1 2018-05-06 Jakub Jelinek <jakub@redhat.com>
4 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
5 the type is addressable. Don't force op into register if it has
8 2018-05-05 Roland McGrath <mcgrathr@google.com>
11 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
12 any section for which we don't know a specific type it should have,
13 regardless of name. Previously this was done only for the exact
14 names ".init_array", ".fini_array", and ".preinit_array".
15 (default_elf_asm_named_section): Add comment about
16 relationship with default_section_type_flags and SECTION_NOTYPE.
17 (get_section): Don't consider it a type conflict if one side has
18 SECTION_NOTYPE and the other doesn't, as long as neither has the
19 SECTION_BSS et al used in the default_section_type_flags logic.
21 2018-05-05 Tom de Vries <tom@codesourcery.com>
24 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
25 (workaround_barsyncs): New function.
26 (nvptx_reorg): Use workaround_barsyncs.
27 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
28 (define_expand "nvptx_membar_cta"): New define_expand.
29 (define_insn "*nvptx_membar_cta"): New insn.
31 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
33 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
34 To improve optimization opportunities.
35 * builtin-types.def: The new needed builtin types for the above.
37 2018-05-04 Richard Biener <rguenther@suse.de>
39 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
40 * gimple-ssa-store-merging.c
41 (imm_store_chain_info::output_merged_store): Remove redundant create,
42 release split_store vector contents on failure.
43 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
44 scalar stmt vector on cache hit.
46 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
48 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
50 * config.gcc (powerpc-xilinx-eabi*): Remove.
51 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
53 (fusion_addis_mem_combo_load): Ditto.
54 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
56 (rs6000_cpu_cpp_builtins): Ditto.
57 * config/rs6000/rs6000-linux.c
58 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
59 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
60 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
62 (rs6000_setup_reg_addr_masks): Ditto.
63 (rs6000_init_hard_regno_mode_ok): Ditto.
64 (rs6000_option_override_internal): Ditto.
65 (legitimate_lo_sum_address_p): Ditto.
66 (rs6000_legitimize_address): Ditto.
67 (rs6000_legitimize_reload_address): Ditto.
68 (rs6000_legitimate_address_p): Ditto.
69 (abi_v4_pass_in_fpr): Ditto.
70 (setup_incoming_varargs): Ditto.
71 (rs6000_gimplify_va_arg): Ditto.
72 (rs6000_split_multireg_move): Ditto.
73 (rs6000_savres_strategy): Ditto.
74 (rs6000_emit_prologue_components): Ditto.
75 (rs6000_emit_epilogue_components): Ditto.
76 (rs6000_emit_prologue): Ditto.
77 (rs6000_emit_epilogue): Ditto.
78 (rs6000_elf_file_end): Ditto.
79 (rs6000_function_value): Ditto.
80 (rs6000_libcall_value): Ditto.
81 * config/rs6000/rs6000.h: Ditto.
82 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
83 (TARGET_MINMAX): ... this. New.
84 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
85 * config/rs6000/rs6000.md: Remove Xilinx FP support.
86 (*movsi_internal1_single): Delete.
87 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
88 mfpu=, mxilinx-fpu): Delete.
89 * config/rs6000/singlefp.h: Delete.
90 * config/rs6000/sysv4.h: Remove Xilinx FP support.
91 * config/rs6000/t-rs6000: Ditto.
92 * config/rs6000/t-xilinx: Delete.
93 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
94 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
95 (VStype_simple): Delete.
96 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
97 * config/rs6000/xfpu.h: Delete.
98 * config/rs6000/xfpu.md: Delete.
99 * config/rs6000/xilinx.h: Delete.
100 * config/rs6000/xilinx.opt: Delete.
101 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
102 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
104 2018-05-04 Tom de Vries <tom@codesourcery.com>
107 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
110 2018-05-04 Richard Biener <rguenther@suse.de>
113 * tree-complex.c (update_complex_assignment): We are always in SSA form.
114 (expand_complex_div_wide): Likewise.
115 (expand_complex_operations_1): Likewise.
116 (expand_complex_libcall): Preserve EH info of the original stmt.
117 (tree_lower_complex): Handle removed blocks.
118 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
119 on complex multiplication and division libcall builtins.
121 2018-05-04 Richard Biener <rguenther@suse.de>
124 * fold-const.c (negate_expr_p): Restrict negation of operand
125 zero of a division to when we know that can happen without
127 (fold_negate_expr_1): Likewise.
129 2018-05-04 Jakub Jelinek <jakub@redhat.com>
132 * real.h (real_nextafter): Declare.
133 * real.c (real_nextafter): New function.
134 * fold-const-call.c (fold_const_nextafter): New function.
135 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
137 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
138 even when arg1_mode is different from arg0_mode.
140 2018-05-03 Nathan Sidwell <nathan@acm.org>
142 * doc/extend.texi (Deprecated Features): Remove
144 (Backwards Compatibility): Likewise.
145 * doc/invoke.texi (C++ Language Options): Likewise.
146 (C++ Dialect Options): Likewise.
148 2018-05-03 Jakub Jelinek <jakub@redhat.com>
151 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
152 _mm512_mask_mullox_epi64): New intrinsics.
154 2018-05-03 Tom de Vries <tom@codesourcery.com>
157 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
158 dump files): Add offload-tree.
160 2018-05-03 Richard Biener <rguenther@suse.de>
162 PR tree-optimization/85615
163 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
164 to loops not nested in BBs loop father to avoid creating multi-entry
167 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
169 PR tree-optimization/70291
170 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
171 arguments. Change return type to tree. Emit libcall as a new
172 statement rather than replacing existing one when inplace_p is true.
173 (expand_complex_multiplication_components): New function.
174 (expand_complex_multiplication): Expand floating-point complex
175 multiplication using the above.
176 (expand_complex_division): Rename inner_type parameter to type.
177 Update expand_complex_libcall call-site.
178 (expand_complex_operations_1): Update expand_complex_multiplication
179 and expand_complex_division call-sites.
181 2018-05-02 Jakub Jelinek <jakub@redhat.com>
184 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
185 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
186 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
187 the highest significant bit of the shift count mask is clear. In
188 check whether and[sq]i3 is needed verify that all significant bits
189 of the shift count other than the highest are set.
191 2018-05-02 Tom de Vries <tom@codesourcery.com>
194 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
195 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
196 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
197 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
198 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
199 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
200 __builtin_goacc_parlevel_size.
202 2018-05-02 Richard Biener <rguenther@suse.de>
204 PR tree-optimization/85597
205 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
206 do not use split vect_get_vec_defs call but call vect_get_slp_defs
209 2018-05-02 Tom de Vries <tom@codesourcery.com>
212 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
213 dump files): Add ltrans-tree.
215 2018-05-02 Tom de Vries <tom@codesourcery.com>
218 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
219 dump files): Add wpa-ipa.
221 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
223 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
224 powerpc*-*-linux*paired* target.
225 * config/rs6000/750cl.h: Delete.
226 * config/rs6000/paired.h: Delete.
227 * config/rs6000/paired.md: Delete.
228 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
230 * config/rs6000/rs6000-builtin.def: Remove paired float support.
231 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
232 comment. Remove paired float support.
233 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
234 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
236 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
237 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
239 * config/rs6000/rs6000.c: Remove paired float support.
240 (paired_expand_vector_init, paired_expand_vector_move,
241 paired_emit_vector_compare, paired_emit_vector_cond_expr,
242 (paired_expand_lv_builtin, paired_expand_stv_builtin,
243 paired_expand_builtin, paired_expand_predicate_builtin,
244 paired_init_builtins): Delete.
245 * config/rs6000/rs6000.h: Remove paired float support.
246 * config/rs6000/rs6000.md: Remove paired float support.
247 (move_from_CR_ov_bit): Delete.
248 * config/rs6000/rs6000.opt (mpaired): Delete.
249 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
250 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
252 2018-05-02 Richard Biener <rguenther@suse.de>
255 * gimplify.c (gimplify_save_expr): When in SSA form allow
256 SAVE_EXPRs to compute to SSA vars.
258 2018-05-02 Jakub Jelinek <jakub@redhat.com>
261 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
262 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
263 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
264 clobber operands[2], instead use a new pseudo. Formatting fixes.
266 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
268 PR tree-optimization/85586
269 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
270 exit early for statements in the same group if the accesses are
273 2018-05-02 Tom de Vries <tom@codesourcery.com>
276 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
279 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
281 PR tree-optimization/85143
282 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
284 2018-05-01 Tom de Vries <tom@codesourcery.com>
287 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
288 not found" error message.
290 2018-05-01 Tom de Vries <tom@codesourcery.com>
293 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
294 * vec.c (test_ordered_remove_if): New function.
295 (vec_c_tests): Call test_ordered_remove_if.
296 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
297 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
298 * tree-vect-patterns.c (vect_pattern_recog_1): Use
299 VEC_ORDERED_REMOVE_IF.
301 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
303 PR tree-optimization/82665
304 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
305 pointer subtraction where arguments come from a memchr call.
307 2018-05-01 Jakub Jelinek <jakub@redhat.com>
309 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
310 --push-state --as-needed and --pop-state instead of --as-needed and
311 --no-as-needed if ld supports it.
312 * configure: Regenerated.
315 * doc/install.texi2html: Replace _002d with - and _002a with * in
316 generated html files using sed.
318 2018-04-30 David Malcolm <dmalcolm@redhat.com>
321 * gcc-rich-location.c (blank_line_before_p): New function.
322 (use_new_line): New function.
323 (gcc_rich_location::add_fixit_insert_formatted): New function.
324 * gcc-rich-location.h
325 (gcc_rich_location::add_fixit_insert_formatted): New function.
327 2018-04-30 David Malcolm <dmalcolm@redhat.com>
329 * selftest.c (assert_streq): Rename "expected" and "actual" to
330 "val1" and "val2". Extend NULL-handling to cover both inputs
331 symmetrically, while still requiring both to be non-NULL for a pass.
332 * selftest.h (assert_streq): Rename "expected" and "actual" to
334 (ASSERT_EQ): Likewise.
335 (ASSERT_EQ_AT): Likewise.
336 (ASSERT_KNOWN_EQ): Likewise.
337 (ASSERT_KNOWN_EQ_AT): Likewise.
338 (ASSERT_NE): Likewise.
339 (ASSERT_MAYBE_NE): Likewise.
340 (ASSERT_MAYBE_NE_AT): Likewise.
341 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
342 the assertion to pass.
343 (ASSERT_STREQ_AT): Likewise.
345 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
347 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
348 interaction with -pie.
350 2018-04-30 David Malcolm <dmalcolm@redhat.com>
352 * selftest.h: Fix alphabetization of per-source-file selftest
355 2018-04-30 Jason Merrill <jason@redhat.com>
357 PR c++/61982 - dead stores to destroyed objects.
358 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
361 2018-04-30 Jason Merrill <jason@redhat.com>
363 * tree.c (build_clobber): New.
364 * tree.h: Declare it.
365 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
367 2018-04-30 David Malcolm <dmalcolm@redhat.com>
369 * diagnostic-show-locus.c (layout::layout): Update for
370 location_get_source_line returning a char_span.
371 (struct char_span): Move to input.h.
372 (struct correction): Update for fields in char_span becoming
374 (struct source_line): Update for location_get_source_line
375 returning a char_span.
376 (layout::print_line): Likewise.
377 * edit-context.c (edited_file::print_content): Likewise.
378 (edited_file::print_diff_hunk): Likewise.
379 (edited_file::print_run_of_changed_lines): Likewise.
380 (edited_file::get_num_lines): Likewise.
381 (edited_line::edited_line): Likewise.
382 * final.c (asm_show_source): Likewise.
383 * input.c (location_get_source_line): Convert return type
384 from const char * to char_span, losing the final "line_len"
386 (dump_location_info): Update for the above.
387 (get_substring_ranges_for_loc): Likewise. Use a char_span
388 when handling the literal within the line.
389 (test_reading_source_line): Update for location_get_source_line
390 returning a char_span.
391 * input.h (class char_span): Move here from
392 diagnostic-show-locus.c, converting from a struct to a class.
393 Make data members private.
394 (char_span::operator bool): New.
395 (char_span::length): New.
396 (char_span::get_buffer): New.
397 (char_span::operator[]): New.
398 (char_span::subspan): Make const.
399 (char_span::xstrdup): New.
400 (location_get_source_line): Convert return type from const char *
401 to char_span, losing the final "line_size" param.
403 2018-04-30 Jan Hubicka <jh@suse.cz>
405 * lto-wrapper.c (ltrans_priorities): New static var.
407 (run_gcc): Read priorities and if doing parallel build order
408 the Makefile by them.
410 2018-04-30 David Malcolm <dmalcolm@redhat.com>
412 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
414 2018-04-30 Richard Biener <rguenther@suse.de>
416 * tree-cfg.c (verify_address): Remove base argument, add
417 flag whether to check TREE_ADDRESSABLE and do that.
418 (verify_expr): Remove.
419 (verify_types_in_gimple_reference): Add pieces from verify_expr.
420 (verify_gimple_assign_single): Likewise.
421 (verify_gimple_switch): Likewise.
422 (verify_expr_location_1): Dereference tp once. Add (disabled)
423 piece from verify_expr.
424 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
426 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
428 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
430 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
432 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
433 (small_data_pattern): Likewise.
434 (arc_rewrite_small_data): Likewise.
435 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
436 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
437 (get_symbol_alignment): New function.
438 (legitimate_small_data_address_p): Likewise.
439 (legitimate_scaled_address): Update, call
440 legitimate_small_data_address_p.
441 (output_sdata): New static variable.
442 (arc_print_operand): Update how we handle small data operands.
443 (arc_print_operand_address): Likewise.
444 (arc_legitimate_address_p): Update, use
445 legitimate_small_data_address_p.
446 (arc_rewrite_small_data_p): Remove.
447 (arc_rewrite_small_data_1): Likewise.
448 (arc_rewrite_small_data): Likewise.
449 (small_data_pattern): Likewise.
450 (compact_sda_memory_operand): Update to use
451 legitimate_small_data_address_p and get_symbol_alignment.
452 (prepare_move_operands): Don't rewite sdata pattern.
453 (prepare_extend_operands): Remove.
454 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
456 (zero_extendqisi2): Likewise.
457 (zero_extendhisi2): Likewise.
458 (extendqihi2): Likewise.
459 (extendqisi2): Likewise.
460 (extendhisi2): Likewise.
464 * config/arc/constraints.md (Usd): Change it to memory constraint.
466 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
468 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
469 as source of std instructions.
470 * config/arc/arc.md (movsi_insn): Update pattern predicate to
471 allow 6-bit constants as source for store instructions.
472 (movdi_insn): Update instruction pattern to allow 6-bit constants
473 as source for store instructions.
475 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
477 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
479 2018-04-30 Nathan Sidwell <nathan@acm.org>
480 Sandra Loosemore <sandra@codesourcery.com>
482 * dumpfile.c (dump_open): Allow '-' for stdout.
483 * doc/invoke.texi (Developer Options): Document dump filename
484 determination early. Document stdin/stdout selection.
486 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
488 Microblaze Target: PIC data text relative
490 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
491 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
493 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
494 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
495 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
496 New addressing mode for data-text relative position indepenedent code.
497 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
498 'ADDRESS_SYMBOLIC_TXT_REL'.
499 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
500 (microblaze_legitimate_pic_operand): Exclude function calls from
501 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
502 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
504 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
505 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
506 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
507 for 'address + offset'.
508 (microblaze_expand_prologue): Add new function prologue call for
510 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
511 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
512 table in case of TARGET_PIC_DATA_TEXT_REL.
513 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
514 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
515 Add new macros 'UNSPEC_TEXT',
516 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
517 + exclude function calls from 'UNSPEC_PLT' in case of data text
519 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
520 new target hook for generating address diff vector tables in case of
522 * doc/tm.texi : Regenerate.
523 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
524 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
525 of addr diff vector generation.
526 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
527 target hook definition.
528 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
529 Add default function for generate_pic_addr_diff_vec -> flag_pic.
530 * doc/invoke.texi (Add new pic option): Add new microblaze pic
531 option for data text relative.
533 2018-04-30 Richard Biener <rguenther@suse.de>
535 * tree-chrec.h (evolution_function_is_constant_p): Remove
537 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
539 2018-04-30 Richard Biener <rguenther@suse.de>
542 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
544 2018-04-30 Richard Biener <rguenther@suse.de>
546 PR tree-optimization/28364
547 PR tree-optimization/85275
548 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
549 copying first exit test.
551 2018-04-28 Mark Wielaard <mark@klomp.org>
553 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
555 (dwarf_AT): Handle DW_AT_addr_base.
556 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
558 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
561 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
562 (*ashl<dwi>3_doubleword_mask_1): Ditto.
563 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
564 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
566 2018-04-28 Richard Biener <rguenther@suse.de>
568 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
569 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
570 to reflect use. Only add interesting stmts.
572 2018-04-27 Martin Jambor <mjambor@suse.cz>
575 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
576 the jump function allows for passing through aggregate values.
578 2018-04-27 David Malcolm <dmalcolm@redhat.com>
580 * input.h (in_system_header_at): Convert from macro to inline
582 (from_macro_expansion_at): Likewise.
