i386/i386.opt (mcet): Change help message to built-in
1168 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
1169 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
1170 enable CET built-in functions.
1172 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
1174 * common/config/i386/i386-common.c
1175 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
1176 OPTION_MASK_ISA_MOVDIRI_UNSET,
1177 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
1178 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
1179 * config.gcc (movdirintrin.h): New header.
1180 * config/i386/cpuid.h (bit_MOVDIRI,
1181 bit_MOVDIR64B): New bits.
1182 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
1184 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
1185 (VOID, PVOID, PCVOID)): New function types.
1186 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
1187 __builtin_ia32_directstoreu_u64,
1188 __builtin_ia32_movdir64b): New builtins.
1189 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
1190 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
1192 (ix86_valid_target_attribute_inner_p): Ditto.
1193 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
1194 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
1195 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
1196 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
1197 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
1198 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
1199 (movdiri<mode>, movdir64b_<mode>): New.
1200 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
1201 * config/i386/immintrin.h: Include movdirintrin.h.
1202 * config/i386/movdirintrin.h: New file.
1203 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
1205 2018-04-19 Richard Biener <rguenther@suse.de>
1208 * cfg.c (clear_bb_flags): When loop state says we have
1209 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
1211 2018-04-19 Richard Biener <rguenther@suse.de>
1213 PR tree-optimization/84737
1214 * tree-vect-data-refs.c (vect_copy_ref_info): New function
1215 copying restrict info.
1216 (vect_setup_realignment): Use it.
1217 * tree-vectorizer.h (vect_copy_ref_info): Declare.
1218 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
1219 the first DR to all generated stores.
1220 (vectorizable_load): Likewise for loads.
1222 2018-04-19 Jakub Jelinek <jakub@redhat.com>
1224 PR tree-optimization/85446
1225 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
1226 the integral and pointer types to have the same precision.
1228 * doc/install.texi: Document --disable-cet being the default and
1231 2018-04-18 Martin Liska <mliska@suse.cz>
1233 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
1236 2018-04-18 Martin Liska <mliska@suse.cz>
1239 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
1242 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
1243 arguments if they are comparable.
1245 2018-04-18 Martin Liska <mliska@suse.cz>
1248 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
1251 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
1254 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
1257 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
1258 ENDBR after calling __morestack.
1260 2018-04-18 David Malcolm <dmalcolm@redhat.com>
1263 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
1264 by using gcc_base_ver to generate a gcc_driver_version, and use
1265 it when generating GCC_DRIVER_NAME.
1266 * configure: Regenerate.
1268 2018-04-18 Jakub Jelinek <jakub@redhat.com>
1271 * config.gcc: Obsolete powerpc*-*-*spe*.
1273 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1276 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
1277 (stabstr_D): Change type of unum from unsigned int to
1278 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
1281 2018-04-17 Jim Wilson <jimw@sifive.com>
1284 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
1285 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
1286 Set arg_pointer_offset after using pretend_args_size.
1288 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1290 PR rtl-optimization/85431
1291 * dse.c (record_store): Ignore zero width stores.
1294 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
1295 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
1296 __builtin_stack_restore rather than after it.
1297 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
1298 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
1299 argument instead of virtual_dynamic_stack_rtx.
1301 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
1303 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
1305 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
1306 Add note to error message to explain internal mapping of overloaded
1307 built-in function name to non-overloaded built-in function name.
1308 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
1311 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
1314 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
1315 where the inputs overlap with the output.
1317 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1319 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
1320 (=v, v) alternative and explicit "memory" attribute.
1321 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
1322 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
1324 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
1325 "sselog1" type instead of "sselog".
1326 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
1327 "sselog". Remove explicit "memory" attribute.
1328 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
1329 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
1331 (vec_extract_hi_v32hi): Merge all alternatives into one, use
1332 "sselog1" type instead of "sselog". Remove explicit "memory"
1334 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
1335 use "sselog1" type instead of "sselog". Remove explicit "memory"
1337 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
1338 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
1340 (vec_extract_hi_v64qi): Merge all alternatives into one, use
1341 "sselog1" type instead of "sselog". Remove explicit "memory"
1343 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
1344 use "sselog1" type instead of "sselog". Remove explicit "memory"
1348 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
1351 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
1354 2018-04-17 Martin Jambor <mjambor@suse.cz>
1357 * ipa-cp.c (create_specialized_node): Call
1358 expand_all_artificial_thunks if necessary.
1360 2018-04-17 Martin Liska <mliska@suse.cz>
1363 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
1364 in message, remote space in between '_G' and '('.
1366 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1369 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
1370 avx512f_vmcmp<mode>3<round_saeonly_name>,
1371 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
1372 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
1373 avx512f_rndscale<mode><round_saeonly_name>,
1374 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
1375 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
1376 Use %<iptr>2 instead of %2 for -masm=intel.
1377 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
1378 avx512f_vcvttss2usi<round_saeonly_name>,
1379 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
1381 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
1382 avx512f_vcvttsd2usi<round_saeonly_name>,
1383 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
1384 Use %q1 instead of %1 for -masm=intel.
1385 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
1386 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
1387 of %3 for -masm=intel.
1388 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
1390 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
1392 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
1394 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
1395 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
1397 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
1399 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
1400 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
1401 %g1 and one with %0 and %1.
1402 (avx512er_vmrcp28<mode><round_saeonly_name>,
1403 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
1405 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
1406 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
1407 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
1408 of %0 and %{%4%} for -masm=intel.
1409 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
1410 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
1411 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
1412 order of %0 and %{%5%}%{z%} for -masm=intel.
1414 2018-04-17 Jan Hubicka <jh@suse.cz>
1417 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
1419 2018-04-17 Martin Liska <mliska@suse.cz>
1422 * multiple_target.c (create_dispatcher_calls): Set apostrophes
1423 for target_clone error message. Make default implementation
1424 clone to be a local declaration.
1425 (separate_attrs): Add new argument and check for an empty
1427 (expand_target_clones): Handle it.
1428 (ipa_target_clone): Make redirection just for target_clones
1431 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
1432 Tom de Vries <tom@codesourcery.com>
1435 * omp-expand.c (expand_oacc_for): Add dummy false branch for
1436 tiled basic blocks without omp continue statements.
1438 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
1441 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
1442 vec_extract expression as having side effects to make sure it gets
1445 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
1448 * config/i386/i386.c (get_builtin_code_for_version): Check
1451 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
1454 * gcc/config.gcc: Support "skylake".
1455 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
1457 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
1458 (processor_target_table): Add "skylake".
1459 (ix86_option_override_internal): Add "skylake".
1460 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
1461 PROCESSOR_CANNONLAKE.
1462 (get_builtin_code_for_version): Fix priority for
1463 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
1464 PROCESSOR_SKYLAKE-AVX512.
1465 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
1466 (processor_type): Add PROCESSOR_SKYLAKE.
1468 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
1469 Jason Merrill <jason@redhat.com>
1472 * convert.c (convert_to_integer_1): Use direct recursion for
1473 enumeral types and types with a precision less than the number
1474 of bits in their mode.
1476 2018-04-16 Julia Koval <julia.koval@intel.com>
1479 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
1480 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
1482 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
1485 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
1486 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
1487 and -mno-direct-move.
1489 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
1492 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
1493 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
1494 Ensure negative shifts result in {0}.
1496 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
1498 PR rtl-optimization/79916
1499 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
1500 regs (if any) to define how to gnerate SD moves when LRA is in
1503 2018-04-13 Jakub Jelinek <jakub@redhat.com>
1505 PR rtl-optimization/85393
1506 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
1507 * except.c (expand_dw2_landing_pad_for_region): Make static.
1508 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
1509 a label and unconditional jump to old_bb, rather than
1510 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
1513 PR rtl-optimization/85376
1514 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
1515 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
1516 instead of a specific value.
1518 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
1519 Bin Cheng <bin.cheng@arm.com>
1521 PR tree-optimization/82965
1522 PR tree-optimization/83991
1523 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
1524 by_profile_only parameter.
1525 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
1526 information if the loop was predicted to iterate too many times.
1527 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
1529 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
1532 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
1535 2018-04-13 Martin Liska <mliska@suse.cz>
1536 Jakub Jelinek <jakub@redhat.com>
1539 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
1540 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
1541 * builtins.c (expand_builtin_memory_copy_args): Use
1542 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
1543 handle dest_addr == pc_rtx.
1545 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
1548 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
1549 asked to not generate direct moves.
1550 (fix_trunc<mode>si2_stfiwx): Similar.
1551 (fix_trunc<mode>si2_internal): Similar.
1553 2018-04-12 Jakub Jelinek <jakub@redhat.com>
1556 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
1557 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
1558 lookup if dest in some wider mode is known to be const0_rtx and
1559 if so, record permanent equivalence for it to be ZERO_EXTEND of
1560 the narrower mode destination.
1562 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
1564 * lto-streamer-out.c (output_function): Revert 259346.
1565 * omp-expand.c (expand_oacc_for): Likewise.
1567 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
1569 PR rtl-optimization/85354
1570 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
1571 * sel-sched.c (sel_global_init): ... here.
1573 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
1576 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
1577 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
1578 mode for PE-COFF targets.
1579 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
1580 (i386_pe_asm_lto_end): Likewise.
1581 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
1582 (TARGET_ASM_LTO_END): Likewise.
1583 * config/i386/winnt.c (saved_debug_info_level): New static variable.
1584 (i386_pe_asm_lto_start): New function.
1585 (i386_pe_asm_lto_end): Likewise.
1587 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
1588 Richard Biener <rguenther@suse.de>
1591 * lto-streamer-out.c (output_function): Fix CFG loop state before
1593 * omp-expand.c (expand_oacc_for): Handle calls to internal
1594 functions like regular functions.
1596 2018-04-12 Richard Biener <rguenther@suse.de>
1599 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
1600 for the early LTO debug to properly generate references to it
1601 during DIE emission. Do not re-use that for the skeleton for
1603 (dwarf2out_early_finish): Likewise.
1605 2018-04-12 Jakub Jelinek <jakub@redhat.com>
1608 * config/i386/sse.md
1609 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
1610 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
1611 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
1612 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
1613 and output is a reg, avoid creating invalid lowpart subreg, but
1614 instead split into a 512-bit move. Don't split if not AVX512VL,
1615 input is xmm16+ reg and output is a mem.
1616 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
1617 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
1618 xmm16+ reg and output is a mem.
1620 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1622 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
1623 also for flag_dwarf2_cfi_asm.
1625 2018-04-12 Jakub Jelinek <jakub@redhat.com>
1627 PR rtl-optimization/85342
1628 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
1629 a bool scalar var inside of the loop instead. Don't try to update
1630 recog_data.operand after failed apply_change_group.
1632 2018-04-12 Tom de Vries <tom@codesourcery.com>
1635 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
1636 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
1637 array with flexible array member as array without given dimension.
1638 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
1639 argument for undefined param to true.
1641 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
1644 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
1645 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
1646 from PowerPC section.
1647 * config/rs6000/sysv4.opt (mcall-): Improve help text.
1648 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
1649 help text that is too long.
1650 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
1651 help text that is too long.
1652 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
1653 help text that is too long.
1655 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
1657 * config/alpha/alpha.md (stack_probe_internal): Rename
1658 from "probe_stack". Update all callers.
1660 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1662 PR rtl-optimization/84566
1663 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
1664 sched_macro_fuse_insns.
1666 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1669 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
1670 (compute_block_dependences): ... from here.
1672 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1674 PR tree-optimization/85331
1675 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
1676 from int to HOST_WIDE_INT.
1678 2018-04-11 Martin Jambor <mjambor@suse.cz>
1681 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
1682 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
1683 not the same as the source val.
1684 (cgraph_edge_brings_value_p): New parameter.
1685 (gather_edges_for_value): Pass destination value to
1686 cgraph_edge_brings_value_p.
1687 (perhaps_add_new_callers): Likewise.
1688 (get_info_about_necessary_edges): Likewise and exclude values brought
1689 only by self-recursive edges.
1690 (create_specialized_node): Redirect only clones of self-calling edges.
1691 (+self_recursive_pass_through_p): New function.
1692 (find_more_scalar_values_for_callers_subset): Use it.
1693 (find_aggregate_values_for_callers_subset): Likewise.
1694 (known_aggs_to_agg_replacement_list): Removed.
1695 (decide_whether_version_node): Re-calculate known constants for all
1696 remaining context clones.
1698 2018-04-11 Richard Biener <rguenther@suse.de>
1701 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
1702 from early DWARF output.
1703 (dwarf2out_early_finish): Output line info unconditionally into
1704 early DWARF and add reference to it.
1706 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1709 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
1710 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
1711 other than V2DFmode using iptr mode attribute.
1712 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
1714 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1716 PR rtl-optimization/84659
1717 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
1719 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1722 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
1724 (output_loc_list): Pass address of a dummy size variable even in the
1725 locview handling loop.
1726 (index_location_lists): Add comment on why skip_loc_list_entry can't
1729 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
1732 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
1735 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
1738 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
1739 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
1740 and -mstring-compare-inline-limit.
1742 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1745 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
1746 for stack clash protection in a register whenever we need it to be in
1749 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1751 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
1752 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
1754 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1757 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
1759 (mlong-double-): Ditto.
1760 * config/rs6000/sysv4.opt (msdata=): Ditto.
1761 (mtls-size=): Ditto.
1763 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
1765 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
1766 erroneous entries for
1767 "vector int vec_ldl (int, long int *)", and
1768 "vector unsigned int vec_ldl (int, unsigned long int *)".
1769 Add comments and entries for
1770 "vector bool char vec_ldl (int, bool char *)",
1771 "vector bool short vec_ldl (int, bool short *)",
1772 "vector bool int vec_ldl (int, bool int *)",
1773 "vector bool long long vec_ldl (int, bool long long *)",
1774 "vector pixel vec_ldl (int, pixel *)",
1775 "vector long long vec_ldl (int, long long *)",
1776 "vector unsigned long long vec_ldl (int, unsigned long long *)".
1777 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
1778 type tree bool_long_long_type_node and correct definition of
1779 bool_V2DI_type_node to make reference to this new type tree.
1780 (rs6000_mangle_type): Replace erroneous reference to
1781 bool_long_type_node with bool_long_long_type_node.
1782 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
1783 comments to emphasize sign distinctions for char and int types and
1784 replace RS6000_BTI_bool_long constant with
1785 RS6000_BTI_bool_long_long constant. Also add comment to restrict
1786 use of RS6000_BTI_pixel.
1787 (bool_long_type_node): Remove this macro definition.
1788 (bool_long_long_type_node): New macro definition
1790 2018-04-10 Jakub Jelinek <jakub@redhat.com>
1792 PR rtl-optimization/85300
1793 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
1794 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
1795 simplify_unary_operation fails.
1797 2018-04-10 Martin Liska <mliska@suse.cz>
1799 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
1800 cgraph_edge and ipa_ref.
1802 2018-04-10 Jakub Jelinek <jakub@redhat.com>
1806 * config/i386/sse.md
1807 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
1808 computation of the VEC_MERGE selector from mask.
1809 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
1810 Fix decoding of the VEC_MERGE selector into mask.
1812 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
1814 PR tree-optimization/85286
1815 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
1817 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
1819 * final.c (final_1): Set insn_last_address as well as
1820 insn_current_address.
1822 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1825 * explow.c (emit_stack_probe): Call validize_mem on memory location
1826 before passing it to gen_probe_stack. Create address operand and
1827 legitimize it for the probe_stack_address case.
1829 2018-04-09 Jan Hubicka <jh@suse.cz>
1832 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
1833 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
1834 * tree.c (free_lang_data_in_type): Fix handling of binfos;
1836 (free_lang_data): Rebuild type inheritance graph.
1838 2018-04-09 Martin Sebor <msebor@redhat.com>
1840 * invoke.texi (-finline-small-functions): Mention other optimization
1842 (-findirect-inlining, -fpartial-inlining): Same.
1843 (-finline-functions-called-once): Same.
1844 (-freorder-blocks-and-partition): Same.
1846 2018-04-09 Jan Hubicka <jh@suse.cz>
1849 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
1850 jumps; choose last target that matches the criteria (i.e.
1851 no partition changes for non-crossing jumps).
1852 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
1853 support for redirecting crossing jumps to non-crossing.
1855 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
1857 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
1858 also for naked functions.
1860 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
1862 * config/arc/arc.md (add_shift): New pattern.
1863 (add_shift2): Likewise.
1864 (sub_shift): Likewise.
1865 (sub_shift_cmp0_noout): Likewise.
1866 (compare_si_ashiftsi): Likewise.
1867 (xbfu_cmp0_noout): New combine pattern.
1868 (xbfu_cmp0"): Likewise.
1869 (movsi_set_cc_insn): Place the predicable variant first.
1870 (commutative_binary_cmp0_noout): Remove clobber.
1871 (commutative_binary_cmp0): New pattern.
1872 (noncommutative_binary_cmp0): Likewise.
1873 (noncommutative_binary_cmp0_noout): Likewise.
1874 (noncommutative_binary_comparison_result_used): Removed.
1875 (rsub_cmp0): New pattern.
1876 (rsub_cmp0_noout): Likewise.
1877 (extzvsi): Changed, keep only meaningful variants.
1878 (SQH, SEZ): New iterators.
1879 (SQH_postfix): New mode attribute.
1880 (SEZ_prefix): New code attribute.
1881 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
1882 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
1883 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
1885 (noncommutative_operator): Check the availability of barrel
1888 2018-04-09 Richard Biener <rguenther@suse.de>
1890 PR tree-optimization/85284
1891 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
1892 Only use the niter constraining form of simple_iv when the exit
1895 2018-04-09 Tom de Vries <tom@codesourcery.com>
1898 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
1899 (define_expand "*memory_barrier"): New define_expand.
1900 (define_insn "memory_barrier"): New insn.
1902 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1904 PR rtl-optimization/80463
1905 PR rtl-optimization/83972
1906 PR rtl-optimization/83480
1908 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
1909 correct producer for the insn.
1910 (tidy_control_flow): Fixup seqnos in case of debug insns.
1912 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1914 PR rtl-optimization/83913
1916 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
1917 different sched-times when merging exprs.
1919 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1921 PR rtl-optimization/83962
1923 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
1924 tidy_fallthru_edge and tidy_control_flow.
1926 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1928 PR rtl-optimization/83530
1930 * sel-sched.c (force_next_insn): New global variable.
1931 (remove_insn_for_debug): When force_next_insn is true, also leave only
1932 next insn in the ready list.
1933 (sel_sched_region): When the region wasn't scheduled, make another pass
1934 over it with force_next_insn set to 1.
1936 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
1938 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
1940 * config/nds32/constants.md (unspec_volatile_element): Add enum values
1941 for interrupt control.
1942 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
1943 functions for interrupt control.
1944 * config/nds32/nds32-intrinsic.md: Likewise.
1945 * config/nds32/nds32_intrinsic.h: Likewise.
1946 * config/nds32/nds32.h (nds32_builtins): Likewise.
1948 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
1950 * config/nds32/nds32.c (nds32_init_machine_status,
1951 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
1952 strict_aligned_p field.
1953 (nds32_expand_to_rtl_hook): New function.
1954 (TARGET_EXPAND_TO_RTL_HOOK): Define.
1955 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
1957 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1958 Chung-Ju Wu <jasonwucj@gmail.com>
1960 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
1961 * config/nds32/nds32-n7.md: New file.
1962 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
1963 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
1965 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
1966 * config/nds32/nds32.md (pipeline_model): Add n7.
1967 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
1968 * config/nds32/pipelines.md: Include n7 settings.
1970 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1971 Chung-Ju Wu <jasonwucj@gmail.com>
1973 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
1974 * config/nds32/nds32-e8.md: New file.
1975 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
1976 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
1978 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
1979 * config/nds32/nds32.md (pipeline_model): Add e8.
1980 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
1981 * config/nds32/pipelines.md: Include e8 settings.
1983 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1984 Chung-Ju Wu <jasonwucj@gmail.com>
1986 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
1987 * config/nds32/nds32-n8.md: New file.
1988 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
1989 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
1991 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
1992 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
1993 * config/nds32/nds32.md (pipeline_model): Add n8.
1994 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
1995 * config/nds32/pipelines.md: Include n8 settings.
1997 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1998 Chung-Ju Wu <jasonwucj@gmail.com>
2000 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
2001 * config/nds32/nds32-n9-2r1w.md: New file.
2002 * config/nds32/nds32-n9-3r2w.md: New file.
2003 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
2004 nds32_register_ports): New or modify for cpu n9.
2005 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
2007 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
2008 * config/nds32/nds32-utils.c: New file.
2009 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
2010 TARGET_MUL_SLOW): Define.
2011 * config/nds32/nds32.md (pipeline_model): New attribute.
2012 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
2013 New options that support cpu n9.
2014 * config/nds32/pipelines.md: Include n9 settings.
2015 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
2017 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
2019 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
2020 information if necessary.
2021 (output_cond_branch_compare_zero): Likewise.
2022 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
2023 (nds32_target_alignment): Refine for alignment.
2024 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
2025 (FUNCTION_BOUNDARY): Modify.
2026 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
2028 * config/nds32/nds32.opt (malways-align, malign-functions): New.
2030 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
2032 * config/nds32/constants.md (unspec_volatile_element): Add values for
2033 TLB operation and data prefetch.
2034 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
2035 functions for TLB operation and data prefetch.
2036 * config/nds32/nds32-intrinsic.md: Likewise.
2037 * config/nds32/nds32_intrinsic.h: Likewise.
2038 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
2039 (nds32_print_operand): Likewise.
2040 * config/nds32/nds32.h (nds32_builtins): Likewise.
2042 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
2043 Andrew Pinski <pinsika@gcc.gnu.org>
2046 * match.pd: Use constant_boolean_node of correct type instead of
2047 boolean_true_node or boolean_false_node for simplifying
2048 pointer comparisons to zero.
2050 2018-04-07 Jakub Jelinek <jakub@redhat.com>
2052 PR tree-optimization/80021
2053 * tree.c (verify_type_variant): Make error call in verify_variant_match
2054 translatable and remove final full stop.
2056 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2058 * config/nds32/constants.md (unspec_volatile_element): Add
2059 UNSPEC_VOLATILE_EH_RETURN.
2060 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
2061 nds32_output_stack_pop): Support dwarf exception handling process.
2062 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
2063 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
2064 exception handling process.
2065 (nds32_compute_stack_frame): Likewise.
2066 (nds32_return_addr_rtx): Likewise.
2067 (nds32_initial_elimination_offset): Likewise.
2068 (nds32_expand_prologue): Likewise.
2069 (nds32_expand_epilogue): Likewise.
2070 (nds32_dynamic_chain_address): New function.
2071 * config/nds32/nds32.h (machine_function): Add fields for dwarf
2073 (DYNAMIC_CHAIN_ADDRESS): Define.
2074 (EH_RETURN_DATA_REGNO): Define.
2075 (EH_RETURN_STACKADJ_RTX): Define.
2076 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
2077 patterns for dwarf exception handling.
2079 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2081 * config/nds32/nds32.h: Clean up obsolete macros.
2083 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2085 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2086 Add enum values for particular instructions.
2087 * config/nds32/nds32-intrinsic.c: Implementation of expanding
2088 particular intrinsic functions.
2089 * config/nds32/nds32-intrinsic.md: Likewise.
2090 * config/nds32/nds32_intrinsic.h: Likewise.
2091 * config/nds32/nds32.h (nds32_builtins): Likewise.
2092 * config/nds32/nds32.md (type): Add pbsad and pbsada.
2093 (btst, ave): New patterns for particular instructions.
2095 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2097 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2098 Add enum values for atomic load/store and memory sync.
2099 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
2101 * config/nds32/nds32-intrinsic.md: Likewise.
2102 * config/nds32/nds32_intrinsic.h: Likewise.
2103 * config/nds32/nds32.h (nds32_builtins): Likewise.
2105 2018-04-07 Jakub Jelinek <jakub@redhat.com>
2107 PR tree-optimization/85257
2108 * fold-const.c (native_encode_vector): If not all elts could fit
2109 and off is -1, return 0 rather than offset.
2110 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
2111 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
2112 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
2113 adjust buffer in native_interpret_expr call.
2115 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2117 * config/nds32/constants.md (unspec_volatile_element): Add cache
2118 control enum values.
2119 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
2120 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
2121 * config/nds32/nds32.c (nds32_cctl_names): New.
2122 (nds32_print_operand): Handle cache control register names.
2123 * config/nds32/nds32.h (nds32_builtins): New enum values.
2124 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
2126 * config/nds32/nds32.md (type): Add mmu.
2127 * config/nds32/pipelines.md (simple_insn): Add mmu.
2129 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2131 * config/nds32/nds32.md (type): Remove call.
2132 * config/nds32/pipelines.md (simple_insn): Likewise.
2134 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2136 * config/nds32/constants.md (unspec_volatile_element): Add
2137 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
2138 UNSPEC_VOLATILE_FMFCFG.
2139 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
2140 description for fmfcfg and fmfcsr.
2141 (bdesc_1arg): Add fmtcsr.
2142 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
2143 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
2144 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
2145 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
2146 unspec_fmfcfg): New patterns.
2147 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
2148 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
2149 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
2150 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
2151 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
2152 __nds32__fmfcfg): Define.
2154 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2156 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
2157 intrinsic register names.
2158 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
2159 intrinsic register enum values and macros.
2161 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2163 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
2164 for load/store addressing form.
2165 (nds32_print_operand_address): Likewise.
2167 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
2170 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
2171 based on LABEL_REF. Remove useless assertion.
2172 (pic_address_needs_scratch): Fix formatting.
2173 (sparc_legitimize_pic_address): Minor tweaks.
2174 (sparc_delegitimize_address): Adjust assertion accordingly.
2175 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
2176 into symbolic_operand.
2177 (movsi_high_pic_label_ref): Likewise.
2178 (movsi_lo_sum_pic_label_ref): Likewise.
2179 (movdi_pic_label_ref): Likewise.
2180 (movdi_high_pic_label_ref): Likewise.
2181 (movdi_lo_sum_pic_label_ref): Likewise.
2183 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
2185 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
2186 custom LIB_SPEC setup.
2188 2018-04-06 Ruslan Bukin <br@bsdpad.com>
2189 Kito Cheng <kito.cheng@gmail.com>
2191 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
2192 * config/riscv/freebsd.h: New.
2194 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
2196 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
2197 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
2200 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
2201 Kito Cheng <kito.cheng@gmail.com>
2203 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
2204 nds32_output_call, nds32_symbol_binds_local_p): New functions.
2205 * config/nds32/nds32-protos.h (nds32_output_call,
2206 nds32_output_return): Declare.
2207 * config/nds32/nds32.md: Refine all the call and return patterns.
2209 2018-04-06 Jakub Jelinek <jakub@redhat.com>
2212 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
2213 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
2215 PR rtl-optimization/84872
2216 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
2217 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
2220 2018-04-06 Tamar Christina <tamar.christina@arm.com>
2222 * expr.c (copy_blkmode_to_reg): Revert 254862.
2223 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
2225 2018-04-06 Richard Biener <rguenther@suse.de>
2228 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
2229 after seeing a component reference with an adjacent field. Treat
2230 refs to arrays at struct end of external decls similar to
2231 refs to unconstrained commons.
2233 2018-04-06 Jakub Jelinek <jakub@redhat.com>
2236 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
2237 look through SAVE_EXPRs with non-side-effects argument. Adjust
2239 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
2242 2018-04-06 Richard Biener <rguenther@suse.de>
2245 * alias.c (find_base_term): New wrapper around find_base_term
2246 unwinding CSELIB_VAL_PTR changes.
2247 (find_base_term): Do not restore CSELIB_VAL_PTR during the
2250 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2252 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
2254 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
2255 constant definitions.
2256 ("nop"): lr 0,0 -> nopr r0
2257 ("nop_lr0", "nop_lr1"): New insn definitions.
2259 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
2261 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
2262 NDS32_V3PUSH_AVAILABLE_P macro.
2264 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
2265 Chung-Ju Wu <jasonwucj@gmail.com>
2267 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
2268 (nds32*-*-*): Add float and fpu_config into supported_defaults.
2269 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
2270 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
2271 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
2272 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
2273 * config/nds32/constraints.md: New constraints and checking for hard
2274 float configuration.
2275 * config/nds32/iterators.md: New mode iterator and attribute for hard
2276 float configuration.
2277 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
2279 * config/nds32/nds32-fpu.md: New file.
2280 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
2281 deal with hard float code generation.
2282 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
2284 (abi_type, float_reg_number): New enum type.
2285 * config/nds32/nds32-predicates.c: New predicates for hard float.
2286 * config/nds32/nds32-protos.h: Declare functions for hard float.
2287 * config/nds32/nds32.c: Implementation for hard float configuration.
2288 * config/nds32/nds32.h: Definitions for hard float configuration.
2289 * config/nds32/nds32.md: Include hard float machine description and
2290 modify patterns for hard float configuration.
2291 * config/nds32/nds32.opt: New options for hard float configuration.
2292 * config/nds32/predicates.md: New predicates for hard float
2295 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
2297 * common/config/nds32/nds32-common.c
2298 (nds32_option_optimization_table): Enable -mreleax-hint by default.
2300 2018-04-05 Jakub Jelinek <jakub@redhat.com>
2303 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
2304 CONSTRUCTOR_ELT (ctor, ...)->value.
2306 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
2309 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
2311 2018-04-05 Tom de Vries <tom@codesourcery.com>
2314 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
2317 2018-04-05 Shiva Chen <shiva0217@gmail.com>
2318 Kito Cheng <kito.cheng@gmail.com>
2320 * config/nds32/constraints.md (U33): Fine-tune checking condition.
2321 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
2322 * config/nds32/nds32.h (nds32_16bit_address_type): Add
2323 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
2325 2018-04-05 Shiva Chen <shiva0217@gmail.com>
2326 Kito Cheng <kito.cheng@gmail.com>
2328 * config/nds32/constraints.md (Ufe): New memory constraint.
2329 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
2330 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
2331 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
2333 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
2334 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
2336 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2338 * config/nds32/nds32.md: Use optimize_size in the condition for
2339 alu-shift instructions.
2341 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2343 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
2345 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2347 * config/nds32/nds32.md (negsi2): Refine pattern.
2349 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
2350 Chung-Ju Wu <jasonwucj@gmail.com>
2352 * config/nds32/iterators.md (shift_rotate): New code iterator.
2353 (shift): New code attribute.
2354 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
2355 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
2356 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
2357 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
2358 bit-wise operations.
2359 (andsi3, *andsi3): Ditto.
2360 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
2361 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
2362 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
2363 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
2364 nds32_ior_operand, nds32_xor_operand): New predicates.
2366 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2368 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
2369 (addsi3, subsi3): ... this.
2371 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2373 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
2375 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2377 * config/nds32/nds32.md: Adjust indention.
2379 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
2381 * config/nds32/nds32.md (feature): New attribute.
2383 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2385 * config/nds32/nds32.md (subtype): New attribute.
2387 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
2390 * config/arm/arm-builtins.c (arm_expand_builtin): Change
2391 expansion to perform a bitwise AND of the argument followed by a
2392 boolean negation of the result.
2394 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
2396 PR rtl-optimization/84878
2397 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
2398 the basic block. Assert the use reference is not artificial and that
2399 it has an associated insn.
2401 2018-04-04 Michael Matz <matz@suse.de>
2403 * builtins.c (compute_objsize): Pass correct operand
2404 to array_at_struct_end_p.
2406 2018-04-04 Richard Biener <rguenther@suse.de>
2409 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
2410 from contexts for DINFO_LEVEL_TERSE and below.
2412 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
2414 * config/nds32/nds32-doubleword.md (move_<mode>): Require
2415 resiter_operand condition.
2416 * config/nds32/nds32.md (*move<mode>): Ditto.
2418 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
2419 Monk Chiang <sh.chiang04@gmail.com>
2421 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
2423 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2425 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
2427 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2428 Kito Cheng <kito.cheng@gmail.com>
2430 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
2431 nds32_cond_code_str, output_cond_branch,
2432 output_cond_branch_compare_zero, 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: New functions.
2438 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
2439 nds32_expand_cstore, nds32_expand_movcc,
2440 nds32_output_cbranchsi4_equality_zero,
2441 nds32_output_cbranchsi4_equality_reg,
2442 nds32_output_cbranchsi4_equality_reg_or_const_int,
2443 nds32_output_cbranchsi4_greater_less_zero): Declare.
2444 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
2445 nds32_rimm11s_operand): New predicates.
2446 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
2447 * config/nds32/nds32.md: Rewrite all the branch and conditional move
2450 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
2452 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
2453 * config/nds32/nds32.md: Ditto.
2454 * config/nds32/pipelines.md: Ditto.
2456 2018-04-04 Richard Biener <rguenther@suse.de>
2458 PR tree-optimization/85168
2459 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
2460 propagating abnormals.
2462 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2464 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
2466 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2467 Kito Cheng <kito.cheng@gmail.com>
2469 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
2470 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
2471 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
2472 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
2473 * config/nds32/nds32.md (sibcall_internal): New.
2474 (sibcall_register): Remove.
2475 (sibcall_immediate): Remove.
2476 (sibcall_value_internal): New.
2477 (sibcall_value_register): Remove.
2478 (sibcall_value_immediate): Remove.
2479 * config/nds32/predicates.md (nds32_general_register_operand): New.
2480 (nds32_call_address_operand): New.
2482 2018-04-03 Jakub Jelinek <jakub@redhat.com>
2484 PR rtl-optimization/85167
2485 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
2486 bb_defs if *split_p, instead preinitialize it to NULL.
2488 PR tree-optimization/85156
2489 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
2490 evaluating the argument multiple times.
2492 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
2494 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
2496 (_mm_cvtpd_ps): Likewise.
2497 (_mm_cvttpd_epi32): Likewise.
2498 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
2499 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
2500 vector, pixel, and bool following altivec.h include.
2502 2018-04-03 Martin Sebor <msebor@redhat.com>
2504 * doc/extend.texi (Common Function Attributes): Clarify.
2505 (const attribute): Likewise.
2506 (pure attribute): Likewise.
2508 2018-04-03 Jakub Jelinek <jakub@redhat.com>
2511 * config/i386/i386.c (ix86_expand_vector_set): Use
2512 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
2514 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
2516 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
2517 instructions when changing rounding bits to preserve precision bits
2518 in the x87 control word.
2520 2018-04-03 Martin Liska <mliska@suse.cz>
2522 PR tree-optimization/82491
2523 * rtl.h (strip_offset_and_add): Replace += suboffset with
2524 poly_uint64 () + suboffset.
2526 2018-03-29 Martin Liska <mliska@suse.cz>
2527 Martin Jambor <mjambor@suse.cz>
2530 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
2531 param_type is not an integral or pointer type.
2533 2018-04-03 Richard Biener <rguenther@suse.de>
2535 * sese.h (recompute_all_dominators): Remove.
2537 2018-04-02 Martin Sebor <msebor@redhat.com>
2539 * doc/invoke.texi (-Wrestrict): Fix typos.
2541 2018-04-02 Jim Wilson <jimw@sifive.com>
2543 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
2544 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
2545 (<optab>di3, <optab>si3_extend): Likewise.
2546 (<optab>si3_mask, <optab>si3_mask_1): New.
2547 (<optab>di3_mask, <optab>di3_mask_1): New.
2548 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
2549 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
2550 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
2552 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
2554 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
2557 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
2559 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
2560 (nds32_canonicalize_comparison): New function.
2562 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
2563 Kito Cheng <kito.cheng@gmail.com>
2564 Kuan-Lin Chen <kuanlinchentw@gmail.com>
2566 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
2567 * config/nds32/constants.md (unspec_volatile_element): Add
2568 UNSPEC_VOLATILE_RELAX_GROUP.
2569 * config/nds32/nds32-relax-opt.c: New file.
2570 * config/nds32/nds32-predicates.c
2571 (nds32_symbol_load_store_p): New function.
2572 * config/nds32/nds32-protos.h
2573 (nds32_symbol_load_store_p): Declare function.
2574 (make_pass_nds32_relax_opt): Declare new rtl pass function.
2575 * config/nds32/nds32.c
2576 (nds32_register_pass): New function to register pass.
2577 (nds32_register_passes): New function to register passes.
2578 * config/nds32/nds32.md (relax_group): New pattern.
2579 * config/nds32/nds32.opt (mrelax-hint): New option.
2580 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
2582 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
2584 * config/nds32/t-nds32: Modify files dependency.
2586 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
2588 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
2589 (PROFILE_HOOK): Define its implementation.
2591 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
2593 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
2594 type and 32-bit size.
2596 2018-04-01 Jakub Jelinek <jakub@redhat.com>
2599 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
2600 (V_128_256): New mode iterator.
2601 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
2602 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
2603 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
2605 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
2606 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
2608 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
2611 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
2612 NaN inputs correctly.
2614 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
2617 * config/rs6000/vsx.md (??r): New mode attribute.
2618 (*vsx_mov<mode>_64bit): Use it.
2619 (*vsx_mov<mode>_32bit): Likewise.
2621 2018-03-30 Martin Sebor <msebor@redhat.com>
2623 PR tree-optimization/84818
2624 * builtins.c (check_access): Use warning_n.
2626 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2629 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
2631 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
2634 2018-03-30 Julia Koval <julia.koval@intel.com>
2637 * x86-tune.def (movx, partial_reg_dependency): Enable for
2640 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
2643 * lra-constraints.c (process_alt_operands): Move setting
2644 this_alternative_matches below.
2646 2018-03-29 Martin Liska <mliska@suse.cz>
2649 * doc/invoke.texi: Document how LTO works with debug info.
2650 Describe auto-load support of binutils. Mention 'x86-64'
2651 as valid option value of -march option.
2653 2018-03-29 Jakub Jelinek <jakub@redhat.com>
2655 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
2658 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
2659 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
2660 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
2663 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
2666 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
2667 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
2668 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
2669 for RS6000_BTM_POWERPC64.
2670 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
2671 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
2672 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
2677 2018-03-28 Carl Love <cel@us.ibm.com>
2680 2017-09-27 Carl Love <cel@us.ibm.com>
2682 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
2683 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
2684 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
2687 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2689 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
2690 instead of __vector bool.
2691 (_mm_max_pu8): Likewise.
2692 (_mm_min_pi16): Likewise.
2694 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
2697 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
2698 (DIVWEUO): Likewise.
2700 (DIVDEUO): Likewise.
2701 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
2702 DIVWEUO and DIVDEUO.
2703 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
2704 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
2705 (div_extend): Likewise.
2706 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
2708 (__builtin_divweuo): Likewise.
2709 (__builtin_divdeo): Likewise.
2710 (__builtin_divdeuo): Likewise.
2712 2018-03-28 Jakub Jelinek <jakub@redhat.com>
2715 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
2716 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
2718 PR tree-optimization/82004
2719 * gimple-match-head.c (optimize_pow_to_exp): New function.
2720 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
2721 Don't fold to exp if optimize_pow_to_exp is false.
2723 2018-03-28 Martin Liska <mliska@suse.cz>
2726 * calls.c (initialize_argument_information): Fix trailing space.
2727 * common.opt: Fix typo and provide better explanation for
2728 -fsanitize-coverage option.
2729 * config/i386/i386.opt: Fix typo.
2731 2018-03-28 Jakub Jelinek <jakub@redhat.com>
2732 Martin Liska <mliska@suse.cz>
2735 * gimplify.c (asan_poison_variable): Don't do the check for
2736 gimplify_omp_ctxp here.
2737 (gimplify_decl_expr): Do it here.
2738 (gimplify_target_expr): Likewise.
2740 2018-03-28 Martin Liska <mliska@suse.cz>
2743 * config/i386/i386.c (ix86_function_arg_advance): Do not call
2744 chkp_type_bounds_count if MPX is not enabled.
2746 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
2748 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
2750 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
2753 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
2754 function to create the function decl for complex long double
2755 multiply and divide for -mabi=ieeelongdouble.
2756 (init_float128_ieee): Call it.
2758 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
2761 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
2762 -fcf-protection=branch -mibt.
2763 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
2765 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2768 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
2770 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
2773 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
2774 extern array declarations.
2776 2018-03-27 Richard Biener <rguenther@suse.de>
2779 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
2780 explicit single_use checks.
2782 2018-03-27 Richard Biener <rguenther@suse.de>
2784 PR tree-optimization/85082
2785 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
2788 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2790 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
2791 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
2792 Turn on fasynchronous-unwind-tables and funwind-tables.
2794 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
2797 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
2798 (*bmi_blsr_<mode>_ccz): Ditto.