583 (from_macro_definition_at): Likewise.
585 2018-04-27 Jeff Law <law@redhat.com>
587 * config.gcc: Mark tile* targets as deprecated/obsolete.
589 2018-04-27 Richard Biener <rguenther@suse.de>
591 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
594 2018-04-27 Richard Biener <rguenther@suse.de>
596 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
598 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
600 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
601 with Yd constraint. Set "preferred_for_speed" attribute from
602 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
604 (*movdi_internal): Ditto.
605 (movti_interunit splitters): Remove
606 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
607 (movdi_interunit splitters): Ditto.
608 * config/i386/constraints.md (Ye): Remove.
609 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
611 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
614 * config/aarch64/constraints.md (Usg): Limit to 31.
617 2018-04-27 Jakub Jelinek <jakub@redhat.com>
619 PR tree-optimization/85529
620 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
621 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
622 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
623 zero extension or masking of the MSB bit.
624 (optimize_range_tests): Add FIRST_BB argument, pass it through
625 to optimize_range_tests_var_bound.
626 (maybe_optimize_range_tests, reassociate_bb): Adjust
627 optimize_range_tests callers.
629 2018-04-26 Richard Biener <rguenther@suse.de>
630 Jakub Jelinek <jakub@redhat.com>
632 * cgraph.h (symbol_table): Just declare debug method here.
633 * symtab.c (symbol_table::debug): Define.
635 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
637 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
639 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
641 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
642 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
643 (*movdi_internal): Substitute Yi and Yj constraint with x
644 and Ym and Yn constraint with y constraint. Update "isa"
645 attribute and set "preferred_for_speed" attribute from
646 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
647 (*movsi_internal): Ditto.
648 (*movdf_internal): Ditto.
649 (*movsf_internal): Ditto.
650 (*zero_extendsidi2): Ditto.
651 * config/i386/sse.md (vec_set<mode>_0): Ditto.
652 (sse2_loadld): Ditto.
653 (*vec_extract<ssevecmodelower>_0): Ditto.
654 (*vec_extractv4si_0_zext_sse4): Ditto.
655 (vec_concatv2di): Ditto.
656 (*vec_dup<mode>): Ditto.
657 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
658 * config/i386/constraints.md (Yi): Remove.
663 2018-04-26 Nathan Sidwell <nathan@acm.org>
665 * dumpfile.c (dump_open): New.
666 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
667 (dump_finish): Detect stdio/stderr by value not name.
669 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
671 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
673 2018-04-26 Tom de Vries <tom@codesourcery.com>
676 * config/nvptx/nvptx.c (verify_neutering_jumps)
677 (verify_neutering_labels): New function
678 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
680 2018-04-26 Tom de Vries <tom@codesourcery.com>
683 * config/nvptx/nvptx.c (needs_neutering_p): New function.
684 (nvptx_single): Use needs_neutering_p to skip over insns that do not
687 2018-04-26 Richard Biener <rguenther@suse.de>
688 Tom de Vries <tom@codesourcery.com>
691 * lto-streamer-out.c (output_function): Fixup loops if required to match
692 discovery done in the reader.
694 2018-04-26 Richard Biener <rguenther@suse.de>
696 PR tree-optimization/85116
697 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
698 have a loop exit from the single latch predecessor. Remove
699 case of header with just condition.
700 (ch_base::copy_headers): Exclude infinite loops from any
702 (pass_ch::execute): Record exits.
704 2018-04-26 Richard Biener <rguenther@suse.de>
706 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
707 prologue cost vector and pass it to vect_get_load_cost.
708 (vect_get_peeling_costs_all_drs): Likewise.
709 (vect_peeling_hash_get_lowest_cost): Likewise.
710 (vect_enhance_data_refs_alignment): Likewise.
712 2018-04-26 Richard Biener <rguenther@suse.de>
715 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
716 checking of integer<->pointer conversions.
717 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
718 sign-/zero-extending pointer types.
719 (expand_omp_for_static_chunk): Likewise.
721 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
722 Jean Lee <xiaoyur347@gmail.com>
724 * config/mips/mips.c (mips_asan_shadow_offset): New function.
725 (TARGET_ASAN_SHADOW_OFFSET): Define.
726 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
727 true for -fsanitize=address.
729 2018-04-25 Mark Wielaard <mark@klomp.org>
731 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
734 2018-04-25 Jakub Jelinek <jakub@redhat.com>
736 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
737 than "alu", remove explicit "memory" and "imm_disp" attributes.
738 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
741 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
742 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
745 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
748 * config/i386/i386.c (ix86_expand_builtin): Change memory
749 operand to XI, extend p0 to Pmode.
750 * config/i386/i386.md: Change unspec volatile and operand
751 1 mode to XI, change operand 0 mode to P.
753 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
755 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
756 GET_MODE_MASK before any checking.
757 (nds32_can_use_bset_p): Likewise.
758 (nds32_can_use_btgl_p): Likewise.
760 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
762 * config/nds32/nds32-doubleword.md: New define_split pattern for
763 illegal register number.
765 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
767 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
769 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
771 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
773 2018-04-25 Richard Biener <rguenther@suse.de>
775 * lto-streamer.h (LTO_major_version): Bump to 8.
777 2018-04-25 Jakub Jelinek <jakub@redhat.com>
779 * BASE-VER: Set to 9.0.0.
781 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
783 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
784 in __abskf2 and __powikf2.
786 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
789 * config/aarch64/constraints.md (Usg, Usj): New constraints.
790 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
791 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
792 Use the above on operand 2. Reindent.
793 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
795 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
798 * common/config/i386/i386-common.c (ix86_handle_option): Don't
800 * config/i386/i386.opt (mcet): Removed.
801 * doc/install.texi: Remove -mcet documentation.
802 * doc/invoke.texi: Likewise.
804 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
807 * doc/install.texi: Remove -mcet from bootstrap-cet.
809 2018-04-24 Jakub Jelinek <jakub@redhat.com>
812 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
813 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
817 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
818 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
819 containing a CONST_VECTOR.
821 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
823 * doc/install.texi: Update newlib dependency for nvptx.
825 2018-04-24 Jakub Jelinek <jakub@redhat.com>
828 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
829 instead of INTVAL when shifting x left.
831 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
833 PR tree-optimization/85478
834 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
835 vect_grouped_store_supported for single element vectors.
837 2018-04-24 Richard Biener <rguenther@suse.de>
840 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
841 load cost increase to the case of non-constant step.
843 2018-04-24 Jakub Jelinek <jakub@redhat.com>
846 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
847 destination if any_malformed_asm.
849 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
852 * expr.c (store_field): In the bitfield case, if the value comes from
853 a function call and is returned in registers by means of a PARALLEL,
854 do not change the mode of the temporary unless BLKmode and VOIDmode.
856 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
858 PR rtl-optimization/85423
859 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
860 dependencies to debug insns when the previous insn is non-debug.
862 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
864 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
865 enums into a single definition.
866 (fls): Fix predicates and printing.
869 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
871 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
872 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
873 and short u6 immediate.
874 (check_if_valid_sleep_operand): Remove.
875 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
877 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
879 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
880 flag_always_save_lp condition.
881 * config/nds32/nds32.opt (malways-save-lp): New option.
883 2018-04-22 Shiva Chen <shiva0217@gmail.com>
885 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
886 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
887 * config/nds32/nds32.h
888 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
889 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
891 2018-04-22 Shiva Chen <shiva0217@gmail.com>
893 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
894 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
896 2018-04-22 Shiva Chen <shiva0217@gmail.com>
897 Chung-Ju Wu <jasonwucj@gmail.com>
899 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
901 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
902 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
904 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
906 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
908 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
910 * config/nds32/nds32-protos.h (nds32_data_alignment,
911 nds32_local_alignment): Declare.
912 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
913 nds32_local_alignment): New functions.
914 (TARGET_CONSTANT_ALIGNMENT): Define.
915 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
917 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
919 * config/nds32/nds32.c
920 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
921 (TARGET_MODES_TIEABLE_P): Likewise.
923 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
925 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
928 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
929 Chung-Ju Wu <jasonwucj@gmail.com>
931 * config/nds32/constants.md (unspec_volatile_element): Add enum values
932 for unaligned access.
933 * config/nds32/nds32-intrinsic.c: Implementation of expanding
935 * config/nds32/nds32-intrinsic.md: Likewise.
936 * config/nds32/nds32_intrinsic.h: Likewise.
937 * config/nds32/nds32.h (nds32_builtins): Likewise.
938 * config/nds32/nds32.opt (munaligned-access): New option.
939 * config/nds32/nds32.c (nds32_asm_file_start): Display
940 flag_unaligned_access status.
942 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
944 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
945 -mno-relax is present.
946 * config/riscv/linux.h (LINK_SPEC): Ditto.
948 2018-04-20 Martin Sebor <msebor@redhat.com>
951 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
953 (gimple_fold_builtin_stxcpy_chk): Same.
954 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
956 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
959 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
960 __powikf2 when long double is IEEE 128-bit.
962 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
964 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
965 step to make sure stack always aligned.
967 2018-04-20 Carl Love <cel@us.ibm.com>
970 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
973 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
974 Tom de Vries <tom@codesourcery.com>
977 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
978 Emit insns for calls too.
979 (nvptx_find_par): Always look for worker-level predecessor insn.
980 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
982 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
983 (nvptx_process_pars): Propagate frames for calls.
985 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
988 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
990 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
991 (ix86_handle_option): Don't handle OPT_mibt.
992 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
994 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
996 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
997 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
998 (ix86_target_macros): Define __CET__ with flag_cf_protection
1000 * config/i386/i386.c (isa2_opts): Remove -mibt.
1001 * config/i386/i386.h (TARGET_IBT): Removed.
1002 (TARGET_IBT_P): Likewise.
1003 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
1004 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
1005 * config/i386/i386.opt (mcet): Update help message.
1008 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
1009 -mcet as an alias for -mshstk.
1011 2018-04-20 Richard Biener <rguenther@suse.de>
1014 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
1015 complexity by forcing a single use of the multiply operand.
1017 2018-04-20 Martin Jambor <mjambor@suse.cz>
1020 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
1021 recursion dependency to only apply to non-clones.
1023 2018-04-20 Martin Jambor <mjambor@suse.cz>
1026 * ipa-cp.c (create_specialized_node): Check that clones of
1027 self-recursive edges exist during IPA-CP.
1029 2018-04-19 Toon Moene <toon@moene.org>
1031 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
1034 2018-04-19 Jakub Jelinek <jakub@redhat.com>
1036 PR tree-optimization/85467
1037 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
1038 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
1039 VECTOR_CST element to type.
1041 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1044 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
1045 * config/i386/i386.md (builtin_setjmp_setup): Removed.
1046 (builtin_longjmp): Likewise.
1047 (save_stack_nonlocal): New pattern.
1048 (restore_stack_nonlocal): Likewise.
1050 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1053 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
1054 Replace ASM_OUTPUT_LABEL with fprintf.
1056 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1059 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
1060 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
1061 * config/i386/i386-c.c (ix86_target_macros_internal): Also
1062 define __IBT__ and __SHSTK__ for -fcf-protection.
1063 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
1065 (ix86_trampoline_init): Likewise.
1066 (x86_output_mi_thunk): Likewise.
1067 (ix86_notrack_prefixed_insn_p): Likewise.
1068 (ix86_option_override_internal): Don't disallow -fcf-protection.
1069 * config/i386/i386.md (rdssp<mode>): Also enable for
1071 (incssp<mode>): Likewise.
1072 (nop_endbr): Likewise.
1073 * config/i386/i386.opt (mcet): Change help message to built-in
1077 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
1078 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
1079 enable CET built-in functions.
1081 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
1083 * common/config/i386/i386-common.c
1084 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
1085 OPTION_MASK_ISA_MOVDIRI_UNSET,
1086 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
1087 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
1088 * config.gcc (movdirintrin.h): New header.
1089 * config/i386/cpuid.h (bit_MOVDIRI,
1090 bit_MOVDIR64B): New bits.
1091 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
1093 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
1094 (VOID, PVOID, PCVOID)): New function types.
1095 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
1096 __builtin_ia32_directstoreu_u64,
1097 __builtin_ia32_movdir64b): New builtins.
1098 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
1099 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
1101 (ix86_valid_target_attribute_inner_p): Ditto.
1102 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
1103 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
1104 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
1105 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
1106 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
1107 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
1108 (movdiri<mode>, movdir64b_<mode>): New.
1109 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
1110 * config/i386/immintrin.h: Include movdirintrin.h.
1111 * config/i386/movdirintrin.h: New file.
1112 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
1114 2018-04-19 Richard Biener <rguenther@suse.de>
1117 * cfg.c (clear_bb_flags): When loop state says we have
1118 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
1120 2018-04-19 Richard Biener <rguenther@suse.de>
1122 PR tree-optimization/84737
1123 * tree-vect-data-refs.c (vect_copy_ref_info): New function
1124 copying restrict info.
1125 (vect_setup_realignment): Use it.
1126 * tree-vectorizer.h (vect_copy_ref_info): Declare.
1127 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
1128 the first DR to all generated stores.
1129 (vectorizable_load): Likewise for loads.
1131 2018-04-19 Jakub Jelinek <jakub@redhat.com>
1133 PR tree-optimization/85446
1134 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
1135 the integral and pointer types to have the same precision.
1137 * doc/install.texi: Document --disable-cet being the default and
1140 2018-04-18 Martin Liska <mliska@suse.cz>
1142 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
1145 2018-04-18 Martin Liska <mliska@suse.cz>
1148 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
1151 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
1152 arguments if they are comparable.
1154 2018-04-18 Martin Liska <mliska@suse.cz>
1157 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
1160 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
1163 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
1166 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
1167 ENDBR after calling __morestack.
1169 2018-04-18 David Malcolm <dmalcolm@redhat.com>
1172 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
1173 by using gcc_base_ver to generate a gcc_driver_version, and use
1174 it when generating GCC_DRIVER_NAME.
1175 * configure: Regenerate.
1177 2018-04-18 Jakub Jelinek <jakub@redhat.com>
1180 * config.gcc: Obsolete powerpc*-*-*spe*.
1182 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1185 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
1186 (stabstr_D): Change type of unum from unsigned int to
1187 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
1190 2018-04-17 Jim Wilson <jimw@sifive.com>
1193 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
1194 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
1195 Set arg_pointer_offset after using pretend_args_size.
1197 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1199 PR rtl-optimization/85431
1200 * dse.c (record_store): Ignore zero width stores.
1203 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
1204 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
1205 __builtin_stack_restore rather than after it.
1206 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
1207 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
1208 argument instead of virtual_dynamic_stack_rtx.
1210 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
1212 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
1214 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
1215 Add note to error message to explain internal mapping of overloaded
1216 built-in function name to non-overloaded built-in function name.
1217 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
1220 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
1223 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
1224 where the inputs overlap with the output.
1226 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1228 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
1229 (=v, v) alternative and explicit "memory" attribute.
1230 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
1231 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
1233 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
1234 "sselog1" type instead of "sselog".
1235 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
1236 "sselog". Remove explicit "memory" attribute.
1237 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
1238 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
1240 (vec_extract_hi_v32hi): Merge all alternatives into one, use
1241 "sselog1" type instead of "sselog". Remove explicit "memory"
1243 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
1244 use "sselog1" type instead of "sselog". Remove explicit "memory"
1246 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
1247 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
1249 (vec_extract_hi_v64qi): Merge all alternatives into one, use
1250 "sselog1" type instead of "sselog". Remove explicit "memory"
1252 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
1253 use "sselog1" type instead of "sselog". Remove explicit "memory"
1257 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
1260 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
1263 2018-04-17 Martin Jambor <mjambor@suse.cz>
1266 * ipa-cp.c (create_specialized_node): Call
1267 expand_all_artificial_thunks if necessary.
1269 2018-04-17 Martin Liska <mliska@suse.cz>
1272 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
1273 in message, remote space in between '_G' and '('.
1275 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1278 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
1279 avx512f_vmcmp<mode>3<round_saeonly_name>,
1280 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
1281 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
1282 avx512f_rndscale<mode><round_saeonly_name>,
1283 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
1284 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
1285 Use %<iptr>2 instead of %2 for -masm=intel.
1286 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
1287 avx512f_vcvttss2usi<round_saeonly_name>,
1288 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
1290 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
1291 avx512f_vcvttsd2usi<round_saeonly_name>,
1292 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
1293 Use %q1 instead of %1 for -masm=intel.
1294 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
1295 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
1296 of %3 for -masm=intel.
1297 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
1299 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
1301 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
1303 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
1304 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
1306 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
1308 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
1309 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
1310 %g1 and one with %0 and %1.
1311 (avx512er_vmrcp28<mode><round_saeonly_name>,
1312 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
1314 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
1315 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
1316 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
1317 of %0 and %{%4%} for -masm=intel.
1318 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
1319 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
1320 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
1321 order of %0 and %{%5%}%{z%} for -masm=intel.
1323 2018-04-17 Jan Hubicka <jh@suse.cz>
1326 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
1328 2018-04-17 Martin Liska <mliska@suse.cz>
1331 * multiple_target.c (create_dispatcher_calls): Set apostrophes
1332 for target_clone error message. Make default implementation
1333 clone to be a local declaration.