2800 2018-03-26 Tom de Vries <tom@codesourcery.com>
2802 PR tree-optimization/85063
2803 * omp-general.c (offloading_function_p): New function. Factor out
2805 * omp-offload.c (pass_omp_target_link::gate): ... here.
2806 * omp-general.h (offloading_function_p): Declare.
2807 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
2808 with attribute omp declare target for offloading functions.
2810 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
2812 PR tree-optimization/84005
2813 * tree-data-ref.h (get_base_for_alignment): Declare.
2814 * tree-data-ref.c (get_base_for_alignment_1): New function.
2815 (get_base_for_alignment): Likewise.
2816 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
2817 get_base_for_alignment to find a suitable base object, instead
2818 of always using drb->base_address.
2820 2018-03-23 Jakub Jelinek <jakub@redhat.com>
2823 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
2824 known size by default.
2826 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
2829 * lra-constraints.c (process_alt_operands): Don't match BLKmode
2830 and non BLKmode operands.
2832 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2835 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
2836 Clean up attributes.
2838 2018-03-23 Richard Biener <rguenther@suse.de>
2841 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
2842 we are going to emit early debug for LTO.
2844 2018-03-23 Jakub Jelinek <jakub@redhat.com>
2847 * function.c (match_asm_constraints_1): Don't optimize if input
2848 doesn't satisfy general_operand predicate for output's mode.
2851 * alias.c (write_dependence_p): Don't require for x_canonicalized
2852 non-VOIDmode if x has VOIDmode.
2855 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
2856 just don't try to optimize it rather than assert it never happens.
2858 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
2860 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
2861 macro expansions for definition of ST_INTERNAL_<mode> and
2862 LD_INTERNAL_<mode> builtins.
2863 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
2865 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
2867 (altivec_expand_st_builtin): Likewise.
2868 (altivec_expand_builtin): Remove calls to deleted functions.
2869 (rs6000_address_for_altivec): Delete this function.
2870 * config/rs6000/vector.md: Remove expands for
2871 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
2873 2018-03-22 Sudakshina Das <sudi.das@arm.com>
2876 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
2877 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
2878 re-computing once computed.
2879 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
2880 (arm_init_machine_status): Initialize
2881 machine->static_chain_stack_bytes.
2883 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
2886 * doc/extend.texi: Add four new prototypes for vec_ld.
2887 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
2888 definitions for more logical presentation.
2889 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
2890 entries for V1TI variants of __builtin_altivec_ld builtin.
2891 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
2892 handling of V1TI variant of LVX icode pattern.
2893 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
2894 (rs6000_gimple_fold_builtin): Likewise.
2895 (altivec_init_builtins): Add code to define
2896 __builtin_altivec_lvx_v1ti function.
2898 2018-03-22 Jakub Jelinek <jakub@redhat.com>
2901 * function.c (match_asm_constraints_1): Don't do the optimization
2902 if input isn't a REG, SUBREG, MEM or constant.
2904 2018-03-22 Tom de Vries <tom@codesourcery.com>
2906 PR tree-optimization/84956
2907 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
2908 bb_has_abnormal_pred.
2910 2018-03-22 Jakub Jelinek <jakub@redhat.com>
2913 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
2914 DECL_INITIAL (decl) to decl at the end.
2915 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
2918 2018-03-21 Joseph Myers <joseph@codesourcery.com>
2920 * doc/extend.texi (__builtin_tgmath): Document when complex
2921 integer types are treated as _Complex _Float64.
2923 2018-03-21 Tom de Vries <tom@codesourcery.com>
2925 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
2927 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2929 PR tree-optimization/84960
2930 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
2931 if it is ENTRY block, move them into single succ of ENTRY in that case.
2933 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
2935 PR tree-optimization/84811
2936 * poly-int.h (poly_span_traits): Remove the T3 parameter and
2937 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
2938 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
2939 (known_subrange_p): Update accordingly. Cast each value involved
2940 in the size comparison, rather than casting the result of the
2943 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2945 PR tree-optimization/84982
2946 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
2947 by flipping the least significant bit rather than all bits from
2948 bitpos to bitpos + bitsize - 1.
2950 2018-03-21 Nathan Sidwell <nathan@acm.org>
2952 * doc/extend.texi (Deprecated Features): Remove mention of
2953 long-deleted deprecations.
2955 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2958 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
2959 * configure: Regenerate.
2961 2018-03-21 Tom de Vries <tom@codesourcery.com>
2963 PR tree-optimization/83126
2964 * tree-parloops.c (num_phis): New function.
2965 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
2967 2018-03-21 Nathan Sidwell <nathan@acm.org>
2969 * doc/extend.texi (Deprecated Features): Update deprecated flags,
2970 mention anon-struct/union members and trailing attributes.
2972 2018-03-21 Bin Cheng <bin.cheng@arm.com>
2974 PR tree-optimization/84969
2975 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
2976 builtin memset partitions if they set different rhs values.
2978 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2980 PR rtl-optimization/84989
2981 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
2982 VEC_DUPLICATE with scalar result mode.
2984 2018-03-21 Martin Liska <mliska@suse.cz>
2987 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
2988 not intended return statement.
2990 2018-03-21 Martin Liska <mliska@suse.cz>
2993 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
2994 (chkp_find_bound_slots_1): Limit number of iterations.
2996 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
2999 * Minor grammar fixes for x86 options.
3001 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3004 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
3005 holding REG_CFA_RESTORE notes, instead turn them into a USE.
3007 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
3010 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
3011 (altivec_lvx_<mode>_1op): Likewise.
3012 (altivec_stvx_<mode>_2op): Likewise.
3013 (altivec_stvx_<mode>_1op): Likewise.
3014 (altivec_lvx_<VM2:mode>): New define_expand.
3015 (altivec_stvx_<VM2:mode>): Likewise.
3016 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
3017 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
3018 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
3019 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
3020 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
3021 (rs6000_gen_lvx): Likewise.
3022 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
3023 (altivec_expand_stv_builtin): Likewise.
3024 (altivec_expand_builtin): Likewise.
3025 * config/rs6000/vector.md: Likewise.
3027 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3030 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
3033 2018-03-20 Richard Biener <rguenther@suse.de>
3036 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
3037 sign-conversions as zero, fall back to standard scalar_stmt
3040 2018-03-20 Martin Liska <mliska@suse.cz>
3043 * predict.c (rebuild_frequencies): Handle case when we have
3044 PROFILE_ABSENT, but flag_guess_branch_prob is false.
3046 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3049 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
3050 flag_section_anchors.
3051 * varasm.c (use_blocks_for_decl_p): Remove hack for
3052 dw2_force_const_mem.
3055 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
3057 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
3058 be created, use lowpart_subreg of operands[0] rather than operands[0]
3060 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
3061 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
3062 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
3063 and n constraint instead of aarch64_shift_imm_di and Usd.
3064 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
3065 (*aarch64_<optab>_reg_minus<mode>3): ... this.
3067 2018-03-20 Sudakshina Das <sudi.das@arm.com>
3070 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
3071 to favor GPR over NEON registers.
3072 (<shift>di3_neon): Likewise.
3074 2018-03-20 Tom de Vries <tom@codesourcery.com>
3077 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
3078 (nvptx_process_pars): Emit bar.sync asap and alap.
3080 2018-03-20 Tom de Vries <tom@codesourcery.com>
3083 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
3084 seen_label if seen_label is already set.
3086 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3089 * config/i386/i386.c (fold_builtin_cpu): For features above 31
3090 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
3091 Use 1U instead of 1. Formatting fixes.
3094 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
3095 instead of TREE_TYPE (s1) for the return value.
3097 2018-03-19 Jakub Jelinek <jakub@redhat.com>
3099 PR tree-optimization/84946
3100 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
3101 bitsize + bitsize in poly_uint64 rather than poly_int64.
3104 * dwarf2asm.c: Include fold-const.c.
3105 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
3106 of decl rather than decl itself.
3108 PR rtl-optimization/84643
3109 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
3111 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
3114 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
3115 calling assemble_variable.
3117 2018-03-19 Sudakshina Das <sudi.das@arm.com>
3120 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
3121 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
3123 2018-03-19 Jim Wilson <jimw@sifive.com>
3126 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
3127 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
3128 (riscv_first_stack_step): Likewise.
3129 (riscv_option_override): Use STACK_BOUNDARY instead of
3131 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
3133 (BIGGEST_ALIGNMENT): Set to 128.
3134 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
3135 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
3138 2018-03-19 Richard Biener <rguenther@suse.de>
3140 PR tree-optimization/84933
3141 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
3142 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
3144 2018-03-19 Richard Biener <rguenther@suse.de>
3146 PR tree-optimization/84859
3147 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
3148 (cond_if_else_store_replacement): Perform sinking operation on
3149 single-store BBs regardless of MAX_STORES_TO_SINK setting.
3150 Generalize what a BB with a single eligible store is.
3152 2018-03-19 Richard Biener <rguenther@suse.de>
3154 PR tree-optimization/84929
3155 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
3156 chrec_is_positive against non-chrec arg.
3158 2018-03-19 Tamar Christina <tamar.christina@arm.com>
3161 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
3163 2018-03-18 Martin Liska <mliska@suse.cz>
3165 PR rtl-optimization/84635
3166 * regrename.c (build_def_use): Use matches_mode only when
3169 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
3171 PR tree-optimization/84913
3172 * tree-vect-loop.c (vectorizable_reduction): Don't try to
3173 vectorize chains of COND_EXPRs.
3175 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3177 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
3179 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3181 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
3183 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3185 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
3187 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
3188 Kito Cheng <kito.cheng@gmail.com>
3190 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
3191 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
3192 (nds32_adjust_reg_alloc_order): New function.
3193 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
3195 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
3197 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
3198 nds32_print_operand, nds32_print_operand_address): Use
3199 HOST_WIDE_INT_PRINT_DEC instead.
3201 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
3203 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
3205 2018-03-17 Jakub Jelinek <jakub@redhat.com>
3208 * config/i386/i386.c (initial_ix86_tune_features,
3209 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
3211 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
3212 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
3213 rather than 1u << ix86_tune. Formatting fix.
3214 (ix86_option_override_internal): Change ix86_arch_mask from
3215 unsigned int to unsigned HOST_WIDE_INT, initialize to
3216 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
3217 (ix86_function_specific_restore): Likewise.
3219 2018-03-16 Jakub Jelinek <jakub@redhat.com>
3222 * postreload.c (reload_combine_recognize_pattern): Perform
3223 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
3224 truncate_int_for_mode the result for the destination's mode.
3227 * hsa-gen.c (mem_type_for_type): Fix comment typo.
3228 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
3230 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
3233 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
3236 * lra-assigns.c (lra_split_hard_reg_for): Don't use
3237 regno_allocno_class_array and sorted_pseudos.
3238 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
3239 insns where regno is used.
3241 2018-03-16 Martin Liska <mliska@suse.cz>
3244 * multiple_target.c (create_dispatcher_calls): Redirect
3245 reference in the symbol table.
3247 2018-03-16 Martin Liska <mliska@suse.cz>
3250 * multiple_target.c (create_dispatcher_calls): Redirect also
3253 2018-03-16 Jakub Jelinek <jakub@redhat.com>
3257 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
3258 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
3259 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
3261 2018-03-16 Julia Koval <julia.koval@intel.com>
3263 * doc/invoke.texi (Skylake Server): Add CLWB.
3264 Cannonlake): Remove CLWB.
3266 2018-03-16 Jakub Jelinek <jakub@redhat.com>
3268 PR tree-optimization/84841
3269 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
3271 (FLOAT_ONE_CONST_TYPE): Define.
3272 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
3273 (sort_by_operand_rank): Put entries with higher constant_type last
3274 rather than first to match comments.
3276 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
3278 * config/nios2/nios2.md (movsi_internal): Fix thinko in
3281 2018-03-15 Jakub Jelinek <jakub@redhat.com>
3284 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
3285 check and use address of target always.
3287 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
3290 * config/i386/i386.c (indirect_thunk_needed): Update comments.
3291 (indirect_thunk_bnd_needed): Likewise.
3292 (indirect_thunks_used): Likewise.
3293 (indirect_thunks_bnd_used): Likewise.
3294 (indirect_return_needed): New.
3295 (indirect_return_bnd_needed): Likewise.
3296 (output_indirect_thunk_function): Add a bool argument for
3298 (output_indirect_thunk_function): Don't generate alias for
3299 function return thunk.
3300 (ix86_code_end): Call output_indirect_thunk_function to generate
3301 function return thunks.
3302 (ix86_output_function_return): Set indirect_return_bnd_needed
3303 and indirect_return_needed instead of indirect_thunk_bnd_needed
3304 and indirect_thunk_needed.
3306 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
3308 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
3309 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
3310 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
3312 2018-03-15 David Malcolm <dmalcolm@redhat.com>
3313 Paul Hua <paul.hua.gm@gmail.com>
3316 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
3318 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
3320 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
3321 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
3324 2018-03-15 Tamar Christina <tamar.christina@arm.com>
3327 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
3328 instead of GET_MODE_SIZE when comparing Units.
3330 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
3333 * varasm.c (hash_section): Return an unchangeble hash value
3334 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
3335 Return !aarch64_can_use_per_function_literal_pools_p ().
3337 2018-03-15 Jakub Jelinek <jakub@redhat.com>
3340 * optabs.c (emit_conditional_move): Pass address of cmode's copy
3341 rather than address of cmode as last argument to prepare_cmp_insn.
3343 2018-03-15 Julia Koval <julia.koval@intel.com>
3345 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
3346 F_AVX512VNNI, F_AVX512BITALG): New.
3348 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
3351 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
3352 insn for floating-point loads and stores.
3354 2018-03-14 Carl Love <cel@us.ibm.com>
3356 * config/rs6000/rs6000-c.c: Add macro definitions for
3357 ALTIVEC_BUILTIN_VEC_PERMXOR.
3358 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
3359 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
3360 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
3361 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
3363 * config/doc/extend.texi: Add prototypes for vec_permxor.
3365 2018-03-14 David Malcolm <dmalcolm@redhat.com>
3368 * diagnostic-show-locus.c (class layout_point): Convert m_line
3369 from int to linenum_type.
3370 (line_span::comparator): Use linenum "compare" function when
3371 comparing line numbers.
3372 (test_line_span): New function.
3373 (layout_range::contains_point): Convert param "row" from int to
3375 (layout_range::intersects_line_p): Likewise.
3376 (layout::will_show_line_p): Likewise.
3377 (layout::print_source_line): Likewise.
3378 (layout::should_print_annotation_line_p): Likewise.
3379 (layout::print_annotation_line): Likewise.
3380 (layout::print_leading_fixits): Likewise.
3381 (layout::annotation_line_showed_range_p): Likewise.
3382 (struct line_corrections): Likewise for field m_row.
3383 (line_corrections::line_corrections): Likewise for param "row".
3384 (layout::print_trailing_fixits): Likewise.
3385 (layout::get_state_at_point): Likewise.
3386 (layout::get_x_bound_for_row): Likewise.
3387 (layout::print_line): Likewise.
3388 (diagnostic_show_locus): Likewise for locals "last_line" and
3390 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
3391 * input.c (selftest::test_linenum_comparisons): New function.
3392 (selftest::input_c_tests): Call it.
3393 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
3394 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
3395 * selftest.h (ASSERT_GT): New macro.
3396 (ASSERT_GT_AT): New macro.
3397 (ASSERT_LT): New macro.
3398 (ASSERT_LT_AT): New macro.
3400 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
3402 PR rtl-optimization/84780
3403 * combine.c (distribute_links): Don't make a link based on pc_rtx.
3405 2018-03-14 Martin Liska <mliska@suse.cz>
3407 * tree.c (record_node_allocation_statistics): Use
3408 get_stats_node_kind.
3409 (get_stats_node_kind): New function extracted from
3410 record_node_allocation_statistics.
3411 (free_node): Use get_stats_node_kind.
3413 2018-03-14 Richard Biener <rguenther@suse.de>
3415 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
3416 that the value-set of ANTIC_IN doesn't grow.
3419 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
3421 (BB_VISITED_WITH_VISITED_SUCCS): New define.
3422 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
3424 2018-03-14 Julia Koval <julia.koval@intel.com>
3426 * config.gcc (icelake-client, icelake-server): New.
3428 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
3429 (initial_ix86_arch_features): Ditto.
3430 (PTA_SKYLAKE): Add SGX.
3431 (PTA_ICELAKE): Remove.
3432 (PTA_ICELAKE_CLIENT): New.
3433 (PTA_ICELAKE_SERVER): New.
3434 (ix86_option_override_internal): Split up icelake on icelake client and
3436 (get_builtin_code_for_version): Ditto.
3437 (fold_builtin_cpu): Ditto.
3438 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
3439 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
3440 * config/i386/i386.h (processor_type): Ditto.
3441 * doc/invoke.texi: Ditto.
3443 2018-03-14 Jakub Jelinek <jakub@redhat.com>
3446 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
3447 INTEGER_CST offset, add it together with bitpos / 8 and
3448 sign extend based on POINTER_SIZE.
3452 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
3455 * config/i386/constraints.md (Yc): New register constraint.
3456 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
3457 Use Yc constraint for alternative 2 of operand 0. Remove
3458 preferred_for_speed attribute.
3460 2018-03-14 Richard Biener <rguenther@suse.de>
3462 PR tree-optimization/84830
3463 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
3464 with the old one to avoid oscillations.
3466 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
3469 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
3471 (assign_by_spills): Return a flag of reload assignment failure.
3472 Do not process the reload assignment failures. Do not spill other
3473 reload pseudos if they has the same reg class. Update n if
3475 (lra_assign): Add a return arg. Set up from the result of
3476 assign_by_spills call.
3477 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
3478 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
3479 usage_insns if it is not NULL.
3480 (spill_hard_reg_in_range): New function.
3481 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
3482 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
3483 function prototypes.
3484 (lra_assign): Change prototype.
3485 * lra.c (lra): Add code to deal with fails by splitting hard reg
3488 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
3490 * config/riscv/riscv.opt (mrelax): New option.
3491 * config/riscv/riscv.c (riscv_file_start): Emit ".option
3492 "norelax" when riscv_mrelax is disabled.
3493 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
3495 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3498 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
3499 reassociation for int modes.
3501 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
3503 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
3504 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
3506 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
3507 * config/aarch64/aarch64-sve.md
3508 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
3509 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
3510 (*extend<mode><Vwide>2): Rename to...
3511 (aarch64_sve_extend<mode><Vwide>2): ...this.
3512 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
3513 renaming the old pattern to...
3514 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
3516 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
3517 define_expand, renaming the old pattern to...
3518 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
3519 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
3520 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
3521 account when deciding which SVE instruction the optab should use.
3522 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
3524 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
3526 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
3527 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
3528 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
3529 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
3530 (tlsdesc_small_<mode>): Turn a define_expand and use
3531 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
3532 (tlsdesc_small_advsimd_<mode>): ...this.
3533 (tlsdesc_small_sve_<mode>): New pattern.
3535 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
3537 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
3538 (UNSPEC_UMUL_HIGHPART): New constants.
3539 (MUL_HIGHPART): New int iteraor.
3540 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
3541 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
3543 (*<su>mul<mode>3_highpart): New define_insn.
3545 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
3548 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
3551 2018-03-13 Martin Liska <mliska@suse.cz>
3554 * (sem_item_optimizer::sem_item_optimizer): Initialize new
3556 (sem_item_optimizer::~sem_item_optimizer): Release it.
3557 (sem_item_optimizer::merge_classes): Register variable aliases.
3558 (sem_item_optimizer::fixup_pt_set): New function.
3559 (sem_item_optimizer::fixup_points_to_sets): Likewise.
3560 * ipa-icf.h: Declare new variables and functions.
3562 2018-03-13 Jakub Jelinek <jakub@redhat.com>
3565 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
3566 integer_pow2p@2 and test integer_pow2p in condition.
3567 (A < 0 ? C : 0): Similarly for @1.
3570 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
3571 characters starting at p contain '\0' character, don't look beyond
3575 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
3576 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
3579 * reg-stack.c (change_stack): Change update_end var from int to
3580 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
3581 also call set_block_for_insn on the newly added insns and rescan.
3584 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
3585 on the last operand.
3588 * tree.c (stabilize_reference_1): Return save_expr (e) for
3589 STATEMENT_LIST even if it doesn't have side-effects.
3591 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
3593 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
3595 2018-03-12 Renlin Li <renlin.li@arm.com>
3597 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
3598 aarch64_output_scalar_simd_mov_immediate.
3600 2018-03-12 Martin Sebor <msebor@redhat.com>
3602 PR tree-optimization/83456
3603 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
3604 for perfectly overlapping calls to memcpy.
3605 (gimple_fold_builtin_memory_chk): Same.
3606 (gimple_fold_builtin_strcpy): Handle no-warning.
3607 (gimple_fold_builtin_stxcpy_chk): Same.
3608 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
3610 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
3612 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
3613 parameter. Use it for SFmode.
3614 (rs6000_function_arg_advance_1): Adjust.
3615 (rs6000_function_arg): Adjust.
3616 (rs6000_gimplify_va_arg): Pass false for that new parameter.
3618 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
3620 PR rtl-optimization/84169
3621 PR rtl-optimization/84780
3622 * combine.c (can_combine_p): Check for a 2-insn combination whether
3623 the destination register is used between the two insns, too.
3625 2018-03-12 Richard Biener <rguenther@suse.de>
3627 PR tree-optimization/84803
3628 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
3629 for refs DR analysis didn't process.
3631 2018-03-12 Richard Biener <rguenther@suse.de>
3633 PR tree-optimization/84777
3634 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
3635 force-vectorize loops ignore whether we are optimizing for size.
3637 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
3639 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
3640 (TARGET_MD_ASM_ADJUST): Define.
3642 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
3643 Kito Cheng <kito.cheng@gmail.com>
3644 Chung-Ju Wu <jasonwucj@gmail.com>
3646 * config/nds32/nds32.c (nds32_compute_stack_frame,
3647 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
3648 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
3649 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
3650 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
3651 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
3652 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
3653 * config/nds32/nds32.md (prologue, epilogue): Use macro
3654 NDS32_V3PUSH_AVAILABLE_P to do checking.
3656 2018-03-11 Jakub Jelinek <jakub@redhat.com>
3659 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
3660 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
3661 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
3662 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
3663 addition of most attributes on !orig_type_die or the attribute not
3664 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
3666 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3667 Chung-Ju Wu <jasonwucj@gmail.com>
3669 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
3671 * config/nds32/nds32.opt (mvh): New option.
3673 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3674 Chung-Ju Wu <jasonwucj@gmail.com>
3676 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
3678 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
3679 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
3682 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3683 Chung-Ju Wu <jasonwucj@gmail.com>
3685 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
3687 * config/nds32/nds32-multiple.md (strlensi): New pattern.
3688 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
3690 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
3691 Kito Cheng <kito.cheng@gmail.com>
3692 Chung-Ju Wu <jasonwucj@gmail.com>
3694 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
3695 UNSPEC_FFMISM and UNSPEC_FLMISM.
3696 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
3697 for ffb, ffmism and flmism.
3698 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
3699 (unspec_ffmism): Ditto.
3700 (unspec_flmism): Ditto.
3701 (nds32_expand_builtin_impl): Check if string extension is available.
3702 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
3703 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
3705 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
3708 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
3711 * lra-assigns.c (assign_by_spills): Return a flag of reload
3712 assignment failure. Do not process the reload assignment
3713 failures. Do not spill other reload pseudos if they has the same
3715 (lra_assign): Add a return arg. Set up from the result of
3716 assign_by_spills call.
3717 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
3718 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
3719 usage_insns if it is not NULL.
3720 (spill_hard_reg_in_range): New function.
3721 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
3722 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
3723 function prototypes.
3724 (lra_assign): Change prototype.
3725 * lra.c (lra): Add code to deal with fails by splitting hard reg
3728 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
3731 * config/i386/i386.opt: Replace Enforcment with Enforcement.
3733 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
3736 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
3737 (dw_val_node): Add val_symbolic_view.
3738 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
3739 (symview_upper_bound): New.
3740 (new_line_info_table): Initialize symviews_since_reset.
3741 (dwarf2out_source_line): Count symviews_since_reset and set
3742 symview_upper_bound.
3743 (dw_val_equal_p): Handle symview.
3744 (add_AT_symview): New.
3745 (print_dw_val): Handle symview.
3746 (attr_checksum, attr_checksum_ordered): Likewise.
3747 (same_dw_val_p, size_of_die): Likewise.
3748 (value_format, output_die): Likewise.
3749 (add_high_low_attributes): Use add_AT_symview for entry_view.
3750 (dwarf2out_finish): Reset symview_upper_bound, clear
3753 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
3756 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
3757 Add strict argument and use it.
3758 (rs6000_split_multireg_move): Update for new strict argument.
3759 (mem_operand_gpr): Disallow all non-offsettable addresses.
3760 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
3762 2018-03-09 Jakub Jelinek <jakub@redhat.com>
3765 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
3766 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
3767 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
3770 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
3771 decl, use remap_type if we want to use the type.
3773 2018-03-09 Martin Sebor <msebor@redhat.com>
3775 PR tree-optimization/84526
3776 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
3778 (builtin_access::generic_overlap): Be prepared to handle non-array
3781 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
3783 PR rtl-optimization/84682
3784 * lra-constraints.c (process_address_1): Check is_address flag
3785 for address constraints.
3786 (process_alt_operands): Likewise.
3787 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
3788 preprocess_constraints.
3789 * recog.h (preprocess_constraints): Add oploc parameter.
3791 * recog.c (preprocess_constraints): Test address_operand for
3792 CT_ADDRESS constraints.
3794 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
3797 * lra-assigns.c (assign_by_spills): Return a flag of reload
3798 assignment failure. Do not process the reload assignment
3799 failures. Do not spill other reload pseudos if they has the same
3801 (lra_assign): Add a return arg. Set up from the result of
3802 assign_by_spills call.
3803 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
3804 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
3805 usage_insns if it is not NULL.
3806 (spill_hard_reg_in_range): New function.
3807 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
3808 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
3809 function prototypes.
3810 (lra_assign): Change prototype.
3811 * lra.c (lra): Add code to deal with fails by splitting hard reg
3814 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3817 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
3818 Accept complain bool parameter. Only emit errors if it is true.
3819 (arm_parse_cpu_option_name): Likewise.
3820 (arm_target_thumb_only): Adjust callers of the above.
3821 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
3822 prototype to take a default true bool parameter.
3823 (arm_parse_arch_option_name): Likewise.
3825 2018-03-09 David Malcolm <dmalcolm@redhat.com>
3826 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
3830 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
3831 * configure: Regenerate.
3832 * configure.ac ("linker --version-script option"): New.
3833 ("linker soname option"): New.
3835 2018-03-09 Richard Biener <rguenther@suse.de>
3837 PR tree-optimization/84775
3838 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
3839 immediate uses of predicate stmts and mark them modified.
3842 PR tree-optimization/84178
3843 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
3845 (version_loop_for_if_conversion): Delay update_ssa call.
3846 (tree_if_conversion): Delay update_ssa until after predicate
3849 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
3852 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
3853 when the function accesses prior frames.
3855 2018-03-08 Jakub Jelinek <jakub@redhat.com>
3858 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
3859 gen_llsym, otherwise call maybe_gen_llsym.
3862 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
3863 has ',' character inside of it.
3865 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3868 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
3869 as clobbering CC_REGNUM.
3871 2018-03-08 Richard Biener <rguenther@suse.de>
3874 * tree-scalar-evolution.c: Include tree-into-ssa.h.
3875 (follow_copies_to_constant): Do not follow SSA names registered
3878 2018-03-08 Richard Biener <rguenther@suse.de>
3880 PR tree-optimization/84178
3881 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
3883 (version_loop_for_if_conversion): Delay update_ssa call.
3884 (tree_if_conversion): Delay update_ssa until after predicate
3887 2018-03-08 David Malcolm <dmalcolm@redhat.com>
3889 PR tree-optimization/84178
3890 * tree-if-conv.c (release_bb_predicate): Remove the
3891 the assertion that the stmts have NULL use_ops.
3892 Discard the statements, asserting that they haven't
3893 yet been added to a BB.
3895 2018-03-08 Richard Biener <rguenther@suse.de>
3897 PR tree-optimization/84746
3898 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
3899 (phi_translate): Pass in destination ANTIC_OUT set.
3900 (phi_translate_1): Likewise. For a simplified result lookup
3901 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
3902 (phi_translate_set): Adjust.
3903 (do_pre_regular_insertion): Likewise.
3904 (do_pre_partial_partial_insertion): Likewise.
3906 2018-03-08 Martin Liska <mliska@suse.cz>
3908 PR gcov-profile/84735
3909 * doc/gcov.texi: Document usage of profile files.
3910 * gcov-io.h: Document changes in the format.
3912 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
3916 * dwarf2out.c (struct dw_line_info_table): Update comments for
3918 (FORCE_RESET_NEXT_VIEW): New.
3919 (FORCE_RESETTING_VIEW_P): New.
3920 (RESETTING_VIEW_P): Check for -1 too.
3921 (ZERO_VIEW_P): Likewise.
3922 (new_line_info_table): Force-reset next view.
3923 (dwarf2out_begin_function): Likewise.
3924 (dwarf2out_source_line): Simplify zero_view_p initialization.
3925 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
3926 view directly. Omit view when omitting .loc at line 0.
3928 2018-03-08 Jakub Jelinek <jakub@redhat.com>
3930 PR tree-optimization/84740
3931 * tree-switch-conversion.c (process_switch): Call build_constructors
3932 only if info.phi_count is non-zero.
3934 PR tree-optimization/84739
3935 * tree-tailcall.c (find_tail_calls): Check call arguments against
3936 DECL_ARGUMENTS (current_function_decl) rather than
3937 DECL_ARGUMENTS (func) when checking for tail recursion.
3939 2018-03-07 Jakub Jelinek <jakub@redhat.com>
3941 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
3942 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
3943 Volker Reichelt's entry and add entries for people that perform
3944 GCC fuzzy testing and report numerous bugs.
3946 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
3949 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
3950 readonly data in sdata, if that is disabled.
3951 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
3952 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
3953 -mreadonly-in-sdata option.
3955 2018-03-07 Martin Sebor <msebor@redhat.com>
3957 PR tree-optimization/84468
3958 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
3959 basic block when looking for nul assignment.
3961 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
3964 * except.h (output_function_exception_table): Adjust prototype.
3965 * except.c (output_function_exception_table): Remove FNNAME parameter
3966 and add SECTION parameter. Ouput one part of the table at a time.
3967 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
3968 the first part of the exception table and emit unwind directives.
3969 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
3970 (i386_pe_seh_cold_init): Likewise.
3971 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
3972 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
3973 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
3974 (ix86_output_call_insn): Emit a nop in one more case for SEH.
3975 * config/i386/winnt.c: Include except.h.
3976 (struct seh_frame_state): Add reg_offset, after_prologue and
3977 in_cold_section fields.
3978 (i386_pe_seh_end_prologue): Set seh->after_prologue.
3979 (i386_pe_seh_cold_init): New function.
3980 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
3981 to seh->in_cold_section.
3982 (seh_emit_push): Record the offset of the push.
3983 (seh_emit_save): Record the offet of the save.
3984 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
3985 Test seh->after_prologue to disregard the epilogue.
3986 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
3987 (i386_pe_end_cold_function): New function.
3989 2018-03-07 Jakub Jelinek <jakub@redhat.com>
3992 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
3993 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
3996 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
3998 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
3999 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
4002 * multiple_target.c: Include tree-inline.h and intl.h.
4003 (expand_target_clones): Diagnose and fail if node->definition and
4004 !tree_versionable_function_p (node->decl).
4006 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
4008 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
4010 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
4011 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
4012 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
4014 2018-03-06 Jakub Jelinek <jakub@redhat.com>
4017 * combine.c (try_combine): Use reg_or_subregno instead of handling
4018 just paradoxical SUBREGs and REGs.
4020 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
4022 * config/arc/arc.c (arc_finalize_pic): Remove function.
4023 (arc_must_save_register): We use single base PIC register, remove
4024 checks to save/restore the PIC register.
4025 (arc_expand_prologue): Likewise.
4026 * config/arc/arc-protos.h (arc_set_default_type_attributes):
4028 (arc_verify_short): Likewise.
4029 (arc_attr_type): Likewise.
4030 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
4031 (walk_stores): Likewise.
4032 (arc_address_cost): Make it static.
4033 (arc_verify_short): Likewise.
4034 (branch_dest): Likewise.
4035 (arc_attr_type): Likewise.
4036 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
4037 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
4038 (arc_final_prescan_insn): Remove inserting the nops due to
4039 hardware hazards. It is done in reorg step.
4040 (insn_length_variant_t): Remove.
4041 (insn_length_parameters_t): Likewise.
4042 (arc_insn_length_parameters): Likewise.
4043 (arc_get_insn_variants): Likewise.
4044 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
4046 2018-03-06 Jakub Jelinek <jakub@redhat.com>
4049 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
4052 PR tree-optimization/84687
4053 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
4055 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
4057 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4059 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
4060 Rename to ppc_speculation_barrier.
4061 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
4062 __builtin_ppc_speculation_barrier.
4064 2018-03-05 Jakub Jelinek <jakub@redhat.com>
4067 * combine.c (combine_simplify_rtx): Don't try to simplify if
4068 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
4071 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
4073 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
4074 to 32 bytes when compiling for POWER9.
4076 2018-03-05 Jakub Jelinek <jakub@redhat.com>
4079 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
4080 regparm >= 3 with no arg reg available also for calls with
4081 flag_force_indirect_call. Pass decl to ix86_function_regparm.
4084 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
4086 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
4088 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
4091 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
4093 2018-03-05 Richard Biener <rguenther@suse.de>
4095 PR tree-optimization/84486
4096 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
4097 When inserting a __builtin_assume_aligned call set the LHS
4098 SSA name alignment info accordingly.
4100 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
4102 PR tree-optimization/84114
4103 * config/aarch64/aarch64.c (aarch64_reassociation_width)
4104 Avoid reassociation of FLOAT_MODE addition.
4106 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
4108 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
4109 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
4110 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
4111 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
4112 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
4113 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
4114 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
4116 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
4117 __builtin_ia32_wbinvd): New builtins.
4118 (SPECIAL_ARGS2): New.
4119 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
4120 (SPECIAL_ARGS2): New.
4121 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
4122 (ix86_valid_target_attribute_inner_p): Ditto.
4123 (ix86_init_mmx_sse_builtins): Add special_args2.
4124 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
4125 TARGET_WBNOINVD_P): New.
4126 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
4127 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
4128 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
4129 * config/i386/immintrin.h (_wbinvd): New intrinsic.
4130 * config/i386/pconfigintrin.h: New file.
4131 * config/i386/wbnoinvdintrin.h: Ditto.
4132 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
4133 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
4135 2018-03-05 Richard Biener <rguenther@suse.de>
4137 PR tree-optimization/84670
4138 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4140 (BB_VISITED_WITH_VISITED_SUCCS): New define.
4141 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4142 (compute_antic_aux): Only assert the number of values in ANTIC_IN
4143 doesn't grow if all successors (recursively) were visited at least
4146 2018-03-05 Richard Biener <rguenther@suse.de>
4148 PR tree-optimization/84650
4149 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
4150 if executed in the loop pipeline.
4152 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
4154 * doc/configfiles.texi (Configuration Files): Move info about
4155 conditionalizing $target-protos.h to...
4156 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
4157 differs from $target-protos.h.
4159 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
4160 Chung-Ju Wu <jasonwucj@gmail.com>
4162 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
4163 * config/nds32/nds32-multiple.md (setmemsi): Define.
4164 * config/nds32/nds32-memory-manipulation.c
4165 (nds32_gen_dup_4_byte_to_word_value): New.
4166 (emit_setmem_word_loop): New.
4167 (emit_setmem_byte_loop): New.
4168 (nds32_expand_setmem_loop): New.
4169 (nds32_expand_setmem_loop_v3m): New.
4170 (nds32_expand_setmem_unroll): New.
4171 (nds32_expand_setmem): New.
4173 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4174 Chung-Ju Wu <jasonwucj@gmail.com>
4176 * config/nds32/nds32-memory-manipulation.c
4177 (nds32_emit_load_store): New.
4178 (nds32_emit_post_inc_load_store): New.
4179 (nds32_emit_mem_move): New.
4180 (nds32_emit_mem_move_block): New.
4181 (nds32_expand_movmemsi_loop_unknown_size): New.
4182 (nds32_expand_movmemsi_loop_known_size): New.
4183 (nds32_expand_movmemsi_loop): New.
4184 (nds32_expand_movmemsi_unroll): New.
4185 (nds32_expand_movmemqi): Rename ...
4186 (nds32_expand_movmemsi): ... to this.
4187 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
4188 (movmemsi): ... to this.
4189 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
4190 (nds32_expand_movmemsi): ... to this.
4192 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4193 Monk Chiang <sh.chiang04@gmail.com>
4194 Chung-Ju Wu <jasonwucj@gmail.com>
4196 * config/nds32/nds32-protos.h
4197 (nds32_expand_load_multiple): New arguments.
4198 (nds32_expand_store_multiple): Ditto.
4199 (nds32_valid_multiple_load_store): Rename ...
4200 (nds32_valid_multiple_load_store_p): ... to this.
4201 * config/nds32/nds32-memory-manipulation.c
4202 (nds32_expand_load_multiple): Refine implementation.
4203 (nds32_expand_store_multiple): Ditto.
4204 * config/nds32/nds32-multiple.md
4205 (load_multiple): Update nds32_expand_load_multiple interface.
4206 (store_multiple): Update nds32_expand_store_multiple interface.
4207 * config/nds32/nds32-predicates.c
4208 (nds32_valid_multiple_load_store): Rename ...
4209 (nds32_valid_multiple_load_store_p): ... to this and refine
4211 * config/nds32/predicates.md
4212 (nds32_load_multiple_and_update_address_operation): New predicate.
4213 (nds32_store_multiple_and_update_address_operation): New predicate.
4215 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4216 Chung-Ju Wu <jasonwucj@gmail.com>
4218 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
4219 (combo): New attribute.
4220 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
4222 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
4224 * config/nds32/nds32.opt: Change -mcmodel= default value.
4226 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
4227 Monk Chiang <sh.chiang04@gmail.com>
4228 Chung-Ju Wu <jasonwucj@gmail.com>
4230 * config/nds32/constants.md (unspec_element): New enum.
4231 * config/nds32/constraints.md (Umw): New constraint.
4232 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
4233 * config/nds32/nds32-intrinsic.md: Likewise.
4234 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
4235 (nds32_valid_smw_lwm_base_p): New.
4236 (nds32_output_smw_single_word): New.
4237 (nds32_output_lmw_single_word): New.
4238 (nds32_expand_unaligned_load): New.
4239 (nds32_expand_unaligned_store): New.
4240 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
4241 (nds32_output_smw_single_word): Declare.
4242 (nds32_output_lmw_single_word): Declare.
4243 (nds32_expand_unaligned_load): Declare.
4244 (nds32_expand_unaligned_store): Declare.
4245 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
4246 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
4247 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
4248 NDS32_BUILTIN_UASTORE_DW.
4249 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
4252 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
4253 Kito Cheng <kito.cheng@gmail.com>
4254 Chung-Ju Wu <jasonwucj@gmail.com>
4256 * config/nds32/nds32-intrinsic.c
4257 (nds32_expand_builtin_null_ftype_reg): Delete.
4258 (nds32_expand_builtin_reg_ftype_imm): Ditto.
4259 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
4260 (nds32_read_argument): New.
4261 (nds32_legitimize_target): Ditto.
4262 (nds32_legitimize_argument): Ditto.
4263 (nds32_check_constant_argument): Ditto.
4264 (nds32_expand_unop_builtin): Ditto.
4265 (nds32_expand_unopimm_builtin): Ditto.
4266 (nds32_expand_binop_builtin): Ditto.
4267 (nds32_builtin_decl_impl): Ditto.
4268 (builtin_description): Ditto.
4269 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
4270 (nds32_init_builtins_impl): Ditto.
4271 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
4272 (nds32_builtin_decl): New.
4273 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
4274 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
4276 2018-03-02 Jeff Law <law@redhat.com>
4278 * reorg.c (stop_search_p): Handle DEBUG_INSN.
4279 (redundant_insn, fill_simple_delay_slots): Likewise.
4280 (fill_slots_from_thread): Likewise.