1334 (separate_attrs): Add new argument and check for an empty
1336 (expand_target_clones): Handle it.
1337 (ipa_target_clone): Make redirection just for target_clones
1340 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
1341 Tom de Vries <tom@codesourcery.com>
1344 * omp-expand.c (expand_oacc_for): Add dummy false branch for
1345 tiled basic blocks without omp continue statements.
1347 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
1350 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
1351 vec_extract expression as having side effects to make sure it gets
1354 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
1357 * config/i386/i386.c (get_builtin_code_for_version): Check
1360 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
1363 * gcc/config.gcc: Support "skylake".
1364 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
1366 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
1367 (processor_target_table): Add "skylake".
1368 (ix86_option_override_internal): Add "skylake".
1369 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
1370 PROCESSOR_CANNONLAKE.
1371 (get_builtin_code_for_version): Fix priority for
1372 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
1373 PROCESSOR_SKYLAKE-AVX512.
1374 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
1375 (processor_type): Add PROCESSOR_SKYLAKE.
1377 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
1378 Jason Merrill <jason@redhat.com>
1381 * convert.c (convert_to_integer_1): Use direct recursion for
1382 enumeral types and types with a precision less than the number
1383 of bits in their mode.
1385 2018-04-16 Julia Koval <julia.koval@intel.com>
1388 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
1389 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
1391 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
1394 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
1395 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
1396 and -mno-direct-move.
1398 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
1401 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
1402 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
1403 Ensure negative shifts result in {0}.
1405 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
1407 PR rtl-optimization/79916
1408 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
1409 regs (if any) to define how to gnerate SD moves when LRA is in
1412 2018-04-13 Jakub Jelinek <jakub@redhat.com>
1414 PR rtl-optimization/85393
1415 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
1416 * except.c (expand_dw2_landing_pad_for_region): Make static.
1417 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
1418 a label and unconditional jump to old_bb, rather than
1419 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
1422 PR rtl-optimization/85376
1423 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
1424 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
1425 instead of a specific value.
1427 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
1428 Bin Cheng <bin.cheng@arm.com>
1430 PR tree-optimization/82965
1431 PR tree-optimization/83991
1432 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
1433 by_profile_only parameter.
1434 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
1435 information if the loop was predicted to iterate too many times.
1436 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
1438 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
1441 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
1444 2018-04-13 Martin Liska <mliska@suse.cz>
1445 Jakub Jelinek <jakub@redhat.com>
1448 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
1449 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
1450 * builtins.c (expand_builtin_memory_copy_args): Use
1451 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
1452 handle dest_addr == pc_rtx.
1454 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
1457 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
1458 asked to not generate direct moves.
1459 (fix_trunc<mode>si2_stfiwx): Similar.
1460 (fix_trunc<mode>si2_internal): Similar.
1462 2018-04-12 Jakub Jelinek <jakub@redhat.com>
1465 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
1466 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
1467 lookup if dest in some wider mode is known to be const0_rtx and
1468 if so, record permanent equivalence for it to be ZERO_EXTEND of
1469 the narrower mode destination.
1471 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
1473 * lto-streamer-out.c (output_function): Revert 259346.
1474 * omp-expand.c (expand_oacc_for): Likewise.
1476 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
1478 PR rtl-optimization/85354
1479 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
1480 * sel-sched.c (sel_global_init): ... here.
1482 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
1485 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
1486 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
1487 mode for PE-COFF targets.
1488 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
1489 (i386_pe_asm_lto_end): Likewise.
1490 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
1491 (TARGET_ASM_LTO_END): Likewise.
1492 * config/i386/winnt.c (saved_debug_info_level): New static variable.
1493 (i386_pe_asm_lto_start): New function.
1494 (i386_pe_asm_lto_end): Likewise.
1496 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
1497 Richard Biener <rguenther@suse.de>
1500 * lto-streamer-out.c (output_function): Fix CFG loop state before
1502 * omp-expand.c (expand_oacc_for): Handle calls to internal
1503 functions like regular functions.
1505 2018-04-12 Richard Biener <rguenther@suse.de>
1508 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
1509 for the early LTO debug to properly generate references to it
1510 during DIE emission. Do not re-use that for the skeleton for
1512 (dwarf2out_early_finish): Likewise.
1514 2018-04-12 Jakub Jelinek <jakub@redhat.com>
1517 * config/i386/sse.md
1518 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
1519 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
1520 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
1521 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
1522 and output is a reg, avoid creating invalid lowpart subreg, but
1523 instead split into a 512-bit move. Don't split if not AVX512VL,
1524 input is xmm16+ reg and output is a mem.
1525 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
1526 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
1527 xmm16+ reg and output is a mem.
1529 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1531 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
1532 also for flag_dwarf2_cfi_asm.
1534 2018-04-12 Jakub Jelinek <jakub@redhat.com>
1536 PR rtl-optimization/85342
1537 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
1538 a bool scalar var inside of the loop instead. Don't try to update
1539 recog_data.operand after failed apply_change_group.
1541 2018-04-12 Tom de Vries <tom@codesourcery.com>
1544 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
1545 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
1546 array with flexible array member as array without given dimension.
1547 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
1548 argument for undefined param to true.
1550 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
1553 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
1554 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
1555 from PowerPC section.
1556 * config/rs6000/sysv4.opt (mcall-): Improve help text.
1557 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
1558 help text that is too long.
1559 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
1560 help text that is too long.
1561 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
1562 help text that is too long.
1564 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
1566 * config/alpha/alpha.md (stack_probe_internal): Rename
1567 from "probe_stack". Update all callers.
1569 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1571 PR rtl-optimization/84566
1572 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
1573 sched_macro_fuse_insns.
1575 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1578 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
1579 (compute_block_dependences): ... from here.
1581 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1583 PR tree-optimization/85331
1584 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
1585 from int to HOST_WIDE_INT.
1587 2018-04-11 Martin Jambor <mjambor@suse.cz>
1590 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
1591 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
1592 not the same as the source val.
1593 (cgraph_edge_brings_value_p): New parameter.
1594 (gather_edges_for_value): Pass destination value to
1595 cgraph_edge_brings_value_p.
1596 (perhaps_add_new_callers): Likewise.
1597 (get_info_about_necessary_edges): Likewise and exclude values brought
1598 only by self-recursive edges.
1599 (create_specialized_node): Redirect only clones of self-calling edges.
1600 (+self_recursive_pass_through_p): New function.
1601 (find_more_scalar_values_for_callers_subset): Use it.
1602 (find_aggregate_values_for_callers_subset): Likewise.
1603 (known_aggs_to_agg_replacement_list): Removed.
1604 (decide_whether_version_node): Re-calculate known constants for all
1605 remaining context clones.
1607 2018-04-11 Richard Biener <rguenther@suse.de>
1610 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
1611 from early DWARF output.
1612 (dwarf2out_early_finish): Output line info unconditionally into
1613 early DWARF and add reference to it.
1615 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1618 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
1619 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
1620 other than V2DFmode using iptr mode attribute.
1621 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
1623 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1625 PR rtl-optimization/84659
1626 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
1628 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1631 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
1633 (output_loc_list): Pass address of a dummy size variable even in the
1634 locview handling loop.
1635 (index_location_lists): Add comment on why skip_loc_list_entry can't
1638 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
1641 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
1644 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
1647 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
1648 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
1649 and -mstring-compare-inline-limit.
1651 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1654 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
1655 for stack clash protection in a register whenever we need it to be in
1658 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1660 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
1661 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
1663 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1666 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
1668 (mlong-double-): Ditto.
1669 * config/rs6000/sysv4.opt (msdata=): Ditto.
1670 (mtls-size=): Ditto.
1672 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
1674 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
1675 erroneous entries for
1676 "vector int vec_ldl (int, long int *)", and
1677 "vector unsigned int vec_ldl (int, unsigned long int *)".
1678 Add comments and entries for
1679 "vector bool char vec_ldl (int, bool char *)",
1680 "vector bool short vec_ldl (int, bool short *)",
1681 "vector bool int vec_ldl (int, bool int *)",
1682 "vector bool long long vec_ldl (int, bool long long *)",
1683 "vector pixel vec_ldl (int, pixel *)",
1684 "vector long long vec_ldl (int, long long *)",
1685 "vector unsigned long long vec_ldl (int, unsigned long long *)".
1686 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
1687 type tree bool_long_long_type_node and correct definition of
1688 bool_V2DI_type_node to make reference to this new type tree.
1689 (rs6000_mangle_type): Replace erroneous reference to
1690 bool_long_type_node with bool_long_long_type_node.
1691 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
1692 comments to emphasize sign distinctions for char and int types and
1693 replace RS6000_BTI_bool_long constant with
1694 RS6000_BTI_bool_long_long constant. Also add comment to restrict
1695 use of RS6000_BTI_pixel.
1696 (bool_long_type_node): Remove this macro definition.
1697 (bool_long_long_type_node): New macro definition
1699 2018-04-10 Jakub Jelinek <jakub@redhat.com>
1701 PR rtl-optimization/85300
1702 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
1703 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
1704 simplify_unary_operation fails.
1706 2018-04-10 Martin Liska <mliska@suse.cz>
1708 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
1709 cgraph_edge and ipa_ref.
1711 2018-04-10 Jakub Jelinek <jakub@redhat.com>
1715 * config/i386/sse.md
1716 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
1717 computation of the VEC_MERGE selector from mask.
1718 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
1719 Fix decoding of the VEC_MERGE selector into mask.
1721 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
1723 PR tree-optimization/85286
1724 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
1726 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
1728 * final.c (final_1): Set insn_last_address as well as
1729 insn_current_address.
1731 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1734 * explow.c (emit_stack_probe): Call validize_mem on memory location
1735 before passing it to gen_probe_stack. Create address operand and
1736 legitimize it for the probe_stack_address case.
1738 2018-04-09 Jan Hubicka <jh@suse.cz>
1741 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
1742 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
1743 * tree.c (free_lang_data_in_type): Fix handling of binfos;
1745 (free_lang_data): Rebuild type inheritance graph.
1747 2018-04-09 Martin Sebor <msebor@redhat.com>
1749 * invoke.texi (-finline-small-functions): Mention other optimization
1751 (-findirect-inlining, -fpartial-inlining): Same.
1752 (-finline-functions-called-once): Same.
1753 (-freorder-blocks-and-partition): Same.
1755 2018-04-09 Jan Hubicka <jh@suse.cz>
1758 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
1759 jumps; choose last target that matches the criteria (i.e.
1760 no partition changes for non-crossing jumps).
1761 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
1762 support for redirecting crossing jumps to non-crossing.
1764 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
1766 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
1767 also for naked functions.
1769 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
1771 * config/arc/arc.md (add_shift): New pattern.
1772 (add_shift2): Likewise.
1773 (sub_shift): Likewise.
1774 (sub_shift_cmp0_noout): Likewise.
1775 (compare_si_ashiftsi): Likewise.
1776 (xbfu_cmp0_noout): New combine pattern.
1777 (xbfu_cmp0"): Likewise.
1778 (movsi_set_cc_insn): Place the predicable variant first.
1779 (commutative_binary_cmp0_noout): Remove clobber.
1780 (commutative_binary_cmp0): New pattern.
1781 (noncommutative_binary_cmp0): Likewise.
1782 (noncommutative_binary_cmp0_noout): Likewise.
1783 (noncommutative_binary_comparison_result_used): Removed.
1784 (rsub_cmp0): New pattern.
1785 (rsub_cmp0_noout): Likewise.
1786 (extzvsi): Changed, keep only meaningful variants.
1787 (SQH, SEZ): New iterators.
1788 (SQH_postfix): New mode attribute.
1789 (SEZ_prefix): New code attribute.
1790 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
1791 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
1792 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
1794 (noncommutative_operator): Check the availability of barrel
1797 2018-04-09 Richard Biener <rguenther@suse.de>
1799 PR tree-optimization/85284
1800 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
1801 Only use the niter constraining form of simple_iv when the exit
1804 2018-04-09 Tom de Vries <tom@codesourcery.com>
1807 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
1808 (define_expand "*memory_barrier"): New define_expand.
1809 (define_insn "memory_barrier"): New insn.
1811 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1813 PR rtl-optimization/80463
1814 PR rtl-optimization/83972
1815 PR rtl-optimization/83480
1817 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
1818 correct producer for the insn.
1819 (tidy_control_flow): Fixup seqnos in case of debug insns.
1821 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1823 PR rtl-optimization/83913
1825 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
1826 different sched-times when merging exprs.
1828 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1830 PR rtl-optimization/83962
1832 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
1833 tidy_fallthru_edge and tidy_control_flow.
1835 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1837 PR rtl-optimization/83530
1839 * sel-sched.c (force_next_insn): New global variable.
1840 (remove_insn_for_debug): When force_next_insn is true, also leave only
1841 next insn in the ready list.
1842 (sel_sched_region): When the region wasn't scheduled, make another pass
1843 over it with force_next_insn set to 1.
1845 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
1847 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
1849 * config/nds32/constants.md (unspec_volatile_element): Add enum values
1850 for interrupt control.
1851 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
1852 functions for interrupt control.
1853 * config/nds32/nds32-intrinsic.md: Likewise.
1854 * config/nds32/nds32_intrinsic.h: Likewise.
1855 * config/nds32/nds32.h (nds32_builtins): Likewise.
1857 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
1859 * config/nds32/nds32.c (nds32_init_machine_status,
1860 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
1861 strict_aligned_p field.
1862 (nds32_expand_to_rtl_hook): New function.
1863 (TARGET_EXPAND_TO_RTL_HOOK): Define.
1864 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
1866 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1867 Chung-Ju Wu <jasonwucj@gmail.com>
1869 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
1870 * config/nds32/nds32-n7.md: New file.
1871 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
1872 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
1874 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
1875 * config/nds32/nds32.md (pipeline_model): Add n7.
1876 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
1877 * config/nds32/pipelines.md: Include n7 settings.
1879 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1880 Chung-Ju Wu <jasonwucj@gmail.com>
1882 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
1883 * config/nds32/nds32-e8.md: New file.
1884 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
1885 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
1887 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
1888 * config/nds32/nds32.md (pipeline_model): Add e8.
1889 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
1890 * config/nds32/pipelines.md: Include e8 settings.
1892 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1893 Chung-Ju Wu <jasonwucj@gmail.com>
1895 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
1896 * config/nds32/nds32-n8.md: New file.
1897 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
1898 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
1900 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
1901 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
1902 * config/nds32/nds32.md (pipeline_model): Add n8.
1903 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
1904 * config/nds32/pipelines.md: Include n8 settings.
1906 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1907 Chung-Ju Wu <jasonwucj@gmail.com>
1909 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
1910 * config/nds32/nds32-n9-2r1w.md: New file.
1911 * config/nds32/nds32-n9-3r2w.md: New file.
1912 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
1913 nds32_register_ports): New or modify for cpu n9.
1914 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
1916 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
1917 * config/nds32/nds32-utils.c: New file.
1918 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
1919 TARGET_MUL_SLOW): Define.
1920 * config/nds32/nds32.md (pipeline_model): New attribute.
1921 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
1922 New options that support cpu n9.
1923 * config/nds32/pipelines.md: Include n9 settings.
1924 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
1926 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
1928 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
1929 information if necessary.
1930 (output_cond_branch_compare_zero): Likewise.
1931 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
1932 (nds32_target_alignment): Refine for alignment.
1933 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
1934 (FUNCTION_BOUNDARY): Modify.
1935 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
1937 * config/nds32/nds32.opt (malways-align, malign-functions): New.
1939 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
1941 * config/nds32/constants.md (unspec_volatile_element): Add values for
1942 TLB operation and data prefetch.
1943 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
1944 functions for TLB operation and data prefetch.
1945 * config/nds32/nds32-intrinsic.md: Likewise.
1946 * config/nds32/nds32_intrinsic.h: Likewise.
1947 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
1948 (nds32_print_operand): Likewise.
1949 * config/nds32/nds32.h (nds32_builtins): Likewise.
1951 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
1952 Andrew Pinski <pinsika@gcc.gnu.org>
1955 * match.pd: Use constant_boolean_node of correct type instead of
1956 boolean_true_node or boolean_false_node for simplifying
1957 pointer comparisons to zero.
1959 2018-04-07 Jakub Jelinek <jakub@redhat.com>
1961 PR tree-optimization/80021
1962 * tree.c (verify_type_variant): Make error call in verify_variant_match
1963 translatable and remove final full stop.
1965 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1967 * config/nds32/constants.md (unspec_volatile_element): Add
1968 UNSPEC_VOLATILE_EH_RETURN.
1969 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
1970 nds32_output_stack_pop): Support dwarf exception handling process.
1971 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
1972 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
1973 exception handling process.
1974 (nds32_compute_stack_frame): Likewise.
1975 (nds32_return_addr_rtx): Likewise.
1976 (nds32_initial_elimination_offset): Likewise.
1977 (nds32_expand_prologue): Likewise.
1978 (nds32_expand_epilogue): Likewise.
1979 (nds32_dynamic_chain_address): New function.
1980 * config/nds32/nds32.h (machine_function): Add fields for dwarf
1982 (DYNAMIC_CHAIN_ADDRESS): Define.