4281 * resource.c (mark_referenced_resources): Likewise.
4282 (mark_set_resources, find_dead_or_set_registers): Likewise.
4284 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4286 * substring-locations.h (format_warning_va): Formatting fix for
4288 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
4290 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
4291 * substring-locations.c: Include intl.h.
4292 (format_warning_va): Turned into small wrapper around
4293 format_warning_n_va, renamed to ...
4294 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
4295 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
4297 (format_warning_at_substring_n): New function.
4298 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
4299 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
4300 format_warning_at_substring with just a shorter name instead of
4301 const function pointer.
4302 (fmtwarn_n): New function.
4303 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
4304 appropriate, get rid of all the fmtstr temporaries, move conditionals
4305 with G_() wrapped string literals directly into fmtwarn arguments,
4306 cast dir.len to (int), formatting fixes.
4308 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
4310 * doc/invoke.texi: Remove "Cilk Plus" references.
4312 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4313 Richard Biener <rguenther@suse.de>
4316 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
4317 for error or warning attributes if CALL_FROM_THUNK_P is set.
4320 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4323 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
4324 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
4327 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
4328 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
4330 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
4331 instead of -1U in last predictors element's probability member.
4333 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
4336 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
4337 arguments if they are comparable.
4339 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
4341 PR tree-optimization/84634
4342 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
4343 masks and masked_loop_p with a single loop_masks, making sure it's
4344 null for bb vectorization.
4346 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
4348 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
4349 (vect_analyze_data_ref_access): Use loop->safe_len rather than
4350 loop->force_vectorize to check whether there is no alias.
4352 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4355 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
4357 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
4359 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
4360 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
4361 instead of a loop around prev_real_insn.
4362 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
4366 * config/i386/i386.c (ix86_print_operand): Use conditional
4367 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
4370 2018-03-02 Richard Biener <rguenther@suse.de>
4372 PR tree-optimization/84427
4373 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
4374 (bitmap_set_subtract_values): Rewrite to handle multiple
4377 (prune_clobbered_mems): Likewise.
4378 (phi_translate): Take edge instead of pred/phiblock.
4379 (phi_translate_1): Likewise.
4380 (phi_translate_set): Likewise. Insert all translated
4381 exprs for a value into the set, keeping possibly multiple
4382 expressions per value.
4383 (compute_antic_aux): Adjust for phi_translate changes.
4384 When intersecting union the expressions and prune those
4385 not in the final value set, keeping possibly multiple
4386 expressions per value. Do not use value-insertion
4387 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
4388 all expressions. Add verification that the value-sets
4389 only shrink during iteration.
4390 (compute_partial_antic_aux): Adjust for the phi_translate changes.
4391 (do_pre_regular_insertion): Likewise.
4392 (do_pre_partial_partial_insertion): Likewise.
4394 2018-03-02 Richard Biener <rguenther@suse.de>
4397 * config/darwin.c (saved_debug_info_level): New static global.
4398 (darwin_asm_lto_start): Disable debug info generation for LTO out.
4399 (darwin_asm_lto_end): Restore debug info generation settings.
4401 2018-03-01 Martin Liska <mliska@suse.cz>
4404 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
4407 2018-03-01 Richard Biener <rguenther@suse.de>
4410 * dwarf2out.c (gen_variable_die): Properly handle late VLA
4411 type annotation with LTO when debug was disabled at compile-time.
4413 2018-03-01 Matthew Fortune <mfortune@gmail.com>
4415 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
4417 (mips_final_postscan_insn): Likewise.
4419 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
4421 PR rtl-optimization/84528
4422 * alias.c (init_alias_target): Add commentary.
4423 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
4424 a unique base value if the frame pointer is not eliminated
4425 to the stack pointer.
4427 2018-03-01 Tom de Vries <tom@codesourcery.com>
4429 PR rtl-optimization/83327
4430 * lra-int.h (hard_regs_spilled_into): Declare.
4431 * lra.c (hard_regs_spilled_into): Define.
4432 (init_reg_info): Init hard_regs_spilled_into.
4433 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
4434 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
4435 (process_bb_lives): Handle hard_regs_spilled_into.
4436 (lra_create_live_ranges_1): Before doing liveness propagation, clear
4437 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
4439 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
4441 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
4442 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
4443 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
4444 * config/rs6000/aix72.h: New file.
4446 2018-02-28 Jakub Jelinek <jakub@redhat.com>
4448 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
4449 instead of warning_at with conditional singular and plural messages
4453 * stor-layout.c (update_alignment_for_field): For
4454 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
4455 && !DECL_PACKED (field), do the alignment update, just use
4456 only desired_align instead of MAX (type_align, desired_align)
4458 (place_field): Don't do known_align < desired_align handling
4459 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
4460 is non-NULL, instead do it after rli->prev_field handling and
4461 only if not within a bitfield word. For DECL_PACKED (field)
4462 use type_align of BITS_PER_UNIT.
4464 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
4466 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
4467 superfluous parentheses and trailing spaces.
4469 2018-02-28 Richard Biener <rguenther@suse.de>
4471 PR tree-optimization/84584
4472 * graphite-scop-detection.c (scop_detection::add_scop): Discard
4473 SCoPs with fake exit edge.
4475 2018-02-28 Martin Liska <mliska@suse.cz>
4478 * timevar.c (timer::print): Fix format to properly print 100%
4481 2018-02-28 Richard Biener <rguenther@suse.de>
4484 * genmatch.c (capture_info::walk_match): Do not mark
4485 captured expressions without operands as expr_p given
4486 they act more like predicates and should be subject to
4487 "lost tail" side-effect preserving.
4489 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
4491 PR rtl-optimization/81611
4492 * auto-inc-dec.c (attempt_change): Move dead note from
4493 mem_insn if it's the next use of regno
4494 (find_address): Take address use of reg holding
4495 non-incremented value. Add parm to limit search to the named
4497 (merge_in_block): Attempt to use a mem insn that is the next
4498 use of the original regno.
4500 2018-02-27 Martin Sebor <msebor@redhat.com>
4503 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
4504 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
4506 2018-02-27 Martin Sebor <msebor@redhat.com>
4508 PR translation/84207
4509 * diagnostic-core.h (warning_n, error_n, inform_n): Change
4510 n argument to unsigned HOST_WIDE_INT.
4511 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
4512 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
4513 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
4514 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
4516 2018-02-27 Richard Biener <rguenther@suse.de>
4518 PR tree-optimization/84512
4519 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
4520 Do not use the estimate returned from record_stmt_cost for
4521 the scalar iteration cost but sum properly using add_stmt_cost.
4523 2018-02-27 Richard Biener <rguenther@suse.de>
4525 PR tree-optimization/84466
4526 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
4527 Adjust last change to less strictly validate use operands.
4529 2018-02-27 Martin Liska <mliska@suse.cz>
4531 PR gcov-profile/84548
4532 * gcov.c (process_file): Allow partial overlap and consider it
4533 also as group functions.
4534 (output_lines): Properly calculate range of lines for a group.
4536 2018-02-27 Martin Liska <mliska@suse.cz>
4538 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
4539 'ggc' suffixes. Change first column width.
4540 (timer::print): Fix formatting of the column.
4542 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
4544 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
4545 preserve inline entry blocks for the sake of debug inline
4546 entry point markers alone.
4547 (remove_unused_locals): Suggest in comments a better place to
4548 force the preservation of inline entry blocks that are
4549 otherwise unused, but do not preserve them.
4551 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
4553 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
4555 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
4558 * config/i386/constraints.md (Bs): Replace
4559 ix86_indirect_branch_register with
4560 TARGET_INDIRECT_BRANCH_REGISTER.
4562 * config/i386/i386.md (indirect_jump): Likewise.
4563 (tablejump): Likewise.
4564 (*sibcall_memory): Likewise.
4565 (*sibcall_value_memory): Likewise.
4566 Peepholes of indirect call and jump via memory: Likewise.
4567 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
4568 (*sibcall_value_GOT_32): Likewise.
4569 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4570 (GOT_memory_operand): Likewise.
4571 (call_insn_operand): Likewise.
4572 (sibcall_insn_operand): Likewise.
4573 (GOT32_symbol_operand): Likewise.
4574 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
4576 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
4578 PR rtl-optimization/83496
4579 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
4580 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
4581 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
4582 redundant insn, if any.
4583 (relax_delay_slots): Likewise.
4584 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
4586 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
4588 PR tree-optimization/83965
4589 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
4590 that grouped statements are part of a reduction chain. Return
4591 true if the statement is not marked as a reduction itself but
4593 (vect_recog_dot_prod_pattern): Don't check whether the statement
4594 is part of a group here.
4595 (vect_recog_sad_pattern): Likewise.
4596 (vect_recog_widen_sum_pattern): Likewise.
4598 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
4601 * final.c (rest_of_clean_state): Also look for calls inside sequences.
4603 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
4606 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
4608 (ix86_output_indirect_function_return): New prototype.
4609 (ix86_split_simple_return_pop_internal): Likewise.
4610 * config/i386/i386.c (indirect_return_via_cx): New.
4611 (indirect_return_via_cx_bnd): Likewise.
4612 (indirect_thunk_name): Handle return va CX_REG.
4613 (output_indirect_thunk_function): Create alias for
4614 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
4615 (ix86_output_indirect_jmp): Remove the bool argument.
4616 (ix86_output_indirect_function_return): New function.
4617 (ix86_split_simple_return_pop_internal): Likewise.
4618 * config/i386/i386.md (*indirect_jump): Don't pass false
4619 to ix86_output_indirect_jmp.
4620 (*tablejump_1): Likewise.
4621 (simple_return_pop_internal): Change it to define_insn_and_split.
4622 Call ix86_split_simple_return_pop_internal to split it for
4624 (simple_return_indirect_internal): Call
4625 ix86_output_indirect_function_return instead of
4626 ix86_output_indirect_jmp.
4628 2018-02-26 Jakub Jelinek <jakub@redhat.com>
4631 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
4632 memset and value initialization afterwards.
4634 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
4636 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
4638 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4641 * common/config/aarch64/aarch64-common.c
4642 (aarch_option_optimization_table[]): Switch
4643 off fomit-frame-pointer
4645 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
4646 Chung-Ju Wu <jasonwucj@gmail.com>
4648 * config/nds32/nds32-multiple.md (load_multiple): Disallow
4650 (store_multiple): Ditto.
4652 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
4654 * config.gcc: Add --with-cpu support for nds32 target.
4655 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
4656 * config/nds32/nds32.opt: Add -mcpu= option.
4658 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
4660 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
4661 isel=yes): Warn for these deprecated options.
4663 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
4665 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
4666 ISA_2_5_MASKS_EMBEDDED.
4668 2018-02-23 Jakub Jelinek <jakub@redhat.com>
4670 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
4671 p->max as pointers rather than using iterative_hash_expr.
4673 2018-02-23 Carl Love <cel@us.ibm.com>
4675 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
4676 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
4678 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
4679 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
4680 P8V_BUILTIN_VEC_VUNSIGNED2.
4682 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
4685 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
4686 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
4688 * lra-constraints.c (curr_insn_transform): Set up
4689 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
4691 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
4692 * lra-eliminations.c (spill_pseudos): Ditto.
4693 (process_insn_for_elimination): Ditto.
4694 * lra-lives.c (reg_early_clobber_p): Use the new macros.
4695 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
4696 LRA_NON_CLOBBERED_ALT.
4698 2018-02-22 Martin Sebor <msebor@redhat.com>
4700 PR tree-optimization/84480
4701 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
4702 to maybe_diag_stxncpy_trunc. Call it.
4703 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
4704 from gimple_fold_builtin_strcpy. Print inlining stack.
4705 (handle_builtin_stxncpy): Print inlining stack.
4706 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
4708 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
4711 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
4712 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
4713 and -fcheck-pointer-bounds are used together.
4714 (indirect_thunk_prefix): New enum.
4715 (indirect_thunk_need_prefix): New function.
4716 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
4717 "_nt" instead of "_bnd" for NOTRACK prefix.
4718 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
4719 (output_indirect_thunk_function): Likewise.
4721 (ix86_code_end): Update output_indirect_thunk_function calls.
4722 (ix86_output_indirect_branch_via_reg): Replace
4723 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
4724 (ix86_output_indirect_branch_via_push): Likewise.
4725 (ix86_output_function_return): Likewise.
4726 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
4727 incompatible with -fcf-protection=branch and
4728 -fcheck-pointer-bounds.
4730 2018-02-22 Steve Ellcey <sellcey@cavium.com>
4733 * config/aarch64/aarch64.c (aarch64_print_address_internal):
4734 Change gcc_assert call to output_operand_lossage.
4736 2018-02-22 Steve Ellcey <sellcey@cavium.com>
4738 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
4740 2018-02-22 DJ Delorie <dj@redhat.com>
4741 Sebastian Perta <sebastian.perta@renesas.com>
4742 Oleg Endo <olegendo@gcc.gnu.org>
4744 * config/rx/rx.c (rx_rtx_costs): New function.
4745 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
4747 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
4749 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
4751 2018-02-22 Martin Liska <mliska@suse.cz>
4754 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
4755 Add "native" as a possible value.
4757 2018-02-22 Martin Liska <mliska@suse.cz>
4760 * config/i386/i386.c (ix86_option_override_internal):
4761 Add "native" as a possible value for -march and -mtune.
4763 2018-02-22 Jakub Jelinek <jakub@redhat.com>
4766 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
4767 to all type variants.
4769 PR tree-optimization/84503
4770 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
4771 width as info->bitpos + info->bitsize - start.
4772 (merged_store_group::merge_overlapping): Simplify width computation.
4773 (check_no_overlap): New function.
4774 (imm_store_chain_info::try_coalesce_bswap): Compute expected
4775 start + width and last_order of the group, fail if check_no_overlap
4777 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
4778 to group if check_no_overlap fails.
4780 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
4782 * config/rs6000/altivec.md: Delete contraint arguments to
4783 define_expand, define_split, and define_peephole2, and in
4784 define_insn_and_split if always unused.
4785 * config/rs6000/darwin.md: Ditto.
4786 * config/rs6000/dfp.md: Ditto.
4787 * config/rs6000/rs6000.md: Ditto.
4788 * config/rs6000/sync.md: Ditto.
4789 * config/rs6000/vector.md: Ditto.
4790 * config/rs6000/vsx.md: Ditto.
4792 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
4794 * config/rs6000/altivec.md: Write output control strings as braced
4795 blocks instead of double-quoted strings.
4796 * config/rs6000/darwin.md: Ditto.
4797 * config/rs6000/rs6000.md: Ditto.
4798 * config/rs6000/vector.md: Ditto.
4799 * config/rs6000/vsx.md: Ditto.
4801 2018-02-21 Jason Merrill <jason@redhat.com>
4803 PR c++/84314 - ICE with templates and fastcall attribute.
4804 * attribs.c (build_type_attribute_qual_variant): Remove assert.
4806 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
4808 * ipa-cp.c (determine_versionability): Fix comment typos.
4810 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
4813 * ipa-cp.c (determine_versionability): Do not version functions caling
4816 2018-02-21 Martin Liska <mliska@suse.cz>
4819 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
4820 Add "native" as a possible value.
4821 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
4822 the macro when native cpu detection is available.
4824 2018-02-21 Martin Liska <mliska@suse.cz>
4827 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
4828 Add "native" as a possible value.
4829 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
4830 when native cpu detection is available.
4832 2018-02-21 Jakub Jelinek <jakub@redhat.com>
4833 Martin Sebor <msebor@redhat.com>
4835 PR tree-optimization/84478
4836 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
4838 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
4839 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
4840 support which is conservatively correct, for 2 only stay conservative
4841 for maxlen. Formatting and comment capitalization fixes. Add STRICT
4842 argument to the 2 argument get_range_strlen, adjust 6 arg
4843 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
4845 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
4846 (gimple_fold_builtin_strlen): Pass true as last argument to
4849 2018-02-20 Martin Sebor <msebor@redhat.com>
4852 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
4853 (builtin_memref::set_base_and_offset): Same. Handle inner references.
4854 (builtin_memref::builtin_memref): Factor out parts into
4855 set_base_and_offset and call it.
4857 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
4860 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
4861 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
4862 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
4863 search at the associated MODE_INT.
4865 2018-02-20 Jeff Law <law@redhat.com>
4868 PR tree-optimization/81592
4870 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
4871 for range data rather than using global data.
4872 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
4873 range data rather than using global data.
4874 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
4875 pass it to children as needed.
4876 (struct directive::fmtresult): Similarly.
4877 (struct directive::set_width): Similarly.
4878 (struct directive::set_precision): Similarly.
4879 (format_integer, format_directive, parse_directive): Similarly.
4880 (format_none): Accept unnamed vr_values parameter.
4881 (format_percent, format_floating, format_character): Similarly.
4882 (format_string, format_plain): Similarly.
4883 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
4884 the EVRP range analyzer for range data rather than using global data.
4885 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
4886 gimple-ssa-evrp-analyze.h
4887 (class sprintf_dom_walker): Add after_dom_children member function.
4888 Add evrp_range_analyzer member.
4889 (sprintf_dom_walker::before_dom_children): Call into the EVRP
4890 range analyzer as needed.
4891 (sprintf_dom_walker::after_dom_children): New member function.
4892 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
4894 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
4895 (evrp_range_analyzer::pop_to_marker): Likewise.
4897 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
4899 PR tree-optimization/84419
4900 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
4901 with the required type if its current type is compatible but
4904 2018-02-20 Jakub Jelinek <jakub@redhat.com>
4907 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
4908 after vectorization.
4910 2018-02-20 Martin Liska <mliska@suse.cz>
4913 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
4914 possible values if we don't have a hint.
4916 2018-02-20 Martin Liska <mliska@suse.cz>
4920 * final.c (shorten_branches): Build align_tab array with one
4922 * opts.c (finish_options): Add alignment option limit check.
4923 (MAX_CODE_ALIGN): Likewise.
4924 (MAX_CODE_ALIGN_VALUE): Likewise.
4925 * doc/invoke.texi: Document maximum allowed option value for
4926 all -falign-* options.
4928 2018-02-19 Jakub Jelinek <jakub@redhat.com>
4931 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
4932 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
4933 * var-tracking.c (emit_note_insn_var_location): Remove all references
4934 to NOTE_INSN_CALL_ARG_LOCATION.
4935 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
4936 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
4937 Use copy_rtx_if_shared.
4938 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
4939 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
4940 (dwarf2out_var_location): Remove handling of
4941 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
4943 * final.c (final_scan_insn): Remove all references to
4944 NOTE_INSN_CALL_ARG_LOCATION.
4945 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
4946 before dumping final insns.
4947 * except.c (emit_note_eh_region_end): Remove all references to
4948 NOTE_INSN_CALL_ARG_LOCATION.
4949 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
4950 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
4951 * config/arc/arc.c (hwloop_optimize): Likewise.
4952 * config/arm/arm.c (create_fix_barrier): Likewise.
4953 * config/s390/s390.c (s390_chunkify_start): Likewise.
4954 * config/sh/sh.c (find_barrier): Likewise.
4955 * config/i386/i386.c (rest_of_insert_endbranch,
4956 ix86_seh_fixup_eh_fallthru): Likewise.
4957 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
4958 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
4959 * config/frv/frv.c (frv_function_prologue): Likewise.
4960 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
4962 (note_outside_basic_block_p): Remove all references to
4963 NOTE_INSN_CALL_ARG_LOCATION.
4964 * gengtype.c (adjust_field_rtx_def): Likewise.
4965 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
4967 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
4968 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
4971 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
4974 PR tree-optimization/84452
4975 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
4976 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
4979 2018-02-19 Martin Liska <mliska@suse.cz>
4982 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
4984 2018-02-19 Martin Liska <mliska@suse.cz>
4985 Richard Sandiford <richard.sandiford@linaro.org>
4987 PR tree-optimization/82491
4988 * gimple-fold.c (get_base_constructor): Make earlier bail out
4991 2018-02-19 Carl Love <cel@us.ibm.com>
4993 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
4994 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
4996 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
4997 P8V_BUILTIN_VEC_NEG.
4999 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
5001 * config/rl78/rl78.md (movdf): New define expand.
5003 2018-02-19 Martin Liska <mliska@suse.cz>
5006 * doc/invoke.texi: Fix typo.
5007 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
5009 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
5011 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
5012 handle rs6000_single_float and rs6000_double_float specially for
5015 2018-02-16 Jeff Law <law@redhat.com>
5017 * config/rx/rx.c (add_pop_cfi_notes): New function.;
5020 2018-02-16 Jakub Jelinek <jakub@redhat.com>
5023 * ipa-inline.c (inline_small_functions): Fix a typo.
5025 2018-02-16 Nathan Sidwell <nathan@acm.org>
5027 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
5029 2018-02-16 Carl Love <cel@us.ibm.com>
5031 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
5032 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
5033 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
5034 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
5035 expansion to P8V_BUILTIN_VEC_FLOAT2.
5037 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
5039 PR rtl-optimization/70023
5040 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
5041 src_regno into account.
5043 2018-02-16 Carl Love <cel@us.ibm.com>
5045 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
5046 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
5047 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
5048 * config/rs6000/rs6000.c: Remove case statements for
5049 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
5050 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
5051 and P9V_BUILTIN_VEC_VINSERT4B.
5052 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
5053 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
5054 * config/rs6000/vsx.md:
5055 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
5058 2018-02-16 Carl Love <cel@us.ibm.com>
5060 * config/rs6000/altivec.h: Add builtin names vec_extract4b
5062 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
5064 * config/rs6000/rs6000-c.c: Add the definitions for
5065 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
5066 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
5067 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
5068 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
5069 definition for insert4b and define insn *insert3b_internal.
5070 * doc/extend.texi: Add documentation for vec_extract4b.
5072 2018-02-16 Nathan Sidwell <nathan@acm.org>
5074 * doc/extend.texi (Backwards Compatibility): Mention friend
5075 injection. Note for-scope is deprecated.
5076 * doc/invoke.texi (-ffriend-injection): Deprecate.
5078 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
5080 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
5081 that moved to I2, also allow destinations that are a paradoxical
5082 subreg (instead of a normal reg).
5084 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
5087 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
5090 2018-02-16 Richard Biener <rguenther@suse.de>
5092 PR tree-optimization/84037
5093 PR tree-optimization/84016
5095 * config/i386/i386.c (ix86_builtin_vectorization_cost):
5096 Adjust vec_construct for the fact we need additional higher latency
5097 128bit inserts for AVX256 and AVX512 vector builds.
5098 (ix86_add_stmt_cost): Scale vector construction cost for
5101 2018-02-16 Richard Biener <rguenther@suse.de>
5103 PR tree-optimization/84417
5104 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
5105 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
5106 (non_rewritable_lvalue_p): Likewise, use poly-ints.
5108 2018-02-16 Martin Liska <mliska@suse.cz>
5111 * internal-fn.def (ASAN_CHECK): Set proper flags.
5112 (ASAN_MARK): Likewise.
5114 2018-02-16 Julia Koval <julia.koval@intel.com>
5116 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
5117 from PTA_CANNONLAKE.
5119 2018-02-16 Jakub Jelinek <jakub@redhat.com>
5122 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
5123 Use ++iter rather than iter++ for std::list iterators.
5124 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
5125 defer deleting them until all nodes in the forest are processed. Do
5126 free even leaf nodes. Change to_process into auto_vec.
5129 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
5130 * vec.h (vec_default_construct): Use memset instead of placement new
5131 if BROKEN_VALUE_INITIALIZATION is defined.
5132 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
5133 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
5136 PR rtl-optimization/83723
5137 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
5138 * lra.c (lra_substitute_pseudo): Likewise. If true, use
5139 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
5141 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
5143 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
5145 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
5147 PR rtl-optimization/81443
5148 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
5149 from inner REGs to paradoxical SUBREGs.
5151 2018-02-16 Richard Biener <rguenther@suse.de>
5153 PR tree-optimization/84399
5154 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
5155 For operands we can analyze at their definition make sure we can
5156 analyze them at each use as well.
5158 2018-02-16 Richard Biener <rguenther@suse.de>
5160 PR tree-optimization/84190
5161 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
5162 volatile accesses if the decl isn't volatile.
5164 2018-02-15 Jason Merrill <jason@redhat.com>
5166 PR c++/84314 - ICE with templates and fastcall attribute.
5167 * attribs.c (build_type_attribute_qual_variant): Don't clobber
5168 TYPE_CANONICAL on an existing type.
5170 2018-02-15 Jakub Jelinek <jakub@redhat.com>
5172 PR tree-optimization/84383
5173 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
5174 dstoff nor call operand_equal_p if dstbase is NULL.
5176 PR tree-optimization/84334
5177 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
5178 also a CONSTANT_CLASS_P, punt.
5180 2018-02-14 Jim Wilson <jimw@sifive.com>
5182 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
5183 first SMALL_OPERAND check. New local min_second_step. Move assert
5184 to where locals are set. Add TARGET_RVC support.
5185 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
5187 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
5189 * doc/invoke.texi: Correct -Wformat-overflow code sample.
5191 2018-02-14 Martin Sebor <msebor@redhat.com>
5193 PR tree-optimization/83698
5194 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
5195 arrays constrain the offset range to their bounds.
5196 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
5197 (builtin_access::overlap): Avoid setting the size of overlap if it's
5199 (maybe_diag_overlap): Also consider arrays when deciding what values
5200 of offsets to include in diagnostics.
5202 2018-02-14 Martin Sebor <msebor@redhat.com>
5205 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
5206 that correspond to the kind of a declaration.
5208 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
5211 * config/pa/pa.md: Load address of PIC label using the linkage table
5212 if the label is nonlocal.
5214 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
5216 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
5217 warning message if user requests -maltivec=be.
5218 * doc/invoke.texi: Document deprecation of -maltivec=be.
5220 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
5223 * config/rs6000/rs6000-c.c: Update definitions for
5224 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
5225 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
5227 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
5230 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
5231 add _get_ssp intrinsics. Remove argument from
5232 __builtin_ia32_rdssp[d|q].
5233 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
5234 * config/i386/i386-builtin.def: Remove argument from
5235 __builtin_ia32_rdssp[d|q].
5236 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
5237 ix86_expand_special_args_builtin for _rdssp[d|q].
5238 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
5239 Clear register before usage.
5240 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
5241 Add documentation for new _get_ssp and _inc_ssp intrinsics.
5243 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
5245 PR tree-optimization/84357
5246 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
5247 operand 1 of an ARRAY_REF too.
5249 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
5252 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
5253 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
5255 (set_of_reg): New struct.
5256 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
5257 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
5258 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
5260 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
5261 Split into bitclr, bitset, bitinvert patterns if appropriate.
5262 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
5263 use rx_fuse_in_memory_bitop.
5264 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
5265 to named insn, correct maximum insn length.
5267 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
5270 * machmode.def: Define a complex mode for PARTIAL_INT.
5271 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
5273 * doc/rtl.texi: Document CSPImode.
5274 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
5276 (msp430_hard_regno_nregs_with_padding): Likewise.
5278 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
5281 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
5283 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
5285 PR rtl-optimization/84169
5286 * combine.c (try_combine): New variable split_i2i3. Set it to true if
5287 we generated a parallel as new i3 and we split that to new i2 and i3
5288 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
5289 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
5290 those to i2, not i1. Partially rewrite this scan code.
5292 2018-02-13 Jakub Jelinek <jakub@redhat.com>
5295 * stor-layout.c (place_field): For variable length fields, adjust
5296 offset_align afterwards not just based on the field's alignment,
5297 but also on the size.
5300 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
5301 of exps and logs in the use_exp2 case.
5303 2018-02-13 Jeff Law <law@redhat.com>
5305 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
5308 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
5311 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
5315 * common.opt (gas-loc-support, gas-locview-support): New.
5316 (ginline-points, ginternal-reset-location-views): New.
5317 * doc/invoke.texi: Document them. Use @itemx where intended.
5318 (gvariable-location-views): Adjust.
5319 * target.def (reset_location_view): New.
5320 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
5321 (TARGET_RESET_LOCATION_VIEW): New.
5322 * doc/tm.texi: Rebuilt.
5323 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
5324 (dwarf2out_default_as_locview_support): New.
5325 (output_asm_line_debug_info): Use option variables.
5326 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
5327 (output_loc_list): Likewise.
5328 (add_high_low_attributes): Check option variables.
5329 Don't output entry view attribute in strict mode.
5330 (gen_inlined_subroutine_die): Check option variables.
5331 (dwarf2out_inline_entry): Likewise.
5332 (init_sections_and_labels): Likewise.
5333 (dwarf2out_early_finish): Likewise.
5334 (maybe_reset_location_view): New, from...
5335 (dwarf2out_var_location): ... here. Call it.
5336 * debug.h (dwarf2out_default_as_loc_support): Declare.
5337 (dwarf2out_default_as_locview_support): Declare.
5338 * hooks.c (hook_int_rtx_insn_0): New.
5339 * hooks.h (hook_int_rtx_insn_0): Declare.
5340 * toplev.c (process_options): Take -gas-loc-support and
5341 -gas-locview-support from dwarf2out. Enable
5342 -gvariable-location-views by default only with locview
5343 assembler support. Enable -ginternal-reset-location-views by
5344 default only if the target defines the corresponding hook.
5345 Enable -ginline-points by default if location views are
5346 enabled; force it disabled if statement frontiers are
5348 * tree-inline.c (expand_call_inline): Check option variables.
5349 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
5351 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
5353 PR tree-optimization/84321
5354 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
5355 handling. Also check whether the anti-range contains any values
5356 that satisfy the mask; switch to a VR_RANGE if not.
5358 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
5361 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
5363 2018-02-13 Martin Jambor <mjambor@suse.cz>
5366 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
5367 of call statements, also set location of a load to a temporary.
5369 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
5371 * config/rl78/rl78.c (add_vector_labels): New function.
5372 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
5373 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
5374 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
5375 which checks that no arguments are passed.
5376 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
5377 * doc/extend.texi: Documentation for the new attribute.
5379 2018-02-13 Andreas Schwab <schwab@suse.de>
5381 * config/riscv/linux.h (CPP_SPEC): Define.
5383 2018-02-13 Jakub Jelinek <jakub@redhat.com>
5386 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
5387 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
5388 OPTION_MASK_ISA_AES as first argument to def_builtin_const
5389 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
5390 instead of OPTION_MASK_ISA_PCLMUL as first argument to
5391 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
5392 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
5393 temporarily for AES and PCLMUL builtins.
5395 PR tree-optimization/84339
5396 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
5397 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
5401 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
5402 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
5403 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
5405 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
5407 * omp-simd-clone.h: New file.
5408 * omp-simd-clone.c: Include omp-simd-clone.h.
5409 (expand_simd_clones): No longer static.
5410 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
5411 cgraph.h and omp-simd-clone.h.
5412 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
5413 (vect_recog_widen_shift_pattern): Formatting fix.
5414 (vect_pattern_recog_1): Don't check optab for calls.
5417 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
5418 operands[2] into a REG before using gen_lowpart on it.
5420 2018-02-12 Jeff Law <law@redhat.com>
5423 * config/sh/sh.c (find_barrier): Consider a sibling call
5426 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
5427 successfully back substituting a reg.
5429 2018-02-12 Richard Biener <rguenther@suse.de>
5431 PR tree-optimization/84037
5432 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
5433 parameter, move visited init to caller.
5434 (vect_slp_analyze_operations): Separate cost from validity
5435 check, initialize visited once for all instances.
5436 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
5438 * tree-vect-stmts.c (vect_model_simple_cost): Make early
5440 (vect_model_promotion_demotion_cost): Likewise.
5441 (vectorizable_bswap): Guard cost modeling with !slp_node
5442 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
5444 (vectorizable_call): Likewise.
5445 (vectorizable_conversion): Likewise.
5446 (vectorizable_assignment): Likewise.
5447 (vectorizable_shift): Likewise.
5448 (vectorizable_operation): Likewise.
5449 (vectorizable_store): Likewise.
5450 (vectorizable_load): Likewise.
5451 (vectorizable_condition): Likewise.
5452 (vectorizable_comparison): Likewise.
5454 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
5457 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
5458 (ASAN_MARK): Fix fnspec to account for return value, change pointer
5459 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
5461 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
5464 * params.def (inline-min-speedup): Increase from 8 to 15.
5465 (max-inline-insns-auto): Decrease from 40 to 30.
5466 * ipa-split.c (consider_split): Add some buffer for function to
5467 be considered inlining candidate.
5468 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
5471 2018-02-12 Richard Biener <rguenther@suse.de>
5473 PR tree-optimization/84037
5474 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
5475 matched stmts if we cannot swap the non-matched ones.
5477 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
5479 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
5480 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
5481 _mm_maskz_scalef_round_ss): New intrinsics.
5482 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
5483 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
5484 __builtin_ia32_scalefss_round): Remove.
5485 (__builtin_ia32_scalefsd_mask_round,
5486 __builtin_ia32_scalefss_mask_round): New intrinsics.
5487 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
5488 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
5489 ((match_operand:VF_128 2 "<round_nimm_predicate>"
5490 "<round_constraint>")): Changed to ...
5491 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
5492 "<round_scalar_constraint>")): ... this.
5493 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
5494 %0, %1, %2<round_op3>}"): Changed to ...
5495 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
5496 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
5497 %2<round_scalar_mask_op3>}"): ... this.
5498 * config/i386/subst.md (round_scalar_nimm_predicate): New.
5500 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
5502 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
5503 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
5504 (_mm_maskz_sqrt_round_ss): New intrinsics.
5505 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
5506 (__builtin_ia32_sqrtsd_mask_round)
5507 (__builtin_ia32_sqrtss_mask_round): New builtins.
5508 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
5509 (__builtin_ia32_sqrtss_round): Remove.
5510 (__builtin_ia32_sqrtsd_mask_round)
5511 (__builtin_ia32_sqrtss_mask_round): New builtins.
5512 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
5513 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
5514 ((match_operand:VF_128 1 "vector_operand"
5515 "xBm,<round_constraint>")): Changed to ...
5516 ((match_operand:VF_128 1 "vector_operand"
5517 "xBm,<round_scalar_constraint>")): ... this.
5518 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
5519 %0, %2, %<iptr>1<round_op3>}): Changed to ...
5520 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
5521 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
5522 %<iptr>1<round_scalar_mask_op3>}): ... this.
5523 ((set_attr "prefix" "<round_prefix>")): Changed to ...
5524 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
5526 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
5529 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
5530 Cast vec_cmpeq result to correct type.
5531 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
5532 Cast vec_cmpgt result to correct type.
5534 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
5536 * final.c (final_scan_insn_1): Renamed from...
5537 (final_scan_insn): ... this. New wrapper, to recover
5538 seen from the outermost call in recursive ones.
5539 * config/sparc/sparc.c (output_return): Drop seen from call.
5540 (output_sibcall): Likewise.
5541 * config/visium/visium.c (output_branch): Likewise.
5543 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
5545 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
5548 2018-02-10 Alan Modra <amodra@gmail.com>
5551 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
5552 Specify LR as an input.
5554 2018-02-10 Jakub Jelinek <jakub@redhat.com>
5557 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
5558 remove_member_access_dummy_vars): New functions.
5559 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
5560 lower_omp_1, execute_lower_omp): Use them.
5562 PR rtl-optimization/84308
5563 * shrink-wrap.c (spread_components): Release todo vector.
5565 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
5567 PR rtl-optimization/57193
5568 * ira-color.c (struct allocno_color_data): Add member
5569 conflict_allocno_hard_prefs.
5570 (update_conflict_allocno_hard_prefs): New.
5571 (bucket_allocno_compare_func): Add a preference based on
5572 conflict_allocno_hard_prefs.
5573 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
5574 (color_allocnos): Remove a dead code. Initiate
5575 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
5577 2018-02-09 Jakub Jelinek <jakub@redhat.com>
5580 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
5581 constraint from =wa to wa. Avoid a subreg on the output operand,
5582 instead use a pseudo and subreg it in a move.
5583 (p9_xxbrd_<mode>): Changed to ...
5584 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
5585 (p9_xxbrd_v2df): New expander.
5586 (p9_xxbrw_<mode>): Changed to ...
5587 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
5588 (p9_xxbrw_v4sf): New expander.
5590 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
5592 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
5594 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
5597 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
5598 multiply in 32-bit mode.
5599 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
5600 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
5603 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
5605 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
5606 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
5607 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
5608 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
5610 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
5613 * dwarf2out.c (is_trivial_indirect_ref): New function.
5614 (dwarf2out_late_global_decl): Do not generate a location
5615 attribute for variables that have a non-trivial DECL_VALUE_EXPR
5616 and that are not defined in the current unit.
5618 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
5620 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
5621 instead of a libcall for UNORDERED.
5623 2018-02-09 Tamar Christina <tamar.christina@arm.com>
5626 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
5627 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
5629 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5632 * config/s390/s390.c (s390_set_current_function): Invoke
5633 s390_indirect_branch_settings also if fndecl didn't change.
5635 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
5637 * config/rs6000/rs6000.md (blockage): Set length to zero.
5639 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
5641 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
5643 2018-02-09 Jakub Jelinek <jakub@redhat.com>
5646 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
5647 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
5651 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
5652 PARALLEL incoming that failed vt_get_decl_and_offset check.
5655 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
5656 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
5658 (get_variable_section): For decls in named .bss* sections pass true as
5659 second argument to bss_initializer_p.
5661 2018-02-09 Marek Polacek <polacek@redhat.com>
5662 Jakub Jelinek <jakub@redhat.com>
5665 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
5666 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
5667 Sync some changes from cxx_fold_indirect_ref.
5669 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
5671 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
5673 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
5674 (BLOCK_INLINE_ENTRY_LABEL): New.
5675 (dwarf2out_var_location): Disregard inline entry markers.
5676 (inline_entry_data): New struct.
5677 (inline_entry_data_hasher): New hashtable type.
5678 (inline_entry_data_hasher::hash): New.
5679 (inline_entry_data_hasher::equal): New.
5680 (inline_entry_data_table): New variable.
5681 (add_high_low_attributes): Add DW_AT_entry_pc and
5682 DW_AT_GNU_entry_view attributes if a pending entry is found
5683 in inline_entry_data_table. Add old entry_pc attribute only
5684 if debug nonbinding markers are disabled.
5685 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
5686 markers are enabled.
5687 (block_within_block_p, dwarf2out_inline_entry): New.
5688 (dwarf2out_finish): Check that no entries remained in
5689 inline_entry_data_table.
5690 * final.c (reemit_insn_block_notes): Handle inline entry notes.
5691 (final_scan_insn, notice_source_line): Likewise.
5692 (rest_of_clean_state): Skip inline entry markers.
5693 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
5695 * gimple.c (gimple_build_debug_inline_entry): New.
5696 * gimple.h (enum gimple_debug_subcode): Add
5697 GIMPLE_DEBUG_INLINE_ENTRY.
5698 (gimple_build_debug_inline_entry): Declare.
5699 (gimple_debug_inline_entry_p): New.
5700 (gimple_debug_nonbind_marker_p): Adjust.
5701 * insn-notes.def (INLINE_ENTRY): New.
5702 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
5703 inline entry marker notes.
5704 (print_insn): Likewise.
5705 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
5706 (INSN_DEBUG_MARKER_KIND): Likewise.
5707 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
5708 * tree-inline.c (expand_call_inline): Build and insert
5709 debug_inline_entry stmt.
5710 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
5711 inline entry blocks early, if nonbind markers are enabled.
5712 (dump_scope_block): Dump fragment info.
5713 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
5714 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
5715 (gimple_build_debug_inline_entry): New.
5716 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
5717 Enable/disable inline entry points too.
5718 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
5719 (DEBUG_INSN): Describe inline entry markers.
5721 * common.opt (gvariable-location-views): New.
5722 (gvariable-location-views=incompat5): New.
5723 * config.in: Rebuilt.
5724 * configure: Rebuilt.
5725 * configure.ac: Test assembler for view support.
5726 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
5727 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
5728 * dwarf2out.c (var_loc_view): New typedef.
5729 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
5730 (dwarf2out_locviews_in_attribute): New.
5731 (dwarf2out_locviews_in_loclist): New.
5732 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
5733 (enum dw_line_info_opcode): Add LI_adv_address.
5734 (struct dw_line_info_table): Add view.
5735 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
5736 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
5737 (zero_view_p): New variable.
5738 (ZERO_VIEW_P): New macro.
5739 (output_asm_line_debug_info): New.
5740 (struct var_loc_node): Add view.
5741 (add_AT_view_list, AT_loc_list): New.
5742 (add_var_loc_to_decl): Add view param. Test it against last.
5743 (new_loc_list): Add view params. Record them.
5744 (AT_loc_list_ptr): Handle loc and view lists.
5745 (view_list_to_loc_list_val_node): New.
5746 (print_dw_val): Handle dw_val_class_view_list.
5747 (size_of_die): Likewise.
5748 (value_format): Likewise.
5749 (loc_list_has_views): New.
5750 (gen_llsym): Set vl_symbol too.
5751 (maybe_gen_llsym, skip_loc_list_entry): New.
5752 (dwarf2out_maybe_output_loclist_view_pair): New.
5753 (output_loc_list): Output view list or entries too.
5754 (output_view_list_offset): New.
5755 (output_die): Handle dw_val_class_view_list.
5756 (output_dwarf_version): New.
5757 (output_compilation_unit_header): Use it.
5758 (output_skeleton_debug_sections): Likewise.
5759 (output_rnglists, output_line_info): Likewise.
5760 (output_pubnames, output_aranges): Update version comments.
5761 (output_one_line_info_table): Output view numbers in asm comments.
5762 (dw_loc_list): Determine current endview, pass it to new_loc_list.
5763 Call maybe_gen_llsym.
5764 (loc_list_from_tree_1): Adjust.
5765 (add_AT_location_description): Create view list attribute if
5766 needed, check it's absent otherwise.
5767 (convert_cfa_to_fb_loc_list): Adjust.
5768 (maybe_emit_file): Call output_asm_line_debug_info for test.
5769 (dwarf2out_var_location): Reset views as needed. Precompute
5770 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
5771 attribute. Set view.
5772 (new_line_info_table): Reset next view.
5773 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
5774 (dwarf2out_source_line): Likewise. Output view resets and labels to
5775 the assembler, or select appropriate line info opcodes.
5776 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
5777 (optimize_string_length): Catch it. Adjust.
5778 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
5779 dw_val_class_view_list, and remove it if no longer needed.
5780 (hash_loc_list): Hash view numbers.
5781 (loc_list_hasher::equal): Compare them.
5782 (optimize_location_lists): Check whether a view list symbol is
5783 needed, and whether the locview attribute is present, and
5784 whether they match. Remove the locview attribute if no longer
5786 (index_location_lists): Call skip_loc_list_entry for test.
5787 (dwarf2out_finish): Call output_asm_line_debug_info for test.
5788 Use output_dwarf_version.
5789 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
5790 (struct dw_val_node): Add val_view_list.
5791 * final.c (SEEN_NEXT_VIEW): New.
5792 (set_next_view_needed): New.
5793 (clear_next_view_needed): New.
5794 (maybe_output_next_view): New.
5795 (final_start_function): Rename to...
5796 (final_start_function_1): ... this. Take pointer to FIRST,
5797 add SEEN parameter. Emit param bindings in the initial view.
5798 (final_start_function): Reintroduce SEEN-less interface.
5799 (final): Rename to...
5800 (final_1): ... this. Take SEEN parameter. Output final pending
5801 next view at the end.
5802 (final): Reintroduce seen-less interface.
5803 (final_scan_insn): Output pending next view before switching
5804 sections or ending a block. Mark the next view as needed when
5805 outputting variable locations. Notify debug backend of section
5806 changes, and of location view changes.
5807 (rest_of_handle_final): Adjust.
5808 * toplev.c (process_options): Autodetect value for debug variable
5809 location views option. Warn on incompat5 without -gdwarf-5.
5810 * doc/invoke.texi (gvariable-location-views): New.
5811 (gvariable-location-views=incompat5): New.
5812 (gno-variable-location-views): New.
5814 2018-02-08 David Malcolm <dmalcolm@redhat.com>
5816 PR tree-optimization/84136
5817 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
5818 that the result of find_edge is non-NULL.
5820 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
5823 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
5824 storing integer register in SImode. Fix cost of 256 and 512
5825 byte aligned SSE register store.
5827 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
5829 * config/i386/i386.c (ix86_multiplication_cost): Fix
5830 multiplication cost for TARGET_AVX512DQ.
5832 2018-02-08 Marek Polacek <polacek@redhat.com>
5834 PR tree-optimization/84238
5835 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
5838 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5840 PR tree-optimization/84265
5841 * tree-vect-stmts.c (vectorizable_store): Don't treat
5842 VMAT_CONTIGUOUS accesses as grouped.
5843 (vectorizable_load): Likewise.
5845 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5847 PR tree-optimization/81635
5848 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
5849 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
5850 (test_round_for_mask): New functions.
5851 (wide_int_cc_tests): Call test_round_for_mask.
5852 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
5853 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
5854 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
5855 range returned by get_range_info.
5857 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
5860 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
5861 * symtab.c: Include builtins.h
5862 (symtab_node::output_to_lto_symbol_table_p): Move here
5863 from lto-streamer-out.c:output_symbol_p.
5864 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
5865 (output_symbol_p): Move all logic to symtab.c
5866 (produce_symtab): Update.
5868 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5870 * config/s390/s390-opts.h (enum indirect_branch): Define.
5871 * config/s390/s390-protos.h (s390_return_addr_from_memory)
5872 (s390_indirect_branch_via_thunk)
5873 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
5874 (enum s390_indirect_branch_type): Define.
5875 * config/s390/s390.c (struct s390_frame_layout, struct
5876 machine_function): Remove.
5877 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
5878 (indirect_branch_table_label_no, indirect_branch_table_name):
5880 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
5881 (enum s390_indirect_branch_option): Define.
5882 (s390_return_addr_from_memory): New function.
5883 (s390_handle_string_attribute): New function.
5884 (s390_attribute_table): Add new attribute handler.
5885 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
5886 (s390_indirect_branch_via_thunk): New function.
5887 (s390_indirect_branch_via_inline_thunk): New function.
5888 (s390_function_ok_for_sibcall): When jumping via thunk disallow
5889 sibling call optimization for non z10 compiles.
5890 (s390_emit_call): Force indirect branch target to be a single
5891 register. Add r1 clobber for non-z10 compiles.
5892 (s390_emit_epilogue): Emit return jump via return_use expander.
5893 (s390_reorg): Handle JUMP_INSNs as execute targets.
5894 (s390_option_override_internal): Perform validity checks for the
5895 new command line options.
5896 (s390_indirect_branch_attrvalue): New function.
5897 (s390_indirect_branch_settings): New function.
5898 (s390_set_current_function): Invoke s390_indirect_branch_settings.
5899 (s390_output_indirect_thunk_function): New function.
5900 (s390_code_end): Implement target hook.
5901 (s390_case_values_threshold): Implement target hook.
5902 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
5904 * config/s390/s390.h (struct s390_frame_layout)
5905 (struct machine_function): Move here from s390.c.
5906 (TARGET_INDIRECT_BRANCH_NOBP_RET)
5907 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
5908 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
5909 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
5910 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
5911 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
5912 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
5913 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
5914 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
5915 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
5916 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
5917 (mnemonic attribute): Add values which aren't recognized
5919 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
5920 pattern for branch conversion. Fix mnemonic attribute.
5921 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
5922 indirect branch via thunk if requested.
5923 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
5924 ("*indirect_jump"): Disable for branch conversion using out of
5926 ("indirect_jump_via_thunk<mode>_z10")
5927 ("indirect_jump_via_thunk<mode>")
5928 ("indirect_jump_via_inlinethunk<mode>_z10")
5929 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
5930 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
5931 ("casesi_jump_via_inlinethunk<mode>_z10")
5932 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
5933 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
5934 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
5935 ("*indirect2_jump"): Disable for branch conversion.
5936 ("casesi_jump"): Turn into expander and expand patterns for branch
5938 ("return_use"): New expander.
5939 ("*return"): Emit return via thunk and rename it to ...
5940 ("*return<mode>"): ... this one.
5941 * config/s390/s390.opt: Add new options and and enum for the
5944 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5946 * lra-constraints.c (match_reload): Unconditionally use
5947 gen_lowpart_SUBREG, rather than selecting between that
5948 and equivalent gen_rtx_SUBREG code.
5950 2018-02-08 Richard Biener <rguenther@suse.de>
5952 PR tree-optimization/84233
5953 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
5954 changed flag instead of boguously re-using phi_inserted.
5956 2018-02-08 Martin Jambor <mjambor@suse.cz>
5958 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
5959 static local variables.
5961 2018-02-08 Richard Biener <rguenther@suse.de>
5963 PR tree-optimization/84278
5964 * tree-vect-stmts.c (vectorizable_store): When looking for
5965 smaller vector types to perform grouped strided loads/stores
5966 make sure the mode is supported by the target.
5967 (vectorizable_load): Likewise.
5969 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
5971 * config/aarch64/aarch64.c (aarch64_components_for_bb):
5972 Increase LDP/STP opportunities by adding adjacent callee-saves.
5974 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
5976 PR rtl-optimization/84068
5977 PR rtl-optimization/83459
5978 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
5980 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
5982 PR tree-optimization/84224
5983 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
5984 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
5987 2018-02-07 Iain Sandoe <iain@codesourcery.com>
5990 * config/rs6000/altivec.md (*restore_world): Remove LR use.
5991 * config/rs6000/predicates.md (restore_world_operation): Adjust op
5992 count, remove one USE.
5994 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
5996 * doc/install.texi (Configuration): Document the
5997 --with-long-double-format={ibm,ieee} PowerPC configuration
6001 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
6002 Convert from define_expand to be define_insn_and_split. Rework
6003 float/double/_Float128 conversions to QI/HI/SImode to work with
6004 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
6005 conversions to QI/HImode types did a store and then a load to
6006 truncate the value. For conversions to VSX registers, don't split
6007 the insn, instead emit the code directly. Use the code iterator
6008 any_fix to combine signed and unsigned conversions.
6009 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
6010 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
6011 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
6012 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
6013 (fix_<mode>di2_hw): Likewise.
6014 (fixuns_<mode>di2_hw): Likewise.
6015 (fix_<mode>si2_hw): Likewise.
6016 (fixuns_<mode>si2_hw): Likewise.
6017 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
6018 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
6019 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
6020 fix<uns>_trunc<SFDF:mode>si2_p8.
6021 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
6023 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
6024 (fix<uns>_<mode>_mem): Likewise.
6025 (fctiw<u>z_<mode>_mem): Likewise.
6026 (fix<uns>_<mode>_mem): Likewise.
6027 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
6028 the register allocator from doing a direct move to the GPRs to do
6029 a store, and instead use the ISA 3.0 store byte/half-word from
6030 vector register instruction. For IEEE 128-bit floating point,
6031 also optimize stores of 32-bit ints.
6032 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
6034 2018-02-07 Alan Hayward <alan.hayward@arm.com>
6036 * genextract.c (push_pathstr_operand): New function to support
6038 (walk_rtx): Call push_pathstr_operand.
6039 (print_path): Support [a-zA-Z].
6041 2018-02-07 Richard Biener <rguenther@suse.de>
6043 PR tree-optimization/84037
6044 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
6045 (cse_and_gimplify_to_preheader): Declare.
6046 (vect_get_place_in_interleaving_chain): Likewise.
6047 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
6049 (_loop_vec_info::~_loop_vec_info): Delete it.
6050 (cse_and_gimplify_to_preheader): New function.
6051 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
6052 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
6053 (vectorizable_load): Likewise. For grouped stores always base
6054 the IV on the first element.
6055 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
6056 condition before gimplifying.
6058 2018-02-07 Jakub Jelinek <jakub@redhat.com>
6060 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
6061 *DIV_EXPR and *MOD_EXPR.
6063 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
6066 * config/i386/i386.c (ix86_option_override_internal): Mask out
6067 the CF_SET bit when checking -fcf-protection.
6069 2018-02-07 Tom de Vries <tom@codesourcery.com>
6072 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
6075 2018-02-07 Richard Biener <rguenther@suse.de>
6077 PR tree-optimization/84204
6078 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
6081 PR tree-optimization/84205
6082 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
6083 special-case isl_ast_op_zdiv_r.
6085 PR tree-optimization/84223
6086 * graphite-scop-detection.c (gather_bbs::before_dom_children):
6087 Only add conditions from within the region.
6088 (gather_bbs::after_dom_children): Adjust.
6090 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
6093 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
6094 * config/avr/avr.md: Only post-reload split REG-REG moves if
6095 either register is GENERAL_REG_P.
6097 2018-02-07 Jakub Jelinek <jakub@redhat.com>
6099 PR tree-optimization/84235
6100 * tree-ssa-scopedtables.c
6101 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
6102 if the subtraction is performed in floating point type where NaNs are
6103 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
6104 build 1. Formatting fix.
6106 2018-02-06 Jakub Jelinek <jakub@redhat.com>
6109 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
6110 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
6111 and skip it regardless of bb boundaries. Use CALL_P macro,
6112 don't test INSN_P (insn) together with CALL_P or JUMP_P check
6113 unnecessarily, formatting fix.
6115 2018-02-06 Michael Collison <michael.collison@arm.com>
6117 * config/arm/thumb2.md:
6118 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
6119 (*thumb_mov_notscc): Ditto.
6121 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
6124 * config/rs6000/rs6000.md (su code attribute): Use "u" for
6125 unsigned_fix, not "s".
6127 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6129 * configure.ac (gcc_fn_eh_frame_ro): New function.
6130 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
6131 correct .eh_frame permissions.
6132 * configure: Regenerate.
6134 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
6136 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
6139 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6141 * config/rs6000/rs6000.c (rs6000_option_override_internal):
6142 Display warning message for -mno-speculate-indirect-jumps.
6144 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
6146 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
6148 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
6150 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
6152 PR tree-optimization/84225
6153 * tree-eh.c (find_trapping_overflow): Only call
6154 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
6156 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6159 * config/i386/i386.c: Reimplement the check of possible options
6160 -mibt/-mshstk conbination. Change error messages.
6161 * doc/invoke.texi: Fix a typo: remove extra '='.
6163 2018-02-06 Marek Polacek <polacek@redhat.com>
6165 PR tree-optimization/84228
6166 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
6168 2018-02-06 Tamar Christina <tamar.christina@arm.com>
6171 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
6172 emitted arch directives.
6173 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
6174 __ARM_FEATURE_COPROC before changing architectures.
6176 2018-02-06 Richard Biener <rguenther@suse.de>
6178 * config/i386/i386.c (print_reg): Fix typo.
6179 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
6181 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
6183 * configure: Regenerate.
6185 2018-02-05 Martin Sebor <msebor@redhat.com>
6187 PR tree-optimization/83369
6188 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
6191 2018-02-05 Martin Liska <mliska@suse.cz>
6193 * doc/invoke.texi: Cherry-pick upstream r323995.
6195 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
6197 * ira.c (ira_init_register_move_cost): Adjust comment.
6199 2018-02-05 Martin Liska <mliska@suse.cz>
6201 PR gcov-profile/84137
6202 * doc/gcov.texi: Fix typo in documentation.
6204 2018-02-05 Martin Liska <mliska@suse.cz>
6206 PR gcov-profile/83879
6207 * doc/gcov.texi: Document necessity of --dynamic-list-data when
6208 using dlopen functionality.
6210 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
6212 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
6213 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
6214 _mm_maskz_range_ss, _mm_mask_range_round_ss,
6215 _mm_maskz_range_round_ss): New intrinsics.
6216 (__builtin_ia32_rangesd128_round)
6217 (__builtin_ia32_rangess128_round): Remove.
6218 (__builtin_ia32_rangesd128_mask_round,
6219 __builtin_ia32_rangess128_mask_round): New builtins.
6220 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
6221 __builtin_ia32_rangess128_round): Remove.
6222 (__builtin_ia32_rangesd128_mask_round,
6223 __builtin_ia32_rangess128_mask_round): New builtins.
6224 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
6225 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
6226 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
6227 "<round_saeonly_constraint>")): Changed to ...
6228 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
6229 "<round_saeonly_scalar_constraint>")): ... this.
6230 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
6231 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
6232 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
6233 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
6234 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
6236 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
6238 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
6240 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
6241 Remove all values except native, 8540 and 8548.
6243 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
6245 * config/i386/i386.c (ix86_output_function_return): Pass
6246 INVALID_REGNUM, instead of -1, as invalid register number to
6247 indirect_thunk_name and output_indirect_thunk.
6249 2018-02-02 Julia Koval <julia.koval@intel.com>
6251 * config.gcc: Add -march=icelake.
6252 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
6253 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
6254 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
6255 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
6256 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
6257 (processor_target_table): Add icelake.
6258 (ix86_option_override_internal): Handle new PTAs.
6259 (get_builtin_code_for_version): Handle icelake.
6260 (M_INTEL_COREI7_ICELAKE): New.
6261 (fold_builtin_cpu): Handle icelake.
6262 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
6263 * doc/invoke.texi: Add -march=icelake.
6265 2018-02-02 Julia Koval <julia.koval@intel.com>
6267 * config/i386/i386.c (ix86_option_override_internal): Change flags type
6268 to wide_int_bitmask.
6269 * wide-int-bitmask.h: New.
6271 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6274 * config/i386/i386.md: Replace Pmode with word_mode in
6275 builtin_setjmp_setup and builtin_longjmp to support x32.
6277 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
6281 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
6283 (rs6000_supported_cpu_names): New static variable.
6284 (linux_cpu_translation_table): Likewise.
6285 (elf_platform) <cpu>: Define new static variable and use it.
6286 Translate kernel AT_PLATFORM name to canonical name if needed.
6287 Error if platform name is unknown.
6289 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
6292 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
6294 2018-02-01 Jeff Law <law@redhat.com>
6297 * config/i386/i386.c (release_scratch_register_on_entry): Add new
6298 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
6299 the scratch if RELEASE_VIA_POP is false.
6300 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
6301 If we have to save a temporary register, decrement SIZE appropriately.
6302 Pass new arguments to release_scratch_register_on_entry.
6303 (ix86_adjust_stack_and_probe): Likewise.
6304 (ix86_emit_probe_stack_range): Pass new arguments to
6305 release_scratch_register_on_entry.
6307 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
6309 PR rtl-optimization/84157
6310 * combine.c (change_zero_ext): Use REG_P predicate in
6311 front of HARD_REGISTER_P predicate.
6313 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
6315 * config/avr/avr.c (avr_option_override): Move disabling of
6316 -fdelete-null-pointer-checks to...
6317 * common/config/avr/avr-common.c (avr_option_optimization_table):
6320 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6322 PR tree-optimization/81635
6323 * tree-data-ref.c (split_constant_offset_1): For types that
6324 wrap on overflow, try to use range info to prove that wrapping
6327 2018-02-01 Renlin Li <renlin.li@arm.com>
6330 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
6332 (aarch64_register_move_cost): Likewise.
6333 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
6335 (REG_CLASS_NAMES): Likewise.
6336 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
6337 TAILCALL_ADDR_REGS. Remove IP registers.
6338 * config/aarch64/aarch64.md (Ucs): Update register constraint.
6340 2018-02-01 Richard Biener <rguenther@suse.de>
6342 * domwalk.h (dom_walker::dom_walker): Add additional constructor
6343 for specifying RPO order and allow NULL for that.
6344 * domwalk.c (dom_walker::dom_walker): Likewise.
6345 (dom_walker::walk): Handle NULL RPO order.
6346 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
6348 (rewrite_update_dom_walker): Likewise.
6349 (mark_def_dom_walker): Likewise.
6351 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6353 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
6354 (aarch64_maybe_expand_sve_subreg_move): Declare.
6355 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
6356 * config/aarch64/predicates.md (aarch64_any_register_operand): New
6358 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
6359 that are semantically a reverse operation.
6360 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
6361 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
6362 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
6364 (aarch64_can_change_mode_class): For big-endian, forbid changes
6365 between two SVE modes if they have different element sizes.
6367 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6369 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
6370 the TImode handling for big-endian targets.
6372 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6374 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
6375 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
6377 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
6378 Remove BSWAP handing for big-endian targets and use the form of
6379 LD1RQ appropariate for the mode.
6381 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6383 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
6384 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
6387 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6390 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
6391 check for operands that need to go through aarch64_sve_reload_be.
6393 2018-02-01 Jakub Jelinek <jakub@redhat.com>
6395 PR tree-optimization/81661
6396 PR tree-optimization/84117
6397 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
6398 * tree-eh.c: Include gimplify.h.
6399 (find_trapping_overflow, replace_trapping_overflow,
6400 rewrite_to_non_trapping_overflow): New functions.
6401 * tree-vect-loop.c: Include tree-eh.h.
6402 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
6403 * tree-data-ref.c: Include tree-eh.h.
6404 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
6406 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
6408 PR rtl-optimization/84123
6409 * combine.c (change_zero_ext): Check if hard register satisfies
6410 can_change_dest_mode before calling gen_lowpart_SUBREG.
6412 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
6415 * ira.c (ira_init_register_move_cost): Remove assert.
6417 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
6419 PR rtl-optimization/84071
6420 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
6421 * doc/tm.texi: Regenerate.
6423 2018-01-31 Richard Biener <rguenther@suse.de>
6425 PR tree-optimization/84132
6426 * tree-data-ref.c (analyze_miv_subscript): Properly
6427 check whether evolution_function_is_affine_multivariate_p
6428 before calling gcd_of_steps_may_divide_p.
6430 2018-01-31 Julia Koval <julia.koval@intel.com>
6433 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
6434 * config/i386/i386.md (rdpid_rex64) New.
6435 (rdpid): Make 32bit only.
6437 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
6440 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
6441 an IDENTIFIER_NODE for FUNCTION_TYPE's.
6443 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
6446 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
6448 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
6450 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
6452 PR rtl-optimization/84071
6453 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
6454 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
6456 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
6458 * config/arc/arc.c (arc_handle_aux_attribute): New function.
6459 (arc_attribute_table): Add 'aux' attribute.
6460 (arc_in_small_data_p): Consider aux like variables.
6461 (arc_is_aux_reg_p): New function.
6462 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
6463 (arc_get_aux_arg): New function.
6464 (prepare_move_operands): Handle aux-register access.
6465 (arc_handle_aux_attribute): New function.
6466 * doc/extend.texi (ARC Variable attributes): Add subsection.
6468 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
6470 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
6471 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
6472 (arc_attribute_table): Add 'uncached' attribute.
6473 (arc_print_operand): Print '.di' flag for uncached memory
6475 (arc_in_small_data_p): Do not consider for small data the uncached
6477 (arc_is_uncached_mem_p): New function.
6478 * config/arc/predicates.md (compact_store_memory_operand): Check
6479 for uncached memory accesses.
6480 (nonvol_nonimm_operand): Likewise.
6481 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
6483 2018-01-31 Jakub Jelinek <jakub@redhat.com>
6486 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
6487 falign-loops=): Add Optimization flag.
6489 2018-01-30 Jeff Law <law@redhat.com>
6492 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
6493 INT_REGISTERS_SAVED. Check it prior to calling
6494 get_scratch_register_on_entry.
6495 (ix86_adjust_stack_and_probe): Similarly.
6496 (ix86_emit_probe_stack_range): Similarly.
6497 (ix86_expand_prologue): Corresponding changes.
6499 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6502 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
6503 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
6505 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
6508 * lra-constraints.c (curr_insn_transform): Process AND in the
6511 2018-01-30 Jakub Jelinek <jakub@redhat.com>
6513 PR rtl-optimization/83986
6514 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
6515 dependence against last_pending_memory_flush in addition to
6518 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
6520 PR tree-optimization/81611
6521 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
6524 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6527 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
6530 2018-01-30 Richard Biener <rguenther@suse.de>
6531 Jakub Jelinek <jakub@redhat.com>
6533 PR tree-optimization/84111
6534 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
6535 inner loops added during recursion, as they don't have up-to-date
6538 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
6541 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
6542 (can_inline_edge_by_limits_p): ... here.
6543 (can_early_inline_edge_p, check_callers,
6544 update_caller_keys, update_callee_keys, recursive_inlining,
6545 add_new_edges_to_heap, speculation_useful_p,
6546 inline_small_functions,
6547 inline_small_functions, flatten_function,
6548 inline_to_all_callers_1): Update.
6550 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
6552 * profile-count.c (profile_count::combine_with_ipa_count): Handle
6555 2018-01-30 Richard Biener <rguenther@suse.de>
6557 PR tree-optimization/83008
6558 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
6559 invariant and constant vector uses in stmts when they need
6562 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6565 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
6566 * configure: Regenerate.
6568 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
6570 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
6572 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
6573 Use gen_rtx_REG rather than gen_lowpart.
6575 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
6577 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
6578 rather than 0 when creating partial subregs.
6580 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
6582 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
6585 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
6588 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
6589 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
6590 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
6591 flags. This restores the settings used before the 2017-07-24.
6592 Turning off pre increment/decrement/modify allows IVOPTS to
6593 optimize DF/SF loops where the index is an int.
6595 2018-01-29 Richard Biener <rguenther@suse.de>
6596 Kelvin Nilsen <kelvin@gcc.gnu.org>
6599 * tree-vect-stmts.c (vectorizable_call): Don't call
6600 targetm.vectorize_builtin_md_vectorized_function if callee is
6603 2018-01-22 Carl Love <cel@us.ibm.com>
6605 * doc/extend.tex: Fix typo in second arg in
6606 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
6608 2018-01-29 Richard Biener <rguenther@suse.de>
6610 PR tree-optimization/84086
6611 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
6612 (flush_ssaname_freelist): When SSA names were released reset
6613 the SCEV hash table.
6615 2018-01-29 Richard Biener <rguenther@suse.de>
6617 PR tree-optimization/84057
6618 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
6619 removed paths when removing edges.
6621 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
6623 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
6624 -mfunction-return=@var{choice}.
6626 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
6629 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
6631 (layout::print_source_line): Likewise.
6632 (test_get_line_width_without_trailing_whitespace): Add test cases.
6634 2018-01-27 Jakub Jelinek <jakub@redhat.com>
6637 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
6640 2018-01-26 Jim Wilson <jimw@sifive.com>
6642 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
6644 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
6647 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6649 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
6650 and CMP + SUB-immediate -> SUBS.
6652 2018-01-26 Martin Sebor <msebor@redhat.com>
6654 PR tree-optimization/83896
6655 * tree-ssa-strlen.c (get_string_len): Rename...
6656 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
6657 Avoid assuming length is constant.
6658 (handle_char_store): Use HOST_WIDE_INT for string length.
6660 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
6663 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
6664 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
6666 2018-01-26 Richard Biener <rguenther@suse.de>
6668 PR rtl-optimization/84003
6669 * dse.c (record_store): Only record redundant stores when
6670 the earlier store aliases at least all accesses the later one does.
6672 2018-01-26 Jakub Jelinek <jakub@redhat.com>
6674 PR rtl-optimization/83985
6675 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
6676 REG_CFA_RESTORE insns.
6677 (delete_unmarked_insns): Don't ignore separate shrink wrapping
6678 REG_CFA_RESTORE insns here.
6681 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
6682 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
6684 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6686 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
6687 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
6688 (arc_init): Likewise.
6689 (arc_override_options): Likewise.
6690 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
6692 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
6694 * config/arc/arc.h (TARGET_DBNZ): Define.
6695 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
6696 properly set the tune attribute.
6697 (dbnz): Use TARGET_DBNZ guard.
6698 * config/arc/arc.opt (mtune): Add core3 option.
6700 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6702 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
6703 recognize new pic like addresses.
6704 (arc_delegitimize_address): Clean up.
6706 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6708 * config/arc/arc-arches.def: Option mrf16 valid for all
6710 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
6711 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
6712 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
6713 * config/arc/arc-tables.opt: Regenerate.
6714 * config/arc/arc.c (arc_conditional_register_usage): Handle
6715 reduced register file case.
6716 (arc_file_start): Set must have build attributes.
6717 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
6719 * config/arc/arc.opt (mrf16): Add new option.
6720 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
6721 * config/arc/genmultilib.awk: Handle new mrf16 option.
6722 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
6723 * config/arc/t-multilib: Regenerate.
6724 * doc/invoke.texi (ARC Options): Document mrf16 option.
6726 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6728 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
6729 * config/arc/arc.c (arc_handle_secure_attribute): New function.
6730 (arc_attribute_table): Add 'secure_call' attribute.
6731 (arc_print_operand): Print secure call operand.
6732 (arc_function_ok_for_sibcall): Don't optimize tail calls when
6734 (arc_is_secure_call_p): New function. * config/arc/arc.md
6735 (call_i): Add support for sjli instruction.
6736 (call_value_i): Likewise.
6737 * config/arc/constraints.md (Csc): New constraint.
6739 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6740 John Eric Martin <John.Martin@emmicro-us.com>
6742 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
6743 * config/arc/arc.c (_arc_jli_section): New struct.
6744 (arc_jli_section): New type.
6745 (rc_jli_sections): New static variable.
6746 (arc_handle_jli_attribute): New function.
6747 (arc_attribute_table): Add jli_always and jli_fixed attribute.
6748 (arc_file_end): New function.
6749 (TARGET_ASM_FILE_END): Define.
6750 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
6751 (arc_add_jli_section): New function.
6752 (jli_call_scan): Likewise.
6753 (arc_reorg): Call jli_call_scan.
6754 (arc_output_addsi): Remove 'S' from printing asm operand.
6755 (arc_is_jli_call_p): New function.
6756 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
6758 (movhi_insn): Likewise.
6759 (movsi_insn): Likewise.
6760 (movsi_set_cc_insn): Likewise.
6761 (loadqi_update): Likewise.
6762 (load_zeroextendqisi_update): Likewise.
6763 (load_signextendqisi_update): Likewise.
6764 (loadhi_update): Likewise.
6765 (load_zeroextendhisi_update): Likewise.
6766 (load_signextendhisi_update): Likewise.
6767 (loadsi_update): Likewise.
6768 (loadsf_update): Likewise.
6769 (movsicc_insn): Likewise.
6770 (bset_insn): Likewise.
6771 (bxor_insn): Likewise.
6772 (bclr_insn): Likewise.
6773 (bmsk_insn): Likewise.
6774 (bicsi3_insn): Likewise.
6775 (cmpsi_cc_c_insn): Likewise.
6776 (movsi_ne): Likewise.
6777 (movsi_cond_exec): Likewise.
6778 (clrsbsi2): Likewise.
6789 (call_i): Remove 'S' asm letter, add jli instruction.
6790 (call_value_i): Likewise.
6791 * config/arc/arc.op (mjli-always): New option.
6792 * config/arc/constraints.md (Cji): New constraint.
6793 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
6795 (subsf3_fpx): Likewise.
6796 (mulsf3_fpx): Likewise.
6797 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
6799 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
6801 * doc/invoke.texi (ARC): Document mjli-always option.
6803 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
6805 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
6806 avoid addition with 0 and use incw and decw where possible.
6808 2018-01-26 Richard Biener <rguenther@suse.de>
6810 PR tree-optimization/81082
6811 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
6812 association if it requires casting to unsigned.
6813 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
6814 from fold_plusminus_mult_expr to catch important cases late when
6815 range info is available.
6817 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6819 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
6820 * configure.ac (hidden_linkonce): New test.
6821 * configure: Regenerate.
6822 * config.in: Regenerate.
6824 2018-01-26 Julia Koval <julia.koval@intel.com>
6826 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
6827 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
6828 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
6829 _mm_mask_bitshuffle_epi64_mask): Fix type.
6830 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
6831 USI_FTYPE_V4DI_V4DI_USI): Remove.
6832 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
6833 __builtin_ia32_vpshufbitqmb256_mask,
6834 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
6835 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
6836 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
6838 2018-01-26 Alan Modra <amodra@gmail.com>
6841 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
6842 UNSPEC_VBPERMQ. Sort other unspecs.
6844 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
6846 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
6848 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
6851 * predict.c (drop_profile): Do not push/pop cfun; update also
6853 (handle_missing_profiles): Fix logic looking for zero profiles.
6855 2018-01-25 Jakub Jelinek <jakub@redhat.com>
6858 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
6859 on functions with #pragma omp declare simd or functions with simd
6861 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
6862 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
6863 Remove trailing \n from warning_at calls.
6865 2018-01-25 Tom de Vries <tom@codesourcery.com>
6868 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
6869 for neutered workers.
6871 2018-01-24 Joseph Myers <joseph@codesourcery.com>
6874 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
6875 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
6877 2018-01-24 Jeff Law <law@redhat.com>
6880 * i386.c (get_probe_interval): Move to earlier point.
6881 (ix86_compute_frame_layout): If -fstack-clash-protection and
6882 the frame is larger than the probe interval, then use pushes
6883 to save registers rather than reg->mem moves.
6884 (ix86_expand_prologue): Remove conditional for int_registers_saved
6887 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
6890 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
6891 min/max for never referenced object.
6893 2018-01-24 Jakub Jelinek <jakub@redhat.com>
6896 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
6898 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
6899 attributes from DECL_ATTRIBUTES (decl) without affecting
6900 DECL_ATTRIBUTES (current_function_decl).
6901 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
6902 functions with non-NULL DECL_ABSTRACT_ORIGIN.
6904 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
6906 PR tree-optimization/83979
6907 * fold-const.c (fold_comparison): Use constant_boolean_node
6908 instead of boolean_{true,false}_node.
6910 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
6912 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
6915 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6917 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
6918 Simplify the clause that sets the length attribute.
6919 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
6920 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
6921 clause that sets the length attribute.
6922 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
6924 2018-01-24 Tom de Vries <tom@codesourcery.com>
6927 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
6928 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
6929 Add strict parameter.
6930 (prevent_branch_around_nothing): Insert dummy insn between branch to
6931 label and label with no ptx insn inbetween.
6932 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
6934 2018-01-24 Tom de Vries <tom@codesourcery.com>
6937 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
6938 for neutered threads in warp.
6939 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
6941 2018-01-24 Richard Biener <rguenther@suse.de>
6943 PR tree-optimization/83176
6944 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
6947 2018-01-24 Richard Biener <rguenther@suse.de>
6949 PR tree-optimization/82819
6950 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
6951 code generating pluses that are no-ops in the target precision.
6953 2018-01-24 Richard Biener <rguenther@suse.de>
6956 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
6958 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
6960 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
6961 to merge probabilities.
6962 * predict.c (probably_never_executed): Also mark as cold functions
6963 with global 0 profile and guessed local profile.
6964 * profile-count.c (profile_probability::combine_with_count): New
6966 * profile-count.h (profile_probability::operator*,
6967 profile_probability::operator*=, profile_probability::operator/,
6968 profile_probability::operator/=): Reduce precision to adjusted
6969 and set value to guessed on contradictory divisions.
6970 (profile_probability::combine_with_freq): Remove.
6971 (profile_probability::combine_wiht_count): Declare.
6972 (profile_count::force_nonzero):: Set to adjusted.
6973 (profile_count::probability_in):: Set quality to adjusted.
6974 * tree-ssa-tail-merge.c (replace_block_by): Use
6977 2018-01-23 Andrew Waterman <andrew@sifive.com>
6978 Jim Wilson <jimw@sifive.com>
6980 * config/riscv/riscv.c (riscv_stack_boundary): New.
6981 (riscv_option_override): Set riscv_stack_boundary. Handle
6982 riscv_preferred_stack_boundary_arg.
6983 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
6984 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
6985 (STACK_BOUNDARY): Set to riscv_stack_boundary.
6986 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
6987 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
6988 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
6990 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
6993 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
6994 of struct ix86_frame.
6995 (ix86_expand_epilogue): Likewise. Add a local variable for
6996 the reg_save_offset field in struct ix86_frame.
6998 2018-01-23 Bin Cheng <bin.cheng@arm.com>
7000 PR tree-optimization/82604
7001 * tree-loop-distribution.c (enum partition_kind): New enum item
7002 PKIND_PARTIAL_MEMSET.
7003 (partition_builtin_p): Support above new enum item.
7004 (generate_code_for_partition): Ditto.
7005 (compute_access_range): Differentiate cases that equality can be
7006 proven at all loops, the innermost loops or no loops.
7007 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
7008 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
7009 (finalize_partitions, distribute_loop): Don't fuse partition of
7010 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
7011 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
7014 2018-01-23 Martin Liska <mliska@suse.cz>
7016 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
7017 order to ignore the predictor.
7018 (PRED_POLYMORPHIC_CALL): Likewise.
7019 (PRED_RECURSIVE_CALL): Likewise.
7021 2018-01-23 Martin Liska <mliska@suse.cz>
7023 * tree-profile.c (tree_profiling): Print function header to
7024 aware reader which function we are working on.
7025 * value-prof.c (gimple_find_values_to_profile): Do not print
7026 not interesting value histograms.
7028 2018-01-23 Martin Liska <mliska@suse.cz>
7030 * profile-count.h (enum profile_quality): Add
7031 profile_uninitialized as the first value. Do not number values
7032 as they are zero based.
7033 (profile_count::verify): Update sanity check.
7034 (profile_probability::verify): Likewise.
7036 2018-01-23 Nathan Sidwell <nathan@acm.org>
7038 * doc/invoke.texi (ffor-scope): Deprecate.
7040 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7042 PR tree-optimization/83510
7043 * domwalk.c (set_all_edges_as_executable): New function.
7044 (dom_walker::dom_walker): Convert bool param
7045 "skip_unreachable_blocks" to enum reachability. Move setup of
7046 edge flags to set_all_edges_as_executable and only do it when
7047 reachability is REACHABLE_BLOCKS.
7048 * domwalk.h (enum dom_walker::reachability): New enum.
7049 (dom_walker::dom_walker): Convert bool param
7050 "skip_unreachable_blocks" to enum reachability.
7051 (set_all_edges_as_executable): New decl.
7052 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
7053 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
7055 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
7056 but converting true to REACHABLE_BLOCKS.
7057 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
7059 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
7060 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
7061 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
7063 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
7064 if check_all_array_refs will be called.
7066 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7068 * tree.c (selftest::test_location_wrappers): Add more test
7071 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7073 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
7074 (selftest::test_bit_in_range): Likewise.
7076 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
7079 * doc/sourcebuild.texi (vect_float): Say that the selector
7080 only describes the situation when -funsafe-math-optimizations is on.
7081 (vect_float_strict): Document.
7083 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
7085 PR tree-optimization/83965
7086 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
7087 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
7088 instead of checking only for a reduction.
7089 (vect_recog_widen_sum_pattern): Likewise.
7091 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7093 * predict.c (probably_never_executed): Only use precise profile info.
7094 (compute_function_frequency): Skip after inlining hack since we now
7095 have quality checking.
7097 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7099 * profile-count.h (profile_probability::very_unlikely,
7100 profile_probability::unlikely, profile_probability::even): Set
7101 precision to guessed.
7103 2018-01-23 Richard Biener <rguenther@suse.de>
7105 PR tree-optimization/83963
7106 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
7107 Properly terminate dominator walk when crossing the exit edge not
7108 when visiting its source block.
7110 2018-01-23 Jakub Jelinek <jakub@redhat.com>
7113 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
7114 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
7116 2018-01-22 Jakub Jelinek <jakub@redhat.com>
7118 PR tree-optimization/83957
7119 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
7120 semicolon after for body surrounded by braces.
7122 PR tree-optimization/83081
7123 * profile-count.h (profile_probability::split): New method.
7124 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
7125 Use profile_probability::split.
7126 (do_compare_rtx_and_jump): Fix adjustment of probabilities
7127 when splitting a single conditional jump into 2.
7129 2018-01-22 David Malcolm <dmalcolm@redhat.com>
7131 PR tree-optimization/69452
7132 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
7135 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7137 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
7138 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
7139 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
7141 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7143 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
7144 * config/rl78/rl78.md (movdi): New define_expand.
7145 * config/rl78/rl78.c (rl78_split_movdi): New function.
7147 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
7150 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
7152 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
7153 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
7154 128-bit to produce an UNSPEC move to get the double word with the
7155 signbit and then a shift directly to do signbit.
7156 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
7157 implementation with a new version that just does either a direct
7158 move or a regular move. Move memory interface to separate insns.
7159 Move insns so they are next to the expander.
7160 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
7161 with signbit move. Split big and little endian case.
7162 (signbit<mode>2_dm_mem_le): Likewise.
7163 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
7164 (signbit<mode>2_dm2): Likewise.
7166 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7168 * config/rl78/rl78.md (anddi3): New define_expand.
7170 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7172 * config/rl78/rl78.md (umindi3): New define_expand.
7174 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7176 * config/rl78/rl78.md (smindi3): New define_expand.
7178 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7180 * config/rl78/rl78.md (smaxdi3): New define_expand.
7182 2018-01-22 Carl Love <cel@us.ibm.com>
7184 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
7185 LVX_V1TI): Add macro expansion.
7186 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
7187 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
7188 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
7189 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
7190 Change check to determine if the instruction is a byte reversing
7191 entry. Fix typo in comment.