1983 (EH_RETURN_DATA_REGNO): Define.
1984 (EH_RETURN_STACKADJ_RTX): Define.
1985 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
1986 patterns for dwarf exception handling.
1988 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1990 * config/nds32/nds32.h: Clean up obsolete macros.
1992 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1994 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1995 Add enum values for particular instructions.
1996 * config/nds32/nds32-intrinsic.c: Implementation of expanding
1997 particular intrinsic functions.
1998 * config/nds32/nds32-intrinsic.md: Likewise.
1999 * config/nds32/nds32_intrinsic.h: Likewise.
2000 * config/nds32/nds32.h (nds32_builtins): Likewise.
2001 * config/nds32/nds32.md (type): Add pbsad and pbsada.
2002 (btst, ave): New patterns for particular instructions.
2004 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2006 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2007 Add enum values for atomic load/store and memory sync.
2008 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
2010 * config/nds32/nds32-intrinsic.md: Likewise.
2011 * config/nds32/nds32_intrinsic.h: Likewise.
2012 * config/nds32/nds32.h (nds32_builtins): Likewise.
2014 2018-04-07 Jakub Jelinek <jakub@redhat.com>
2016 PR tree-optimization/85257
2017 * fold-const.c (native_encode_vector): If not all elts could fit
2018 and off is -1, return 0 rather than offset.
2019 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
2020 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
2021 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
2022 adjust buffer in native_interpret_expr call.
2024 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2026 * config/nds32/constants.md (unspec_volatile_element): Add cache
2027 control enum values.
2028 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
2029 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
2030 * config/nds32/nds32.c (nds32_cctl_names): New.
2031 (nds32_print_operand): Handle cache control register names.
2032 * config/nds32/nds32.h (nds32_builtins): New enum values.
2033 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
2035 * config/nds32/nds32.md (type): Add mmu.
2036 * config/nds32/pipelines.md (simple_insn): Add mmu.
2038 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2040 * config/nds32/nds32.md (type): Remove call.
2041 * config/nds32/pipelines.md (simple_insn): Likewise.
2043 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2045 * config/nds32/constants.md (unspec_volatile_element): Add
2046 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
2047 UNSPEC_VOLATILE_FMFCFG.
2048 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
2049 description for fmfcfg and fmfcsr.
2050 (bdesc_1arg): Add fmtcsr.
2051 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
2052 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
2053 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
2054 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
2055 unspec_fmfcfg): New patterns.
2056 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
2057 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
2058 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
2059 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
2060 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
2061 __nds32__fmfcfg): Define.
2063 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2065 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
2066 intrinsic register names.
2067 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
2068 intrinsic register enum values and macros.
2070 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2072 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
2073 for load/store addressing form.
2074 (nds32_print_operand_address): Likewise.
2076 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
2079 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
2080 based on LABEL_REF. Remove useless assertion.
2081 (pic_address_needs_scratch): Fix formatting.
2082 (sparc_legitimize_pic_address): Minor tweaks.
2083 (sparc_delegitimize_address): Adjust assertion accordingly.
2084 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
2085 into symbolic_operand.
2086 (movsi_high_pic_label_ref): Likewise.
2087 (movsi_lo_sum_pic_label_ref): Likewise.
2088 (movdi_pic_label_ref): Likewise.
2089 (movdi_high_pic_label_ref): Likewise.
2090 (movdi_lo_sum_pic_label_ref): Likewise.
2092 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
2094 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
2095 custom LIB_SPEC setup.
2097 2018-04-06 Ruslan Bukin <br@bsdpad.com>
2098 Kito Cheng <kito.cheng@gmail.com>
2100 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
2101 * config/riscv/freebsd.h: New.
2103 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
2105 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
2106 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
2109 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
2110 Kito Cheng <kito.cheng@gmail.com>
2112 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
2113 nds32_output_call, nds32_symbol_binds_local_p): New functions.
2114 * config/nds32/nds32-protos.h (nds32_output_call,
2115 nds32_output_return): Declare.
2116 * config/nds32/nds32.md: Refine all the call and return patterns.
2118 2018-04-06 Jakub Jelinek <jakub@redhat.com>
2121 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
2122 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
2124 PR rtl-optimization/84872
2125 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
2126 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
2129 2018-04-06 Tamar Christina <tamar.christina@arm.com>
2131 * expr.c (copy_blkmode_to_reg): Revert 254862.
2132 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
2134 2018-04-06 Richard Biener <rguenther@suse.de>
2137 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
2138 after seeing a component reference with an adjacent field. Treat
2139 refs to arrays at struct end of external decls similar to
2140 refs to unconstrained commons.
2142 2018-04-06 Jakub Jelinek <jakub@redhat.com>
2145 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
2146 look through SAVE_EXPRs with non-side-effects argument. Adjust
2148 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
2151 2018-04-06 Richard Biener <rguenther@suse.de>
2154 * alias.c (find_base_term): New wrapper around find_base_term
2155 unwinding CSELIB_VAL_PTR changes.
2156 (find_base_term): Do not restore CSELIB_VAL_PTR during the
2159 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2161 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
2163 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
2164 constant definitions.
2165 ("nop"): lr 0,0 -> nopr r0
2166 ("nop_lr0", "nop_lr1"): New insn definitions.
2168 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
2170 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
2171 NDS32_V3PUSH_AVAILABLE_P macro.
2173 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
2174 Chung-Ju Wu <jasonwucj@gmail.com>
2176 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
2177 (nds32*-*-*): Add float and fpu_config into supported_defaults.
2178 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
2179 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
2180 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
2181 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
2182 * config/nds32/constraints.md: New constraints and checking for hard
2183 float configuration.
2184 * config/nds32/iterators.md: New mode iterator and attribute for hard
2185 float configuration.
2186 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
2188 * config/nds32/nds32-fpu.md: New file.
2189 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
2190 deal with hard float code generation.
2191 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
2193 (abi_type, float_reg_number): New enum type.
2194 * config/nds32/nds32-predicates.c: New predicates for hard float.
2195 * config/nds32/nds32-protos.h: Declare functions for hard float.
2196 * config/nds32/nds32.c: Implementation for hard float configuration.
2197 * config/nds32/nds32.h: Definitions for hard float configuration.
2198 * config/nds32/nds32.md: Include hard float machine description and
2199 modify patterns for hard float configuration.
2200 * config/nds32/nds32.opt: New options for hard float configuration.
2201 * config/nds32/predicates.md: New predicates for hard float
2204 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
2206 * common/config/nds32/nds32-common.c
2207 (nds32_option_optimization_table): Enable -mreleax-hint by default.
2209 2018-04-05 Jakub Jelinek <jakub@redhat.com>
2212 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
2213 CONSTRUCTOR_ELT (ctor, ...)->value.
2215 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
2218 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
2220 2018-04-05 Tom de Vries <tom@codesourcery.com>
2223 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
2226 2018-04-05 Shiva Chen <shiva0217@gmail.com>
2227 Kito Cheng <kito.cheng@gmail.com>
2229 * config/nds32/constraints.md (U33): Fine-tune checking condition.
2230 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
2231 * config/nds32/nds32.h (nds32_16bit_address_type): Add
2232 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
2234 2018-04-05 Shiva Chen <shiva0217@gmail.com>
2235 Kito Cheng <kito.cheng@gmail.com>
2237 * config/nds32/constraints.md (Ufe): New memory constraint.
2238 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
2239 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
2240 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
2242 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
2243 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
2245 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2247 * config/nds32/nds32.md: Use optimize_size in the condition for
2248 alu-shift instructions.
2250 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2252 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
2254 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2256 * config/nds32/nds32.md (negsi2): Refine pattern.
2258 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
2259 Chung-Ju Wu <jasonwucj@gmail.com>
2261 * config/nds32/iterators.md (shift_rotate): New code iterator.
2262 (shift): New code attribute.
2263 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
2264 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
2265 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
2266 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
2267 bit-wise operations.
2268 (andsi3, *andsi3): Ditto.
2269 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
2270 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
2271 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
2272 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
2273 nds32_ior_operand, nds32_xor_operand): New predicates.
2275 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2277 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
2278 (addsi3, subsi3): ... this.
2280 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2282 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
2284 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2286 * config/nds32/nds32.md: Adjust indention.
2288 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
2290 * config/nds32/nds32.md (feature): New attribute.
2292 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2294 * config/nds32/nds32.md (subtype): New attribute.
2296 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
2299 * config/arm/arm-builtins.c (arm_expand_builtin): Change
2300 expansion to perform a bitwise AND of the argument followed by a
2301 boolean negation of the result.
2303 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
2305 PR rtl-optimization/84878
2306 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
2307 the basic block. Assert the use reference is not artificial and that
2308 it has an associated insn.
2310 2018-04-04 Michael Matz <matz@suse.de>
2312 * builtins.c (compute_objsize): Pass correct operand
2313 to array_at_struct_end_p.
2315 2018-04-04 Richard Biener <rguenther@suse.de>
2318 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
2319 from contexts for DINFO_LEVEL_TERSE and below.
2321 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
2323 * config/nds32/nds32-doubleword.md (move_<mode>): Require
2324 resiter_operand condition.
2325 * config/nds32/nds32.md (*move<mode>): Ditto.
2327 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
2328 Monk Chiang <sh.chiang04@gmail.com>
2330 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
2332 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2334 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
2336 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2337 Kito Cheng <kito.cheng@gmail.com>
2339 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
2340 nds32_cond_code_str, output_cond_branch,
2341 output_cond_branch_compare_zero, nds32_expand_cbranch,
2342 nds32_expand_cstore, nds32_expand_movcc,
2343 nds32_output_cbranchsi4_equality_zero,
2344 nds32_output_cbranchsi4_equality_reg,
2345 nds32_output_cbranchsi4_equality_reg_or_const_int,
2346 nds32_output_cbranchsi4_greater_less_zero: New functions.
2347 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
2348 nds32_expand_cstore, nds32_expand_movcc,
2349 nds32_output_cbranchsi4_equality_zero,
2350 nds32_output_cbranchsi4_equality_reg,
2351 nds32_output_cbranchsi4_equality_reg_or_const_int,
2352 nds32_output_cbranchsi4_greater_less_zero): Declare.
2353 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
2354 nds32_rimm11s_operand): New predicates.
2355 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
2356 * config/nds32/nds32.md: Rewrite all the branch and conditional move
2359 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
2361 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
2362 * config/nds32/nds32.md: Ditto.
2363 * config/nds32/pipelines.md: Ditto.
2365 2018-04-04 Richard Biener <rguenther@suse.de>
2367 PR tree-optimization/85168
2368 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
2369 propagating abnormals.
2371 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2373 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
2375 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2376 Kito Cheng <kito.cheng@gmail.com>
2378 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
2379 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
2380 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
2381 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
2382 * config/nds32/nds32.md (sibcall_internal): New.
2383 (sibcall_register): Remove.
2384 (sibcall_immediate): Remove.
2385 (sibcall_value_internal): New.
2386 (sibcall_value_register): Remove.
2387 (sibcall_value_immediate): Remove.
2388 * config/nds32/predicates.md (nds32_general_register_operand): New.
2389 (nds32_call_address_operand): New.
2391 2018-04-03 Jakub Jelinek <jakub@redhat.com>
2393 PR rtl-optimization/85167
2394 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
2395 bb_defs if *split_p, instead preinitialize it to NULL.
2397 PR tree-optimization/85156
2398 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
2399 evaluating the argument multiple times.
2401 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
2403 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
2405 (_mm_cvtpd_ps): Likewise.
2406 (_mm_cvttpd_epi32): Likewise.
2407 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
2408 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
2409 vector, pixel, and bool following altivec.h include.
2411 2018-04-03 Martin Sebor <msebor@redhat.com>
2413 * doc/extend.texi (Common Function Attributes): Clarify.
2414 (const attribute): Likewise.
2415 (pure attribute): Likewise.
2417 2018-04-03 Jakub Jelinek <jakub@redhat.com>
2420 * config/i386/i386.c (ix86_expand_vector_set): Use
2421 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
2423 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
2425 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
2426 instructions when changing rounding bits to preserve precision bits
2427 in the x87 control word.
2429 2018-04-03 Martin Liska <mliska@suse.cz>
2431 PR tree-optimization/82491
2432 * rtl.h (strip_offset_and_add): Replace += suboffset with
2433 poly_uint64 () + suboffset.
2435 2018-03-29 Martin Liska <mliska@suse.cz>
2436 Martin Jambor <mjambor@suse.cz>
2439 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
2440 param_type is not an integral or pointer type.
2442 2018-04-03 Richard Biener <rguenther@suse.de>
2444 * sese.h (recompute_all_dominators): Remove.
2446 2018-04-02 Martin Sebor <msebor@redhat.com>
2448 * doc/invoke.texi (-Wrestrict): Fix typos.
2450 2018-04-02 Jim Wilson <jimw@sifive.com>
2452 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
2453 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
2454 (<optab>di3, <optab>si3_extend): Likewise.
2455 (<optab>si3_mask, <optab>si3_mask_1): New.
2456 (<optab>di3_mask, <optab>di3_mask_1): New.
2457 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
2458 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
2459 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
2461 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
2463 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
2466 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
2468 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
2469 (nds32_canonicalize_comparison): New function.
2471 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
2472 Kito Cheng <kito.cheng@gmail.com>
2473 Kuan-Lin Chen <kuanlinchentw@gmail.com>
2475 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
2476 * config/nds32/constants.md (unspec_volatile_element): Add
2477 UNSPEC_VOLATILE_RELAX_GROUP.
2478 * config/nds32/nds32-relax-opt.c: New file.
2479 * config/nds32/nds32-predicates.c
2480 (nds32_symbol_load_store_p): New function.
2481 * config/nds32/nds32-protos.h
2482 (nds32_symbol_load_store_p): Declare function.
2483 (make_pass_nds32_relax_opt): Declare new rtl pass function.
2484 * config/nds32/nds32.c
2485 (nds32_register_pass): New function to register pass.
2486 (nds32_register_passes): New function to register passes.
2487 * config/nds32/nds32.md (relax_group): New pattern.
2488 * config/nds32/nds32.opt (mrelax-hint): New option.
2489 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
2491 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
2493 * config/nds32/t-nds32: Modify files dependency.
2495 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
2497 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
2498 (PROFILE_HOOK): Define its implementation.
2500 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
2502 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
2503 type and 32-bit size.
2505 2018-04-01 Jakub Jelinek <jakub@redhat.com>
2508 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
2509 (V_128_256): New mode iterator.
2510 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
2511 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
2512 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
2514 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
2515 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
2517 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
2520 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
2521 NaN inputs correctly.
2523 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
2526 * config/rs6000/vsx.md (??r): New mode attribute.
2527 (*vsx_mov<mode>_64bit): Use it.
2528 (*vsx_mov<mode>_32bit): Likewise.
2530 2018-03-30 Martin Sebor <msebor@redhat.com>
2532 PR tree-optimization/84818
2533 * builtins.c (check_access): Use warning_n.
2535 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2538 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
2540 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
2543 2018-03-30 Julia Koval <julia.koval@intel.com>
2546 * x86-tune.def (movx, partial_reg_dependency): Enable for
2549 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
2552 * lra-constraints.c (process_alt_operands): Move setting
2553 this_alternative_matches below.
2555 2018-03-29 Martin Liska <mliska@suse.cz>
2558 * doc/invoke.texi: Document how LTO works with debug info.
2559 Describe auto-load support of binutils. Mention 'x86-64'
2560 as valid option value of -march option.
2562 2018-03-29 Jakub Jelinek <jakub@redhat.com>
2564 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
2567 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
2568 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
2569 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
2572 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
2575 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
2576 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
2577 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
2578 for RS6000_BTM_POWERPC64.
2579 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
2580 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
2581 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
2586 2018-03-28 Carl Love <cel@us.ibm.com>
2589 2017-09-27 Carl Love <cel@us.ibm.com>
2591 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
2592 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
2593 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
2596 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2598 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
2599 instead of __vector bool.
2600 (_mm_max_pu8): Likewise.
2601 (_mm_min_pi16): Likewise.
2603 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
2606 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
2607 (DIVWEUO): Likewise.
2609 (DIVDEUO): Likewise.
2610 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
2611 DIVWEUO and DIVDEUO.
2612 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
2613 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
2614 (div_extend): Likewise.
2615 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
2617 (__builtin_divweuo): Likewise.
2618 (__builtin_divdeo): Likewise.
2619 (__builtin_divdeuo): Likewise.
2621 2018-03-28 Jakub Jelinek <jakub@redhat.com>
2624 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
2625 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
2627 PR tree-optimization/82004
2628 * gimple-match-head.c (optimize_pow_to_exp): New function.
2629 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
2630 Don't fold to exp if optimize_pow_to_exp is false.
2632 2018-03-28 Martin Liska <mliska@suse.cz>
2635 * calls.c (initialize_argument_information): Fix trailing space.
2636 * common.opt: Fix typo and provide better explanation for
2637 -fsanitize-coverage option.
2638 * config/i386/i386.opt: Fix typo.
2640 2018-03-28 Jakub Jelinek <jakub@redhat.com>
2641 Martin Liska <mliska@suse.cz>
2644 * gimplify.c (asan_poison_variable): Don't do the check for
2645 gimplify_omp_ctxp here.