7192 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
7193 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
7194 Add def_builtin calls for new builtins.
7195 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
7196 Add define_insn expansion.
7198 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7200 * config/rl78/rl78.md (umaxdi3): New define_expand.
7202 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7204 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
7205 for non-QImode registers.
7207 2018-01-22 Richard Biener <rguenther@suse.de>
7209 PR tree-optimization/83963
7210 * graphite-scop-detection.c (scop_detection::get_sese): Delay
7211 including the loop exit block.
7212 (scop_detection::merge_sese): Likewise.
7213 (scop_detection::add_scop): Do it here instead.
7215 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7217 * doc/sourcebuild.texi (arm_softfloat): Document.
7219 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
7222 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
7223 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
7224 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
7226 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7227 David Edelsohn <dje.gcc@gmail.com>
7230 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7231 Change "crset eq" to "crset 2".
7232 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7233 (*call_indirect_aix<mode>_nospec): Likewise.
7234 (*call_value_indirect_aix<mode>_nospec): Likewise.
7235 (*call_indirect_elfv2<mode>_nospec): Likewise.
7236 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
7237 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
7238 change assembly output from . to $.
7239 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7240 (indirect_jump<mode>_nospec): Change assembly output from . to $.
7241 (*tablejump<mode>_internal1_nospec): Likewise.
7243 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
7246 * config/sh/sh_optimize_sett_clrt.cc:
7247 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
7249 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
7251 PR tree-optimization/83940
7252 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
7253 offset_dt to vect_constant_def rather than vect_unknown_def_type.
7254 (vect_check_load_store_mask): Add a mask_dt_out parameter and
7255 use it to pass back the definition type.
7256 (vect_check_store_rhs): Likewise rhs_dt_out.
7257 (vect_build_gather_load_calls): Add a mask_dt argument and use
7258 it instead of a call to vect_is_simple_use.
7259 (vectorizable_store): Update calls to vect_check_load_store_mask
7260 and vect_check_store_rhs. Use the dt returned by the latter instead
7261 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
7262 instead of calls to vect_is_simple_use. Pass the scalar rather
7263 than the vector operand to vect_is_simple_use when handling
7264 second and subsequent copies of an rhs value.
7265 (vectorizable_load): Update calls to vect_check_load_store_mask
7266 and vect_build_gather_load_calls. Use the cached mask_dt and
7267 gs_info.offset_dt instead of calls to vect_is_simple_use.
7269 2018-01-20 Jakub Jelinek <jakub@redhat.com>
7272 * tree-emutls.c: Include gimplify.h.
7273 (lower_emutls_2): New function.
7274 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
7275 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
7276 it before further processing.
7279 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
7280 UINTVAL (trueop1) instead of INTVAL (op1).
7282 2018-01-19 Jakub Jelinek <jakub@redhat.com>
7286 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
7287 INCOMING_FRAME_SP_OFFSET if not defined.
7288 (scan_trace): Add ENTRY argument. If true and
7289 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
7290 emit a note to adjust the CFA offset.
7291 (create_cfi_notes): Adjust scan_trace callers.
7292 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
7293 INCOMING_FRAME_SP_OFFSET in the CIE.
7294 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
7295 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
7297 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
7298 * doc/tm.texi: Regenerated.
7300 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7302 PR rtl-optimization/83147
7303 * lra-constraints.c (remove_inheritance_pseudos): Use
7304 lra_substitute_pseudo_within_insn.
7306 2018-01-19 Tom de Vries <tom@codesourcery.com>
7307 Cesar Philippidis <cesar@codesourcery.com>
7310 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
7312 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
7315 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
7316 spaces for function labels.
7318 2018-01-19 Martin Liska <mliska@suse.cz>
7320 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
7321 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
7322 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
7323 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
7324 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
7325 (PRED_CONST_RETURN): Change from 69 to 65.
7326 (PRED_NULL_RETURN): Change from 91 to 71.
7327 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
7328 (PRED_LOOP_GUARD): Change from 66 to 73.
7330 2018-01-19 Martin Liska <mliska@suse.cz>
7332 * predict.c (predict_insn_def): Add new assert.
7333 (struct branch_predictor): Change type to signed integer.
7334 (test_prediction_value_range): Amend test to cover
7336 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
7337 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
7338 (PRED_LOOP_ITERATIONS_MAX): Likewise.
7339 (PRED_LOOP_IV_COMPARE): Likewise.
7340 * predict.h (PROB_UNINITIALIZED): Define new constant.
7342 2018-01-19 Martin Liska <mliska@suse.cz>
7344 * predict.c (dump_prediction): Add new format for
7345 analyze_brprob.py script which is enabled with -details
7347 * profile-count.h (precise_p): New function.
7349 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
7351 PR tree-optimization/83922
7352 * tree-vect-loop.c (vect_verify_full_masking): Return false if
7353 there are no statements that need masking.
7354 (vect_active_double_reduction_p): New function.
7355 (vect_analyze_loop_operations): Use it when handling phis that
7356 are not in the loop header.
7358 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
7360 PR tree-optimization/83914
7361 * tree-vect-loop.c (vectorizable_induction): Don't convert
7362 init_expr or apply the peeling adjustment for inductions
7363 that are nested within the vectorized loop.
7365 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7367 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
7370 2018-01-18 Jakub Jelinek <jakub@redhat.com>
7374 * function.h (gimplify_parameters): Add gimple_seq * argument.
7375 * function.c: Include gimple.h and options.h.
7376 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
7377 for the added local temporaries if needed.
7378 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
7379 if there are any parameter cleanups, wrap whole body into a
7380 try/finally with the cleanups.
7382 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
7385 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
7386 Use GET_MODE_CLASS for scalar floating point.
7388 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
7392 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
7393 Fix call of call_cgraph_insertion_hooks.
7395 2018-01-18 Martin Sebor <msebor@redhat.com>
7397 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
7399 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
7402 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
7405 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
7408 * common.opt: (-ffile-prefix-map): New option.
7409 * opts.c (common_handle_option): Defer it.
7410 * opts-global.c (handle_common_deferred_options): Handle it.
7411 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
7412 * file-prefix-map.h: New file.
7413 (remap_debug_filename, add_debug_prefix_map): ...here.
7414 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
7415 * final.c (debug_prefix_map, add_debug_prefix_map
7416 remap_debug_filename): Move to...
7417 * file-prefix-map.c: New file.
7418 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
7419 generalize, get rid of alloca(), use strrchr() instead of strchr().
7420 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
7421 Implement in terms of add_prefix_map().
7422 (remap_macro_filename, remap_debug_filename): Implement in term of
7424 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
7425 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
7426 * dbxout.c: Include file-prefix-map.h.
7427 * varasm.c: Likewise.
7428 * vmsdbgout.c: Likewise.
7429 * xcoffout.c: Likewise.
7430 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
7431 * doc/cppopts.texi (-fmacro-prefix-map): Document.
7432 * doc/invoke.texi (-ffile-prefix-map): Document.
7433 (-fdebug-prefix-map): Update description.
7435 2018-01-18 Martin Liska <mliska@suse.cz>
7437 * config/i386/i386.c (indirect_thunk_name): Document that also
7439 (output_indirect_thunk): Document why both instructions
7440 (pause and lfence) are generated.
7442 2018-01-18 Richard Biener <rguenther@suse.de>
7444 PR tree-optimization/83887
7445 * graphite-scop-detection.c
7446 (scop_detection::get_nearest_dom_with_single_entry): Remove.
7447 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
7448 (scop_detection::merge_sese): Re-implement with a flood-fill
7449 algorithm that properly finds a SESE region if it exists.
7451 2018-01-18 Jakub Jelinek <jakub@redhat.com>
7454 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
7455 pointer_diff optimizations use view_convert instead of convert.
7457 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7459 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7460 Generate different code for -mno-speculate-indirect-jumps.
7461 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7462 (*call_indirect_aix<mode>): Disable for
7463 -mno-speculate-indirect-jumps.
7464 (*call_indirect_aix<mode>_nospec): New define_insn.
7465 (*call_value_indirect_aix<mode>): Disable for
7466 -mno-speculate-indirect-jumps.
7467 (*call_value_indirect_aix<mode>_nospec): New define_insn.
7468 (*sibcall_nonlocal_sysv<mode>): Generate different code for
7469 -mno-speculate-indirect-jumps.
7470 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7472 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
7474 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
7475 long double type, set the flags for noting the default long double
7476 type, even if we don't pass or return a long double type.
7478 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
7481 * ipa-inline.c (flatten_function): Do not overwrite final inlining
7484 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
7486 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
7487 support for merge[hl].
7488 (fold_mergehl_helper): New helper function.
7489 (tree-vector-builder.h): New #include for tree_vector_builder usage.
7490 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
7491 (altivec_vmrglw_direct): Add xxmrglw insn.
7493 2018-01-17 Andrew Waterman <andrew@sifive.com>
7495 * config/riscv/riscv.c (riscv_conditional_register_usage): If
7496 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
7498 2018-01-17 David Malcolm <dmalcolm@redhat.com>
7501 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
7502 call the lto_location_cache before reading the
7503 DECL_SOURCE_LOCATION of the types.
7505 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
7506 Richard Sandiford <richard.sandiford@linaro.org>
7508 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
7509 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
7510 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
7511 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
7512 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
7514 * config/aarch64/constraints.md (aarch64_movti_operand):
7516 * config/aarch64/predicates.md (Uti): Add new constraint.
7518 2018-01-17 Carl Love <cel@us.ibm.com>
7520 * config/rs6000/vsx.md (define_expand xl_len_r,
7521 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
7522 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
7524 (define_expand, define_insn): Move the shift left from the
7525 define_insn to the define_expand for lxvl and stxvl instructions.
7526 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
7527 and XL_LEN_R definitions to PURE.
7529 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
7531 * config/i386/i386.c (indirect_thunk_name): Declare regno
7532 as unsigned int. Compare regno with INVALID_REGNUM.
7533 (output_indirect_thunk): Ditto.
7534 (output_indirect_thunk_function): Ditto.
7535 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
7536 in the call to output_indirect_thunk_function.
7538 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
7541 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
7542 rather than the size of inner_type to determine the stack slot size
7543 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
7545 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
7548 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
7551 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
7553 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
7554 endian Linux systems to optionally enable multilibs for selecting
7555 the long double type if the user configured an explicit type.
7556 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
7557 have no long double multilibs if not defined.
7558 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
7559 warn if the user used -mabi={ieee,ibm}longdouble and we built
7560 multilibs for long double.
7561 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
7562 appropriate multilib option.
7563 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
7565 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
7566 for building long double multilibs.
7567 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
7569 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
7571 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
7574 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
7576 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
7579 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
7582 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
7585 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
7587 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
7588 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
7590 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
7592 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
7593 different rtl trees depending on TARGET_64BIT.
7594 (rs6000_gen_lvx): Likewise.
7596 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
7598 * config/visium/visium.md (nop): Tweak comment.
7599 (hazard_nop): Likewise.
7601 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7603 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
7604 -mspeculate-indirect-jumps.
7605 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
7606 for -mno-speculate-indirect-jumps.
7607 (*call_indirect_elfv2<mode>_nospec): New define_insn.
7608 (*call_value_indirect_elfv2<mode>): Disable for
7609 -mno-speculate-indirect-jumps.
7610 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
7611 (indirect_jump): Emit different RTL for
7612 -mno-speculate-indirect-jumps.
7613 (*indirect_jump<mode>): Disable for
7614 -mno-speculate-indirect-jumps.
7615 (*indirect_jump<mode>_nospec): New define_insn.
7616 (tablejump): Emit different RTL for
7617 -mno-speculate-indirect-jumps.
7618 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
7619 (tablejumpsi_nospec): New define_expand.
7620 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
7621 (tablejumpdi_nospec): New define_expand.
7622 (*tablejump<mode>_internal1): Disable for
7623 -mno-speculate-indirect-jumps.
7624 (*tablejump<mode>_internal1_nospec): New define_insn.
7625 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
7628 2018-01-16 Artyom Skrobov tyomitch@gmail.com
7630 * caller-save.c (insert_save): Drop unnecessary parameter. All
7633 2018-01-16 Jakub Jelinek <jakub@redhat.com>
7634 Richard Biener <rguenth@suse.de>
7637 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
7638 return early, inline manually is_gimple_sizepos. Make sure if we
7639 call gimplify_expr we don't end up with a gimple constant.
7640 * tree.c (variably_modified_type_p): Don't return true for
7641 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
7642 * gimplify.h (is_gimple_sizepos): Remove.
7644 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7646 PR tree-optimization/83857
7647 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
7648 vectorizable_live_operation for pure SLP statements.
7649 (vectorizable_live_operation): Handle PHIs.
7651 2018-01-16 Richard Biener <rguenther@suse.de>
7653 PR tree-optimization/83867
7654 * tree-vect-stmts.c (vect_transform_stmt): Precompute
7655 nested_in_vect_loop_p since the scalar stmt may get invalidated.
7657 2018-01-16 Jakub Jelinek <jakub@redhat.com>
7660 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
7661 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
7662 If off is not INTEGER_CST, issue a may not be aligned warning
7663 rather than isn't aligned. Use isn%'t rather than isn't.
7664 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
7666 <case MULT_EXPR>: Improve the case when bottom and one of the
7667 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
7668 operand, in that case check if the other operand is multiple of
7669 bottom divided by the INTEGER_CST operand.
7671 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7674 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
7675 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
7676 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
7677 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
7678 * config/pa/pa.c (pa_function_arg_advance): Likewise.
7679 (pa_function_arg, pa_arg_partial_bytes): Likewise.
7680 (pa_function_arg_size): New function.
7682 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7684 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
7685 in a separate statement.
7687 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7689 PR tree-optimization/83847
7690 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
7691 group gathers and scatters.
7693 2018-01-16 Jakub Jelinek <jakub@redhat.com>
7695 PR rtl-optimization/86620
7696 * params.def (max-sched-ready-insns): Bump minimum value to 1.
7698 PR rtl-optimization/83213
7699 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
7700 to last if both are JUMP_INSNs.
7702 PR tree-optimization/83843
7703 * gimple-ssa-store-merging.c
7704 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
7705 store_immediate_info for bswap/nop orig_stores.
7707 2018-01-15 Andrew Waterman <andrew@sifive.com>
7709 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
7711 <UDIV>: Increase cost if !TARGET_DIV.
7713 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
7715 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
7716 (define_attr "cr_logical_3op"): New.
7717 (cceq_ior_compare): Adjust.
7718 (cceq_ior_compare_complement): Adjust.
7719 (*cceq_rev_compare): Adjust.
7720 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
7721 (is_cracked_insn): Adjust.
7722 (insn_must_be_first_in_group): Adjust.
7723 * config/rs6000/40x.md: Adjust.
7724 * config/rs6000/440.md: Adjust.
7725 * config/rs6000/476.md: Adjust.
7726 * config/rs6000/601.md: Adjust.
7727 * config/rs6000/603.md: Adjust.
7728 * config/rs6000/6xx.md: Adjust.
7729 * config/rs6000/7450.md: Adjust.
7730 * config/rs6000/7xx.md: Adjust.
7731 * config/rs6000/8540.md: Adjust.
7732 * config/rs6000/cell.md: Adjust.
7733 * config/rs6000/e300c2c3.md: Adjust.
7734 * config/rs6000/e500mc.md: Adjust.
7735 * config/rs6000/e500mc64.md: Adjust.
7736 * config/rs6000/e5500.md: Adjust.
7737 * config/rs6000/e6500.md: Adjust.
7738 * config/rs6000/mpc.md: Adjust.
7739 * config/rs6000/power4.md: Adjust.
7740 * config/rs6000/power5.md: Adjust.
7741 * config/rs6000/power6.md: Adjust.
7742 * config/rs6000/power7.md: Adjust.
7743 * config/rs6000/power8.md: Adjust.
7744 * config/rs6000/power9.md: Adjust.
7745 * config/rs6000/rs64.md: Adjust.
7746 * config/rs6000/titan.md: Adjust.
7748 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7750 * config/i386/predicates.md (indirect_branch_operand): Rewrite
7751 ix86_indirect_branch_register logic.
7753 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7755 * config/i386/constraints.md (Bs): Update
7756 ix86_indirect_branch_register check. Don't check
7757 ix86_indirect_branch_register with GOT_memory_operand.
7759 * config/i386/predicates.md (GOT_memory_operand): Don't check
7760 ix86_indirect_branch_register here.
7761 (GOT32_symbol_operand): Likewise.
7763 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7765 * config/i386/predicates.md (constant_call_address_operand):
7766 Rewrite ix86_indirect_branch_register logic.
7767 (sibcall_insn_operand): Likewise.
7769 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7771 * config/i386/constraints.md (Bs): Replace
7772 ix86_indirect_branch_thunk_register with
7773 ix86_indirect_branch_register.
7775 * config/i386/i386.md (indirect_jump): Likewise.
7776 (tablejump): Likewise.
7777 (*sibcall_memory): Likewise.
7778 (*sibcall_value_memory): Likewise.
7779 Peepholes of indirect call and jump via memory: Likewise.
7780 * config/i386/i386.opt: Likewise.
7781 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7782 (GOT_memory_operand): Likewise.
7783 (call_insn_operand): Likewise.
7784 (sibcall_insn_operand): Likewise.
7785 (GOT32_symbol_operand): Likewise.
7787 2018-01-15 Jakub Jelinek <jakub@redhat.com>
7790 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
7791 type rather than type addr's type points to.
7792 (expand_omp_atomic_mutex): Likewise.
7793 (expand_omp_atomic): Likewise.
7795 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7798 * config/i386/i386.c (output_indirect_thunk_function): Use
7799 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
7800 for __x86_return_thunk.
7802 2018-01-15 Richard Biener <rguenther@suse.de>
7805 * expmed.c (extract_bit_field_1): Fix typo.
7807 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7810 * config/arm/iterators.md (VF): New mode iterator.
7811 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
7812 Remove integer-related logic from pattern.
7813 (neon_vabd<mode>_3): Likewise.
7815 2018-01-15 Jakub Jelinek <jakub@redhat.com>
7818 * common.opt (fstrict-overflow): No longer an alias.
7819 (fwrapv-pointer): New option.
7820 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
7821 also for pointer types based on flag_wrapv_pointer.
7822 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
7823 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
7824 opts->x_flag_wrapv got set.
7825 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
7826 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
7827 POINTER_TYPE_OVERFLOW_UNDEFINED.
7828 * match.pd: Likewise in address comparison pattern.
7829 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
7831 2018-01-15 Richard Biener <rguenther@suse.de>
7834 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
7835 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
7836 Reset type names to their identifier if their TYPE_DECL doesn't
7837 have linkage (and thus is used for ODR and devirt).
7838 (save_debug_info_for_decl): Remove.
7839 (save_debug_info_for_type): Likewise.
7840 (add_tree_to_fld_list): Adjust.
7841 * tree-pretty-print.c (dump_generic_node): Make dumping of
7842 type names more robust.
7844 2018-01-15 Richard Biener <rguenther@suse.de>
7846 * BASE-VER: Bump to 8.0.1.
7848 2018-01-14 Martin Sebor <msebor@redhat.com>
7851 * builtins.c (check_access): Avoid warning when the no-warning bit
7854 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
7856 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
7857 * ira-color (allocno_hard_regs_compare): Likewise.
7859 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
7862 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
7863 Use .pushsection/.popsection.
7865 2018-01-14 Martin Sebor <msebor@redhat.com>
7868 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
7870 2018-01-14 Jakub Jelinek <jakub@redhat.com>
7872 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
7873 entry from extra_headers.
7874 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
7875 extra_headers, make the list bitwise identical to the i?86-*-* one.
7877 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7879 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
7880 -mcmodel=large with -mindirect-branch=thunk,
7881 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
7882 -mfunction-return=thunk-extern.
7883 * doc/invoke.texi: Document -mcmodel=large is incompatible with
7884 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
7885 -mfunction-return=thunk and -mfunction-return=thunk-extern.
7887 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7889 * config/i386/i386.c (print_reg): Print the name of the full
7890 integer register without '%'.
7891 (ix86_print_operand): Handle 'V'.
7892 * doc/extend.texi: Document 'V' modifier.
7894 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7896 * config/i386/constraints.md (Bs): Disallow memory operand for
7897 -mindirect-branch-register.
7899 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7900 (GOT_memory_operand): Likewise.
7901 (call_insn_operand): Likewise.
7902 (sibcall_insn_operand): Likewise.
7903 (GOT32_symbol_operand): Likewise.
7904 * config/i386/i386.md (indirect_jump): Call convert_memory_address
7905 for -mindirect-branch-register.
7906 (tablejump): Likewise.
7907 (*sibcall_memory): Likewise.
7908 (*sibcall_value_memory): Likewise.
7909 Disallow peepholes of indirect call and jump via memory for
7910 -mindirect-branch-register.
7911 (*call_pop): Replace m with Bw.
7912 (*call_value_pop): Likewise.
7913 (*sibcall_pop_memory): Replace m with Bs.
7914 * config/i386/i386.opt (mindirect-branch-register): New option.
7915 * doc/invoke.texi: Document -mindirect-branch-register option.
7917 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7919 * config/i386/i386-protos.h (ix86_output_function_return): New.
7920 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
7921 set function_return_type.
7922 (indirect_thunk_name): Add ret_p to indicate thunk for function
7924 (output_indirect_thunk_function): Pass false to
7925 indirect_thunk_name.
7926 (ix86_output_indirect_branch_via_reg): Likewise.
7927 (ix86_output_indirect_branch_via_push): Likewise.
7928 (output_indirect_thunk_function): Create alias for function
7929 return thunk if regno < 0.
7930 (ix86_output_function_return): New function.
7931 (ix86_handle_fndecl_attribute): Handle function_return.
7932 (ix86_attribute_table): Add function_return.
7933 * config/i386/i386.h (machine_function): Add
7934 function_return_type.
7935 * config/i386/i386.md (simple_return_internal): Use
7936 ix86_output_function_return.
7937 (simple_return_internal_long): Likewise.
7938 * config/i386/i386.opt (mfunction-return=): New option.
7939 (indirect_branch): Mention -mfunction-return=.
7940 * doc/extend.texi: Document function_return function attribute.
7941 * doc/invoke.texi: Document -mfunction-return= option.
7943 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7945 * config/i386/i386-opts.h (indirect_branch): New.
7946 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
7947 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
7948 with local indirect jump when converting indirect call and jump.
7949 (ix86_set_indirect_branch_type): New.
7950 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
7951 (indirectlabelno): New.
7952 (indirect_thunk_needed): Likewise.
7953 (indirect_thunk_bnd_needed): Likewise.
7954 (indirect_thunks_used): Likewise.
7955 (indirect_thunks_bnd_used): Likewise.
7956 (INDIRECT_LABEL): Likewise.
7957 (indirect_thunk_name): Likewise.
7958 (output_indirect_thunk): Likewise.
7959 (output_indirect_thunk_function): Likewise.
7960 (ix86_output_indirect_branch_via_reg): Likewise.
7961 (ix86_output_indirect_branch_via_push): Likewise.
7962 (ix86_output_indirect_branch): Likewise.
7963 (ix86_output_indirect_jmp): Likewise.
7964 (ix86_code_end): Call output_indirect_thunk_function if needed.
7965 (ix86_output_call_insn): Call ix86_output_indirect_branch if
7967 (ix86_handle_fndecl_attribute): Handle indirect_branch.
7968 (ix86_attribute_table): Add indirect_branch.
7969 * config/i386/i386.h (machine_function): Add indirect_branch_type
7970 and has_local_indirect_jump.
7971 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
7973 (tablejump): Likewise.
7974 (*indirect_jump): Use ix86_output_indirect_jmp.
7975 (*tablejump_1): Likewise.
7976 (simple_return_indirect_internal): Likewise.
7977 * config/i386/i386.opt (mindirect-branch=): New option.
7978 (indirect_branch): New.
7981 (thunk-inline): Likewise.
7982 (thunk-extern): Likewise.
7983 * doc/extend.texi: Document indirect_branch function attribute.
7984 * doc/invoke.texi: Document -mindirect-branch= option.
7986 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
7989 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
7991 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
7993 * ipa-inline.c (want_inline_small_function_p): Return false if
7994 inlining has already failed with CIF_FINAL_ERROR.
7995 (update_caller_keys): Call want_inline_small_function_p before
7997 (update_callee_keys): Likewise.
7999 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
8001 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
8003 (rs6000_quadword_masked_address_p): Likewise.
8004 (quad_aligned_load_p): Likewise.
8005 (quad_aligned_store_p): Likewise.
8006 (const_load_sequence_p): Add comment to describe the outer-most loop.
8007 (mimic_memory_attributes_and_flags): New function.
8008 (rs6000_gen_stvx): Likewise.
8009 (replace_swapped_aligned_store): Likewise.
8010 (rs6000_gen_lvx): Likewise.
8011 (replace_swapped_aligned_load): Likewise.
8012 (replace_swapped_load_constant): Capitalize argument name in
8013 comment describing this function.
8014 (rs6000_analyze_swaps): Add a third pass to search for vector loads
8015 and stores that access quad-word aligned addresses and replace
8016 with stvx or lvx instructions when appropriate.
8017 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
8018 New function prototype.
8019 (rs6000_quadword_masked_address_p): Likewise.
8020 (rs6000_gen_lvx): Likewise.
8021 (rs6000_gen_stvx): Likewise.
8022 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
8023 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
8024 when memory address is aligned.
8025 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
8026 this split to select lvx instruction when memory address is aligned.
8027 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
8028 instruction when memory address is aligned.
8029 (*vsx_le_perm_load_v16qi): Likewise.
8030 (four unnamed splitters): Modify to select the stvx instruction
8031 when memory is aligned.
8033 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
8035 * predict.c (determine_unlikely_bbs): Handle correctly BBs
8036 which appears in the queue multiple times.
8038 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8039 Alan Hayward <alan.hayward@arm.com>
8040 David Sherwood <david.sherwood@arm.com>
8042 * tree-vectorizer.h (vec_lower_bound): New structure.
8043 (_loop_vec_info): Add check_nonzero and lower_bounds.
8044 (LOOP_VINFO_CHECK_NONZERO): New macro.
8045 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
8046 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
8047 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
8048 fields. Make seg_len the distance travelled, not including the
8050 (dr_direction_indicator): Declare.
8051 (dr_zero_step_indicator): Likewise.
8052 (dr_known_forward_stride_p): Likewise.
8053 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
8055 (runtime_alias_check_p): Allow runtime alias checks with
8057 (operator ==): Compare access_size and align.
8058 (prune_runtime_alias_test_list): Rework for new distinction between
8059 the access_size and seg_len.
8060 (create_intersect_range_checks_index): Likewise. Cope with polynomial
8062 (get_segment_min_max): New function.
8063 (create_intersect_range_checks): Use it.
8064 (dr_step_indicator): New function.
8065 (dr_direction_indicator): Likewise.
8066 (dr_zero_step_indicator): Likewise.
8067 (dr_known_forward_stride_p): Likewise.
8068 * tree-loop-distribution.c (data_ref_segment_size): Return
8069 DR_STEP * (niters - 1).
8070 (compute_alias_check_pairs): Update call to the dr_with_seg_len
8072 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
8073 (vect_preserves_scalar_order_p): New function, split out from...
8074 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
8075 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
8076 (vect_vfa_access_size): New function.
8077 (vect_vfa_align): Likewise.
8078 (vect_compile_time_alias): Take access_size_a and access_b arguments.
8079 (dump_lower_bound): New function.
8080 (vect_check_lower_bound): Likewise.
8081 (vect_small_gap_p): Likewise.
8082 (vectorizable_with_step_bound_p): Likewise.
8083 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
8084 depencies if the vectorization factor is 1. Convert the checks
8085 for nonzero steps into checks on the bounds of DR_STEP. Try using
8086 a bunds check for variable steps if the minimum required step is
8087 relatively small. Update calls to the dr_with_seg_len
8088 constructor and to vect_compile_time_alias.
8089 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
8091 (vect_loop_versioning): Call it.
8092 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
8094 (vect_estimate_min_profitable_iters): Account for any bounds checks.
8096 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8097 Alan Hayward <alan.hayward@arm.com>
8098 David Sherwood <david.sherwood@arm.com>
8100 * doc/sourcebuild.texi (vect_scatter_store): Document.
8101 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
8103 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
8105 * genopinit.c (main): Add supports_vec_scatter_store and
8106 supports_vec_scatter_store_cached to target_optabs.
8107 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
8108 IFN_MASK_SCATTER_STORE.
8109 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
8111 * internal-fn.h (internal_store_fn_p): Declare.
8112 (internal_fn_stored_value_index): Likewise.
8113 * internal-fn.c (scatter_store_direct): New macro.
8114 (expand_scatter_store_optab_fn): New function.
8115 (direct_scatter_store_optab_supported_p): New macro.
8116 (internal_store_fn_p): New function.
8117 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
8118 IFN_MASK_SCATTER_STORE.
8119 (internal_fn_mask_index): Likewise.
8120 (internal_fn_stored_value_index): New function.
8121 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
8123 * optabs-query.h (supports_vec_scatter_store_p): Declare.
8124 * optabs-query.c (supports_vec_scatter_store_p): New function.
8125 * tree-vectorizer.h (vect_get_store_rhs): Declare.
8126 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
8127 true for scatter stores.
8128 (vect_gather_scatter_fn_p): Handle scatter stores too.
8129 (vect_check_gather_scatter): Consider using scatter stores if
8130 supports_vec_scatter_store_p.
8131 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
8133 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
8134 internal_fn_stored_value_index.
8135 (check_load_store_masking): Handle scatter stores too.
8136 (vect_get_store_rhs): Make public.
8137 (vectorizable_call): Use internal_store_fn_p.
8138 (vectorizable_store): Handle scatter store internal functions.
8139 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
8140 when deciding whether the end of the group has been reached.
8141 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
8142 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
8143 (mask_scatter_store<mode>): New insns.
8145 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8146 Alan Hayward <alan.hayward@arm.com>
8147 David Sherwood <david.sherwood@arm.com>
8149 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
8150 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
8151 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
8153 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
8154 Use vect_truncate_gather_scatter_offset if we can't treat the
8155 operation as a normal gather load or scatter store.
8156 (get_group_load_store_type): Take the gather_scatter_info
8157 as argument. Try using a gather load or scatter store for
8158 single-element groups.
8159 (get_load_store_type): Update calls to get_group_load_store_type
8160 and vect_use_strided_gather_scatters_p.
8162 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8163 Alan Hayward <alan.hayward@arm.com>
8164 David Sherwood <david.sherwood@arm.com>
8166 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
8167 optional tree argument.
8168 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
8170 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
8171 but continue to use the current value as a fallback.
8172 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
8173 to compare the updates.
8174 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
8175 (get_load_store_type): Use it when handling a strided access.
8176 (vect_get_strided_load_store_ops): New function.
8177 (vect_get_data_ptr_increment): Likewise.
8178 (vectorizable_load): Handle strided gather loads. Always pass
8179 a step to vect_create_data_ref_ptr and bump_vector_ptr.
8181 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8182 Alan Hayward <alan.hayward@arm.com>
8183 David Sherwood <david.sherwood@arm.com>
8185 * doc/md.texi (gather_load@var{m}): Document.
8186 (mask_gather_load@var{m}): Likewise.
8187 * genopinit.c (main): Add supports_vec_gather_load and
8188 supports_vec_gather_load_cached to target_optabs.
8189 * optabs-tree.c (init_tree_optimization_optabs): Use
8190 ggc_cleared_alloc to allocate target_optabs.
8191 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
8192 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
8194 * internal-fn.h (internal_load_fn_p): Declare.
8195 (internal_gather_scatter_fn_p): Likewise.
8196 (internal_fn_mask_index): Likewise.
8197 (internal_gather_scatter_fn_supported_p): Likewise.
8198 * internal-fn.c (gather_load_direct): New macro.
8199 (expand_gather_load_optab_fn): New function.
8200 (direct_gather_load_optab_supported_p): New macro.
8201 (direct_internal_fn_optab): New function.
8202 (internal_load_fn_p): Likewise.
8203 (internal_gather_scatter_fn_p): Likewise.
8204 (internal_fn_mask_index): Likewise.
8205 (internal_gather_scatter_fn_supported_p): Likewise.
8206 * optabs-query.c (supports_at_least_one_mode_p): New function.
8207 (supports_vec_gather_load_p): Likewise.
8208 * optabs-query.h (supports_vec_gather_load_p): Declare.
8209 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
8210 and memory_type field.
8211 (NUM_PATTERNS): Bump to 15.
8212 * tree-vect-data-refs.c: Include internal-fn.h.
8213 (vect_gather_scatter_fn_p): New function.
8214 (vect_describe_gather_scatter_call): Likewise.
8215 (vect_check_gather_scatter): Try using internal functions for
8216 gather loads. Recognize existing calls to a gather load function.
8217 (vect_analyze_data_refs): Consider using gather loads if
8218 supports_vec_gather_load_p.
8219 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
8220 (vect_get_gather_scatter_offset_type): Likewise.
8221 (vect_convert_mask_for_vectype): Likewise.
8222 (vect_add_conversion_to_patterm): Likewise.
8223 (vect_try_gather_scatter_pattern): Likewise.
8224 (vect_recog_gather_scatter_pattern): New pattern recognizer.
8225 (vect_vect_recog_func_ptrs): Add it.
8226 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
8227 internal_fn_mask_index and internal_gather_scatter_fn_p.
8228 (check_load_store_masking): Take the gather_scatter_info as an
8229 argument and handle gather loads.
8230 (vect_get_gather_scatter_ops): New function.
8231 (vectorizable_call): Check internal_load_fn_p.
8232 (vectorizable_load): Likewise. Handle gather load internal
8234 (vectorizable_store): Update call to check_load_store_masking.
8235 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
8236 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
8237 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
8238 (aarch64_gather_scale_operand_d): New predicates.
8239 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
8240 (mask_gather_load<mode>): New insns.
8242 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8243 Alan Hayward <alan.hayward@arm.com>
8244 David Sherwood <david.sherwood@arm.com>
8246 * optabs.def (fold_left_plus_optab): New optab.
8247 * doc/md.texi (fold_left_plus_@var{m}): Document.
8248 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
8249 * internal-fn.c (fold_left_direct): Define.
8250 (expand_fold_left_optab_fn): Likewise.
8251 (direct_fold_left_optab_supported_p): Likewise.
8252 * fold-const-call.c (fold_const_fold_left): New function.
8253 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
8254 * tree-parloops.c (valid_reduction_p): New function.
8255 (gather_scalar_reductions): Use it.
8256 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
8257 (vect_finish_replace_stmt): Declare.
8258 * tree-vect-loop.c (fold_left_reduction_fn): New function.
8259 (needs_fold_left_reduction_p): New function, split out from...
8260 (vect_is_simple_reduction): ...here. Accept reductions that
8261 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
8262 (vect_force_simple_reduction): Also store the reduction type in
8263 the assignment's STMT_VINFO_REDUC_TYPE.
8264 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
8265 (merge_with_identity): New function.
8266 (vect_expand_fold_left): Likewise.
8267 (vectorize_fold_left_reduction): Likewise.
8268 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
8269 scalar phi in place for it. Check for target support and reject
8270 cases that would reassociate the operation. Defer the transform
8271 phase to vectorize_fold_left_reduction.
8272 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
8273 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
8274 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
8276 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8278 * tree-if-conv.c (predicate_mem_writes): Remove redundant
8279 call to ifc_temp_var.
8281 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8282 Alan Hayward <alan.hayward@arm.com>
8283 David Sherwood <david.sherwood@arm.com>
8285 * target.def (legitimize_address_displacement): Take the original
8286 offset as a poly_int.
8287 * targhooks.h (default_legitimize_address_displacement): Update
8289 * targhooks.c (default_legitimize_address_displacement): Likewise.
8290 * doc/tm.texi: Regenerate.
8291 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
8292 as an argument, moving assert of ad->disp == ad->disp_term to...
8293 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
8294 Try calling targetm.legitimize_address_displacement before expanding
8295 the address rather than afterwards, and adjust for the new interface.
8296 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
8297 Match the new hook interface. Handle SVE addresses.
8298 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
8301 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8303 * Makefile.in (OBJS): Add early-remat.o.
8304 * target.def (select_early_remat_modes): New hook.
8305 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
8306 * doc/tm.texi: Regenerate.
8307 * targhooks.h (default_select_early_remat_modes): Declare.
8308 * targhooks.c (default_select_early_remat_modes): New function.
8309 * timevar.def (TV_EARLY_REMAT): New timevar.
8310 * passes.def (pass_early_remat): New pass.
8311 * tree-pass.h (make_pass_early_remat): Declare.
8312 * early-remat.c: New file.
8313 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
8315 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
8317 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8318 Alan Hayward <alan.hayward@arm.com>
8319 David Sherwood <david.sherwood@arm.com>
8321 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
8322 vfm1 with a bound_epilog parameter.
8323 (vect_do_peeling): Update calls accordingly, and move the prologue
8324 call earlier in the function. Treat the base bound_epilog as 0 for
8325 fully-masked loops and retain vf - 1 for other loops. Add 1 to
8326 this base when peeling for gaps.
8327 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
8328 with fully-masked loops.
8329 (vect_estimate_min_profitable_iters): Handle the single peeled
8330 iteration in that case.
8332 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8333 Alan Hayward <alan.hayward@arm.com>
8334 David Sherwood <david.sherwood@arm.com>
8336 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
8337 single-element interleaving even if the size is not a power of 2.
8338 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
8339 accesses for single-element interleaving if the group size is
8342 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8343 Alan Hayward <alan.hayward@arm.com>
8344 David Sherwood <david.sherwood@arm.com>
8346 * doc/md.texi (fold_extract_last_@var{m}): Document.
8347 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
8348 * optabs.def (fold_extract_last_optab): New optab.
8349 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
8350 * internal-fn.c (fold_extract_direct): New macro.
8351 (expand_fold_extract_optab_fn): Likewise.
8352 (direct_fold_extract_optab_supported_p): Likewise.
8353 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
8354 * tree-vect-loop.c (vect_model_reduction_cost): Handle
8355 EXTRACT_LAST_REDUCTION.
8356 (get_initial_def_for_reduction): Do not create an initial vector
8357 for EXTRACT_LAST_REDUCTION reductions.
8358 (vectorizable_reduction): Leave the scalar phi in place for
8359 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
8360 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
8361 epilogue code for EXTRACT_LAST_REDUCTION and defer the
8362 transform phase to vectorizable_condition.
8363 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
8365 (vect_finish_stmt_generation): ...here.
8366 (vect_finish_replace_stmt): New function.
8367 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
8368 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
8370 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
8372 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8373 Alan Hayward <alan.hayward@arm.com>
8374 David Sherwood <david.sherwood@arm.com>
8376 * doc/md.texi (extract_last_@var{m}): Document.
8377 * optabs.def (extract_last_optab): New optab.
8378 * internal-fn.def (EXTRACT_LAST): New internal function.
8379 * internal-fn.c (cond_unary_direct): New macro.
8380 (expand_cond_unary_optab_fn): Likewise.
8381 (direct_cond_unary_optab_supported_p): Likewise.
8382 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
8383 loops using EXTRACT_LAST.
8384 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
8385 (extract_last_<mode>): ...this optab.
8386 (vec_extract<mode><Vel>): Update accordingly.
8388 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8389 Alan Hayward <alan.hayward@arm.com>
8390 David Sherwood <david.sherwood@arm.com>
8392 * target.def (empty_mask_is_expensive): New hook.
8393 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
8394 * doc/tm.texi: Regenerate.
8395 * targhooks.h (default_empty_mask_is_expensive): Declare.
8396 * targhooks.c (default_empty_mask_is_expensive): New function.
8397 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
8398 if the target says that empty masks are expensive.
8399 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
8401 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
8403 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8404 Alan Hayward <alan.hayward@arm.com>
8405 David Sherwood <david.sherwood@arm.com>
8407 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
8408 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
8409 (vect_use_loop_mask_for_alignment_p): New function.
8410 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
8411 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
8412 niters_skip argument. Make sure that the first niters_skip elements
8413 of the first iteration are inactive.
8414 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
8415 Update call to vect_set_loop_masks_directly.
8416 (get_misalign_in_elems): New function, split out from...
8417 (vect_gen_prolog_loop_niters): ...here.
8418 (vect_update_init_of_dr): Take a code argument that specifies whether
8419 the adjustment should be added or subtracted.
8420 (vect_update_init_of_drs): Likewise.
8421 (vect_prepare_for_masked_peels): New function.
8422 (vect_do_peeling): Skip prologue peeling if we're using a mask
8423 instead. Update call to vect_update_inits_of_drs.
8424 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
8426 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
8427 alignment. Do not include the number of peeled iterations in
8428 the minimum threshold in that case.
8429 (vectorizable_induction): Adjust the start value down by
8430 LOOP_VINFO_MASK_SKIP_NITERS iterations.
8431 (vect_transform_loop): Call vect_prepare_for_masked_peels.
8432 Take the number of skipped iterations into account when calculating
8434 * tree-vect-stmts.c (vect_gen_while_not): New function.
8436 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8437 Alan Hayward <alan.hayward@arm.com>
8438 David Sherwood <david.sherwood@arm.com>
8440 * doc/sourcebuild.texi (vect_fully_masked): Document.
8441 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
8443 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
8445 (vect_analyze_loop_2): ...here. Don't check the vectorization
8446 factor against the number of loop iterations if the loop is
8449 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8450 Alan Hayward <alan.hayward@arm.com>
8451 David Sherwood <david.sherwood@arm.com>
8453 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
8454 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
8455 (dump_groups): Update accordingly.
8456 (iv_use::mem_type): New member variable.
8457 (address_p): New function.
8458 (record_use): Add a mem_type argument and initialize the new
8460 (record_group_use): Add a mem_type argument. Use address_p.
8461 Remove obsolete null checks of base_object. Update call to record_use.
8462 (find_interesting_uses_op): Update call to record_group_use.
8463 (find_interesting_uses_cond): Likewise.
8464 (find_interesting_uses_address): Likewise.
8465 (get_mem_type_for_internal_fn): New function.
8466 (find_address_like_use): Likewise.
8467 (find_interesting_uses_stmt): Try find_address_like_use before
8468 calling find_interesting_uses_op.
8469 (addr_offset_valid_p): Use the iv mem_type field as the type
8470 of the addressed memory.
8471 (add_autoinc_candidates): Likewise.
8472 (get_address_cost): Likewise.
8473 (split_small_address_groups_p): Use address_p.
8474 (split_address_groups): Likewise.
8475 (add_iv_candidate_for_use): Likewise.
8476 (autoinc_possible_for_pair): Likewise.
8477 (rewrite_groups): Likewise.
8478 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
8479 (determine_group_iv_cost): Update after split of USE_ADDRESS.
8480 (get_alias_ptr_type_for_ptr_address): New function.
8481 (rewrite_use_address): Rewrite address uses in calls that were
8482 identified by find_address_like_use.
8484 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8485 Alan Hayward <alan.hayward@arm.com>
8486 David Sherwood <david.sherwood@arm.com>
8488 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
8490 * gimple-expr.h (is_gimple_addressable: Likewise.
8491 * gimple-expr.c (is_gimple_address): Likewise.
8492 * internal-fn.c (expand_call_mem_ref): New function.
8493 (expand_mask_load_optab_fn): Use it.
8494 (expand_mask_store_optab_fn): Likewise.
8496 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8497 Alan Hayward <alan.hayward@arm.com>
8498 David Sherwood <david.sherwood@arm.com>
8500 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
8501 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
8502 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
8503 (cond_umax@var{mode}): Document.
8504 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
8505 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
8506 (cond_umin_optab, cond_umax_optab): New optabs.
8507 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
8508 (COND_IOR, COND_XOR): New internal functions.
8509 * internal-fn.h (get_conditional_internal_fn): Declare.
8510 * internal-fn.c (cond_binary_direct): New macro.
8511 (expand_cond_binary_optab_fn): Likewise.
8512 (direct_cond_binary_optab_supported_p): Likewise.
8513 (get_conditional_internal_fn): New function.
8514 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
8515 Cope with reduction statements that are vectorized as calls rather
8517 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
8518 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
8519 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
8520 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
8521 (UNSPEC_COND_EOR): New unspecs.
8522 (optab): Add mappings for them.
8523 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
8524 (sve_int_op, sve_fp_op): New int attributes.
8526 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8527 Alan Hayward <alan.hayward@arm.com>
8528 David Sherwood <david.sherwood@arm.com>
8530 * optabs.def (while_ult_optab): New optab.
8531 * doc/md.texi (while_ult@var{m}@var{n}): Document.
8532 * internal-fn.def (WHILE_ULT): New internal function.
8533 * internal-fn.h (direct_internal_fn_supported_p): New override
8534 that takes two types as argument.
8535 * internal-fn.c (while_direct): New macro.
8536 (expand_while_optab_fn): New function.
8537 (convert_optab_supported_p): Likewise.
8538 (direct_while_optab_supported_p): New macro.
8539 * wide-int.h (wi::udiv_ceil): New function.
8540 * tree-vectorizer.h (rgroup_masks): New structure.
8541 (vec_loop_masks): New typedef.
8542 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
8544 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
8545 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
8546 (vect_max_vf): New function.
8547 (slpeel_make_loop_iterate_ntimes): Delete.
8548 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
8549 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
8550 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
8551 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
8552 internal-fn.h, stor-layout.h and optabs-query.h.
8553 (vect_set_loop_mask): New function.
8554 (add_preheader_seq): Likewise.
8555 (add_header_seq): Likewise.
8556 (interleave_supported_p): Likewise.
8557 (vect_maybe_permute_loop_masks): Likewise.
8558 (vect_set_loop_masks_directly): Likewise.
8559 (vect_set_loop_condition_masked): Likewise.
8560 (vect_set_loop_condition_unmasked): New function, split out from
8561 slpeel_make_loop_iterate_ntimes.
8562 (slpeel_make_loop_iterate_ntimes): Rename to..
8563 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
8564 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
8565 (vect_do_peeling): Update call accordingly.
8566 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
8568 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
8569 mask_compare_type, can_fully_mask_p and fully_masked_p.
8570 (release_vec_loop_masks): New function.
8571 (_loop_vec_info): Use it to free the loop masks.
8572 (can_produce_all_loop_masks_p): New function.
8573 (vect_get_max_nscalars_per_iter): Likewise.
8574 (vect_verify_full_masking): Likewise.
8575 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
8576 retries, and free the mask rgroups before retrying. Check loop-wide
8577 reasons for disallowing fully-masked loops. Make the final decision
8578 about whether use a fully-masked loop or not.
8579 (vect_estimate_min_profitable_iters): Do not assume that peeling
8580 for the number of iterations will be needed for fully-masked loops.
8581 (vectorizable_reduction): Disable fully-masked loops.
8582 (vectorizable_live_operation): Likewise.
8583 (vect_halve_mask_nunits): New function.
8584 (vect_double_mask_nunits): Likewise.
8585 (vect_record_loop_mask): Likewise.
8586 (vect_get_loop_mask): Likewise.
8587 (vect_transform_loop): Handle the case in which the final loop
8588 iteration might handle a partial vector. Call vect_set_loop_condition
8589 instead of slpeel_make_loop_iterate_ntimes.
8590 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
8591 (check_load_store_masking): New function.
8592 (prepare_load_store_mask): Likewise.
8593 (vectorizable_store): Handle fully-masked loops.
8594 (vectorizable_load): Likewise.
8595 (supportable_widening_operation): Use vect_halve_mask_nunits for
8597 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
8598 (vect_gen_while): New function.
8599 * config/aarch64/aarch64.md (umax<mode>3): New expander.
8600 (aarch64_uqdec<mode>): New insn.
8602 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8603 Alan Hayward <alan.hayward@arm.com>
8604 David Sherwood <david.sherwood@arm.com>
8606 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
8607 (reduc_xor_scal_optab): New optabs.
8608 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
8609 (reduc_xor_scal_@var{m}): Document.
8610 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
8611 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
8613 * fold-const-call.c (fold_const_call): Handle them.
8614 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
8615 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
8616 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
8617 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
8618 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
8619 (UNSPEC_XORV): New unspecs.
8620 (optab): Add entries for them.
8621 (BITWISEV): New int iterator.
8622 (bit_reduc_op): New int attributes.
8624 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8625 Alan Hayward <alan.hayward@arm.com>
8626 David Sherwood <david.sherwood@arm.com>
8628 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
8629 * internal-fn.def (VEC_SHL_INSERT): New internal function.
8630 * optabs.def (vec_shl_insert_optab): New optab.
8631 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
8632 (duplicate_and_interleave): Likewise.
8633 * tree-vect-loop.c: Include internal-fn.h.
8634 (neutral_op_for_slp_reduction): New function, split out from
8635 get_initial_defs_for_reduction.
8636 (get_initial_def_for_reduction): Handle option 2 for variable-length
8637 vectors by loading the neutral value into a vector and then shifting
8638 the initial value into element 0.
8639 (get_initial_defs_for_reduction): Replace the code argument with
8640 the neutral value calculated by neutral_op_for_slp_reduction.
8641 Use gimple_build_vector for constant-length vectors.
8642 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
8643 but the first group_size elements have a neutral value.
8644 Use duplicate_and_interleave otherwise.
8645 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
8646 Update call to get_initial_defs_for_reduction. Handle SLP
8647 reductions for variable-length vectors by creating one vector
8648 result for each scalar result, with the elements associated
8649 with other scalar results stubbed out with the neutral value.
8650 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
8651 Require IFN_VEC_SHL_INSERT for double reductions on
8652 variable-length vectors, or SLP reductions that have
8653 a neutral value. Require can_duplicate_and_interleave_p
8654 support for variable-length unchained SLP reductions if there
8655 is no neutral value, such as for MIN/MAX reductions. Also require
8656 the number of vector elements to be a multiple of the number of
8657 SLP statements when doing variable-length unchained SLP reductions.
8658 Update call to vect_create_epilog_for_reduction.
8659 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
8660 and remove initial values.
8661 (duplicate_and_interleave): Make public.
8662 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
8663 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
8665 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8666 Alan Hayward <alan.hayward@arm.com>
8667 David Sherwood <david.sherwood@arm.com>
8669 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
8670 (can_duplicate_and_interleave_p): New function.
8671 (vect_get_and_check_slp_defs): Take the vector of statements
8672 rather than just the current one. Remove excess parentheses.
8673 Restriction rejectinon of vect_constant_def and vect_external_def
8674 for variable-length vectors to boolean types, or types for which
8675 can_duplicate_and_interleave_p is false.
8676 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
8677 (duplicate_and_interleave): New function.
8678 (vect_get_constant_vectors): Use gimple_build_vector for
8679 constant-length vectors and suitable variable-length constant
8680 vectors. Use duplicate_and_interleave for other variable-length
8681 vectors. Don't defer the update when inserting new statements.
8683 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8684 Alan Hayward <alan.hayward@arm.com>
8685 David Sherwood <david.sherwood@arm.com>
8687 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
8688 min_profitable_iters doesn't go negative.
8690 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8691 Alan Hayward <alan.hayward@arm.com>
8692 David Sherwood <david.sherwood@arm.com>
8694 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
8695 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
8696 * optabs.def (vec_mask_load_lanes_optab): New optab.
8697 (vec_mask_store_lanes_optab): Likewise.
8698 * internal-fn.def (MASK_LOAD_LANES): New internal function.
8699 (MASK_STORE_LANES): Likewise.
8700 * internal-fn.c (mask_load_lanes_direct): New macro.
8701 (mask_store_lanes_direct): Likewise.
8702 (expand_mask_load_optab_fn): Handle masked operations.
8703 (expand_mask_load_lanes_optab_fn): New macro.
8704 (expand_mask_store_optab_fn): Handle masked operations.
8705 (expand_mask_store_lanes_optab_fn): New macro.
8706 (direct_mask_load_lanes_optab_supported_p): Likewise.
8707 (direct_mask_store_lanes_optab_supported_p): Likewise.
8708 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
8710 (vect_load_lanes_supported): Likewise.
8711 * tree-vect-data-refs.c (strip_conversion): New function.
8712 (can_group_stmts_p): Likewise.
8713 (vect_analyze_data_ref_accesses): Use it instead of checking
8714 for a pair of assignments.
8715 (vect_store_lanes_supported): Take a masked_p parameter.
8716 (vect_load_lanes_supported): Likewise.
8717 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
8718 vect_store_lanes_supported and vect_load_lanes_supported.
8719 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
8720 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
8721 parameter. Don't allow gaps for masked accesses.
8722 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
8723 and vect_load_lanes_supported.
8724 (get_load_store_type): Take a masked_p parameter and update
8725 call to get_group_load_store_type.
8726 (vectorizable_store): Update call to get_load_store_type.
8727 Handle IFN_MASK_STORE_LANES.
8728 (vectorizable_load): Update call to get_load_store_type.
8729 Handle IFN_MASK_LOAD_LANES.
8731 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8732 Alan Hayward <alan.hayward@arm.com>
8733 David Sherwood <david.sherwood@arm.com>
8735 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
8737 * config/aarch64/aarch64-protos.h
8738 (aarch64_sve_struct_memory_operand_p): Declare.
8739 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
8740 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
8741 (VPRED, vpred): Handle SVE structure modes.
8742 * config/aarch64/constraints.md (Utx): New constraint.
8743 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
8744 (aarch64_sve_struct_nonimmediate_operand): New predicates.
8745 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
8746 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
8747 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
8748 structure modes. Split into pieces after RA.
8749 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
8750 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
8752 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
8753 SVE structure modes.
8754 (aarch64_classify_address): Likewise.
8755 (sizetochar): Move earlier in file.
8756 (aarch64_print_operand): Handle SVE register lists.
8757 (aarch64_array_mode): New function.
8758 (aarch64_sve_struct_memory_operand_p): Likewise.
8759 (TARGET_ARRAY_MODE): Redefine.
8761 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8762 Alan Hayward <alan.hayward@arm.com>
8763 David Sherwood <david.sherwood@arm.com>
8765 * target.def (array_mode): New target hook.
8766 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
8767 * doc/tm.texi: Regenerate.
8768 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
8769 * hooks.c (hook_optmode_mode_uhwi_none): New function.
8770 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
8772 * stor-layout.c (mode_for_array): Likewise. Support polynomial
8775 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8776 Alan Hayward <alan.hayward@arm.com>
8777 David Sherwood <david.sherwood@arm.com>
8779 * fold-const.c (fold_binary_loc): Check the argument types
8780 rather than the result type when testing for a vector operation.
8782 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8784 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
8785 * doc/tm.texi: Regenerate.
8787 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8788 Alan Hayward <alan.hayward@arm.com>
8789 David Sherwood <david.sherwood@arm.com>
8791 * doc/invoke.texi (-msve-vector-bits=): Document new option.
8792 (sve): Document new AArch64 extension.
8793 * doc/md.texi (w): Extend the description of the AArch64
8794 constraint to include SVE vectors.
8795 (Upl, Upa): Document new AArch64 predicate constraints.
8796 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
8798 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
8799 (msve-vector-bits=): New option.
8800 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
8801 SVE when these are disabled.
8802 (sve): New extension.
8803 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
8804 modes. Adjust their number of units based on aarch64_sve_vg.
8805 (MAX_BITSIZE_MODE_ANY_MODE): Define.
8806 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
8807 aarch64_addr_query_type.
8808 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
8809 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
8810 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
8811 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
8812 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
8813 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
8814 (aarch64_simd_imm_zero_p): Delete.
8815 (aarch64_check_zero_based_sve_index_immediate): Declare.
8816 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
8817 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
8818 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
8819 (aarch64_sve_float_mul_immediate_p): Likewise.
8820 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
8822 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
8823 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
8824 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
8825 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
8826 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
8827 (aarch64_regmode_natural_size): Likewise.
8828 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
8829 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
8831 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
8832 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
8833 for VG and the SVE predicate registers.
8834 (V_ALIASES): Add a "z"-prefixed alias.
8835 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
8836 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
8837 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
8838 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
8839 (REG_CLASS_NAMES): Add entries for them.
8840 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
8841 and the predicate registers.
8842 (aarch64_sve_vg): Declare.
8843 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
8844 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
8845 (REGMODE_NATURAL_SIZE): Define.
8846 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
8848 * config/aarch64/aarch64.c: Include cfgrtl.h.
8849 (simd_immediate_info): Add a constructor for series vectors,
8850 and an associated step field.
8851 (aarch64_sve_vg): New variable.
8852 (aarch64_dbx_register_number): Handle VG and the predicate registers.
8853 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
8854 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
8855 (VEC_ANY_DATA, VEC_STRUCT): New constants.
8856 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
8857 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
8858 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
8859 (aarch64_get_mask_mode): New functions.
8860 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
8861 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
8862 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
8863 predicate modes and predicate registers. Explicitly restrict
8864 GPRs to modes of 16 bytes or smaller. Only allow FP registers
8865 to store a vector mode if it is recognized by
8866 aarch64_classify_vector_mode.
8867 (aarch64_regmode_natural_size): New function.
8868 (aarch64_hard_regno_caller_save_mode): Return the original mode
8870 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
8871 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
8872 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
8873 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
8875 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
8876 does not overlap dest if the function is frame-related. Handle
8878 (aarch64_split_add_offset): New function.
8879 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
8880 them aarch64_add_offset.
8881 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
8882 and update call to aarch64_sub_sp.
8883 (aarch64_add_cfa_expression): New function.
8884 (aarch64_expand_prologue): Pass extra temporary registers to the
8885 functions above. Handle the case in which we need to emit new
8886 DW_CFA_expressions for registers that were originally saved
8887 relative to the stack pointer, but now have to be expressed
8888 relative to the frame pointer.
8889 (aarch64_output_mi_thunk): Pass extra temporary registers to the
8891 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
8892 IP0 and IP1 values for SVE frames.
8893 (aarch64_expand_vec_series): New function.
8894 (aarch64_expand_sve_widened_duplicate): Likewise.
8895 (aarch64_expand_sve_const_vector): Likewise.
8896 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
8897 Handle SVE constants. Use emit_move_insn to move a force_const_mem
8898 into the register, rather than emitting a SET directly.
8899 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
8900 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
8901 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
8902 (offset_9bit_signed_scaled_p): New functions.
8903 (aarch64_replicate_bitmask_imm): New function.
8904 (aarch64_bitmask_imm): Use it.
8905 (aarch64_cannot_force_const_mem): Reject expressions involving
8906 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
8907 (aarch64_classify_index): Handle SVE indices, by requiring
8908 a plain register index with a scale that matches the element size.
8909 (aarch64_classify_address): Handle SVE addresses. Assert that
8910 the mode of the address is VOIDmode or an integer mode.
8911 Update call to aarch64_classify_symbol.
8912 (aarch64_classify_symbolic_expression): Update call to
8913 aarch64_classify_symbol.
8914 (aarch64_const_vec_all_in_range_p): New function.
8915 (aarch64_print_vector_float_operand): Likewise.
8916 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
8917 "vN" for FP registers with SVE modes. Handle (const ...) vectors
8918 and the FP immediates 1.0 and 0.5.
8919 (aarch64_print_address_internal): Handle SVE addresses.
8920 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
8921 (aarch64_regno_regclass): Handle predicate registers.
8922 (aarch64_secondary_reload): Handle big-endian reloads of SVE
8924 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
8925 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
8926 (aarch64_convert_sve_vector_bits): New function.
8927 (aarch64_override_options): Use it to handle -msve-vector-bits=.
8928 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
8930 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
8931 Handle SVE vector and predicate modes. Accept VL-based constants
8932 that need only one temporary register, and VL offsets that require
8933 no temporary registers.
8934 (aarch64_conditional_register_usage): Mark the predicate registers
8935 as fixed if SVE isn't available.
8936 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
8937 Return true for SVE vector and predicate modes.
8938 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
8939 rather than an unsigned int. Handle SVE modes.
8940 (aarch64_preferred_simd_mode): Update call accordingly. Handle
8942 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
8944 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
8945 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
8946 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
8947 (aarch64_sve_float_mul_immediate_p): New functions.
8948 (aarch64_sve_valid_immediate): New function.
8949 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
8950 Explicitly reject structure modes. Check for INDEX constants.
8951 Handle PTRUE and PFALSE constants.
8952 (aarch64_check_zero_based_sve_index_immediate): New function.
8953 (aarch64_simd_imm_zero_p): Delete.
8954 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
8955 vector modes. Accept constants in the range of CNT[BHWD].
8956 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
8957 ask for an Advanced SIMD mode.
8958 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
8959 (aarch64_simd_vector_alignment): Handle SVE predicates.
8960 (aarch64_vectorize_preferred_vector_alignment): New function.
8961 (aarch64_simd_vector_alignment_reachable): Use it instead of
8963 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
8964 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
8966 (MAX_VECT_LEN): Delete.
8967 (expand_vec_perm_d): Add a vec_flags field.
8968 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
8969 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
8970 (aarch64_evpc_ext): Don't apply a big-endian lane correction
8972 (aarch64_evpc_rev): Rename to...
8973 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
8974 (aarch64_evpc_rev_global): New function.
8975 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
8976 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
8978 (aarch64_evpc_sve_tbl): New function.
8979 (aarch64_expand_vec_perm_const_1): Update after rename of
8980 aarch64_evpc_rev. Handle SVE permutes too, trying
8981 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
8982 than aarch64_evpc_tbl.
8983 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
8984 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
8985 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
8986 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
8987 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
8988 (aarch64_expand_sve_vcond): New functions.
8989 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
8990 of aarch64_vector_mode_p.
8991 (aarch64_dwarf_poly_indeterminate_value): New function.
8992 (aarch64_compute_pressure_classes): Likewise.
8993 (aarch64_can_change_mode_class): Likewise.
8994 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
8995 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
8996 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
8997 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
8998 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
8999 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
9000 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
9001 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
9003 (Dn, Dl, Dr): Accept const as well as const_vector.
9004 (Dz): Likewise. Compare against CONST0_RTX.
9005 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
9006 of "vector" where appropriate.
9007 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
9008 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
9009 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
9010 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
9011 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
9012 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
9013 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
9014 (v_int_equiv): Extend to SVE modes.
9015 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
9017 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
9018 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
9019 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
9020 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
9021 (SVE_COND_FP_CMP): New int iterators.
9022 (perm_hilo): Handle the new unpack unspecs.
9023 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
9025 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
9026 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
9027 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
9028 (aarch64_equality_operator, aarch64_constant_vector_operand)
9029 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
9030 (aarch64_sve_nonimmediate_operand): Likewise.
9031 (aarch64_sve_general_operand): Likewise.
9032 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
9033 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
9034 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
9035 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
9036 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
9037 (aarch64_sve_float_arith_immediate): Likewise.
9038 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
9039 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
9040 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
9041 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
9042 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
9043 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
9044 (aarch64_sve_float_arith_operand): Likewise.
9045 (aarch64_sve_float_arith_with_sub_operand): Likewise.
9046 (aarch64_sve_float_mul_operand): Likewise.
9047 (aarch64_sve_vec_perm_operand): Likewise.
9048 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
9049 (aarch64_mov_operand): Accept const_poly_int and const_vector.
9050 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
9051 as well as const_vector.
9052 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
9053 in file. Use CONST0_RTX and CONSTM1_RTX.
9054 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
9055 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
9056 Use aarch64_simd_imm_zero.
9057 * config/aarch64/aarch64-sve.md: New file.
9058 * config/aarch64/aarch64.md: Include it.
9059 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
9060 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
9061 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
9062 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
9063 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
9064 (sve): New attribute.
9065 (enabled): Disable instructions with the sve attribute unless
9067 (movqi, movhi): Pass CONST_POLY_INT operaneds through
9068 aarch64_expand_mov_immediate.
9069 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
9070 CNT[BHSD] immediates.
9071 (movti): Split CONST_POLY_INT moves into two halves.
9072 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
9073 Split additions that need a temporary here if the destination
9074 is the stack pointer.
9075 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
9076 (*add<mode>3_poly_1): New instruction.
9077 (set_clobber_cc): New expander.
9079 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9081 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
9082 parameter and use it instead of GET_MODE_SIZE (innermode). Use
9083 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
9084 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
9085 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
9086 Change innermode from fixed_mode_size to machine_mode.
9087 (simplify_subreg): Update call accordingly. Handle a constant-sized
9088 subreg of a variable-length CONST_VECTOR.
9090 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9091 Alan Hayward <alan.hayward@arm.com>
9092 David Sherwood <david.sherwood@arm.com>
9094 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
9095 (add_offset_to_base): New function, split out from...
9096 (create_mem_ref): ...here. When handling a scale other than 1,
9097 check first whether the address is valid without the offset.
9098 Add it into the base if so, leaving the index and scale as-is.
9100 2018-01-12 Jakub Jelinek <jakub@redhat.com>
9103 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
9104 fold_for_warn before checking if arg2 is INTEGER_CST.
9106 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
9108 * config/rs6000/predicates.md (load_multiple_operation): Delete.
9109 (store_multiple_operation): Delete.
9110 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
9111 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
9112 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
9113 guarded by TARGET_STRING.
9114 (rs6000_output_load_multiple): Delete.
9115 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
9116 OPTION_MASK_STRING / TARGET_STRING handling.
9117 (print_operand) <'N', 'O'>: Add comment that these are unused now.
9118 (const rs6000_opt_masks) <"string">: Change mask to 0.
9119 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
9120 (MASK_STRING): Delete.
9121 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
9123 (load_multiple): Delete.
9130 (store_multiple): Delete.
9137 (movmemsi_8reg): Delete.
9138 (corresponding unnamed define_insn): Delete.
9139 (movmemsi_6reg): Delete.
9140 (corresponding unnamed define_insn): Delete.
9141 (movmemsi_4reg): Delete.
9142 (corresponding unnamed define_insn): Delete.
9143 (movmemsi_2reg): Delete.
9144 (corresponding unnamed define_insn): Delete.
9145 (movmemsi_1reg): Delete.
9146 (corresponding unnamed define_insn): Delete.
9147 * config/rs6000/rs6000.opt (mno-string): New.
9148 (mstring): Replace by deprecation warning stub.
9149 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
9151 2018-01-12 Jakub Jelinek <jakub@redhat.com>
9153 * regrename.c (regrename_do_replace): If replacing the same
9154 reg multiple times, try to reuse last created gen_raw_REG.
9157 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
9158 main to workaround a bug in GDB.
9160 2018-01-12 Tom de Vries <tom@codesourcery.com>
9163 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
9165 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
9167 PR rtl-optimization/80481
9168 * ira-color.c (get_cap_member): New function.
9169 (allocnos_conflict_by_live_ranges_p): Use it.
9170 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
9171 (setup_slot_coalesced_allocno_live_ranges): Ditto.
9173 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
9176 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
9177 (*saddl_se_1): Ditto.
9179 (*ssubl_se_1): Ditto.
9181 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9183 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
9184 rather than wi::to_widest for DR_INITs.
9185 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
9186 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
9187 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
9189 (vect_analyze_group_access_1): Note that here.
9191 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9193 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
9194 polynomial type sizes.
9196 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9198 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
9199 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
9200 (gimple_add_tmp_var): Likewise.
9202 2018-01-12 Martin Liska <mliska@suse.cz>
9204 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
9205 (gimple_alloc_sizes): Likewise.
9206 (dump_gimple_statistics): Use PRIu64 in printf format.
9207 * gimple.h: Change uint64_t to int.
9209 2018-01-12 Martin Liska <mliska@suse.cz>
9211 * tree-core.h: Use uint64_t instead of int.
9212 * tree.c (tree_node_counts): Likewise.
9213 (tree_node_sizes): Likewise.
9214 (dump_tree_statistics): Use PRIu64 in printf format.
9216 2018-01-12 Martin Liska <mliska@suse.cz>
9218 * Makefile.in: As qsort_chk is implemented in vec.c, add
9219 vec.o to linkage of gencfn-macros.
9220 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
9221 passing the info to record_node_allocation_statistics.
9222 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
9224 * ggc-common.c (struct ggc_usage): Add operator== and use
9225 it in operator< and compare function.
9226 * mem-stats.h (struct mem_usage): Likewise.
9227 * vec.c (struct vec_usage): Remove operator< and compare
9228 function. Can be simply inherited.
9230 2018-01-12 Martin Jambor <mjambor@suse.cz>
9233 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
9234 * tree-ssa-math-opts.c: Include domwalk.h.
9235 (convert_mult_to_fma_1): New function.
9236 (fma_transformation_info): New type.
9237 (fma_deferring_state): Likewise.
9238 (cancel_fma_deferring): New function.
9239 (result_of_phi): Likewise.
9240 (last_fma_candidate_feeds_initial_phi): Likewise.
9241 (convert_mult_to_fma): Added deferring logic, split actual
9242 transformation to convert_mult_to_fma_1.
9243 (math_opts_dom_walker): New type.
9244 (math_opts_dom_walker::after_dom_children): New method, body moved
9245 here from pass_optimize_widening_mul::execute, added deferring logic
9247 (pass_optimize_widening_mul::execute): Moved most of code to
9248 math_opts_dom_walker::after_dom_children.
9249 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
9250 * config/i386/i386.c (ix86_option_override_internal): Added
9251 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
9253 2018-01-12 Richard Biener <rguenther@suse.de>
9256 * dwarf2out.c (gen_variable_die): Do not reset old_die for
9257 inline instance vars.
9259 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
9262 * config/rx/rx.c (rx_is_restricted_memory_address):
9265 2018-01-12 Richard Biener <rguenther@suse.de>
9267 PR tree-optimization/80846
9268 * target.def (split_reduction): New target hook.
9269 * targhooks.c (default_split_reduction): New function.
9270 * targhooks.h (default_split_reduction): Declare.
9271 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
9272 target requests first reduce vectors by combining low and high
9274 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
9275 (get_vectype_for_scalar_type_and_size): Export.
9276 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
9277 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
9278 * doc/tm.texi: Regenerate.
9279 * config/i386/i386.c (ix86_split_reduction): Implement
9280 TARGET_VECTORIZE_SPLIT_REDUCTION.
9282 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
9285 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
9286 in PIC mode except for TARGET_VXWORKS_RTP.
9287 * config/sparc/sparc.c: Include cfgrtl.h.
9288 (TARGET_INIT_PIC_REG): Define.
9289 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
9290 (sparc_pic_register_p): New predicate.
9291 (sparc_legitimate_address_p): Use it.
9292 (sparc_legitimize_pic_address): Likewise.
9293 (sparc_delegitimize_address): Likewise.
9294 (sparc_mode_dependent_address_p): Likewise.
9295 (gen_load_pcrel_sym): Remove 4th parameter.
9296 (load_got_register): Adjust call to above. Remove obsolete stuff.
9297 (sparc_expand_prologue): Do not call load_got_register here.
9298 (sparc_flat_expand_prologue): Likewise.
9299 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
9300 (sparc_use_pseudo_pic_reg): New function.
9301 (sparc_init_pic_reg): Likewise.
9302 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
9303 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
9305 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
9307 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
9308 Add item for branch_cost.
9310 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
9312 PR rtl-optimization/83565
9313 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
9314 not extend the result to a larger mode for rotate operations.
9315 (num_sign_bit_copies1): Likewise.
9317 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9320 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
9322 Use values-Xc.o for -pedantic.
9323 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
9325 2018-01-12 Martin Liska <mliska@suse.cz>
9328 * ipa-devirt.c (final_warning_record::grow_type_warnings):
9330 (possible_polymorphic_call_targets): Use it.
9331 (ipa_devirt): Likewise.
9333 2018-01-12 Martin Liska <mliska@suse.cz>
9335 * profile-count.h (enum profile_quality): Use 0 as invalid
9336 enum value of profile_quality.
9338 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
9340 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
9341 -mext-string options.
9343 2018-01-12 Richard Biener <rguenther@suse.de>
9345 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
9346 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
9347 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
9349 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
9351 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
9353 * configure.ac (--with-long-double-format): Add support for the
9354 configuration option to change the default long double format on
9356 * config.gcc (powerpc*-linux*-*): Likewise.
9357 * configure: Regenerate.
9358 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
9359 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
9360 used without modification.
9362 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9364 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
9365 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
9366 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
9367 MISC_BUILTIN_SPEC_BARRIER.
9368 (rs6000_init_builtins): Likewise.
9369 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
9371 (speculation_barrier): New define_insn.
9372 * doc/extend.texi: Document __builtin_speculation_barrier.
9374 2018-01-11 Jakub Jelinek <jakub@redhat.com>
9377 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
9378 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
9379 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
9381 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
9382 integral modes instead of "ss" and "sd".
9383 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
9384 vectors with 32-bit and 64-bit elements.
9385 (vecdupssescalarmodesuffix): New mode attribute.
9386 (vec_dup<mode>): Use it.
9388 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
9391 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
9392 frame if argument is passed on stack.
9394 2018-01-11 Jakub Jelinek <jakub@redhat.com>
9397 * ree.c (combine_reaching_defs): Optimize also
9398 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
9399 reg2=any_extend(exp); reg1=reg2;, formatting fix.
9401 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
9404 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
9406 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
9409 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
9410 after they are computed.
9412 2018-01-11 Bin Cheng <bin.cheng@arm.com>
9414 PR tree-optimization/83695
9415 * gimple-loop-linterchange.cc
9416 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
9417 reset cached scev information after interchange.
9418 (pass_linterchange::execute): Remove call to scev_reset_htab.
9420 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9422 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
9423 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
9424 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
9425 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
9426 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
9427 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
9428 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
9429 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
9430 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
9431 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
9432 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
9433 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
9434 (V_lane_reg): Likewise.
9435 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
9437 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
9438 (vfmal_lane_low<mode>_intrinsic,
9439 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
9440 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
9441 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
9442 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
9443 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
9444 vfmsl_lane_high<mode>_intrinsic): New define_insns.
9446 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9448 * config/arm/arm-cpus.in (fp16fml): New feature.
9449 (ALL_SIMD): Add fp16fml.
9450 (armv8.2-a): Add fp16fml as an option.
9451 (armv8.3-a): Likewise.
9452 (armv8.4-a): Add fp16fml as part of fp16.
9453 * config/arm/arm.h (TARGET_FP16FML): Define.
9454 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
9456 * config/arm/arm-modes.def (V2HF): Define.
9457 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
9458 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
9459 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
9460 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
9461 vfmsl_low, vfmsl_high): New set of builtins.
9462 * config/arm/iterators.md (PLUSMINUS): New code iterator.
9463 (vfml_op): New code attribute.
9464 (VFMLHALVES): New int iterator.
9465 (VFML, VFMLSEL): New mode attributes.
9466 (V_reg): Define mapping for V2HF.
9467 (V_hi, V_lo): New mode attributes.
9468 (VF_constraint): Likewise.
9469 (vfml_half, vfml_half_selector): New int attributes.
9470 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
9472 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
9473 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
9475 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
9476 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
9477 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
9478 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
9480 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
9481 Document new effective target and option set.
9483 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9485 * config/arm/arm-cpus.in (armv8_4): New feature.
9486 (ARMv8_4a): New fgroup.
9487 (armv8.4-a): New arch.
9488 * config/arm/arm-tables.opt: Regenerate.
9489 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
9490 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
9491 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
9492 Add matching rules for -march=armv8.4-a and extensions.
9493 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
9495 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
9498 * config/rx/rx.md (BW): New mode attribute.
9499 (sync_lock_test_and_setsi): Add mode suffix to insn output.
9501 2018-01-11 Richard Biener <rguenther@suse.de>
9503 PR tree-optimization/83435
9504 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
9505 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
9506 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
9508 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9509 Alan Hayward <alan.hayward@arm.com>
9510 David Sherwood <david.sherwood@arm.com>
9512 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
9514 (aarch64_classify_address): Initialize it. Track polynomial offsets.
9515 (aarch64_print_address_internal): Use it to check for a zero offset.
9517 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9518 Alan Hayward <alan.hayward@arm.com>
9519 David Sherwood <david.sherwood@arm.com>
9521 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
9522 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
9523 Return a poly_int64 rather than a HOST_WIDE_INT.
9524 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
9525 rather than a HOST_WIDE_INT.
9526 * config/aarch64/aarch64.h (aarch64_frame): Protect with
9527 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
9528 hard_fp_offset, frame_size, initial_adjust, callee_offset and
9529 final_offset from HOST_WIDE_INT to poly_int64.
9530 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
9531 to_constant when getting the number of units in an Advanced SIMD
9533 (aarch64_builtin_vectorized_function): Check for a constant number
9535 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
9537 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
9538 attribute instead of GET_MODE_NUNITS.
9539 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
9540 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
9541 GET_MODE_SIZE for fixed-size registers.
9542 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
9543 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
9544 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
9545 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
9546 (aarch64_print_operand, aarch64_print_address_internal)
9547 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
9548 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
9549 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
9550 Handle polynomial GET_MODE_SIZE.
9551 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
9552 wider than SImode without modification.
9553 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
9554 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
9555 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
9556 passing and returning SVE modes.
9557 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
9558 rather than GEN_INT.
9559 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
9560 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
9561 (aarch64_allocate_and_probe_stack_space): Likewise.
9562 (aarch64_layout_frame): Cope with polynomial offsets.
9563 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
9564 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
9566 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
9567 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
9568 poly_int64 rather than a HOST_WIDE_INT.
9569 (aarch64_get_separate_components, aarch64_process_components)
9570 (aarch64_expand_prologue, aarch64_expand_epilogue)
9571 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
9572 (aarch64_anchor_offset): New function, split out from...
9573 (aarch64_legitimize_address): ...here.
9574 (aarch64_builtin_vectorization_cost): Handle polynomial
9575 TYPE_VECTOR_SUBPARTS.
9576 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
9578 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
9579 number of elements from the PARALLEL rather than the mode.
9580 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
9581 rather than GET_MODE_BITSIZE.
9582 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
9583 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
9584 (aarch64_expand_vec_perm_const_1): Handle polynomial
9585 d->perm.length () and d->perm elements.
9586 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
9587 Apply to_constant to d->perm elements.
9588 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
9589 polynomial CONST_VECTOR_NUNITS.
9590 (aarch64_move_pointer): Take amount as a poly_int64 rather
9592 (aarch64_progress_pointer): Avoid temporary variable.
9593 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
9594 the mode attribute instead of GET_MODE.
9596 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9597 Alan Hayward <alan.hayward@arm.com>
9598 David Sherwood <david.sherwood@arm.com>
9600 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
9601 x exists before using it.
9602 (aarch64_add_constant_internal): Rename to...
9603 (aarch64_add_offset_1): ...this. Replace regnum with separate
9604 src and dest rtxes. Handle the case in which they're different,
9605 including when the offset is zero. Replace scratchreg with an rtx.
9606 Use 2 additions if there is no spare register into which we can
9607 move a 16-bit constant.
9608 (aarch64_add_constant): Delete.
9609 (aarch64_add_offset): Replace reg with separate src and dest
9610 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
9611 Use aarch64_add_offset_1.
9612 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
9613 an rtx rather than an int. Take the delta as a poly_int64
9614 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
9615 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
9616 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
9617 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
9618 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
9620 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
9621 aarch64_add_constant.
9623 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9625 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
9626 Use scalar_float_mode.
9628 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9630 * config/aarch64/aarch64-simd.md
9631 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
9632 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
9633 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
9634 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
9635 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
9636 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
9637 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
9638 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
9639 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
9640 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
9642 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9645 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
9646 targ_options->x_arm_arch_string is non NULL.
9648 2018-01-11 Tamar Christina <tamar.christina@arm.com>
9650 * config/aarch64/aarch64.h
9651 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
9653 2018-01-11 Sudakshina Das <sudi.das@arm.com>
9656 * expmed.c (emit_store_flag_force): Swap if const op0
9657 and change VOIDmode to mode of op0.
9659 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9661 PR rtl-optimization/83761
9662 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
9663 than bytes to mode_for_size.
9665 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9668 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
9669 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
9672 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9675 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
9676 when in layout mode.
9677 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
9678 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
9681 2018-01-10 Michael Collison <michael.collison@arm.com>
9683 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
9684 * config/aarch64/aarch64-option-extension.def: Add
9685 AARCH64_OPT_EXTENSION of 'fp16fml'.
9686 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9687 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
9688 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
9689 * config/aarch64/constraints.md (Ui7): New constraint.
9690 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
9691 (VFMLA_SEL_W): Ditto.
9694 (VFMLA16_LOW): New int iterator.
9695 (VFMLA16_HIGH): Ditto.
9696 (UNSPEC_FMLAL): New unspec.
9697 (UNSPEC_FMLSL): Ditto.
9698 (UNSPEC_FMLAL2): Ditto.
9699 (UNSPEC_FMLSL2): Ditto.
9700 (f16mac): New code attribute.