2646 (gimplify_decl_expr): Do it here.
2647 (gimplify_target_expr): Likewise.
2649 2018-03-28 Martin Liska <mliska@suse.cz>
2652 * config/i386/i386.c (ix86_function_arg_advance): Do not call
2653 chkp_type_bounds_count if MPX is not enabled.
2655 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
2657 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
2659 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
2662 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
2663 function to create the function decl for complex long double
2664 multiply and divide for -mabi=ieeelongdouble.
2665 (init_float128_ieee): Call it.
2667 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
2670 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
2671 -fcf-protection=branch -mibt.
2672 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
2674 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2677 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
2679 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
2682 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
2683 extern array declarations.
2685 2018-03-27 Richard Biener <rguenther@suse.de>
2688 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
2689 explicit single_use checks.
2691 2018-03-27 Richard Biener <rguenther@suse.de>
2693 PR tree-optimization/85082
2694 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
2697 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2699 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
2700 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
2701 Turn on fasynchronous-unwind-tables and funwind-tables.
2703 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
2706 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
2707 (*bmi_blsr_<mode>_ccz): Ditto.
2709 2018-03-26 Tom de Vries <tom@codesourcery.com>
2711 PR tree-optimization/85063
2712 * omp-general.c (offloading_function_p): New function. Factor out
2714 * omp-offload.c (pass_omp_target_link::gate): ... here.
2715 * omp-general.h (offloading_function_p): Declare.
2716 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
2717 with attribute omp declare target for offloading functions.
2719 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
2721 PR tree-optimization/84005
2722 * tree-data-ref.h (get_base_for_alignment): Declare.
2723 * tree-data-ref.c (get_base_for_alignment_1): New function.
2724 (get_base_for_alignment): Likewise.
2725 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
2726 get_base_for_alignment to find a suitable base object, instead
2727 of always using drb->base_address.
2729 2018-03-23 Jakub Jelinek <jakub@redhat.com>
2732 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
2733 known size by default.
2735 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
2738 * lra-constraints.c (process_alt_operands): Don't match BLKmode
2739 and non BLKmode operands.
2741 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2744 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
2745 Clean up attributes.
2747 2018-03-23 Richard Biener <rguenther@suse.de>
2750 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
2751 we are going to emit early debug for LTO.
2753 2018-03-23 Jakub Jelinek <jakub@redhat.com>
2756 * function.c (match_asm_constraints_1): Don't optimize if input
2757 doesn't satisfy general_operand predicate for output's mode.
2760 * alias.c (write_dependence_p): Don't require for x_canonicalized
2761 non-VOIDmode if x has VOIDmode.
2764 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
2765 just don't try to optimize it rather than assert it never happens.
2767 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
2769 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
2770 macro expansions for definition of ST_INTERNAL_<mode> and
2771 LD_INTERNAL_<mode> builtins.
2772 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
2774 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
2776 (altivec_expand_st_builtin): Likewise.
2777 (altivec_expand_builtin): Remove calls to deleted functions.
2778 (rs6000_address_for_altivec): Delete this function.
2779 * config/rs6000/vector.md: Remove expands for
2780 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
2782 2018-03-22 Sudakshina Das <sudi.das@arm.com>
2785 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
2786 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
2787 re-computing once computed.
2788 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
2789 (arm_init_machine_status): Initialize
2790 machine->static_chain_stack_bytes.
2792 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
2795 * doc/extend.texi: Add four new prototypes for vec_ld.
2796 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
2797 definitions for more logical presentation.
2798 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
2799 entries for V1TI variants of __builtin_altivec_ld builtin.
2800 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
2801 handling of V1TI variant of LVX icode pattern.
2802 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
2803 (rs6000_gimple_fold_builtin): Likewise.
2804 (altivec_init_builtins): Add code to define
2805 __builtin_altivec_lvx_v1ti function.
2807 2018-03-22 Jakub Jelinek <jakub@redhat.com>
2810 * function.c (match_asm_constraints_1): Don't do the optimization
2811 if input isn't a REG, SUBREG, MEM or constant.
2813 2018-03-22 Tom de Vries <tom@codesourcery.com>
2815 PR tree-optimization/84956
2816 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
2817 bb_has_abnormal_pred.
2819 2018-03-22 Jakub Jelinek <jakub@redhat.com>
2822 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
2823 DECL_INITIAL (decl) to decl at the end.
2824 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
2827 2018-03-21 Joseph Myers <joseph@codesourcery.com>
2829 * doc/extend.texi (__builtin_tgmath): Document when complex
2830 integer types are treated as _Complex _Float64.
2832 2018-03-21 Tom de Vries <tom@codesourcery.com>
2834 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
2836 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2838 PR tree-optimization/84960
2839 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
2840 if it is ENTRY block, move them into single succ of ENTRY in that case.
2842 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
2844 PR tree-optimization/84811
2845 * poly-int.h (poly_span_traits): Remove the T3 parameter and
2846 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
2847 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
2848 (known_subrange_p): Update accordingly. Cast each value involved
2849 in the size comparison, rather than casting the result of the
2852 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2854 PR tree-optimization/84982
2855 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
2856 by flipping the least significant bit rather than all bits from
2857 bitpos to bitpos + bitsize - 1.
2859 2018-03-21 Nathan Sidwell <nathan@acm.org>
2861 * doc/extend.texi (Deprecated Features): Remove mention of
2862 long-deleted deprecations.
2864 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2867 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
2868 * configure: Regenerate.
2870 2018-03-21 Tom de Vries <tom@codesourcery.com>
2872 PR tree-optimization/83126
2873 * tree-parloops.c (num_phis): New function.
2874 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
2876 2018-03-21 Nathan Sidwell <nathan@acm.org>
2878 * doc/extend.texi (Deprecated Features): Update deprecated flags,
2879 mention anon-struct/union members and trailing attributes.
2881 2018-03-21 Bin Cheng <bin.cheng@arm.com>
2883 PR tree-optimization/84969
2884 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
2885 builtin memset partitions if they set different rhs values.
2887 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2889 PR rtl-optimization/84989
2890 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
2891 VEC_DUPLICATE with scalar result mode.
2893 2018-03-21 Martin Liska <mliska@suse.cz>
2896 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
2897 not intended return statement.
2899 2018-03-21 Martin Liska <mliska@suse.cz>
2902 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
2903 (chkp_find_bound_slots_1): Limit number of iterations.
2905 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
2908 * Minor grammar fixes for x86 options.
2910 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2913 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
2914 holding REG_CFA_RESTORE notes, instead turn them into a USE.
2916 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
2919 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
2920 (altivec_lvx_<mode>_1op): Likewise.
2921 (altivec_stvx_<mode>_2op): Likewise.
2922 (altivec_stvx_<mode>_1op): Likewise.
2923 (altivec_lvx_<VM2:mode>): New define_expand.
2924 (altivec_stvx_<VM2:mode>): Likewise.
2925 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
2926 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
2927 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
2928 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
2929 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
2930 (rs6000_gen_lvx): Likewise.
2931 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
2932 (altivec_expand_stv_builtin): Likewise.
2933 (altivec_expand_builtin): Likewise.
2934 * config/rs6000/vector.md: Likewise.
2936 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2939 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
2942 2018-03-20 Richard Biener <rguenther@suse.de>
2945 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
2946 sign-conversions as zero, fall back to standard scalar_stmt
2949 2018-03-20 Martin Liska <mliska@suse.cz>
2952 * predict.c (rebuild_frequencies): Handle case when we have
2953 PROFILE_ABSENT, but flag_guess_branch_prob is false.
2955 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2958 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
2959 flag_section_anchors.
2960 * varasm.c (use_blocks_for_decl_p): Remove hack for
2961 dw2_force_const_mem.
2964 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
2966 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
2967 be created, use lowpart_subreg of operands[0] rather than operands[0]
2969 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
2970 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
2971 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
2972 and n constraint instead of aarch64_shift_imm_di and Usd.
2973 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
2974 (*aarch64_<optab>_reg_minus<mode>3): ... this.
2976 2018-03-20 Sudakshina Das <sudi.das@arm.com>
2979 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
2980 to favor GPR over NEON registers.
2981 (<shift>di3_neon): Likewise.
2983 2018-03-20 Tom de Vries <tom@codesourcery.com>
2986 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
2987 (nvptx_process_pars): Emit bar.sync asap and alap.
2989 2018-03-20 Tom de Vries <tom@codesourcery.com>
2992 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
2993 seen_label if seen_label is already set.
2995 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2998 * config/i386/i386.c (fold_builtin_cpu): For features above 31
2999 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
3000 Use 1U instead of 1. Formatting fixes.
3003 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
3004 instead of TREE_TYPE (s1) for the return value.
3006 2018-03-19 Jakub Jelinek <jakub@redhat.com>
3008 PR tree-optimization/84946
3009 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
3010 bitsize + bitsize in poly_uint64 rather than poly_int64.
3013 * dwarf2asm.c: Include fold-const.c.
3014 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
3015 of decl rather than decl itself.
3017 PR rtl-optimization/84643
3018 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
3020 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
3023 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
3024 calling assemble_variable.
3026 2018-03-19 Sudakshina Das <sudi.das@arm.com>
3029 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
3030 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
3032 2018-03-19 Jim Wilson <jimw@sifive.com>
3035 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
3036 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
3037 (riscv_first_stack_step): Likewise.
3038 (riscv_option_override): Use STACK_BOUNDARY instead of
3040 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
3042 (BIGGEST_ALIGNMENT): Set to 128.
3043 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
3044 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
3047 2018-03-19 Richard Biener <rguenther@suse.de>
3049 PR tree-optimization/84933
3050 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
3051 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
3053 2018-03-19 Richard Biener <rguenther@suse.de>
3055 PR tree-optimization/84859
3056 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
3057 (cond_if_else_store_replacement): Perform sinking operation on
3058 single-store BBs regardless of MAX_STORES_TO_SINK setting.
3059 Generalize what a BB with a single eligible store is.
3061 2018-03-19 Richard Biener <rguenther@suse.de>
3063 PR tree-optimization/84929
3064 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
3065 chrec_is_positive against non-chrec arg.
3067 2018-03-19 Tamar Christina <tamar.christina@arm.com>
3070 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
3072 2018-03-18 Martin Liska <mliska@suse.cz>
3074 PR rtl-optimization/84635
3075 * regrename.c (build_def_use): Use matches_mode only when
3078 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
3080 PR tree-optimization/84913
3081 * tree-vect-loop.c (vectorizable_reduction): Don't try to
3082 vectorize chains of COND_EXPRs.
3084 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3086 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
3088 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3090 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
3092 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3094 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
3096 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
3097 Kito Cheng <kito.cheng@gmail.com>
3099 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
3100 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
3101 (nds32_adjust_reg_alloc_order): New function.
3102 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
3104 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
3106 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
3107 nds32_print_operand, nds32_print_operand_address): Use
3108 HOST_WIDE_INT_PRINT_DEC instead.
3110 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
3112 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
3114 2018-03-17 Jakub Jelinek <jakub@redhat.com>
3117 * config/i386/i386.c (initial_ix86_tune_features,
3118 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
3120 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
3121 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
3122 rather than 1u << ix86_tune. Formatting fix.
3123 (ix86_option_override_internal): Change ix86_arch_mask from
3124 unsigned int to unsigned HOST_WIDE_INT, initialize to
3125 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
3126 (ix86_function_specific_restore): Likewise.
3128 2018-03-16 Jakub Jelinek <jakub@redhat.com>
3131 * postreload.c (reload_combine_recognize_pattern): Perform
3132 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
3133 truncate_int_for_mode the result for the destination's mode.
3136 * hsa-gen.c (mem_type_for_type): Fix comment typo.
3137 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
3139 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
3142 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
3145 * lra-assigns.c (lra_split_hard_reg_for): Don't use
3146 regno_allocno_class_array and sorted_pseudos.
3147 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
3148 insns where regno is used.
3150 2018-03-16 Martin Liska <mliska@suse.cz>
3153 * multiple_target.c (create_dispatcher_calls): Redirect
3154 reference in the symbol table.
3156 2018-03-16 Martin Liska <mliska@suse.cz>
3159 * multiple_target.c (create_dispatcher_calls): Redirect also
3162 2018-03-16 Jakub Jelinek <jakub@redhat.com>
3166 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
3167 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
3168 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
3170 2018-03-16 Julia Koval <julia.koval@intel.com>
3172 * doc/invoke.texi (Skylake Server): Add CLWB.
3173 Cannonlake): Remove CLWB.
3175 2018-03-16 Jakub Jelinek <jakub@redhat.com>
3177 PR tree-optimization/84841
3178 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
3180 (FLOAT_ONE_CONST_TYPE): Define.
3181 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
3182 (sort_by_operand_rank): Put entries with higher constant_type last
3183 rather than first to match comments.
3185 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
3187 * config/nios2/nios2.md (movsi_internal): Fix thinko in
3190 2018-03-15 Jakub Jelinek <jakub@redhat.com>
3193 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
3194 check and use address of target always.
3196 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
3199 * config/i386/i386.c (indirect_thunk_needed): Update comments.
3200 (indirect_thunk_bnd_needed): Likewise.
3201 (indirect_thunks_used): Likewise.
3202 (indirect_thunks_bnd_used): Likewise.
3203 (indirect_return_needed): New.
3204 (indirect_return_bnd_needed): Likewise.
3205 (output_indirect_thunk_function): Add a bool argument for
3207 (output_indirect_thunk_function): Don't generate alias for
3208 function return thunk.
3209 (ix86_code_end): Call output_indirect_thunk_function to generate
3210 function return thunks.
3211 (ix86_output_function_return): Set indirect_return_bnd_needed
3212 and indirect_return_needed instead of indirect_thunk_bnd_needed
3213 and indirect_thunk_needed.
3215 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
3217 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
3218 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
3219 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
3221 2018-03-15 David Malcolm <dmalcolm@redhat.com>
3222 Paul Hua <paul.hua.gm@gmail.com>
3225 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
3227 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
3229 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
3230 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
3233 2018-03-15 Tamar Christina <tamar.christina@arm.com>
3236 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
3237 instead of GET_MODE_SIZE when comparing Units.
3239 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
3242 * varasm.c (hash_section): Return an unchangeble hash value
3243 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
3244 Return !aarch64_can_use_per_function_literal_pools_p ().
3246 2018-03-15 Jakub Jelinek <jakub@redhat.com>
3249 * optabs.c (emit_conditional_move): Pass address of cmode's copy
3250 rather than address of cmode as last argument to prepare_cmp_insn.
3252 2018-03-15 Julia Koval <julia.koval@intel.com>
3254 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
3255 F_AVX512VNNI, F_AVX512BITALG): New.
3257 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
3260 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
3261 insn for floating-point loads and stores.
3263 2018-03-14 Carl Love <cel@us.ibm.com>
3265 * config/rs6000/rs6000-c.c: Add macro definitions for
3266 ALTIVEC_BUILTIN_VEC_PERMXOR.
3267 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
3268 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
3269 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
3270 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
3272 * config/doc/extend.texi: Add prototypes for vec_permxor.
3274 2018-03-14 David Malcolm <dmalcolm@redhat.com>
3277 * diagnostic-show-locus.c (class layout_point): Convert m_line
3278 from int to linenum_type.
3279 (line_span::comparator): Use linenum "compare" function when
3280 comparing line numbers.
3281 (test_line_span): New function.
3282 (layout_range::contains_point): Convert param "row" from int to
3284 (layout_range::intersects_line_p): Likewise.
3285 (layout::will_show_line_p): Likewise.
3286 (layout::print_source_line): Likewise.
3287 (layout::should_print_annotation_line_p): Likewise.
3288 (layout::print_annotation_line): Likewise.
3289 (layout::print_leading_fixits): Likewise.
3290 (layout::annotation_line_showed_range_p): Likewise.
3291 (struct line_corrections): Likewise for field m_row.
3292 (line_corrections::line_corrections): Likewise for param "row".
3293 (layout::print_trailing_fixits): Likewise.
3294 (layout::get_state_at_point): Likewise.
3295 (layout::get_x_bound_for_row): Likewise.
3296 (layout::print_line): Likewise.
3297 (diagnostic_show_locus): Likewise for locals "last_line" and
3299 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
3300 * input.c (selftest::test_linenum_comparisons): New function.
3301 (selftest::input_c_tests): Call it.
3302 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
3303 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
3304 * selftest.h (ASSERT_GT): New macro.
3305 (ASSERT_GT_AT): New macro.
3306 (ASSERT_LT): New macro.
3307 (ASSERT_LT_AT): New macro.
3309 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
3311 PR rtl-optimization/84780
3312 * combine.c (distribute_links): Don't make a link based on pc_rtx.
3314 2018-03-14 Martin Liska <mliska@suse.cz>
3316 * tree.c (record_node_allocation_statistics): Use
3317 get_stats_node_kind.
3318 (get_stats_node_kind): New function extracted from
3319 record_node_allocation_statistics.
3320 (free_node): Use get_stats_node_kind.
3322 2018-03-14 Richard Biener <rguenther@suse.de>
3324 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
3325 that the value-set of ANTIC_IN doesn't grow.