9701 * config/aarch64/aarch64-simd-builtins.def
9702 (aarch64_fmlal_lowv2sf): Ditto.
9703 (aarch64_fmlsl_lowv2sf): Ditto.
9704 (aarch64_fmlalq_lowv4sf): Ditto.
9705 (aarch64_fmlslq_lowv4sf): Ditto.
9706 (aarch64_fmlal_highv2sf): Ditto.
9707 (aarch64_fmlsl_highv2sf): Ditto.
9708 (aarch64_fmlalq_highv4sf): Ditto.
9709 (aarch64_fmlslq_highv4sf): Ditto.
9710 (aarch64_fmlal_lane_lowv2sf): Ditto.
9711 (aarch64_fmlsl_lane_lowv2sf): Ditto.
9712 (aarch64_fmlal_laneq_lowv2sf): Ditto.
9713 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
9714 (aarch64_fmlalq_lane_lowv4sf): Ditto.
9715 (aarch64_fmlsl_lane_lowv4sf): Ditto.
9716 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
9717 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
9718 (aarch64_fmlal_lane_highv2sf): Ditto.
9719 (aarch64_fmlsl_lane_highv2sf): Ditto.
9720 (aarch64_fmlal_laneq_highv2sf): Ditto.
9721 (aarch64_fmlsl_laneq_highv2sf): Ditto.
9722 (aarch64_fmlalq_lane_highv4sf): Ditto.
9723 (aarch64_fmlsl_lane_highv4sf): Ditto.
9724 (aarch64_fmlalq_laneq_highv4sf): Ditto.
9725 (aarch64_fmlsl_laneq_highv4sf): Ditto.
9726 * config/aarch64/aarch64-simd.md:
9727 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
9728 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
9729 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
9730 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
9731 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
9732 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
9733 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
9734 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
9735 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
9736 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
9737 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
9738 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
9739 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
9740 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
9741 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
9742 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
9743 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
9744 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
9745 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
9746 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
9747 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
9748 (vfmlsl_low_u32): Ditto.
9749 (vfmlalq_low_u32): Ditto.
9750 (vfmlslq_low_u32): Ditto.
9751 (vfmlal_high_u32): Ditto.
9752 (vfmlsl_high_u32): Ditto.
9753 (vfmlalq_high_u32): Ditto.
9754 (vfmlslq_high_u32): Ditto.
9755 (vfmlal_lane_low_u32): Ditto.
9756 (vfmlsl_lane_low_u32): Ditto.
9757 (vfmlal_laneq_low_u32): Ditto.
9758 (vfmlsl_laneq_low_u32): Ditto.
9759 (vfmlalq_lane_low_u32): Ditto.
9760 (vfmlslq_lane_low_u32): Ditto.
9761 (vfmlalq_laneq_low_u32): Ditto.
9762 (vfmlslq_laneq_low_u32): Ditto.
9763 (vfmlal_lane_high_u32): Ditto.
9764 (vfmlsl_lane_high_u32): Ditto.
9765 (vfmlal_laneq_high_u32): Ditto.
9766 (vfmlsl_laneq_high_u32): Ditto.
9767 (vfmlalq_lane_high_u32): Ditto.
9768 (vfmlslq_lane_high_u32): Ditto.
9769 (vfmlalq_laneq_high_u32): Ditto.
9770 (vfmlslq_laneq_high_u32): Ditto.
9771 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
9772 (AARCH64_FL_FOR_ARCH8_4): New.
9773 (AARCH64_ISA_F16FML): New ISA flag.
9774 (TARGET_F16FML): New feature flag for fp16fml.
9775 (doc/invoke.texi): Document new fp16fml option.
9777 2018-01-10 Michael Collison <michael.collison@arm.com>
9779 * config/aarch64/aarch64-builtins.c:
9780 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
9781 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9782 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
9783 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
9784 (AARCH64_ISA_SHA3): New ISA flag.
9785 (TARGET_SHA3): New feature flag for sha3.
9786 * config/aarch64/iterators.md (sha512_op): New int attribute.
9787 (CRYPTO_SHA512): New int iterator.
9788 (UNSPEC_SHA512H): New unspec.
9789 (UNSPEC_SHA512H2): Ditto.
9790 (UNSPEC_SHA512SU0): Ditto.
9791 (UNSPEC_SHA512SU1): Ditto.
9792 * config/aarch64/aarch64-simd-builtins.def
9793 (aarch64_crypto_sha512hqv2di): New builtin.
9794 (aarch64_crypto_sha512h2qv2di): Ditto.
9795 (aarch64_crypto_sha512su0qv2di): Ditto.
9796 (aarch64_crypto_sha512su1qv2di): Ditto.
9797 (aarch64_eor3qv8hi): Ditto.
9798 (aarch64_rax1qv2di): Ditto.
9799 (aarch64_xarqv2di): Ditto.
9800 (aarch64_bcaxqv8hi): Ditto.
9801 * config/aarch64/aarch64-simd.md:
9802 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
9803 (aarch64_crypto_sha512su0qv2di): Ditto.
9804 (aarch64_crypto_sha512su1qv2di): Ditto.
9805 (aarch64_eor3qv8hi): Ditto.
9806 (aarch64_rax1qv2di): Ditto.
9807 (aarch64_xarqv2di): Ditto.
9808 (aarch64_bcaxqv8hi): Ditto.
9809 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
9810 (vsha512h2q_u64): Ditto.
9811 (vsha512su0q_u64): Ditto.
9812 (vsha512su1q_u64): Ditto.
9813 (veor3q_u16): Ditto.
9814 (vrax1q_u64): Ditto.
9816 (vbcaxq_u16): Ditto.
9817 * config/arm/types.md (crypto_sha512): New type attribute.
9818 (crypto_sha3): Ditto.
9819 (doc/invoke.texi): Document new sha3 option.
9821 2018-01-10 Michael Collison <michael.collison@arm.com>
9823 * config/aarch64/aarch64-builtins.c:
9824 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
9825 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9826 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
9827 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
9828 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
9829 (AARCH64_ISA_SM4): New ISA flag.
9830 (TARGET_SM4): New feature flag for sm4.
9831 * config/aarch64/aarch64-simd-builtins.def
9832 (aarch64_sm3ss1qv4si): Ditto.
9833 (aarch64_sm3tt1aq4si): Ditto.
9834 (aarch64_sm3tt1bq4si): Ditto.
9835 (aarch64_sm3tt2aq4si): Ditto.
9836 (aarch64_sm3tt2bq4si): Ditto.
9837 (aarch64_sm3partw1qv4si): Ditto.
9838 (aarch64_sm3partw2qv4si): Ditto.
9839 (aarch64_sm4eqv4si): Ditto.
9840 (aarch64_sm4ekeyqv4si): Ditto.
9841 * config/aarch64/aarch64-simd.md:
9842 (aarch64_sm3ss1qv4si): Ditto.
9843 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
9844 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
9845 (aarch64_sm4eqv4si): Ditto.
9846 (aarch64_sm4ekeyqv4si): Ditto.
9847 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
9848 (sm3part_op): Ditto.
9849 (CRYPTO_SM3TT): Ditto.
9850 (CRYPTO_SM3PART): Ditto.
9851 (UNSPEC_SM3SS1): New unspec.
9852 (UNSPEC_SM3TT1A): Ditto.
9853 (UNSPEC_SM3TT1B): Ditto.
9854 (UNSPEC_SM3TT2A): Ditto.
9855 (UNSPEC_SM3TT2B): Ditto.
9856 (UNSPEC_SM3PARTW1): Ditto.
9857 (UNSPEC_SM3PARTW2): Ditto.
9858 (UNSPEC_SM4E): Ditto.
9859 (UNSPEC_SM4EKEY): Ditto.
9860 * config/aarch64/constraints.md (Ui2): New constraint.
9861 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
9862 * config/arm/types.md (crypto_sm3): New type attribute.
9863 (crypto_sm4): Ditto.
9864 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
9865 (vsm3tt1aq_u32): Ditto.
9866 (vsm3tt1bq_u32): Ditto.
9867 (vsm3tt2aq_u32): Ditto.
9868 (vsm3tt2bq_u32): Ditto.
9869 (vsm3partw1q_u32): Ditto.
9870 (vsm3partw2q_u32): Ditto.
9871 (vsm4eq_u32): Ditto.
9872 (vsm4ekeyq_u32): Ditto.
9873 (doc/invoke.texi): Document new sm4 option.
9875 2018-01-10 Michael Collison <michael.collison@arm.com>
9877 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
9878 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
9879 (AARCH64_FL_FOR_ARCH8_4): New.
9880 (AARCH64_FL_V8_4): New flag.
9881 (doc/invoke.texi): Document new armv8.4-a option.
9883 2018-01-10 Michael Collison <michael.collison@arm.com>
9885 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9886 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
9887 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
9888 * config/aarch64/aarch64-option-extension.def: Add
9889 AARCH64_OPT_EXTENSION of 'sha2'.
9890 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
9891 (crypto): Disable sha2 and aes if crypto disabled.
9892 (crypto): Enable aes and sha2 if enabled.
9893 (simd): Disable sha2 and aes if simd disabled.
9894 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
9896 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
9897 (TARGET_SHA2): New feature flag for sha2.
9898 (TARGET_AES): New feature flag for aes.
9899 * config/aarch64/aarch64-simd.md:
9900 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
9901 conditional on TARGET_AES.
9902 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
9903 (aarch64_crypto_sha1hsi): Make pattern conditional
9905 (aarch64_crypto_sha1hv4si): Ditto.
9906 (aarch64_be_crypto_sha1hv4si): Ditto.
9907 (aarch64_crypto_sha1su1v4si): Ditto.
9908 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
9909 (aarch64_crypto_sha1su0v4si): Ditto.
9910 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
9911 (aarch64_crypto_sha256su0v4si): Ditto.
9912 (aarch64_crypto_sha256su1v4si): Ditto.
9913 (doc/invoke.texi): Document new aes and sha2 options.
9915 2018-01-10 Martin Sebor <msebor@redhat.com>
9917 PR tree-optimization/83781
9918 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
9921 2018-01-11 Martin Sebor <msebor@gmail.com>
9922 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9924 PR tree-optimization/83501
9925 PR tree-optimization/81703
9927 * tree-ssa-strlen.c (get_string_cst): Rename...
9928 (get_string_len): ...to this. Handle global constants.
9929 (handle_char_store): Adjust.
9931 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
9932 Jim Wilson <jimw@sifive.com>
9934 * config/riscv/riscv-protos.h (riscv_output_return): New.
9935 * config/riscv/riscv.c (struct machine_function): New naked_p field.
9936 (riscv_attribute_table, riscv_output_return),
9937 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
9938 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
9939 (riscv_compute_frame_info): Only compute frame->mask if not a naked
9941 (riscv_expand_prologue): Add early return for naked function.
9942 (riscv_expand_epilogue): Likewise.
9943 (riscv_function_ok_for_sibcall): Return false for naked function.
9944 (riscv_set_current_function): New.
9945 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
9946 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
9947 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
9948 * doc/extend.texi (RISC-V Function Attributes): New.
9950 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
9952 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
9953 check for 128-bit long double before checking TCmode.
9954 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
9955 128-bit long doubles before checking TFmode or TCmode.
9956 (FLOAT128_IBM_P): Likewise.
9958 2018-01-10 Martin Sebor <msebor@redhat.com>
9960 PR tree-optimization/83671
9961 * builtins.c (c_strlen): Unconditionally return zero for the empty
9963 Use -Warray-bounds for warnings.
9964 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
9965 for non-constant array indices with COMPONENT_REF, arrays of
9966 arrays, and pointers to arrays.
9967 (gimple_fold_builtin_strlen): Determine and set length range for
9968 non-constant character arrays.
9970 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
9973 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
9976 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
9978 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
9980 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
9983 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
9984 VECTOR_MEM_ALTIVEC_OR_VSX_P.
9985 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
9986 indexed_or_indirect_operand predicate.
9987 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
9988 (*vsx_le_perm_load_v8hi): Likewise.
9989 (*vsx_le_perm_load_v16qi): Likewise.
9990 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
9991 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
9992 (*vsx_le_perm_store_v8hi): Likewise.
9993 (*vsx_le_perm_store_v16qi): Likewise.
9994 (eight unnamed splitters): Likewise.
9996 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
9998 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
9999 * config/rs6000/emmintrin.h: Likewise.
10000 * config/rs6000/mmintrin.h: Likewise.
10001 * config/rs6000/xmmintrin.h: Likewise.
10003 2018-01-10 David Malcolm <dmalcolm@redhat.com>
10006 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
10008 * tree.c (tree_nop_conversion): Return true for location wrapper
10010 (maybe_wrap_with_location): New function.
10011 (selftest::check_strip_nops): New function.
10012 (selftest::test_location_wrappers): New function.
10013 (selftest::tree_c_tests): Call it.
10014 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
10015 (maybe_wrap_with_location): New decl.
10016 (EXPR_LOCATION_WRAPPER_P): New macro.
10017 (location_wrapper_p): New inline function.
10018 (tree_strip_any_location_wrapper): New inline function.
10020 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
10023 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
10024 stack_realign_offset for the largest alignment of stack slot
10026 (ix86_find_max_used_stack_alignment): New function.
10027 (ix86_finalize_stack_frame_flags): Use it. Set
10028 max_used_stack_alignment if we don't realign stack.
10029 * config/i386/i386.h (machine_function): Add
10030 max_used_stack_alignment.
10032 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
10034 * config/arm/arm.opt (-mbranch-cost): New option.
10035 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
10038 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
10041 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
10042 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
10044 2018-01-10 Richard Biener <rguenther@suse.de>
10047 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
10048 early out so it also covers the case where we have a non-NULL
10051 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
10053 PR tree-optimization/83753
10054 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
10055 for non-strided grouped accesses if the number of elements is 1.
10057 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10060 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
10061 * i386.h (TARGET_USE_GATHER): Define.
10062 * x86-tune.def (X86_TUNE_USE_GATHER): New.
10064 2018-01-10 Martin Liska <mliska@suse.cz>
10067 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
10068 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
10070 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
10071 CLEANUP_NO_PARTITIONING is not set.
10073 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
10075 * doc/rtl.texi: Remove documentation of (const ...) wrappers
10076 for vectors, as a partial revert of r254296.
10077 * rtl.h (const_vec_p): Delete.
10078 (const_vec_duplicate_p): Don't test for vector CONSTs.
10079 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
10080 * expmed.c (make_tree): Likewise.
10083 * common.md (E, F): Use CONSTANT_P instead of checking for
10085 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
10086 checking for CONST_VECTOR.
10088 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10090 PR middle-end/83575
10091 * predict.c (force_edge_cold): Handle in more sane way edges
10092 with no prediction.
10094 2018-01-09 Carl Love <cel@us.ibm.com>
10096 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
10098 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
10099 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
10100 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
10101 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
10102 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
10103 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
10104 * config/rs6000/rs6000-protos.h: Add extern defition for
10105 rs6000_generate_float2_double_code.
10106 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
10108 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
10109 (float2_v2df): Add define_expand.
10111 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
10114 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
10115 op_mode in the force_to_mode call.
10117 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10119 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
10120 instead of checking each element individually.
10121 (aarch64_evpc_uzp): Likewise.
10122 (aarch64_evpc_zip): Likewise.
10123 (aarch64_evpc_ext): Likewise.
10124 (aarch64_evpc_rev): Likewise.
10125 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
10126 instead of checking each element individually. Return true without
10128 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
10129 whether all selected elements come from the same input, instead of
10130 checking each element individually. Remove calls to gen_rtx_REG,
10131 start_sequence and end_sequence and instead assert that no rtl is
10134 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10136 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
10137 order of HIGH and CONST checks.
10139 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10141 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
10142 if the destination isn't an SSA_NAME.
10144 2018-01-09 Richard Biener <rguenther@suse.de>
10146 PR tree-optimization/83668
10147 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
10149 (canonicalize_loop_form): ... here, renamed from ...
10150 (canonicalize_loop_closed_ssa_form): ... this and amended to
10151 swap successor edges for loop exit blocks to make us use
10152 the RPO order we need for initial schedule generation.
10154 2018-01-09 Joseph Myers <joseph@codesourcery.com>
10156 PR tree-optimization/64811
10157 * match.pd: When optimizing comparisons with Inf, avoid
10158 introducing or losing exceptions from comparisons with NaN.
10160 2018-01-09 Martin Liska <mliska@suse.cz>
10163 * asan.c (shadow_mem_size): Add gcc_assert.
10165 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
10167 Don't save registers in main().
10170 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
10171 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
10172 * config/avr/avr.c (avr_set_current_function): Don't error if
10173 naked, OS_task or OS_main are specified at the same time.
10174 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
10176 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
10178 * common/config/avr/avr-common.c (avr_option_optimization_table):
10179 Switch on -mmain-is-OS_task for optimizing compilations.
10181 2018-01-09 Richard Biener <rguenther@suse.de>
10183 PR tree-optimization/83572
10184 * graphite.c: Include cfganal.h.
10185 (graphite_transform_loops): Connect infinite loops to exit
10186 and remove fake edges at the end.
10188 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10190 * ipa-inline.c (edge_badness): Revert accidental checkin.
10192 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10195 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
10196 symbols; not inline clones.
10198 2018-01-09 Jakub Jelinek <jakub@redhat.com>
10201 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
10202 hard registers. Formatting fixes.
10204 PR preprocessor/83722
10205 * gcc.c (try_generate_repro): Pass
10206 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
10207 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
10210 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
10211 Kito Cheng <kito.cheng@gmail.com>
10213 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
10214 (riscv_leaf_function_p): Delete.
10215 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
10217 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10219 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
10221 (do_ifelse): New function.
10222 (do_isel): New function.
10223 (do_sub3): New function.
10224 (do_add3): New function.
10225 (do_load_mask_compare): New function.
10226 (do_overlap_load_compare): New function.
10227 (expand_compare_loop): New function.
10228 (expand_block_compare): Call expand_compare_loop() when appropriate.
10229 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
10230 option description.
10231 (-mblock-compare-inline-loop-limit): New option.
10233 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10236 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
10237 Reverse order of second and third operands in first alternative.
10238 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
10239 of first and second elements in UNSPEC_VPERMR vector.
10240 (altivec_expand_vec_perm_le): Likewise.
10242 2018-01-08 Jeff Law <law@redhat.com>
10244 PR rtl-optimizatin/81308
10245 * tree-switch-conversion.c (cfg_altered): New file scoped static.
10246 (process_switch): If group_case_labels makes a change, then set
10248 (pass_convert_switch::execute): If a switch is converted, then
10249 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
10251 PR rtl-optimization/81308
10252 * recog.c (split_all_insns): Conditionally cleanup the CFG after
10255 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
10257 PR target/83663 - Revert r255946
10258 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
10259 generation for cases where splatting a value is not useful.
10260 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
10261 across a vec_duplicate and a paradoxical subreg forming a vector
10262 mode to a vec_concat.
10264 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10266 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
10267 -march=armv8.3-a variants.
10268 * config/arm/t-multilib: Likewise.
10269 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
10271 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10273 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
10275 (cceq_ior_compare_complement): Give it a name so I can use it, and
10276 change boolean_or_operator predicate to boolean_operator so it can
10277 be used to generate a crand.
10278 (eqne): New code iterator.
10279 (bd/bd_neg): New code_attrs.
10280 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
10281 a single define_insn.
10282 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
10283 decrement (bdnzt/bdnzf/bdzt/bdzf).
10284 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
10285 with the new names of the branch decrement patterns, and added the
10286 names of the branch decrement conditional patterns.
10288 2018-01-08 Richard Biener <rguenther@suse.de>
10290 PR tree-optimization/83563
10291 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
10294 2018-01-08 Richard Biener <rguenther@suse.de>
10296 PR middle-end/83713
10297 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
10299 2018-01-08 Richard Biener <rguenther@suse.de>
10301 PR tree-optimization/83685
10302 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
10303 references to abnormals.
10305 2018-01-08 Richard Biener <rguenther@suse.de>
10308 * dwarf2out.c (output_indirect_strings): Handle empty
10309 skeleton_debug_str_hash.
10310 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
10312 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
10314 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
10315 (emit_store_direct): Likewise.
10316 (arc_trampoline_adjust_address): Likewise.
10317 (arc_asm_trampoline_template): New function.
10318 (arc_initialize_trampoline): Use asm_trampoline_template.
10319 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
10320 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
10321 * config/arc/arc.md (flush_icache): Delete pattern.
10323 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
10325 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
10326 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
10329 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
10332 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
10333 by not USED_FOR_TARGET.
10334 (make_pass_resolve_sw_modes): Likewise.
10336 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
10338 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
10341 2018-01-08 Richard Biener <rguenther@suse.de>
10343 PR middle-end/83580
10344 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
10346 2018-01-08 Richard Biener <rguenther@suse.de>
10348 PR middle-end/83517
10349 * match.pd ((t * 2) / 2) -> t): Add missing :c.
10351 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
10353 PR middle-end/81897
10354 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
10355 basic blocks with a small number of successors.
10356 (convert_control_dep_chain_into_preds): Improve handling of
10358 (dump_predicates): Split apart into...
10359 (dump_pred_chain): ...here...
10360 (dump_pred_info): ...and here.
10361 (can_one_predicate_be_invalidated_p): Add debugging printfs.
10362 (can_chain_union_be_invalidated_p): Improve check for invalidation
10364 (uninit_uses_cannot_happen): Avoid unnecessary if
10365 convert_control_dep_chain_into_preds yielded nothing.
10367 2018-01-06 Martin Sebor <msebor@redhat.com>
10369 PR tree-optimization/83640
10370 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
10371 subtracting negative offset from size.
10372 (builtin_access::overlap): Adjust offset bounds of the access to fall
10373 within the size of the object if possible.
10375 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
10377 PR rtl-optimization/83699
10378 * expmed.c (extract_bit_field_1): Restrict the vector usage of
10379 extract_bit_field_as_subreg to cases in which the extracted
10380 value is also a vector.
10382 * lra-constraints.c (process_alt_operands): Test for the equivalence
10383 substitutions when detecting a possible reload cycle.
10385 2018-01-06 Jakub Jelinek <jakub@redhat.com>
10388 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
10389 by default if flag_selective_schedling{,2}. Formatting fixes.
10391 PR rtl-optimization/83682
10392 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
10393 if it has non-VECTOR_MODE element mode.
10394 (vec_duplicate_p): Likewise.
10396 PR middle-end/83694
10397 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
10398 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
10400 2018-01-05 Jakub Jelinek <jakub@redhat.com>
10403 * config/i386/i386-builtin.def
10404 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
10405 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
10406 Require also OPTION_MASK_ISA_AVX512F in addition to
10407 OPTION_MASK_ISA_GFNI.
10408 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
10409 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
10410 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
10411 to OPTION_MASK_ISA_GFNI.
10412 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
10413 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
10414 OPTION_MASK_ISA_AVX512BW.
10415 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
10416 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
10417 addition to OPTION_MASK_ISA_GFNI.
10418 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
10419 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
10420 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
10421 to OPTION_MASK_ISA_GFNI.
10422 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
10423 a requirement for all ISAs rather than any of them with a few
10425 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
10427 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
10428 bitmasks to be enabled with 3 exceptions, instead of requiring any
10429 enabled ISA with lots of exceptions.
10430 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
10431 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
10432 Change avx512bw in isa attribute to avx512f.
10433 * config/i386/sgxintrin.h: Add license boilerplate.
10434 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
10435 to __AVX512F__ and __AVX512VL to __AVX512VL__.
10436 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
10437 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
10439 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
10440 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
10441 temporarily sse2 rather than sse if not enabled already.
10444 * config/i386/sse.md (VI248_VLBW): Rename to ...
10445 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
10446 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
10447 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
10448 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
10449 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
10450 mode iterator instead of VI248_VLBW.
10452 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
10454 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
10455 (record_modified): Skip clobbers; add debug output.
10456 (param_change_prob): Use sreal frequencies.
10458 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
10460 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
10461 punt for user-aligned variables.
10463 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
10465 * tree-chrec.c (chrec_contains_symbols): Return true for
10468 2018-01-05 Sudakshina Das <sudi.das@arm.com>
10471 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
10472 of (x|y) == x for BICS pattern.
10474 2018-01-05 Jakub Jelinek <jakub@redhat.com>
10476 PR tree-optimization/83605
10477 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
10478 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
10481 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
10483 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
10484 * config/epiphany/rtems.h: New file.
10486 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10487 Uros Bizjak <ubizjak@gmail.com>
10490 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
10491 QIreg_operand instead of register_operand predicate.
10492 * config/i386/i386.c (ix86_rop_should_change_byte_p,
10493 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
10494 comments instead of -fmitigate[-_]rop.
10496 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10499 * cgraphunit.c (symbol_table::compile): Switch to text_section
10500 before calling assembly_start debug hook.
10501 * run-rtl-passes.c (run_rtl_passes): Likewise.
10504 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10506 * tree-vrp.c (extract_range_from_binary_expr_1): Check
10507 range_int_cst_p rather than !symbolic_range_p before calling
10508 extract_range_from_multiplicative_op_1.
10510 2018-01-04 Jeff Law <law@redhat.com>
10512 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
10513 redundant test in assertion.
10515 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10517 * doc/rtl.texi: Document machine_mode wrapper classes.
10519 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10521 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
10522 using tree_to_uhwi.
10524 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10526 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
10527 the VEC_PERM_EXPR fold to fail.
10529 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10532 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
10533 to switched_sections.
10535 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10538 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
10539 test for d.testing.
10541 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
10544 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
10545 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
10547 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10550 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
10551 is BLKmode and bitpos not zero or mode change is needed.
10553 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10556 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
10559 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
10562 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
10563 instead of MULT rtx. Update all corresponding splitters.
10564 (*saddl_se): Ditto.
10565 (*ssub<modesuffix>): Ditto.
10566 (*ssubl_se): Ditto.
10567 (*cmp_sadd_di): Update split patterns.
10568 (*cmp_sadd_si): Ditto.
10569 (*cmp_sadd_sidi): Ditto.
10570 (*cmp_ssub_di): Ditto.
10571 (*cmp_ssub_si): Ditto.
10572 (*cmp_ssub_sidi): Ditto.
10573 * config/alpha/predicates.md (const23_operand): New predicate.
10574 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
10575 Look for ASHIFT, not MULT inner operand.
10576 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
10578 2018-01-04 Martin Liska <mliska@suse.cz>
10580 PR gcov-profile/83669
10581 * gcov.c (output_intermediate_file): Add version to intermediate
10583 * doc/gcov.texi: Document new field 'version' in intermediate
10584 file format. Fix location of '-k' option of gcov command.
10586 2018-01-04 Martin Liska <mliska@suse.cz>
10589 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
10591 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10593 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
10595 2018-01-03 Martin Sebor <msebor@redhat.com>
10597 PR tree-optimization/83655
10598 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
10599 checking calls with invalid arguments.
10601 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10603 * tree-vect-stmts.c (vect_get_store_rhs): New function.
10604 (vectorizable_mask_load_store): Delete.
10605 (vectorizable_call): Return false for masked loads and stores.
10606 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
10607 instead of gimple_assign_rhs1.
10608 (vectorizable_load): Handle IFN_MASK_LOAD.
10609 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
10611 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10613 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
10615 (vectorizable_mask_load_store): ...here.
10616 (vectorizable_load): ...and here.
10618 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10620 * tree-vect-stmts.c (vect_build_all_ones_mask)
10621 (vect_build_zero_merge_argument): New functions, split out from...
10622 (vectorizable_load): ...here.
10624 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10626 * tree-vect-stmts.c (vect_check_store_rhs): New function,
10628 (vectorizable_mask_load_store): ...here.
10629 (vectorizable_store): ...and here.
10631 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10633 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
10635 (vectorizable_mask_load_store): ...here.
10637 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10639 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
10640 (vect_model_store_cost): Take a vec_load_store_type instead of a
10642 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
10643 (vect_model_store_cost): Take a vec_load_store_type instead of a
10645 (vectorizable_mask_load_store): Update accordingly.
10646 (vectorizable_store): Likewise.
10647 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
10649 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10651 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
10652 IFN_MASK_LOAD calls here rather than...
10653 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
10655 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10656 Alan Hayward <alan.hayward@arm.com>
10657 David Sherwood <david.sherwood@arm.com>
10659 * expmed.c (extract_bit_field_1): For vector extracts,
10660 fall back to extract_bit_field_as_subreg if vec_extract
10663 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10664 Alan Hayward <alan.hayward@arm.com>
10665 David Sherwood <david.sherwood@arm.com>
10667 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
10668 they are variable or constant sized.
10669 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
10670 slots for constant-sized data.
10672 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10673 Alan Hayward <alan.hayward@arm.com>
10674 David Sherwood <david.sherwood@arm.com>
10676 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
10677 handling COND_EXPRs with boolean comparisons, try to find a better
10678 basis for the mask type than the boolean itself.
10680 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10682 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
10683 is calculated and how it can be overridden.
10684 * genmodes.c (max_bitsize_mode_any_mode): New variable.
10685 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
10687 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
10690 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10691 Alan Hayward <alan.hayward@arm.com>
10692 David Sherwood <david.sherwood@arm.com>
10694 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
10695 Remove the mode argument.
10696 (aarch64_simd_valid_immediate): Remove the mode and inverse
10698 * config/aarch64/iterators.md (bitsize): New iterator.
10699 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
10700 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
10701 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
10702 aarch64_simd_valid_immediate.
10703 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
10704 (aarch64_reg_or_bic_imm): Likewise.
10705 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
10706 with an insn_type enum and msl with a modifier_type enum.
10707 Replace element_width with a scalar_mode. Change the shift
10708 to unsigned int. Add constructors for scalar_float_mode and
10709 scalar_int_mode elements.
10710 (aarch64_vect_float_const_representable_p): Delete.
10711 (aarch64_can_const_movi_rtx_p)
10712 (aarch64_simd_scalar_immediate_valid_for_move)
10713 (aarch64_simd_make_constant): Update call to
10714 aarch64_simd_valid_immediate.
10715 (aarch64_advsimd_valid_immediate_hs): New function.
10716 (aarch64_advsimd_valid_immediate): Likewise.
10717 (aarch64_simd_valid_immediate): Remove mode and inverse
10718 arguments. Rewrite to use the above. Use const_vec_duplicate_p
10719 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
10720 and aarch64_float_const_representable_p on the result.
10721 (aarch64_output_simd_mov_immediate): Remove mode argument.
10722 Update call to aarch64_simd_valid_immediate and use of
10723 simd_immediate_info.
10724 (aarch64_output_scalar_simd_mov_immediate): Update call
10727 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10728 Alan Hayward <alan.hayward@arm.com>
10729 David Sherwood <david.sherwood@arm.com>
10731 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
10732 (mode_nunits): Likewise CONST_MODE_NUNITS.
10733 * machmode.def (ADJUST_NUNITS): Document.
10734 * genmodes.c (mode_data::need_nunits_adj): New field.
10735 (blank_mode): Update accordingly.
10736 (adj_nunits): New variable.
10737 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
10739 (emit_mode_size_inline): Set need_bytesize_adj for all modes
10740 listed in adj_nunits.
10741 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
10742 listed in adj_nunits. Don't emit case statements for such modes.
10743 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
10744 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
10745 nothing if adj_nunits is nonnull.
10746 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
10747 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
10748 (emit_mode_fbit): Update use of print_maybe_const_decl.
10749 (emit_move_size): Likewise. Treat the array as non-const
10751 (emit_mode_adjustments): Handle adj_nunits.
10753 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10755 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
10756 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
10757 (VECTOR_MODES): Use it.
10758 (make_vector_modes): Take the prefix as an argument.
10760 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10761 Alan Hayward <alan.hayward@arm.com>
10762 David Sherwood <david.sherwood@arm.com>
10764 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
10765 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
10766 for MODE_VECTOR_BOOL.
10767 * machmode.def (VECTOR_BOOL_MODE): Document.
10768 * genmodes.c (VECTOR_BOOL_MODE): New macro.
10769 (make_vector_bool_mode): New function.
10770 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
10772 * lto-streamer-in.c (lto_input_mode_table): Likewise.
10773 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
10775 * stor-layout.c (int_mode_for_mode): Likewise.
10776 * tree.c (build_vector_type_for_mode): Likewise.
10777 * varasm.c (output_constant_pool_2): Likewise.
10778 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
10779 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
10780 for MODE_VECTOR_BOOL.
10781 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
10782 of mode class checks.
10783 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
10784 instead of a list of mode class checks.
10785 (expand_vector_scalar_condition): Likewise.
10786 (type_for_widest_vector_mode): Handle BImode as an inner mode.
10788 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10789 Alan Hayward <alan.hayward@arm.com>
10790 David Sherwood <david.sherwood@arm.com>
10792 * machmode.h (mode_size): Change from unsigned short to
10794 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
10795 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
10796 or if measurement_type is not polynomial.
10797 (fixed_size_mode::includes_p): Check for constant-sized modes.
10798 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
10799 return a poly_uint16 rather than an unsigned short.
10800 (emit_mode_size): Change the type of mode_size from unsigned short
10801 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
10802 (emit_mode_adjustments): Cope with polynomial vector sizes.
10803 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10805 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10807 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
10808 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
10809 * caller-save.c (setup_save_areas): Likewise.
10810 (replace_reg_with_saved_mem): Likewise.
10811 * calls.c (emit_library_call_value_1): Likewise.
10812 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
10813 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
10814 (gen_lowpart_for_combine): Likewise.
10815 * convert.c (convert_to_integer_1): Likewise.
10816 * cse.c (equiv_constant, cse_insn): Likewise.
10817 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
10818 (cselib_subst_to_values): Likewise.
10819 * dce.c (word_dce_process_block): Likewise.
10820 * df-problems.c (df_word_lr_mark_ref): Likewise.
10821 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
10822 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
10823 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
10824 (rtl_for_decl_location): Likewise.
10825 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
10826 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
10827 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
10828 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
10829 (expand_expr_real_1): Likewise.
10830 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
10831 (pad_below): Likewise.
10832 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
10833 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
10834 * ira.c (get_subreg_tracking_sizes): Likewise.
10835 * ira-build.c (ira_create_allocno_objects): Likewise.
10836 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
10837 (ira_sort_regnos_for_alter_reg): Likewise.
10838 * ira-costs.c (record_operand_costs): Likewise.
10839 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
10840 (resolve_simple_move): Likewise.
10841 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
10842 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
10843 (lra_constraints): Likewise.
10844 (CONST_POOL_OK_P): Reject variable-sized modes.
10845 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
10846 (add_pseudo_to_slot, lra_spill): Likewise.
10847 * omp-low.c (omp_clause_aligned_alignment): Likewise.
10848 * optabs-query.c (get_best_extraction_insn): Likewise.
10849 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
10850 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
10851 (expand_mult_highpart, valid_multiword_target_p): Likewise.
10852 * recog.c (offsettable_address_addr_space_p): Likewise.
10853 * regcprop.c (maybe_mode_change): Likewise.
10854 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
10855 * regrename.c (build_def_use): Likewise.
10856 * regstat.c (dump_reg_info): Likewise.
10857 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
10858 (find_reloads, find_reloads_subreg_address): Likewise.
10859 * reload1.c (eliminate_regs_1): Likewise.
10860 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
10861 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
10862 (simplify_binary_operation_1, simplify_subreg): Likewise.
10863 * targhooks.c (default_function_arg_padding): Likewise.
10864 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
10865 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
10866 (verify_gimple_assign_ternary): Likewise.
10867 * tree-inline.c (estimate_move_cost): Likewise.
10868 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10869 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
10870 (get_address_cost_ainc): Likewise.
10871 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
10872 (vect_supportable_dr_alignment): Likewise.
10873 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
10874 (vectorizable_reduction): Likewise.
10875 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
10876 (vectorizable_operation, vectorizable_load): Likewise.
10877 * tree.c (build_same_sized_truth_vector_type): Likewise.
10878 * valtrack.c (cleanup_auto_inc_dec): Likewise.
10879 * var-tracking.c (emit_note_insn_var_location): Likewise.
10880 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
10881 (ADDR_VEC_ALIGN): Likewise.
10883 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10884 Alan Hayward <alan.hayward@arm.com>
10885 David Sherwood <david.sherwood@arm.com>
10887 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
10889 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
10890 or if measurement_type is polynomial.
10891 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
10892 * combine.c (make_extraction): Likewise.
10893 * dse.c (find_shift_sequence): Likewise.
10894 * dwarf2out.c (mem_loc_descriptor): Likewise.
10895 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
10896 (extract_bit_field, extract_low_bits): Likewise.
10897 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
10898 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
10899 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
10900 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
10901 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
10902 * reload.c (find_reloads): Likewise.
10903 * reload1.c (alter_reg): Likewise.
10904 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
10905 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
10906 * tree-if-conv.c (predicate_mem_writes): Likewise.
10907 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
10908 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
10909 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
10910 * valtrack.c (dead_debug_insert_temp): Likewise.
10911 * varasm.c (mergeable_constant_section): Likewise.
10912 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
10914 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10915 Alan Hayward <alan.hayward@arm.com>
10916 David Sherwood <david.sherwood@arm.com>
10918 * expr.c (expand_assignment): Cope with polynomial mode sizes
10919 when assigning to a CONCAT.
10921 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10922 Alan Hayward <alan.hayward@arm.com>
10923 David Sherwood <david.sherwood@arm.com>
10925 * machmode.h (mode_precision): Change from unsigned short to
10927 (mode_to_precision): Return a poly_uint16 rather than an unsigned
10929 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
10930 or if measurement_type is not polynomial.
10931 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
10932 in which the mode is already known to be a scalar_int_mode.
10933 * genmodes.c (emit_mode_precision): Change the type of mode_precision
10934 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
10936 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10937 for GET_MODE_PRECISION.
10938 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10939 for GET_MODE_PRECISION.
10940 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
10942 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
10943 (expand_field_assignment, make_extraction): Likewise.
10944 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
10945 (get_last_value): Likewise.
10946 * convert.c (convert_to_integer_1): Likewise.
10947 * cse.c (cse_insn): Likewise.
10948 * expr.c (expand_expr_real_1): Likewise.
10949 * lra-constraints.c (simplify_operand_subreg): Likewise.
10950 * optabs-query.c (can_atomic_load_p): Likewise.
10951 * optabs.c (expand_atomic_load): Likewise.
10952 (expand_atomic_store): Likewise.
10953 * ree.c (combine_reaching_defs): Likewise.
10954 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
10955 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
10956 * tree.h (type_has_mode_precision_p): Likewise.
10957 * ubsan.c (instrument_si_overflow): Likewise.
10959 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10960 Alan Hayward <alan.hayward@arm.com>
10961 David Sherwood <david.sherwood@arm.com>
10963 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
10964 polynomial numbers of units.
10965 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
10966 (valid_vector_subparts_p): New function.
10967 (build_vector_type): Remove temporary shim and take the number
10968 of units as a poly_uint64 rather than an int.
10969 (build_opaque_vector_type): Take the number of units as a
10970 poly_uint64 rather than an int.
10971 * tree.c (build_vector_from_ctor): Handle polynomial
10972 TYPE_VECTOR_SUBPARTS.
10973 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
10974 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
10975 (build_vector_from_val): If the number of units is variable,
10976 use build_vec_duplicate_cst for constant operands and
10977 VEC_DUPLICATE_EXPR otherwise.
10978 (make_vector_type): Remove temporary is_constant ().
10979 (build_vector_type, build_opaque_vector_type): Take the number of
10980 units as a poly_uint64 rather than an int.
10981 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
10983 * cfgexpand.c (expand_debug_expr): Likewise.
10984 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
10985 (store_constructor, expand_expr_real_1): Likewise.
10986 (const_scalar_mask_from_tree): Likewise.
10987 * fold-const-call.c (fold_const_reduction): Likewise.
10988 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
10989 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
10990 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
10991 (fold_relational_const): Likewise.
10992 (native_interpret_vector): Likewise. Change the size from an
10993 int to an unsigned int.
10994 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
10995 TYPE_VECTOR_SUBPARTS.
10996 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
10997 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
10998 duplicating a non-constant operand into a variable-length vector.
10999 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
11000 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
11001 * ipa-icf.c (sem_variable::equals): Likewise.
11002 * match.pd: Likewise.
11003 * omp-simd-clone.c (simd_clone_subparts): Likewise.
11004 * print-tree.c (print_node): Likewise.
11005 * stor-layout.c (layout_type): Likewise.
11006 * targhooks.c (default_builtin_vectorization_cost): Likewise.
11007 * tree-cfg.c (verify_gimple_comparison): Likewise.
11008 (verify_gimple_assign_binary): Likewise.
11009 (verify_gimple_assign_ternary): Likewise.
11010 (verify_gimple_assign_single): Likewise.
11011 * tree-pretty-print.c (dump_generic_node): Likewise.
11012 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11013 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
11014 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
11015 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
11016 (vect_shift_permute_load_chain): Likewise.
11017 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
11018 (expand_vector_condition, optimize_vector_constructor): Likewise.
11019 (lower_vec_perm, get_compute_type): Likewise.
11020 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11021 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
11022 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
11023 (vect_recog_mask_conversion_pattern): Likewise.
11024 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
11025 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
11026 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11027 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
11028 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
11029 (vectorizable_shift, vectorizable_operation, vectorizable_store)
11030 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
11031 (supportable_widening_operation): Likewise.
11032 (supportable_narrowing_operation): Likewise.
11033 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
11035 * varasm.c (output_constant): Likewise.
11037 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11038 Alan Hayward <alan.hayward@arm.com>
11039 David Sherwood <david.sherwood@arm.com>
11041 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
11042 so that both the length == 3 and length != 3 cases set up their
11043 own permute vectors. Add comments explaining why we know the
11044 number of elements is constant.
11045 (vect_permute_load_chain): Likewise.
11047 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11048 Alan Hayward <alan.hayward@arm.com>
11049 David Sherwood <david.sherwood@arm.com>
11051 * machmode.h (mode_nunits): Change from unsigned char to
11053 (ONLY_FIXED_SIZE_MODES): New macro.
11054 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
11055 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
11056 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
11058 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
11059 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
11060 or if measurement_type is not polynomial.
11061 * genmodes.c (ZERO_COEFFS): New macro.
11062 (emit_mode_nunits_inline): Make mode_nunits_inline return a
11064 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
11065 Use ZERO_COEFFS when emitting initializers.
11066 * data-streamer.h (bp_pack_poly_value): New function.
11067 (bp_unpack_poly_value): Likewise.
11068 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11069 for GET_MODE_NUNITS.
11070 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11071 for GET_MODE_NUNITS.
11072 * tree.c (make_vector_type): Remove temporary shim and make
11073 the real function take the number of units as a poly_uint64
11074 rather than an int.
11075 (build_vector_type_for_mode): Handle polynomial nunits.
11076 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
11077 * emit-rtl.c (const_vec_series_p_1): Likewise.
11078 (gen_rtx_CONST_VECTOR): Likewise.
11079 * fold-const.c (test_vec_duplicate_folding): Likewise.
11080 * genrecog.c (validate_pattern): Likewise.
11081 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
11082 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
11083 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
11084 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
11085 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
11086 * rtlanal.c (subreg_get_info): Likewise.
11087 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11088 (vect_grouped_load_supported): Likewise.
11089 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
11090 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
11091 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
11092 (simplify_const_unary_operation, simplify_binary_operation_1)
11093 (simplify_const_binary_operation, simplify_ternary_operation)
11094 (test_vector_ops_duplicate, test_vector_ops): Likewise.
11095 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
11096 instead of CONST_VECTOR_NUNITS.
11097 * varasm.c (output_constant_pool_2): Likewise.
11098 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
11099 explicit-encoded elements in the XVEC for variable-length vectors.
11101 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11103 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
11105 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11106 Alan Hayward <alan.hayward@arm.com>
11107 David Sherwood <david.sherwood@arm.com>
11109 * coretypes.h (fixed_size_mode): Declare.
11110 (fixed_size_mode_pod): New typedef.
11111 * builtins.h (target_builtins::x_apply_args_mode)
11112 (target_builtins::x_apply_result_mode): Change type to
11113 fixed_size_mode_pod.
11114 * builtins.c (apply_args_size, apply_result_size, result_vector)
11115 (expand_builtin_apply_args_1, expand_builtin_apply)
11116 (expand_builtin_return): Update accordingly.
11118 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11120 * cse.c (hash_rtx_cb): Hash only the encoded elements.
11121 * cselib.c (cselib_hash_rtx): Likewise.
11122 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
11123 CONST_VECTOR encoding.
11125 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11126 Jeff Law <law@redhat.com>
11129 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
11130 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
11131 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
11132 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
11135 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
11136 explicitly probe *sp in a noreturn function if there were any callee
11137 register saves or frame pointer is needed.
11139 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11142 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
11143 BLKmode for ternary, binary or unary expressions.
11146 * var-tracking.c (delete_vta_debug_insn): New inline function.
11147 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
11148 insns from get_insns () to NULL instead of each bb separately.
11149 Use delete_vta_debug_insn. No longer static.
11150 (vt_debug_insns_local, variable_tracking_main_1): Adjust
11151 delete_vta_debug_insns callers.
11152 * rtl.h (delete_vta_debug_insns): Declare.
11153 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
11154 instead of variable_tracking_main.
11156 2018-01-03 Martin Sebor <msebor@redhat.com>
11158 PR tree-optimization/83603
11159 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
11160 arguments past the endof the argument list in functions declared
11161 without a prototype.
11162 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
11163 Avoid checking when arguments are null.
11165 2018-01-03 Martin Sebor <msebor@redhat.com>
11168 * doc/extend.texi (attribute const): Fix a typo.
11169 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
11170 issuing -Wsuggest-attribute for void functions.
11172 2018-01-03 Martin Sebor <msebor@redhat.com>
11174 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
11175 offset_int::from instead of wide_int::to_shwi.
11176 (maybe_diag_overlap): Remove assertion.
11177 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
11178 * gimple-ssa-sprintf.c (format_directive): Same.
11179 (parse_directive): Same.
11180 (sprintf_dom_walker::compute_format_length): Same.
11181 (try_substitute_return_value): Same.
11183 2018-01-03 Jeff Law <law@redhat.com>
11185 PR middle-end/83654
11186 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
11187 non-constant residual for zero at runtime and avoid probing in
11188 that case. Reorganize code for trailing problem to mirror handling
11191 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11193 PR tree-optimization/83501
11194 * tree-ssa-strlen.c (get_string_cst): New.
11195 (handle_char_store): Call get_string_cst.
11197 2018-01-03 Martin Liska <mliska@suse.cz>
11199 PR tree-optimization/83593
11200 * tree-ssa-strlen.c: Include tree-cfg.h.
11201 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
11202 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
11203 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
11205 (strlen_dom_walker::before_dom_children): Call
11206 gimple_purge_dead_eh_edges. Dump tranformation with details
11208 (strlen_dom_walker::before_dom_children): Update call by adding
11209 new argument cleanup_eh.
11210 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
11212 2018-01-03 Martin Liska <mliska@suse.cz>
11215 * cif-code.def (VARIADIC_THUNK): New enum value.
11216 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
11219 2018-01-03 Jan Beulich <jbeulich@suse.com>
11221 * sse.md (mov<mode>_internal): Tighten condition for when to use
11222 vmovdqu<ssescalarsize> for TI and OI modes.
11224 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11226 Update copyright years.
11228 2018-01-03 Martin Liska <mliska@suse.cz>
11231 * ipa-visibility.c (function_and_variable_visibility): Skip
11232 functions with noipa attribure.
11234 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11236 * gcc.c (process_command): Update copyright notice dates.
11237 * gcov-dump.c (print_version): Ditto.
11238 * gcov.c (print_version): Ditto.
11239 * gcov-tool.c (print_version): Ditto.
11240 * gengtype.c (create_file): Ditto.
11241 * doc/cpp.texi: Bump @copying's copyright year.
11242 * doc/cppinternals.texi: Ditto.
11243 * doc/gcc.texi: Ditto.
11244 * doc/gccint.texi: Ditto.
11245 * doc/gcov.texi: Ditto.
11246 * doc/install.texi: Ditto.
11247 * doc/invoke.texi: Ditto.
11249 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11251 * vector-builder.h (vector_builder::m_full_nelts): Change from
11252 unsigned int to poly_uint64.
11253 (vector_builder::full_nelts): Update prototype accordingly.
11254 (vector_builder::new_vector): Likewise.
11255 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
11256 (vector_builder::operator ==): Likewise.
11257 (vector_builder::finalize): Likewise.
11258 * int-vector-builder.h (int_vector_builder::int_vector_builder):
11259 Take the number of elements as a poly_uint64 rather than an
11261 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
11262 from unsigned int to poly_uint64.
11263 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
11264 (vec_perm_indices::new_vector): Likewise.
11265 (vec_perm_indices::length): Likewise.
11266 (vec_perm_indices::nelts_per_input): Likewise.
11267 (vec_perm_indices::input_nelts): Likewise.
11268 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
11269 number of elements per input as a poly_uint64 rather than an
11270 unsigned int. Use the original encoding for variable-length
11271 vectors, rather than clamping each individual element.
11272 For the second and subsequent elements in each pattern,
11273 clamp the step and base before clamping their sum.
11274 (vec_perm_indices::series_p): Handle polynomial element counts.
11275 (vec_perm_indices::all_in_range_p): Likewise.
11276 (vec_perm_indices_to_tree): Likewise.
11277 (vec_perm_indices_to_rtx): Likewise.
11278 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
11279 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
11280 (tree_vector_builder::new_binary_operation): Handle polynomial
11281 element counts. Return false if we need to know the number
11282 of elements at compile time.
11283 * fold-const.c (fold_vec_perm): Punt if the number of elements
11284 isn't known at compile time.
11286 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11288 * vec-perm-indices.h (vec_perm_builder): Change element type
11289 from HOST_WIDE_INT to poly_int64.
11290 (vec_perm_indices::element_type): Update accordingly.
11291 (vec_perm_indices::clamp): Handle polynomial element_types.
11292 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
11293 (vec_perm_indices::all_in_range_p): Likewise.
11294 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
11296 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
11297 polynomial vec_perm_indices element types.
11298 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
11299 * fold-const.c (fold_vec_perm): Likewise.
11300 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
11301 * tree-vect-generic.c (lower_vec_perm): Likewise.
11302 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
11303 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
11304 element type to HOST_WIDE_INT.
11306 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11307 Alan Hayward <alan.hayward@arm.com>
11308 David Sherwood <david.sherwood@arm.com>
11310 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
11311 rather than an int. Use plus_constant.
11312 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
11313 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
11315 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11316 Alan Hayward <alan.hayward@arm.com>
11317 David Sherwood <david.sherwood@arm.com>
11319 * calls.c (emit_call_1, expand_call): Change struct_value_size from
11320 a HOST_WIDE_INT to a poly_int64.
11322 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11323 Alan Hayward <alan.hayward@arm.com>
11324 David Sherwood <david.sherwood@arm.com>
11326 * calls.c (load_register_parameters): Cope with polynomial
11327 mode sizes. Require a constant size for BLKmode parameters
11328 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
11329 forces a parameter to be padded at the lsb end in order to
11330 fill a complete number of words, require the parameter size
11331 to be ordered wrt UNITS_PER_WORD.
11333 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11334 Alan Hayward <alan.hayward@arm.com>
11335 David Sherwood <david.sherwood@arm.com>
11337 * reload1.c (spill_stack_slot_width): Change element type
11338 from unsigned int to poly_uint64_pod.
11339 (alter_reg): Treat mode sizes as polynomial.
11341 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11342 Alan Hayward <alan.hayward@arm.com>
11343 David Sherwood <david.sherwood@arm.com>
11345 * reload.c (complex_word_subreg_p): New function.
11346 (reload_inner_reg_of_subreg, push_reload): Use it.
11348 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11349 Alan Hayward <alan.hayward@arm.com>
11350 David Sherwood <david.sherwood@arm.com>
11352 * lra-constraints.c (process_alt_operands): Reject matched
11353 operands whose sizes aren't ordered.
11354 (match_reload): Refer to this check here.
11356 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11357 Alan Hayward <alan.hayward@arm.com>
11358 David Sherwood <david.sherwood@arm.com>
11360 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
11361 that the mode size is in the set {1, 2, 4, 8, 16}.
11363 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11364 Alan Hayward <alan.hayward@arm.com>
11365 David Sherwood <david.sherwood@arm.com>
11367 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
11368 Use plus_constant instead of gen_rtx_PLUS.
11370 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11371 Alan Hayward <alan.hayward@arm.com>
11372 David Sherwood <david.sherwood@arm.com>
11374 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
11375 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
11376 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
11377 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
11378 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
11379 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
11380 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
11381 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
11382 * config/i386/i386.c (ix86_push_rounding): ...this new function.
11383 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
11385 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
11386 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
11387 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
11388 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
11389 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
11390 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
11391 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
11392 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
11393 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
11394 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
11396 * expr.c (emit_move_resolve_push): Treat the input and result
11397 of PUSH_ROUNDING as a poly_int64.
11398 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
11399 (emit_push_insn): Likewise.
11400 * lra-eliminations.c (mark_not_eliminable): Likewise.
11401 * recog.c (push_operand): Likewise.
11402 * reload1.c (elimination_effects): Likewise.
11403 * rtlanal.c (nonzero_bits1): Likewise.
11404 * calls.c (store_one_arg): Likewise. Require the padding to be
11405 known at compile time.
11407 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11408 Alan Hayward <alan.hayward@arm.com>
11409 David Sherwood <david.sherwood@arm.com>
11411 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
11412 Use plus_constant instead of gen_rtx_PLUS.
11414 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11415 Alan Hayward <alan.hayward@arm.com>
11416 David Sherwood <david.sherwood@arm.com>
11418 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
11419 rather than an int.
11421 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11422 Alan Hayward <alan.hayward@arm.com>
11423 David Sherwood <david.sherwood@arm.com>
11425 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
11426 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
11427 via stack temporaries. Treat the mode size as polynomial too.
11429 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11430 Alan Hayward <alan.hayward@arm.com>
11431 David Sherwood <david.sherwood@arm.com>
11433 * expr.c (expand_expr_real_2): When handling conversions involving
11434 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
11435 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
11436 as a poly_uint64 too.
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 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
11444 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11445 Alan Hayward <alan.hayward@arm.com>
11446 David Sherwood <david.sherwood@arm.com>
11448 * combine.c (can_change_dest_mode): Handle polynomial
11449 REGMODE_NATURAL_SIZE.
11450 * expmed.c (store_bit_field_1): Likewise.
11451 * expr.c (store_constructor): Likewise.
11452 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
11453 and polynomial REGMODE_NATURAL_SIZE.
11454 (gen_lowpart_common): Likewise.
11455 * reginfo.c (record_subregs_of_mode): Likewise.
11456 * rtlanal.c (read_modify_subreg_p): Likewise.
11458 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11459 Alan Hayward <alan.hayward@arm.com>
11460 David Sherwood <david.sherwood@arm.com>
11462 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
11463 numbers of 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 * match.pd: Cope with polynomial numbers of vector elements.
11471 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11472 Alan Hayward <alan.hayward@arm.com>
11473 David Sherwood <david.sherwood@arm.com>
11475 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
11476 in a POINTER_PLUS_EXPR.
11478 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11479 Alan Hayward <alan.hayward@arm.com>
11480 David Sherwood <david.sherwood@arm.com>
11482 * omp-simd-clone.c (simd_clone_subparts): New function.
11483 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
11484 (ipa_simd_modify_function_body): Likewise.
11486 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11487 Alan Hayward <alan.hayward@arm.com>
11488 David Sherwood <david.sherwood@arm.com>
11490 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
11491 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
11492 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
11493 (expand_vector_condition, vector_element): Likewise.
11494 (subparts_gt): New function.
11495 (get_compute_type): Use subparts_gt.
11496 (count_type_subparts): Delete.
11497 (expand_vector_operations_1): Use subparts_gt instead of
11498 count_type_subparts.
11500 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11501 Alan Hayward <alan.hayward@arm.com>
11502 David Sherwood <david.sherwood@arm.com>
11504 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
11505 (vect_compile_time_alias): ...this new function. Do the calculation
11506 on poly_ints rather than trees.
11507 (vect_prune_runtime_alias_test_list): Update call accordingly.
11509 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11510 Alan Hayward <alan.hayward@arm.com>
11511 David Sherwood <david.sherwood@arm.com>
11513 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
11515 (vect_schedule_slp_instance): Likewise.
11517 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11518 Alan Hayward <alan.hayward@arm.com>
11519 David Sherwood <david.sherwood@arm.com>
11521 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
11522 constant and extern definitions for variable-length vectors.
11523 (vect_get_constant_vectors): Note that the number of units
11524 is known to be constant.
11526 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11527 Alan Hayward <alan.hayward@arm.com>
11528 David Sherwood <david.sherwood@arm.com>
11530 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
11531 of units as polynomial. Choose between WIDE and NARROW based
11534 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11535 Alan Hayward <alan.hayward@arm.com>
11536 David Sherwood <david.sherwood@arm.com>
11538 * tree-vect-stmts.c (simd_clone_subparts): New function.
11539 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
11541 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11542 Alan Hayward <alan.hayward@arm.com>
11543 David Sherwood <david.sherwood@arm.com>
11545 * tree-vect-stmts.c (vectorizable_call): Treat the number of
11546 vectors as polynomial. Use build_index_vector for
11547 IFN_GOMP_SIMD_LANE.
11549 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11550 Alan Hayward <alan.hayward@arm.com>
11551 David Sherwood <david.sherwood@arm.com>
11553 * tree-vect-stmts.c (get_load_store_type): Treat the number of
11554 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
11555 for variable-length vectors.
11556 (vectorizable_mask_load_store): Treat the number of units as
11557 polynomial, asserting that it is constant if the condition has
11558 already been enforced.
11559 (vectorizable_store, vectorizable_load): Likewise.
11561 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11562 Alan Hayward <alan.hayward@arm.com>
11563 David Sherwood <david.sherwood@arm.com>
11565 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
11566 of units as polynomial. Punt if we can't tell at compile time
11567 which vector contains the final result.
11569 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11570 Alan Hayward <alan.hayward@arm.com>
11571 David Sherwood <david.sherwood@arm.com>
11573 * tree-vect-loop.c (vectorizable_induction): Treat the number
11574 of units as polynomial. Punt on SLP inductions. Use an integer
11575 VEC_SERIES_EXPR for variable-length integer reductions. Use a
11576 cast of such a series for variable-length floating-point
11579 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11580 Alan Hayward <alan.hayward@arm.com>
11581 David Sherwood <david.sherwood@arm.com>
11583 * tree.h (build_index_vector): Declare.
11584 * tree.c (build_index_vector): New function.
11585 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
11586 of units as polynomial, forcibly converting it to a constant if
11587 vectorizable_reduction has already enforced the condition.
11588 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
11589 to create a {1,2,3,...} vector.
11590 (vectorizable_reduction): Treat the number of units as polynomial.
11591 Choose vectype_in based on the largest scalar element size rather
11592 than the smallest number of units. Enforce the restrictions
11595 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11596 Alan Hayward <alan.hayward@arm.com>
11597 David Sherwood <david.sherwood@arm.com>
11599 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
11600 number of units as polynomial.
11602 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11603 Alan Hayward <alan.hayward@arm.com>
11604 David Sherwood <david.sherwood@arm.com>
11606 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
11607 * target.def (autovectorize_vector_sizes): Return the vector sizes
11608 by pointer, using vector_sizes rather than a bitmask.
11609 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
11610 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
11611 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
11613 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
11614 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
11615 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
11616 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
11617 * omp-general.c (omp_max_vf): Likewise.
11618 * omp-low.c (omp_clause_aligned_alignment): Likewise.
11619 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
11620 * tree-vect-loop.c (vect_analyze_loop): Likewise.
11621 * tree-vect-slp.c (vect_slp_bb): Likewise.
11622 * doc/tm.texi: Regenerate.
11623 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
11625 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
11626 the vector size as a poly_uint64 rather than an unsigned int.
11627 (current_vector_size): Change from an unsigned int to a poly_uint64.
11628 (get_vectype_for_scalar_type): Update accordingly.
11629 * tree.h (build_truth_vector_type): Take the size and number of
11630 units as a poly_uint64 rather than an unsigned int.
11631 (build_vector_type): Add a temporary overload that takes
11632 the number of units as a poly_uint64 rather than an unsigned int.
11633 * tree.c (make_vector_type): Likewise.
11634 (build_truth_vector_type): Take the number of units as a poly_uint64
11635 rather than an unsigned int.
11637 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11638 Alan Hayward <alan.hayward@arm.com>
11639 David Sherwood <david.sherwood@arm.com>
11641 * target.def (get_mask_mode): Take the number of units and length
11642 as poly_uint64s rather than unsigned ints.
11643 * targhooks.h (default_get_mask_mode): Update accordingly.
11644 * targhooks.c (default_get_mask_mode): Likewise.
11645 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
11646 * doc/tm.texi: Regenerate.
11648 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11649 Alan Hayward <alan.hayward@arm.com>
11650 David Sherwood <david.sherwood@arm.com>
11652 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
11653 * omp-general.c (omp_max_vf): Likewise.
11654 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
11655 (expand_omp_simd): Handle polynomial safelen.
11656 * omp-low.c (omplow_simd_context): Add a default constructor.
11657 (omplow_simd_context::max_vf): Change from int to poly_uint64.
11658 (lower_rec_simd_input_clauses): Update accordingly.
11659 (lower_rec_input_clauses): Likewise.
11661 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11662 Alan Hayward <alan.hayward@arm.com>
11663 David Sherwood <david.sherwood@arm.com>
11665 * tree-vectorizer.h (vect_nunits_for_cost): New function.
11666 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
11667 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
11668 (vect_analyze_slp_cost): Likewise.
11669 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
11670 (vect_model_load_cost): Likewise.
11672 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11673 Alan Hayward <alan.hayward@arm.com>
11674 David Sherwood <david.sherwood@arm.com>
11676 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
11677 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
11678 from an unsigned int * to a poly_uint64_pod *.
11679 (calculate_unrolling_factor): New function.
11680 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
11682 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11683 Alan Hayward <alan.hayward@arm.com>
11684 David Sherwood <david.sherwood@arm.com>
11686 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
11687 from an unsigned int to a poly_uint64.
11688 (_loop_vec_info::slp_unrolling_factor): Likewise.
11689 (_loop_vec_info::vectorization_factor): Change from an int
11691 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
11692 (vect_get_num_vectors): New function.
11693 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
11694 (vect_get_num_copies): Use vect_get_num_vectors.
11695 (vect_analyze_data_ref_dependences): Change max_vf from an int *
11696 to an unsigned int *.
11697 (vect_analyze_data_refs): Change min_vf from an int * to a
11699 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
11700 than an unsigned HOST_WIDE_INT.
11701 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
11702 (vect_analyze_data_ref_dependence): Change max_vf from an int *
11703 to an unsigned int *.
11704 (vect_analyze_data_ref_dependences): Likewise.
11705 (vect_compute_data_ref_alignment): Handle polynomial vf.
11706 (vect_enhance_data_refs_alignment): Likewise.
11707 (vect_prune_runtime_alias_test_list): Likewise.
11708 (vect_shift_permute_load_chain): Likewise.
11709 (vect_supportable_dr_alignment): Likewise.
11710 (dependence_distance_ge_vf): Take the vectorization factor as a
11711 poly_uint64 rather than an unsigned HOST_WIDE_INT.
11712 (vect_analyze_data_refs): Change min_vf from an int * to a
11714 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
11715 vfm1 as a poly_uint64 rather than an int. Make the same change
11716 for the returned bound_scalar.
11717 (vect_gen_vector_loop_niters): Handle polynomial vf.
11718 (vect_do_peeling): Likewise. Update call to
11719 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
11720 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
11722 * tree-vect-loop.c (vect_determine_vectorization_factor)
11723 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
11724 (vect_get_known_peeling_cost): Likewise.
11725 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
11726 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
11727 (vect_transform_loop): Likewise. Use the lowest possible VF when
11728 updating the upper bounds of the loop.
11729 (vect_min_worthwhile_factor): Make static. Return an unsigned int
11730 rather than an int.
11731 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
11732 polynomial unroll factors.
11733 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
11734 (vect_make_slp_decision): Likewise.
11735 (vect_supported_load_permutation_p): Likewise, and polynomial
11737 (vect_analyze_slp_cost): Handle polynomial vf.
11738 (vect_slp_analyze_node_operations): Likewise.
11739 (vect_slp_analyze_bb_1): Likewise.
11740 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
11741 than an unsigned HOST_WIDE_INT.
11742 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
11743 (vectorizable_load): Handle polynomial vf.
11744 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
11746 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
11748 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11749 Alan Hayward <alan.hayward@arm.com>
11750 David Sherwood <david.sherwood@arm.com>
11752 * match.pd: Handle bit operations involving three constants
11753 and try to fold one pair.
11755 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11757 * tree-vect-loop-manip.c: Include gimple-fold.h.
11758 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
11759 niters_maybe_zero parameters. Handle other cases besides a step of 1.
11760 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
11761 Add a path that uses a step of VF instead of 1, but disable it
11763 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
11764 and niters_no_overflow parameters. Update calls to
11765 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
11766 Create a new SSA name if the latter choses to use a ste other
11767 than zero, and return it via niters_vector_mult_vf_var.
11768 * tree-vect-loop.c (vect_transform_loop): Update calls to
11769 vect_do_peeling, vect_gen_vector_loop_niters and
11770 slpeel_make_loop_iterate_ntimes.
11771 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
11772 (vect_gen_vector_loop_niters): Update declarations after above changes.
11774 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
11776 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
11777 128-bit round to integer instructions.
11778 (ceil<mode>2): Likewise.
11779 (btrunc<mode>2): Likewise.
11780 (round<mode>2): Likewise.
11782 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11784 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
11785 unaligned VSX load/store on P8/P9.
11786 (expand_block_clear): Allow the use of unaligned VSX
11787 load/store on P8/P9.
11789 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11791 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
11793 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
11794 swap associated with both a load and a store.
11796 2018-01-02 Andrew Waterman <andrew@sifive.com>
11798 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
11799 * config/riscv/riscv.md (clear_cache): Use it.
11801 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
11803 * web.c: Remove out-of-date comment.
11805 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11807 * expr.c (fixup_args_size_notes): Check that any existing
11808 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
11809 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
11810 (emit_single_push_insn): ...here.
11812 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11814 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
11815 (const_vector_encoded_nelts): New function.
11816 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
11817 (const_vector_int_elt, const_vector_elt): Declare.
11818 * emit-rtl.c (const_vector_int_elt_1): New function.
11819 (const_vector_elt): Likewise.
11820 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
11821 of CONST_VECTOR_ELT.
11823 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11825 * expr.c: Include rtx-vector-builder.h.
11826 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
11827 directly on the tree encoding.
11828 (const_vector_from_tree): Likewise.
11829 * optabs.c: Include rtx-vector-builder.h.
11830 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
11831 sequence of "u" values.
11832 * vec-perm-indices.c: Include rtx-vector-builder.h.
11833 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
11834 directly on the vec_perm_indices encoding.
11836 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11838 * doc/rtl.texi (const_vector): Describe new encoding scheme.
11839 * Makefile.in (OBJS): Add rtx-vector-builder.o.
11840 * rtx-vector-builder.h: New file.
11841 * rtx-vector-builder.c: Likewise.
11842 * rtl.h (rtx_def::u2): Add a const_vector field.
11843 (CONST_VECTOR_NPATTERNS): New macro.
11844 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
11845 (CONST_VECTOR_DUPLICATE_P): Likewise.
11846 (CONST_VECTOR_STEPPED_P): Likewise.
11847 (CONST_VECTOR_ENCODED_ELT): Likewise.
11848 (const_vec_duplicate_p): Check for a duplicated vector encoding.
11849 (unwrap_const_vec_duplicate): Likewise.
11850 (const_vec_series_p): Check for a non-duplicated vector encoding.
11851 Say that the function only returns true for integer vectors.
11852 * emit-rtl.c: Include rtx-vector-builder.h.
11853 (gen_const_vec_duplicate_1): Delete.
11854 (gen_const_vector): Call gen_const_vec_duplicate instead of
11855 gen_const_vec_duplicate_1.
11856 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
11857 (gen_const_vec_duplicate): Use rtx_vector_builder.
11858 (gen_const_vec_series): Likewise.
11859 (gen_rtx_CONST_VECTOR): Likewise.
11860 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
11861 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
11862 Build a new vector rather than modifying a CONST_VECTOR in-place.
11863 (handle_special_swappables): Update call accordingly.
11864 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
11865 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
11866 Build a new vector rather than modifying a CONST_VECTOR in-place.
11867 (handle_special_swappables): Update call accordingly.
11869 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11871 * simplify-rtx.c (simplify_const_binary_operation): Use
11872 CONST_VECTOR_ELT instead of XVECEXP.
11874 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11876 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
11877 the selector elements to be different from the data elements
11878 if the selector is a VECTOR_CST.
11879 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
11880 ssizetype for the selector.
11882 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11884 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
11885 before testing each element individually.
11886 * tree-vect-generic.c (lower_vec_perm): Likewise.
11888 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11890 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
11891 * selftest-run-tests.c (selftest::run_tests): Call it.
11892 * vector-builder.h (vector_builder::operator ==): New function.
11893 (vector_builder::operator !=): Likewise.
11894 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
11895 (vec_perm_indices::all_from_input_p): New function.
11896 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
11897 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
11898 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
11899 instead of reading the VECTOR_CST directly. Detect whether both
11900 vector inputs are the same before constructing the vec_perm_indices,
11901 and update the number of inputs argument accordingly. Use the
11902 utility functions added above. Only construct sel2 if we need to.
11904 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11906 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
11907 the broadcast of the low byte.
11908 (expand_mult_highpart): Use an explicit encoding for the permutes.
11909 * optabs-query.c (can_mult_highpart_p): Likewise.
11910 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
11911 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11912 (vectorizable_bswap): Likewise.
11913 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
11914 explicit encoding for the power-of-2 permutes.
11915 (vect_permute_store_chain): Likewise.
11916 (vect_grouped_load_supported): Likewise.
11917 (vect_permute_load_chain): Likewise.
11919 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11921 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
11922 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
11923 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
11924 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
11925 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
11926 (vect_gen_perm_mask_any): Likewise.
11928 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11930 * int-vector-builder.h: New file.
11931 * vec-perm-indices.h: Include int-vector-builder.h.
11932 (vec_perm_indices): Redefine as an int_vector_builder.
11933 (auto_vec_perm_indices): Delete.
11934 (vec_perm_builder): Redefine as a stand-alone class.
11935 (vec_perm_indices::vec_perm_indices): New function.
11936 (vec_perm_indices::clamp): Likewise.
11937 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
11938 (vec_perm_indices::new_vector): New function.
11939 (vec_perm_indices::new_expanded_vector): Update for new
11940 vec_perm_indices class.
11941 (vec_perm_indices::rotate_inputs): New function.
11942 (vec_perm_indices::all_in_range_p): Operate directly on the
11943 encoded form, without computing elided elements.
11944 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
11945 encoding. Update for new vec_perm_indices class.
11946 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
11947 the given vec_perm_builder.
11948 (expand_vec_perm_var): Update vec_perm_builder constructor.
11949 (expand_mult_highpart): Use vec_perm_builder instead of
11950 auto_vec_perm_indices.
11951 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
11952 vec_perm_indices instead of auto_vec_perm_indices. Use a single
11953 or double series encoding as appropriate.
11954 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
11955 vec_perm_indices instead of auto_vec_perm_indices.
11956 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11957 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11958 (vect_permute_store_chain): Likewise.
11959 (vect_grouped_load_supported): Likewise.
11960 (vect_permute_load_chain): Likewise.
11961 (vect_shift_permute_load_chain): Likewise.
11962 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
11963 (vect_transform_slp_perm_load): Likewise.
11964 (vect_schedule_slp_instance): Likewise.
11965 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11966 (vectorizable_mask_load_store): Likewise.
11967 (vectorizable_bswap): Likewise.
11968 (vectorizable_store): Likewise.
11969 (vectorizable_load): Likewise.
11970 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
11971 vec_perm_indices instead of auto_vec_perm_indices. Use
11972 tree_to_vec_perm_builder to read the vector from a tree.
11973 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
11974 vec_perm_builder instead of a vec_perm_indices.
11975 (have_whole_vector_shift): Use vec_perm_builder and
11976 vec_perm_indices instead of auto_vec_perm_indices. Leave the
11977 truncation to calc_vec_perm_mask_for_shift.
11978 (vect_create_epilog_for_reduction): Likewise.
11979 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
11980 from auto_vec_perm_indices to vec_perm_indices.
11981 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
11982 instead of changing individual elements.
11983 (aarch64_vectorize_vec_perm_const): Use new_vector to install
11984 the vector in d.perm.
11985 * config/arm/arm.c (expand_vec_perm_d::perm): Change
11986 from auto_vec_perm_indices to vec_perm_indices.
11987 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
11988 instead of changing individual elements.
11989 (arm_vectorize_vec_perm_const): Use new_vector to install
11990 the vector in d.perm.
11991 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
11992 Update vec_perm_builder constructor.
11993 (rs6000_expand_interleave): Likewise.
11994 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
11995 (rs6000_expand_interleave): Likewise.
11997 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11999 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
12000 to qimode could truncate the indices.
12001 * optabs.c (expand_vec_perm_var): Likewise.
12003 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12005 * Makefile.in (OBJS): Add vec-perm-indices.o.
12006 * vec-perm-indices.h: New file.
12007 * vec-perm-indices.c: Likewise.
12008 * target.h (vec_perm_indices): Replace with a forward class
12010 (auto_vec_perm_indices): Move to vec-perm-indices.h.
12011 * optabs.h: Include vec-perm-indices.h.
12012 (expand_vec_perm): Delete.
12013 (selector_fits_mode_p, expand_vec_perm_var): Declare.
12014 (expand_vec_perm_const): Declare.
12015 * target.def (vec_perm_const_ok): Replace with...
12016 (vec_perm_const): ...this new hook.
12017 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
12018 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
12019 * doc/tm.texi: Regenerate.
12020 * optabs.def (vec_perm_const): Delete.
12021 * doc/md.texi (vec_perm_const): Likewise.
12022 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
12023 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
12024 expand_vec_perm for constant permutation vectors. Assert that
12025 the mode of variable permutation vectors is the integer equivalent
12026 of the mode that is being permuted.
12027 * optabs-query.h (selector_fits_mode_p): Declare.
12028 * optabs-query.c: Include vec-perm-indices.h.
12029 (selector_fits_mode_p): New function.
12030 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
12031 is defined, instead of checking whether the vec_perm_const_optab
12032 exists. Use targetm.vectorize.vec_perm_const instead of
12033 targetm.vectorize.vec_perm_const_ok. Check whether the indices
12034 fit in the vector mode before using a variable permute.
12035 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
12036 vec_perm_indices instead of an rtx.
12037 (expand_vec_perm): Replace with...
12038 (expand_vec_perm_const): ...this new function. Take the selector
12039 as a vec_perm_indices rather than an rtx. Also take the mode of
12040 the selector. Update call to shift_amt_for_vec_perm_mask.
12041 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
12042 Use vec_perm_indices::new_expanded_vector to expand the original
12043 selector into bytes. Check whether the indices fit in the vector
12044 mode before using a variable permute.
12045 (expand_vec_perm_var): Make global.
12046 (expand_mult_highpart): Use expand_vec_perm_const.
12047 * fold-const.c: Includes vec-perm-indices.h.
12048 * tree-ssa-forwprop.c: Likewise.
12049 * tree-vect-data-refs.c: Likewise.
12050 * tree-vect-generic.c: Likewise.
12051 * tree-vect-loop.c: Likewise.
12052 * tree-vect-slp.c: Likewise.
12053 * tree-vect-stmts.c: Likewise.
12054 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
12056 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
12057 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
12058 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
12059 (aarch64_vectorize_vec_perm_const): ...this new function.
12060 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12061 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12062 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
12063 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
12064 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12065 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12066 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
12068 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
12069 check for NEON modes.
12070 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
12071 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
12072 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
12073 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
12075 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
12076 the old VEC_PERM_CONST conditions.
12077 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
12078 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
12079 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
12080 (ia64_vectorize_vec_perm_const_ok): Merge into...
12081 (ia64_vectorize_vec_perm_const): ...this new function.
12082 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
12083 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
12084 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
12085 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
12086 * config/mips/mips.c (mips_expand_vec_perm_const)
12087 (mips_vectorize_vec_perm_const_ok): Merge into...
12088 (mips_vectorize_vec_perm_const): ...this new function.
12089 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
12090 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
12091 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
12092 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
12093 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
12094 (rs6000_expand_vec_perm_const): Delete.
12095 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
12097 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12098 (altivec_expand_vec_perm_const_le): Take each operand individually.
12099 Operate on constant selectors rather than rtxes.
12100 (altivec_expand_vec_perm_const): Likewise. Update call to
12101 altivec_expand_vec_perm_const_le.
12102 (rs6000_expand_vec_perm_const): Delete.
12103 (rs6000_vectorize_vec_perm_const_ok): Delete.
12104 (rs6000_vectorize_vec_perm_const): New function.
12105 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
12106 an element count and rtx array.
12107 (rs6000_expand_extract_even): Update call accordingly.
12108 (rs6000_expand_interleave): Likewise.
12109 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
12110 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
12111 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
12112 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
12113 (rs6000_expand_vec_perm_const): Delete.
12114 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12115 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12116 (altivec_expand_vec_perm_const_le): Take each operand individually.
12117 Operate on constant selectors rather than rtxes.
12118 (altivec_expand_vec_perm_const): Likewise. Update call to
12119 altivec_expand_vec_perm_const_le.
12120 (rs6000_expand_vec_perm_const): Delete.
12121 (rs6000_vectorize_vec_perm_const_ok): Delete.
12122 (rs6000_vectorize_vec_perm_const): New function. Remove stray
12123 reference to the SPE evmerge intructions.
12124 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
12125 an element count and rtx array.
12126 (rs6000_expand_extract_even): Update call accordingly.
12127 (rs6000_expand_interleave): Likewise.
12128 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
12129 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
12131 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12133 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12135 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
12136 vector mode and that that mode matches the mode of the data
12138 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
12139 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
12140 directly using expand_vec_perm_1 when forcing selectors into
12142 (expand_vec_perm_var): New function, split out from expand_vec_perm.
12144 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12146 * optabs-query.h (can_vec_perm_p): Delete.
12147 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
12148 * optabs-query.c (can_vec_perm_p): Split into...
12149 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
12150 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
12151 particular selector is valid.
12152 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12153 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12154 (vect_grouped_load_supported): Likewise.
12155 (vect_shift_permute_load_chain): Likewise.
12156 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
12157 (vect_transform_slp_perm_load): Likewise.
12158 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12159 (vectorizable_bswap): Likewise.
12160 (vect_gen_perm_mask_checked): Likewise.
12161 * fold-const.c (fold_ternary_loc): Likewise. Don't take
12162 implementations of variable permutation vectors into account
12163 when deciding which selector to use.
12164 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
12165 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
12166 with a false third argument.
12167 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
12168 to test whether the constant selector is valid and can_vec_perm_var_p
12169 to test whether a variable selector is valid.
12171 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12173 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
12174 * optabs-query.c (can_vec_perm_p): Likewise.
12175 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
12176 instead of vec_perm_indices.
12177 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
12178 (vect_gen_perm_mask_checked): Likewise,
12179 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
12180 (vect_gen_perm_mask_checked): Likewise,
12182 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12184 * optabs-query.h (qimode_for_vec_perm): Declare.
12185 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
12186 (qimode_for_vec_perm): ...this new function.
12187 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
12189 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12191 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
12192 does not have a conditional at the top.
12194 2018-01-02 Richard Biener <rguenther@suse.de>
12196 * ipa-inline.c (big_speedup_p): Fix expression.
12198 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
12201 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
12204 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
12208 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
12209 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
12210 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
12211 cond_taken_branch_cost 3->4.
12213 2018-01-01 Jakub Jelinek <jakub@redhat.com>
12215 PR tree-optimization/83581
12216 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
12217 TODO_cleanup_cfg if any changes have been made.
12219 PR middle-end/83608
12220 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
12221 convert_modes if target mode has the right side, but different mode
12224 PR middle-end/83609
12225 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
12226 last argument when extracting from CONCAT. If either from_real or
12227 from_imag is NULL, use expansion through memory. If result is not
12228 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
12229 the parts directly to inner mode, if even that fails, use expansion
12232 PR middle-end/83623
12233 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
12234 check for bswap in mode rather than HImode and use that in expand_unop
12237 Copyright (C) 2018 Free Software Foundation, Inc.
12239 Copying and distribution of this file, with or without modification,
12240 are permitted in any medium without royalty provided the copyright
12241 notice and this notice are preserved.