3328 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
3330 (BB_VISITED_WITH_VISITED_SUCCS): New define.
3331 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
3333 2018-03-14 Julia Koval <julia.koval@intel.com>
3335 * config.gcc (icelake-client, icelake-server): New.
3337 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
3338 (initial_ix86_arch_features): Ditto.
3339 (PTA_SKYLAKE): Add SGX.
3340 (PTA_ICELAKE): Remove.
3341 (PTA_ICELAKE_CLIENT): New.
3342 (PTA_ICELAKE_SERVER): New.
3343 (ix86_option_override_internal): Split up icelake on icelake client and
3345 (get_builtin_code_for_version): Ditto.
3346 (fold_builtin_cpu): Ditto.
3347 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
3348 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
3349 * config/i386/i386.h (processor_type): Ditto.
3350 * doc/invoke.texi: Ditto.
3352 2018-03-14 Jakub Jelinek <jakub@redhat.com>
3355 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
3356 INTEGER_CST offset, add it together with bitpos / 8 and
3357 sign extend based on POINTER_SIZE.
3361 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
3364 * config/i386/constraints.md (Yc): New register constraint.
3365 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
3366 Use Yc constraint for alternative 2 of operand 0. Remove
3367 preferred_for_speed attribute.
3369 2018-03-14 Richard Biener <rguenther@suse.de>
3371 PR tree-optimization/84830
3372 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
3373 with the old one to avoid oscillations.
3375 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
3378 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
3380 (assign_by_spills): Return a flag of reload assignment failure.
3381 Do not process the reload assignment failures. Do not spill other
3382 reload pseudos if they has the same reg class. Update n if
3384 (lra_assign): Add a return arg. Set up from the result of
3385 assign_by_spills call.
3386 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
3387 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
3388 usage_insns if it is not NULL.
3389 (spill_hard_reg_in_range): New function.
3390 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
3391 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
3392 function prototypes.
3393 (lra_assign): Change prototype.
3394 * lra.c (lra): Add code to deal with fails by splitting hard reg
3397 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
3399 * config/riscv/riscv.opt (mrelax): New option.
3400 * config/riscv/riscv.c (riscv_file_start): Emit ".option
3401 "norelax" when riscv_mrelax is disabled.
3402 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
3404 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3407 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
3408 reassociation for int modes.
3410 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
3412 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
3413 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
3415 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
3416 * config/aarch64/aarch64-sve.md
3417 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
3418 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
3419 (*extend<mode><Vwide>2): Rename to...
3420 (aarch64_sve_extend<mode><Vwide>2): ...this.
3421 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
3422 renaming the old pattern to...
3423 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
3425 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
3426 define_expand, renaming the old pattern to...
3427 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
3428 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
3429 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
3430 account when deciding which SVE instruction the optab should use.
3431 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
3433 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
3435 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
3436 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
3437 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
3438 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
3439 (tlsdesc_small_<mode>): Turn a define_expand and use
3440 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
3441 (tlsdesc_small_advsimd_<mode>): ...this.
3442 (tlsdesc_small_sve_<mode>): New pattern.
3444 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
3446 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
3447 (UNSPEC_UMUL_HIGHPART): New constants.
3448 (MUL_HIGHPART): New int iteraor.
3449 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
3450 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
3452 (*<su>mul<mode>3_highpart): New define_insn.
3454 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
3457 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
3460 2018-03-13 Martin Liska <mliska@suse.cz>
3463 * (sem_item_optimizer::sem_item_optimizer): Initialize new
3465 (sem_item_optimizer::~sem_item_optimizer): Release it.
3466 (sem_item_optimizer::merge_classes): Register variable aliases.
3467 (sem_item_optimizer::fixup_pt_set): New function.
3468 (sem_item_optimizer::fixup_points_to_sets): Likewise.
3469 * ipa-icf.h: Declare new variables and functions.
3471 2018-03-13 Jakub Jelinek <jakub@redhat.com>
3474 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
3475 integer_pow2p@2 and test integer_pow2p in condition.
3476 (A < 0 ? C : 0): Similarly for @1.
3479 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
3480 characters starting at p contain '\0' character, don't look beyond
3484 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
3485 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
3488 * reg-stack.c (change_stack): Change update_end var from int to
3489 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
3490 also call set_block_for_insn on the newly added insns and rescan.
3493 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
3494 on the last operand.
3497 * tree.c (stabilize_reference_1): Return save_expr (e) for
3498 STATEMENT_LIST even if it doesn't have side-effects.
3500 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
3502 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
3504 2018-03-12 Renlin Li <renlin.li@arm.com>
3506 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
3507 aarch64_output_scalar_simd_mov_immediate.
3509 2018-03-12 Martin Sebor <msebor@redhat.com>
3511 PR tree-optimization/83456
3512 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
3513 for perfectly overlapping calls to memcpy.
3514 (gimple_fold_builtin_memory_chk): Same.
3515 (gimple_fold_builtin_strcpy): Handle no-warning.
3516 (gimple_fold_builtin_stxcpy_chk): Same.
3517 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
3519 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
3521 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
3522 parameter. Use it for SFmode.
3523 (rs6000_function_arg_advance_1): Adjust.
3524 (rs6000_function_arg): Adjust.
3525 (rs6000_gimplify_va_arg): Pass false for that new parameter.
3527 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
3529 PR rtl-optimization/84169
3530 PR rtl-optimization/84780
3531 * combine.c (can_combine_p): Check for a 2-insn combination whether
3532 the destination register is used between the two insns, too.
3534 2018-03-12 Richard Biener <rguenther@suse.de>
3536 PR tree-optimization/84803
3537 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
3538 for refs DR analysis didn't process.
3540 2018-03-12 Richard Biener <rguenther@suse.de>
3542 PR tree-optimization/84777
3543 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
3544 force-vectorize loops ignore whether we are optimizing for size.
3546 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
3548 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
3549 (TARGET_MD_ASM_ADJUST): Define.
3551 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
3552 Kito Cheng <kito.cheng@gmail.com>
3553 Chung-Ju Wu <jasonwucj@gmail.com>
3555 * config/nds32/nds32.c (nds32_compute_stack_frame,
3556 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
3557 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
3558 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
3559 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
3560 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
3561 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
3562 * config/nds32/nds32.md (prologue, epilogue): Use macro
3563 NDS32_V3PUSH_AVAILABLE_P to do checking.
3565 2018-03-11 Jakub Jelinek <jakub@redhat.com>
3568 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
3569 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
3570 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
3571 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
3572 addition of most attributes on !orig_type_die or the attribute not
3573 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
3575 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3576 Chung-Ju Wu <jasonwucj@gmail.com>
3578 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
3580 * config/nds32/nds32.opt (mvh): New option.
3582 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3583 Chung-Ju Wu <jasonwucj@gmail.com>
3585 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
3587 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
3588 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
3591 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3592 Chung-Ju Wu <jasonwucj@gmail.com>
3594 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
3596 * config/nds32/nds32-multiple.md (strlensi): New pattern.
3597 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
3599 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
3600 Kito Cheng <kito.cheng@gmail.com>
3601 Chung-Ju Wu <jasonwucj@gmail.com>
3603 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
3604 UNSPEC_FFMISM and UNSPEC_FLMISM.
3605 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
3606 for ffb, ffmism and flmism.
3607 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
3608 (unspec_ffmism): Ditto.
3609 (unspec_flmism): Ditto.
3610 (nds32_expand_builtin_impl): Check if string extension is available.
3611 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
3612 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
3614 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
3617 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
3620 * lra-assigns.c (assign_by_spills): Return a flag of reload
3621 assignment failure. Do not process the reload assignment
3622 failures. Do not spill other reload pseudos if they has the same
3624 (lra_assign): Add a return arg. Set up from the result of
3625 assign_by_spills call.
3626 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
3627 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
3628 usage_insns if it is not NULL.
3629 (spill_hard_reg_in_range): New function.
3630 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
3631 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
3632 function prototypes.
3633 (lra_assign): Change prototype.
3634 * lra.c (lra): Add code to deal with fails by splitting hard reg
3637 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
3640 * config/i386/i386.opt: Replace Enforcment with Enforcement.
3642 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
3645 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
3646 (dw_val_node): Add val_symbolic_view.
3647 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
3648 (symview_upper_bound): New.
3649 (new_line_info_table): Initialize symviews_since_reset.
3650 (dwarf2out_source_line): Count symviews_since_reset and set
3651 symview_upper_bound.
3652 (dw_val_equal_p): Handle symview.
3653 (add_AT_symview): New.
3654 (print_dw_val): Handle symview.
3655 (attr_checksum, attr_checksum_ordered): Likewise.
3656 (same_dw_val_p, size_of_die): Likewise.
3657 (value_format, output_die): Likewise.
3658 (add_high_low_attributes): Use add_AT_symview for entry_view.
3659 (dwarf2out_finish): Reset symview_upper_bound, clear
3662 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
3665 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
3666 Add strict argument and use it.
3667 (rs6000_split_multireg_move): Update for new strict argument.
3668 (mem_operand_gpr): Disallow all non-offsettable addresses.
3669 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
3671 2018-03-09 Jakub Jelinek <jakub@redhat.com>
3674 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
3675 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
3676 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
3679 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
3680 decl, use remap_type if we want to use the type.
3682 2018-03-09 Martin Sebor <msebor@redhat.com>
3684 PR tree-optimization/84526
3685 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
3687 (builtin_access::generic_overlap): Be prepared to handle non-array
3690 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
3692 PR rtl-optimization/84682
3693 * lra-constraints.c (process_address_1): Check is_address flag
3694 for address constraints.
3695 (process_alt_operands): Likewise.
3696 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
3697 preprocess_constraints.
3698 * recog.h (preprocess_constraints): Add oploc parameter.
3700 * recog.c (preprocess_constraints): Test address_operand for
3701 CT_ADDRESS constraints.
3703 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
3706 * lra-assigns.c (assign_by_spills): Return a flag of reload
3707 assignment failure. Do not process the reload assignment
3708 failures. Do not spill other reload pseudos if they has the same
3710 (lra_assign): Add a return arg. Set up from the result of
3711 assign_by_spills call.
3712 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
3713 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
3714 usage_insns if it is not NULL.
3715 (spill_hard_reg_in_range): New function.
3716 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
3717 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
3718 function prototypes.
3719 (lra_assign): Change prototype.
3720 * lra.c (lra): Add code to deal with fails by splitting hard reg
3723 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3726 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
3727 Accept complain bool parameter. Only emit errors if it is true.
3728 (arm_parse_cpu_option_name): Likewise.
3729 (arm_target_thumb_only): Adjust callers of the above.
3730 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
3731 prototype to take a default true bool parameter.
3732 (arm_parse_arch_option_name): Likewise.
3734 2018-03-09 David Malcolm <dmalcolm@redhat.com>
3735 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
3739 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
3740 * configure: Regenerate.
3741 * configure.ac ("linker --version-script option"): New.
3742 ("linker soname option"): New.
3744 2018-03-09 Richard Biener <rguenther@suse.de>
3746 PR tree-optimization/84775
3747 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
3748 immediate uses of predicate stmts and mark them modified.
3751 PR tree-optimization/84178
3752 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
3754 (version_loop_for_if_conversion): Delay update_ssa call.
3755 (tree_if_conversion): Delay update_ssa until after predicate
3758 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
3761 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
3762 when the function accesses prior frames.
3764 2018-03-08 Jakub Jelinek <jakub@redhat.com>
3767 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
3768 gen_llsym, otherwise call maybe_gen_llsym.
3771 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
3772 has ',' character inside of it.
3774 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3777 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
3778 as clobbering CC_REGNUM.
3780 2018-03-08 Richard Biener <rguenther@suse.de>
3783 * tree-scalar-evolution.c: Include tree-into-ssa.h.
3784 (follow_copies_to_constant): Do not follow SSA names registered
3787 2018-03-08 Richard Biener <rguenther@suse.de>
3789 PR tree-optimization/84178
3790 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
3792 (version_loop_for_if_conversion): Delay update_ssa call.
3793 (tree_if_conversion): Delay update_ssa until after predicate
3796 2018-03-08 David Malcolm <dmalcolm@redhat.com>
3798 PR tree-optimization/84178
3799 * tree-if-conv.c (release_bb_predicate): Remove the
3800 the assertion that the stmts have NULL use_ops.
3801 Discard the statements, asserting that they haven't
3802 yet been added to a BB.
3804 2018-03-08 Richard Biener <rguenther@suse.de>
3806 PR tree-optimization/84746
3807 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
3808 (phi_translate): Pass in destination ANTIC_OUT set.
3809 (phi_translate_1): Likewise. For a simplified result lookup
3810 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
3811 (phi_translate_set): Adjust.
3812 (do_pre_regular_insertion): Likewise.
3813 (do_pre_partial_partial_insertion): Likewise.
3815 2018-03-08 Martin Liska <mliska@suse.cz>
3817 PR gcov-profile/84735
3818 * doc/gcov.texi: Document usage of profile files.
3819 * gcov-io.h: Document changes in the format.
3821 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
3825 * dwarf2out.c (struct dw_line_info_table): Update comments for
3827 (FORCE_RESET_NEXT_VIEW): New.
3828 (FORCE_RESETTING_VIEW_P): New.
3829 (RESETTING_VIEW_P): Check for -1 too.
3830 (ZERO_VIEW_P): Likewise.
3831 (new_line_info_table): Force-reset next view.
3832 (dwarf2out_begin_function): Likewise.
3833 (dwarf2out_source_line): Simplify zero_view_p initialization.
3834 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
3835 view directly. Omit view when omitting .loc at line 0.
3837 2018-03-08 Jakub Jelinek <jakub@redhat.com>
3839 PR tree-optimization/84740
3840 * tree-switch-conversion.c (process_switch): Call build_constructors
3841 only if info.phi_count is non-zero.
3843 PR tree-optimization/84739
3844 * tree-tailcall.c (find_tail_calls): Check call arguments against
3845 DECL_ARGUMENTS (current_function_decl) rather than
3846 DECL_ARGUMENTS (func) when checking for tail recursion.
3848 2018-03-07 Jakub Jelinek <jakub@redhat.com>
3850 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
3851 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
3852 Volker Reichelt's entry and add entries for people that perform
3853 GCC fuzzy testing and report numerous bugs.
3855 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
3858 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
3859 readonly data in sdata, if that is disabled.
3860 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
3861 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
3862 -mreadonly-in-sdata option.
3864 2018-03-07 Martin Sebor <msebor@redhat.com>
3866 PR tree-optimization/84468
3867 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
3868 basic block when looking for nul assignment.
3870 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
3873 * except.h (output_function_exception_table): Adjust prototype.
3874 * except.c (output_function_exception_table): Remove FNNAME parameter
3875 and add SECTION parameter. Ouput one part of the table at a time.
3876 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
3877 the first part of the exception table and emit unwind directives.
3878 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
3879 (i386_pe_seh_cold_init): Likewise.
3880 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
3881 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
3882 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
3883 (ix86_output_call_insn): Emit a nop in one more case for SEH.
3884 * config/i386/winnt.c: Include except.h.
3885 (struct seh_frame_state): Add reg_offset, after_prologue and
3886 in_cold_section fields.
3887 (i386_pe_seh_end_prologue): Set seh->after_prologue.
3888 (i386_pe_seh_cold_init): New function.
3889 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
3890 to seh->in_cold_section.
3891 (seh_emit_push): Record the offset of the push.
3892 (seh_emit_save): Record the offet of the save.
3893 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
3894 Test seh->after_prologue to disregard the epilogue.
3895 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
3896 (i386_pe_end_cold_function): New function.
3898 2018-03-07 Jakub Jelinek <jakub@redhat.com>
3901 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
3902 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
3905 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
3907 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
3908 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
3911 * multiple_target.c: Include tree-inline.h and intl.h.
3912 (expand_target_clones): Diagnose and fail if node->definition and
3913 !tree_versionable_function_p (node->decl).
3915 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
3917 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
3919 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
3920 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
3921 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
3923 2018-03-06 Jakub Jelinek <jakub@redhat.com>
3926 * combine.c (try_combine): Use reg_or_subregno instead of handling
3927 just paradoxical SUBREGs and REGs.
3929 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
3931 * config/arc/arc.c (arc_finalize_pic): Remove function.
3932 (arc_must_save_register): We use single base PIC register, remove
3933 checks to save/restore the PIC register.
3934 (arc_expand_prologue): Likewise.
3935 * config/arc/arc-protos.h (arc_set_default_type_attributes):
3937 (arc_verify_short): Likewise.
3938 (arc_attr_type): Likewise.
3939 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
3940 (walk_stores): Likewise.
3941 (arc_address_cost): Make it static.
3942 (arc_verify_short): Likewise.
3943 (branch_dest): Likewise.
3944 (arc_attr_type): Likewise.
3945 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
3946 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
3947 (arc_final_prescan_insn): Remove inserting the nops due to
3948 hardware hazards. It is done in reorg step.
3949 (insn_length_variant_t): Remove.
3950 (insn_length_parameters_t): Likewise.
3951 (arc_insn_length_parameters): Likewise.
3952 (arc_get_insn_variants): Likewise.
3953 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
3955 2018-03-06 Jakub Jelinek <jakub@redhat.com>
3958 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
3961 PR tree-optimization/84687
3962 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
3964 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
3966 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3968 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
3969 Rename to ppc_speculation_barrier.
3970 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
3971 __builtin_ppc_speculation_barrier.
3973 2018-03-05 Jakub Jelinek <jakub@redhat.com>
3976 * combine.c (combine_simplify_rtx): Don't try to simplify if
3977 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
3980 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
3982 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
3983 to 32 bytes when compiling for POWER9.
3985 2018-03-05 Jakub Jelinek <jakub@redhat.com>
3988 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
3989 regparm >= 3 with no arg reg available also for calls with
3990 flag_force_indirect_call. Pass decl to ix86_function_regparm.
3993 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
3995 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
3997 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
4000 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
4002 2018-03-05 Richard Biener <rguenther@suse.de>
4004 PR tree-optimization/84486
4005 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
4006 When inserting a __builtin_assume_aligned call set the LHS
4007 SSA name alignment info accordingly.
4009 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
4011 PR tree-optimization/84114
4012 * config/aarch64/aarch64.c (aarch64_reassociation_width)
4013 Avoid reassociation of FLOAT_MODE addition.
4015 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
4017 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
4018 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
4019 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
4020 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
4021 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
4022 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
4023 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
4025 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
4026 __builtin_ia32_wbinvd): New builtins.
4027 (SPECIAL_ARGS2): New.
4028 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
4029 (SPECIAL_ARGS2): New.
4030 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
4031 (ix86_valid_target_attribute_inner_p): Ditto.
4032 (ix86_init_mmx_sse_builtins): Add special_args2.
4033 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
4034 TARGET_WBNOINVD_P): New.
4035 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
4036 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
4037 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
4038 * config/i386/immintrin.h (_wbinvd): New intrinsic.
4039 * config/i386/pconfigintrin.h: New file.
4040 * config/i386/wbnoinvdintrin.h: Ditto.
4041 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
4042 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
4044 2018-03-05 Richard Biener <rguenther@suse.de>
4046 PR tree-optimization/84670
4047 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4049 (BB_VISITED_WITH_VISITED_SUCCS): New define.
4050 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4051 (compute_antic_aux): Only assert the number of values in ANTIC_IN
4052 doesn't grow if all successors (recursively) were visited at least
4055 2018-03-05 Richard Biener <rguenther@suse.de>
4057 PR tree-optimization/84650
4058 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
4059 if executed in the loop pipeline.
4061 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
4063 * doc/configfiles.texi (Configuration Files): Move info about
4064 conditionalizing $target-protos.h to...
4065 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
4066 differs from $target-protos.h.
4068 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
4069 Chung-Ju Wu <jasonwucj@gmail.com>
4071 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
4072 * config/nds32/nds32-multiple.md (setmemsi): Define.
4073 * config/nds32/nds32-memory-manipulation.c
4074 (nds32_gen_dup_4_byte_to_word_value): New.
4075 (emit_setmem_word_loop): New.
4076 (emit_setmem_byte_loop): New.
4077 (nds32_expand_setmem_loop): New.
4078 (nds32_expand_setmem_loop_v3m): New.
4079 (nds32_expand_setmem_unroll): New.
4080 (nds32_expand_setmem): New.
4082 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4083 Chung-Ju Wu <jasonwucj@gmail.com>
4085 * config/nds32/nds32-memory-manipulation.c
4086 (nds32_emit_load_store): New.
4087 (nds32_emit_post_inc_load_store): New.
4088 (nds32_emit_mem_move): New.
4089 (nds32_emit_mem_move_block): New.
4090 (nds32_expand_movmemsi_loop_unknown_size): New.
4091 (nds32_expand_movmemsi_loop_known_size): New.
4092 (nds32_expand_movmemsi_loop): New.
4093 (nds32_expand_movmemsi_unroll): New.
4094 (nds32_expand_movmemqi): Rename ...
4095 (nds32_expand_movmemsi): ... to this.
4096 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
4097 (movmemsi): ... to this.
4098 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
4099 (nds32_expand_movmemsi): ... to this.
4101 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4102 Monk Chiang <sh.chiang04@gmail.com>
4103 Chung-Ju Wu <jasonwucj@gmail.com>
4105 * config/nds32/nds32-protos.h
4106 (nds32_expand_load_multiple): New arguments.
4107 (nds32_expand_store_multiple): Ditto.
4108 (nds32_valid_multiple_load_store): Rename ...
4109 (nds32_valid_multiple_load_store_p): ... to this.
4110 * config/nds32/nds32-memory-manipulation.c
4111 (nds32_expand_load_multiple): Refine implementation.
4112 (nds32_expand_store_multiple): Ditto.
4113 * config/nds32/nds32-multiple.md
4114 (load_multiple): Update nds32_expand_load_multiple interface.
4115 (store_multiple): Update nds32_expand_store_multiple interface.
4116 * config/nds32/nds32-predicates.c
4117 (nds32_valid_multiple_load_store): Rename ...
4118 (nds32_valid_multiple_load_store_p): ... to this and refine
4120 * config/nds32/predicates.md
4121 (nds32_load_multiple_and_update_address_operation): New predicate.
4122 (nds32_store_multiple_and_update_address_operation): New predicate.
4124 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4125 Chung-Ju Wu <jasonwucj@gmail.com>
4127 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
4128 (combo): New attribute.
4129 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
4131 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
4133 * config/nds32/nds32.opt: Change -mcmodel= default value.
4135 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
4136 Monk Chiang <sh.chiang04@gmail.com>
4137 Chung-Ju Wu <jasonwucj@gmail.com>
4139 * config/nds32/constants.md (unspec_element): New enum.
4140 * config/nds32/constraints.md (Umw): New constraint.
4141 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
4142 * config/nds32/nds32-intrinsic.md: Likewise.
4143 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
4144 (nds32_valid_smw_lwm_base_p): New.
4145 (nds32_output_smw_single_word): New.
4146 (nds32_output_lmw_single_word): New.
4147 (nds32_expand_unaligned_load): New.
4148 (nds32_expand_unaligned_store): New.
4149 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
4150 (nds32_output_smw_single_word): Declare.
4151 (nds32_output_lmw_single_word): Declare.
4152 (nds32_expand_unaligned_load): Declare.
4153 (nds32_expand_unaligned_store): Declare.
4154 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
4155 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
4156 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
4157 NDS32_BUILTIN_UASTORE_DW.
4158 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
4161 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
4162 Kito Cheng <kito.cheng@gmail.com>
4163 Chung-Ju Wu <jasonwucj@gmail.com>
4165 * config/nds32/nds32-intrinsic.c
4166 (nds32_expand_builtin_null_ftype_reg): Delete.
4167 (nds32_expand_builtin_reg_ftype_imm): Ditto.
4168 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
4169 (nds32_read_argument): New.
4170 (nds32_legitimize_target): Ditto.
4171 (nds32_legitimize_argument): Ditto.
4172 (nds32_check_constant_argument): Ditto.
4173 (nds32_expand_unop_builtin): Ditto.
4174 (nds32_expand_unopimm_builtin): Ditto.
4175 (nds32_expand_binop_builtin): Ditto.
4176 (nds32_builtin_decl_impl): Ditto.
4177 (builtin_description): Ditto.
4178 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
4179 (nds32_init_builtins_impl): Ditto.
4180 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
4181 (nds32_builtin_decl): New.
4182 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
4183 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
4185 2018-03-02 Jeff Law <law@redhat.com>
4187 * reorg.c (stop_search_p): Handle DEBUG_INSN.
4188 (redundant_insn, fill_simple_delay_slots): Likewise.
4189 (fill_slots_from_thread): Likewise.
4190 * resource.c (mark_referenced_resources): Likewise.
4191 (mark_set_resources, find_dead_or_set_registers): Likewise.
4193 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4195 * substring-locations.h (format_warning_va): Formatting fix for
4197 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
4199 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
4200 * substring-locations.c: Include intl.h.
4201 (format_warning_va): Turned into small wrapper around
4202 format_warning_n_va, renamed to ...
4203 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
4204 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
4206 (format_warning_at_substring_n): New function.
4207 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
4208 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
4209 format_warning_at_substring with just a shorter name instead of
4210 const function pointer.
4211 (fmtwarn_n): New function.
4212 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
4213 appropriate, get rid of all the fmtstr temporaries, move conditionals
4214 with G_() wrapped string literals directly into fmtwarn arguments,
4215 cast dir.len to (int), formatting fixes.
4217 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
4219 * doc/invoke.texi: Remove "Cilk Plus" references.
4221 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4222 Richard Biener <rguenther@suse.de>
4225 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
4226 for error or warning attributes if CALL_FROM_THUNK_P is set.
4229 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4232 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
4233 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
4236 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
4237 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
4239 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
4240 instead of -1U in last predictors element's probability member.
4242 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
4245 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
4246 arguments if they are comparable.
4248 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
4250 PR tree-optimization/84634
4251 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
4252 masks and masked_loop_p with a single loop_masks, making sure it's
4253 null for bb vectorization.
4255 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
4257 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
4258 (vect_analyze_data_ref_access): Use loop->safe_len rather than
4259 loop->force_vectorize to check whether there is no alias.
4261 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4264 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
4266 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
4268 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
4269 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
4270 instead of a loop around prev_real_insn.
4271 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
4275 * config/i386/i386.c (ix86_print_operand): Use conditional
4276 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
4279 2018-03-02 Richard Biener <rguenther@suse.de>
4281 PR tree-optimization/84427
4282 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
4283 (bitmap_set_subtract_values): Rewrite to handle multiple
4286 (prune_clobbered_mems): Likewise.
4287 (phi_translate): Take edge instead of pred/phiblock.
4288 (phi_translate_1): Likewise.
4289 (phi_translate_set): Likewise. Insert all translated
4290 exprs for a value into the set, keeping possibly multiple
4291 expressions per value.
4292 (compute_antic_aux): Adjust for phi_translate changes.
4293 When intersecting union the expressions and prune those
4294 not in the final value set, keeping possibly multiple
4295 expressions per value. Do not use value-insertion
4296 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
4297 all expressions. Add verification that the value-sets
4298 only shrink during iteration.
4299 (compute_partial_antic_aux): Adjust for the phi_translate changes.
4300 (do_pre_regular_insertion): Likewise.
4301 (do_pre_partial_partial_insertion): Likewise.
4303 2018-03-02 Richard Biener <rguenther@suse.de>
4306 * config/darwin.c (saved_debug_info_level): New static global.
4307 (darwin_asm_lto_start): Disable debug info generation for LTO out.
4308 (darwin_asm_lto_end): Restore debug info generation settings.
4310 2018-03-01 Martin Liska <mliska@suse.cz>
4313 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
4316 2018-03-01 Richard Biener <rguenther@suse.de>
4319 * dwarf2out.c (gen_variable_die): Properly handle late VLA
4320 type annotation with LTO when debug was disabled at compile-time.
4322 2018-03-01 Matthew Fortune <mfortune@gmail.com>
4324 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
4326 (mips_final_postscan_insn): Likewise.
4328 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
4330 PR rtl-optimization/84528
4331 * alias.c (init_alias_target): Add commentary.
4332 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
4333 a unique base value if the frame pointer is not eliminated
4334 to the stack pointer.
4336 2018-03-01 Tom de Vries <tom@codesourcery.com>
4338 PR rtl-optimization/83327
4339 * lra-int.h (hard_regs_spilled_into): Declare.
4340 * lra.c (hard_regs_spilled_into): Define.
4341 (init_reg_info): Init hard_regs_spilled_into.
4342 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
4343 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
4344 (process_bb_lives): Handle hard_regs_spilled_into.
4345 (lra_create_live_ranges_1): Before doing liveness propagation, clear
4346 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
4348 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
4350 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
4351 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
4352 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
4353 * config/rs6000/aix72.h: New file.
4355 2018-02-28 Jakub Jelinek <jakub@redhat.com>
4357 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
4358 instead of warning_at with conditional singular and plural messages
4362 * stor-layout.c (update_alignment_for_field): For
4363 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
4364 && !DECL_PACKED (field), do the alignment update, just use
4365 only desired_align instead of MAX (type_align, desired_align)
4367 (place_field): Don't do known_align < desired_align handling
4368 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
4369 is non-NULL, instead do it after rli->prev_field handling and
4370 only if not within a bitfield word. For DECL_PACKED (field)
4371 use type_align of BITS_PER_UNIT.
4373 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
4375 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
4376 superfluous parentheses and trailing spaces.
4378 2018-02-28 Richard Biener <rguenther@suse.de>
4380 PR tree-optimization/84584
4381 * graphite-scop-detection.c (scop_detection::add_scop): Discard
4382 SCoPs with fake exit edge.
4384 2018-02-28 Martin Liska <mliska@suse.cz>
4387 * timevar.c (timer::print): Fix format to properly print 100%
4390 2018-02-28 Richard Biener <rguenther@suse.de>
4393 * genmatch.c (capture_info::walk_match): Do not mark
4394 captured expressions without operands as expr_p given
4395 they act more like predicates and should be subject to
4396 "lost tail" side-effect preserving.
4398 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
4400 PR rtl-optimization/81611
4401 * auto-inc-dec.c (attempt_change): Move dead note from
4402 mem_insn if it's the next use of regno
4403 (find_address): Take address use of reg holding
4404 non-incremented value. Add parm to limit search to the named
4406 (merge_in_block): Attempt to use a mem insn that is the next
4407 use of the original regno.
4409 2018-02-27 Martin Sebor <msebor@redhat.com>
4412 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
4413 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
4415 2018-02-27 Martin Sebor <msebor@redhat.com>
4417 PR translation/84207
4418 * diagnostic-core.h (warning_n, error_n, inform_n): Change
4419 n argument to unsigned HOST_WIDE_INT.
4420 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
4421 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
4422 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
4423 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
4425 2018-02-27 Richard Biener <rguenther@suse.de>
4427 PR tree-optimization/84512
4428 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
4429 Do not use the estimate returned from record_stmt_cost for
4430 the scalar iteration cost but sum properly using add_stmt_cost.
4432 2018-02-27 Richard Biener <rguenther@suse.de>
4434 PR tree-optimization/84466
4435 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
4436 Adjust last change to less strictly validate use operands.
4438 2018-02-27 Martin Liska <mliska@suse.cz>
4440 PR gcov-profile/84548
4441 * gcov.c (process_file): Allow partial overlap and consider it
4442 also as group functions.
4443 (output_lines): Properly calculate range of lines for a group.
4445 2018-02-27 Martin Liska <mliska@suse.cz>
4447 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
4448 'ggc' suffixes. Change first column width.
4449 (timer::print): Fix formatting of the column.
4451 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
4453 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
4454 preserve inline entry blocks for the sake of debug inline
4455 entry point markers alone.
4456 (remove_unused_locals): Suggest in comments a better place to
4457 force the preservation of inline entry blocks that are
4458 otherwise unused, but do not preserve them.
4460 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
4462 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
4464 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
4467 * config/i386/constraints.md (Bs): Replace
4468 ix86_indirect_branch_register with
4469 TARGET_INDIRECT_BRANCH_REGISTER.
4471 * config/i386/i386.md (indirect_jump): Likewise.
4472 (tablejump): Likewise.
4473 (*sibcall_memory): Likewise.
4474 (*sibcall_value_memory): Likewise.
4475 Peepholes of indirect call and jump via memory: Likewise.
4476 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
4477 (*sibcall_value_GOT_32): Likewise.
4478 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4479 (GOT_memory_operand): Likewise.
4480 (call_insn_operand): Likewise.
4481 (sibcall_insn_operand): Likewise.
4482 (GOT32_symbol_operand): Likewise.
4483 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
4485 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
4487 PR rtl-optimization/83496
4488 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
4489 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
4490 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
4491 redundant insn, if any.
4492 (relax_delay_slots): Likewise.
4493 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
4495 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
4497 PR tree-optimization/83965
4498 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
4499 that grouped statements are part of a reduction chain. Return
4500 true if the statement is not marked as a reduction itself but
4502 (vect_recog_dot_prod_pattern): Don't check whether the statement
4503 is part of a group here.
4504 (vect_recog_sad_pattern): Likewise.
4505 (vect_recog_widen_sum_pattern): Likewise.
4507 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
4510 * final.c (rest_of_clean_state): Also look for calls inside sequences.
4512 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
4515 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
4517 (ix86_output_indirect_function_return): New prototype.
4518 (ix86_split_simple_return_pop_internal): Likewise.
4519 * config/i386/i386.c (indirect_return_via_cx): New.
4520 (indirect_return_via_cx_bnd): Likewise.
4521 (indirect_thunk_name): Handle return va CX_REG.
4522 (output_indirect_thunk_function): Create alias for
4523 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
4524 (ix86_output_indirect_jmp): Remove the bool argument.
4525 (ix86_output_indirect_function_return): New function.
4526 (ix86_split_simple_return_pop_internal): Likewise.
4527 * config/i386/i386.md (*indirect_jump): Don't pass false
4528 to ix86_output_indirect_jmp.
4529 (*tablejump_1): Likewise.
4530 (simple_return_pop_internal): Change it to define_insn_and_split.
4531 Call ix86_split_simple_return_pop_internal to split it for
4533 (simple_return_indirect_internal): Call
4534 ix86_output_indirect_function_return instead of
4535 ix86_output_indirect_jmp.
4537 2018-02-26 Jakub Jelinek <jakub@redhat.com>
4540 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
4541 memset and value initialization afterwards.
4543 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
4545 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
4547 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4550 * common/config/aarch64/aarch64-common.c
4551 (aarch_option_optimization_table[]): Switch
4552 off fomit-frame-pointer
4554 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
4555 Chung-Ju Wu <jasonwucj@gmail.com>
4557 * config/nds32/nds32-multiple.md (load_multiple): Disallow
4559 (store_multiple): Ditto.
4561 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
4563 * config.gcc: Add --with-cpu support for nds32 target.
4564 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
4565 * config/nds32/nds32.opt: Add -mcpu= option.
4567 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
4569 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
4570 isel=yes): Warn for these deprecated options.
4572 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
4574 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
4575 ISA_2_5_MASKS_EMBEDDED.
4577 2018-02-23 Jakub Jelinek <jakub@redhat.com>
4579 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
4580 p->max as pointers rather than using iterative_hash_expr.
4582 2018-02-23 Carl Love <cel@us.ibm.com>
4584 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
4585 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
4587 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
4588 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
4589 P8V_BUILTIN_VEC_VUNSIGNED2.
4591 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
4594 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
4595 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
4597 * lra-constraints.c (curr_insn_transform): Set up
4598 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
4600 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
4601 * lra-eliminations.c (spill_pseudos): Ditto.
4602 (process_insn_for_elimination): Ditto.
4603 * lra-lives.c (reg_early_clobber_p): Use the new macros.
4604 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
4605 LRA_NON_CLOBBERED_ALT.
4607 2018-02-22 Martin Sebor <msebor@redhat.com>
4609 PR tree-optimization/84480
4610 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
4611 to maybe_diag_stxncpy_trunc. Call it.
4612 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
4613 from gimple_fold_builtin_strcpy. Print inlining stack.
4614 (handle_builtin_stxncpy): Print inlining stack.
4615 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
4617 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
4620 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
4621 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
4622 and -fcheck-pointer-bounds are used together.
4623 (indirect_thunk_prefix): New enum.
4624 (indirect_thunk_need_prefix): New function.
4625 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
4626 "_nt" instead of "_bnd" for NOTRACK prefix.
4627 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
4628 (output_indirect_thunk_function): Likewise.
4630 (ix86_code_end): Update output_indirect_thunk_function calls.
4631 (ix86_output_indirect_branch_via_reg): Replace
4632 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
4633 (ix86_output_indirect_branch_via_push): Likewise.
4634 (ix86_output_function_return): Likewise.
4635 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
4636 incompatible with -fcf-protection=branch and
4637 -fcheck-pointer-bounds.
4639 2018-02-22 Steve Ellcey <sellcey@cavium.com>
4642 * config/aarch64/aarch64.c (aarch64_print_address_internal):
4643 Change gcc_assert call to output_operand_lossage.
4645 2018-02-22 Steve Ellcey <sellcey@cavium.com>
4647 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
4649 2018-02-22 DJ Delorie <dj@redhat.com>
4650 Sebastian Perta <sebastian.perta@renesas.com>
4651 Oleg Endo <olegendo@gcc.gnu.org>
4653 * config/rx/rx.c (rx_rtx_costs): New function.
4654 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
4656 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
4658 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
4660 2018-02-22 Martin Liska <mliska@suse.cz>
4663 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
4664 Add "native" as a possible value.
4666 2018-02-22 Martin Liska <mliska@suse.cz>
4669 * config/i386/i386.c (ix86_option_override_internal):
4670 Add "native" as a possible value for -march and -mtune.
4672 2018-02-22 Jakub Jelinek <jakub@redhat.com>
4675 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
4676 to all type variants.
4678 PR tree-optimization/84503
4679 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
4680 width as info->bitpos + info->bitsize - start.
4681 (merged_store_group::merge_overlapping): Simplify width computation.
4682 (check_no_overlap): New function.
4683 (imm_store_chain_info::try_coalesce_bswap): Compute expected
4684 start + width and last_order of the group, fail if check_no_overlap
4686 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
4687 to group if check_no_overlap fails.
4689 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
4691 * config/rs6000/altivec.md: Delete contraint arguments to
4692 define_expand, define_split, and define_peephole2, and in
4693 define_insn_and_split if always unused.
4694 * config/rs6000/darwin.md: Ditto.
4695 * config/rs6000/dfp.md: Ditto.
4696 * config/rs6000/rs6000.md: Ditto.
4697 * config/rs6000/sync.md: Ditto.
4698 * config/rs6000/vector.md: Ditto.
4699 * config/rs6000/vsx.md: Ditto.
4701 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
4703 * config/rs6000/altivec.md: Write output control strings as braced
4704 blocks instead of double-quoted strings.
4705 * config/rs6000/darwin.md: Ditto.
4706 * config/rs6000/rs6000.md: Ditto.
4707 * config/rs6000/vector.md: Ditto.
4708 * config/rs6000/vsx.md: Ditto.
4710 2018-02-21 Jason Merrill <jason@redhat.com>
4712 PR c++/84314 - ICE with templates and fastcall attribute.
4713 * attribs.c (build_type_attribute_qual_variant): Remove assert.
4715 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
4717 * ipa-cp.c (determine_versionability): Fix comment typos.
4719 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
4722 * ipa-cp.c (determine_versionability): Do not version functions caling
4725 2018-02-21 Martin Liska <mliska@suse.cz>
4728 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
4729 Add "native" as a possible value.
4730 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
4731 the macro when native cpu detection is available.
4733 2018-02-21 Martin Liska <mliska@suse.cz>
4736 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
4737 Add "native" as a possible value.
4738 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
4739 when native cpu detection is available.
4741 2018-02-21 Jakub Jelinek <jakub@redhat.com>
4742 Martin Sebor <msebor@redhat.com>
4744 PR tree-optimization/84478
4745 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
4747 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
4748 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
4749 support which is conservatively correct, for 2 only stay conservative
4750 for maxlen. Formatting and comment capitalization fixes. Add STRICT
4751 argument to the 2 argument get_range_strlen, adjust 6 arg
4752 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
4754 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
4755 (gimple_fold_builtin_strlen): Pass true as last argument to
4758 2018-02-20 Martin Sebor <msebor@redhat.com>
4761 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
4762 (builtin_memref::set_base_and_offset): Same. Handle inner references.
4763 (builtin_memref::builtin_memref): Factor out parts into
4764 set_base_and_offset and call it.
4766 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
4769 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
4770 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
4771 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
4772 search at the associated MODE_INT.
4774 2018-02-20 Jeff Law <law@redhat.com>
4777 PR tree-optimization/81592
4779 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
4780 for range data rather than using global data.
4781 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
4782 range data rather than using global data.
4783 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
4784 pass it to children as needed.
4785 (struct directive::fmtresult): Similarly.
4786 (struct directive::set_width): Similarly.
4787 (struct directive::set_precision): Similarly.
4788 (format_integer, format_directive, parse_directive): Similarly.
4789 (format_none): Accept unnamed vr_values parameter.
4790 (format_percent, format_floating, format_character): Similarly.
4791 (format_string, format_plain): Similarly.
4792 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
4793 the EVRP range analyzer for range data rather than using global data.
4794 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
4795 gimple-ssa-evrp-analyze.h
4796 (class sprintf_dom_walker): Add after_dom_children member function.
4797 Add evrp_range_analyzer member.
4798 (sprintf_dom_walker::before_dom_children): Call into the EVRP
4799 range analyzer as needed.
4800 (sprintf_dom_walker::after_dom_children): New member function.
4801 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
4803 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
4804 (evrp_range_analyzer::pop_to_marker): Likewise.
4806 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
4808 PR tree-optimization/84419
4809 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
4810 with the required type if its current type is compatible but
4813 2018-02-20 Jakub Jelinek <jakub@redhat.com>
4816 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
4817 after vectorization.
4819 2018-02-20 Martin Liska <mliska@suse.cz>
4822 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
4823 possible values if we don't have a hint.
4825 2018-02-20 Martin Liska <mliska@suse.cz>
4829 * final.c (shorten_branches): Build align_tab array with one
4831 * opts.c (finish_options): Add alignment option limit check.
4832 (MAX_CODE_ALIGN): Likewise.
4833 (MAX_CODE_ALIGN_VALUE): Likewise.
4834 * doc/invoke.texi: Document maximum allowed option value for
4835 all -falign-* options.
4837 2018-02-19 Jakub Jelinek <jakub@redhat.com>
4840 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
4841 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
4842 * var-tracking.c (emit_note_insn_var_location): Remove all references
4843 to NOTE_INSN_CALL_ARG_LOCATION.
4844 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
4845 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
4846 Use copy_rtx_if_shared.
4847 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
4848 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
4849 (dwarf2out_var_location): Remove handling of
4850 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
4852 * final.c (final_scan_insn): Remove all references to
4853 NOTE_INSN_CALL_ARG_LOCATION.
4854 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
4855 before dumping final insns.
4856 * except.c (emit_note_eh_region_end): Remove all references to
4857 NOTE_INSN_CALL_ARG_LOCATION.
4858 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
4859 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
4860 * config/arc/arc.c (hwloop_optimize): Likewise.
4861 * config/arm/arm.c (create_fix_barrier): Likewise.
4862 * config/s390/s390.c (s390_chunkify_start): Likewise.
4863 * config/sh/sh.c (find_barrier): Likewise.
4864 * config/i386/i386.c (rest_of_insert_endbranch,
4865 ix86_seh_fixup_eh_fallthru): Likewise.
4866 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
4867 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
4868 * config/frv/frv.c (frv_function_prologue): Likewise.
4869 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
4871 (note_outside_basic_block_p): Remove all references to
4872 NOTE_INSN_CALL_ARG_LOCATION.
4873 * gengtype.c (adjust_field_rtx_def): Likewise.
4874 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
4876 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
4877 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
4880 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
4883 PR tree-optimization/84452
4884 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
4885 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
4888 2018-02-19 Martin Liska <mliska@suse.cz>
4891 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
4893 2018-02-19 Martin Liska <mliska@suse.cz>
4894 Richard Sandiford <richard.sandiford@linaro.org>
4896 PR tree-optimization/82491
4897 * gimple-fold.c (get_base_constructor): Make earlier bail out
4900 2018-02-19 Carl Love <cel@us.ibm.com>
4902 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
4903 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
4905 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
4906 P8V_BUILTIN_VEC_NEG.
4908 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
4910 * config/rl78/rl78.md (movdf): New define expand.
4912 2018-02-19 Martin Liska <mliska@suse.cz>
4915 * doc/invoke.texi: Fix typo.
4916 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
4918 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
4920 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
4921 handle rs6000_single_float and rs6000_double_float specially for
4924 2018-02-16 Jeff Law <law@redhat.com>
4926 * config/rx/rx.c (add_pop_cfi_notes): New function.;
4929 2018-02-16 Jakub Jelinek <jakub@redhat.com>
4932 * ipa-inline.c (inline_small_functions): Fix a typo.
4934 2018-02-16 Nathan Sidwell <nathan@acm.org>
4936 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
4938 2018-02-16 Carl Love <cel@us.ibm.com>
4940 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
4941 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
4942 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
4943 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
4944 expansion to P8V_BUILTIN_VEC_FLOAT2.
4946 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
4948 PR rtl-optimization/70023
4949 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
4950 src_regno into account.
4952 2018-02-16 Carl Love <cel@us.ibm.com>
4954 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
4955 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
4956 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
4957 * config/rs6000/rs6000.c: Remove case statements for
4958 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
4959 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
4960 and P9V_BUILTIN_VEC_VINSERT4B.
4961 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
4962 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
4963 * config/rs6000/vsx.md:
4964 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
4967 2018-02-16 Carl Love <cel@us.ibm.com>
4969 * config/rs6000/altivec.h: Add builtin names vec_extract4b
4971 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
4973 * config/rs6000/rs6000-c.c: Add the definitions for
4974 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
4975 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
4976 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
4977 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
4978 definition for insert4b and define insn *insert3b_internal.
4979 * doc/extend.texi: Add documentation for vec_extract4b.
4981 2018-02-16 Nathan Sidwell <nathan@acm.org>
4983 * doc/extend.texi (Backwards Compatibility): Mention friend
4984 injection. Note for-scope is deprecated.
4985 * doc/invoke.texi (-ffriend-injection): Deprecate.
4987 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
4989 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
4990 that moved to I2, also allow destinations that are a paradoxical
4991 subreg (instead of a normal reg).
4993 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
4996 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
4999 2018-02-16 Richard Biener <rguenther@suse.de>
5001 PR tree-optimization/84037
5002 PR tree-optimization/84016
5004 * config/i386/i386.c (ix86_builtin_vectorization_cost):
5005 Adjust vec_construct for the fact we need additional higher latency
5006 128bit inserts for AVX256 and AVX512 vector builds.
5007 (ix86_add_stmt_cost): Scale vector construction cost for
5010 2018-02-16 Richard Biener <rguenther@suse.de>
5012 PR tree-optimization/84417
5013 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
5014 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
5015 (non_rewritable_lvalue_p): Likewise, use poly-ints.
5017 2018-02-16 Martin Liska <mliska@suse.cz>
5020 * internal-fn.def (ASAN_CHECK): Set proper flags.
5021 (ASAN_MARK): Likewise.
5023 2018-02-16 Julia Koval <julia.koval@intel.com>
5025 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
5026 from PTA_CANNONLAKE.
5028 2018-02-16 Jakub Jelinek <jakub@redhat.com>
5031 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
5032 Use ++iter rather than iter++ for std::list iterators.
5033 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
5034 defer deleting them until all nodes in the forest are processed. Do
5035 free even leaf nodes. Change to_process into auto_vec.
5038 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
5039 * vec.h (vec_default_construct): Use memset instead of placement new
5040 if BROKEN_VALUE_INITIALIZATION is defined.
5041 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
5042 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
5045 PR rtl-optimization/83723
5046 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
5047 * lra.c (lra_substitute_pseudo): Likewise. If true, use
5048 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
5050 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
5052 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
5054 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
5056 PR rtl-optimization/81443
5057 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
5058 from inner REGs to paradoxical SUBREGs.
5060 2018-02-16 Richard Biener <rguenther@suse.de>
5062 PR tree-optimization/84399
5063 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
5064 For operands we can analyze at their definition make sure we can
5065 analyze them at each use as well.
5067 2018-02-16 Richard Biener <rguenther@suse.de>
5069 PR tree-optimization/84190
5070 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
5071 volatile accesses if the decl isn't volatile.
5073 2018-02-15 Jason Merrill <jason@redhat.com>
5075 PR c++/84314 - ICE with templates and fastcall attribute.
5076 * attribs.c (build_type_attribute_qual_variant): Don't clobber
5077 TYPE_CANONICAL on an existing type.
5079 2018-02-15 Jakub Jelinek <jakub@redhat.com>
5081 PR tree-optimization/84383
5082 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
5083 dstoff nor call operand_equal_p if dstbase is NULL.
5085 PR tree-optimization/84334
5086 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
5087 also a CONSTANT_CLASS_P, punt.
5089 2018-02-14 Jim Wilson <jimw@sifive.com>
5091 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
5092 first SMALL_OPERAND check. New local min_second_step. Move assert
5093 to where locals are set. Add TARGET_RVC support.
5094 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
5096 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
5098 * doc/invoke.texi: Correct -Wformat-overflow code sample.
5100 2018-02-14 Martin Sebor <msebor@redhat.com>
5102 PR tree-optimization/83698
5103 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
5104 arrays constrain the offset range to their bounds.
5105 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
5106 (builtin_access::overlap): Avoid setting the size of overlap if it's
5108 (maybe_diag_overlap): Also consider arrays when deciding what values
5109 of offsets to include in diagnostics.
5111 2018-02-14 Martin Sebor <msebor@redhat.com>
5114 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
5115 that correspond to the kind of a declaration.
5117 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
5120 * config/pa/pa.md: Load address of PIC label using the linkage table
5121 if the label is nonlocal.
5123 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
5125 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
5126 warning message if user requests -maltivec=be.
5127 * doc/invoke.texi: Document deprecation of -maltivec=be.
5129 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
5132 * config/rs6000/rs6000-c.c: Update definitions for
5133 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
5134 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
5136 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
5139 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
5140 add _get_ssp intrinsics. Remove argument from
5141 __builtin_ia32_rdssp[d|q].
5142 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
5143 * config/i386/i386-builtin.def: Remove argument from
5144 __builtin_ia32_rdssp[d|q].
5145 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
5146 ix86_expand_special_args_builtin for _rdssp[d|q].
5147 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
5148 Clear register before usage.
5149 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
5150 Add documentation for new _get_ssp and _inc_ssp intrinsics.
5152 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
5154 PR tree-optimization/84357
5155 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
5156 operand 1 of an ARRAY_REF too.
5158 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
5161 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
5162 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
5164 (set_of_reg): New struct.
5165 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
5166 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
5167 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
5169 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
5170 Split into bitclr, bitset, bitinvert patterns if appropriate.
5171 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
5172 use rx_fuse_in_memory_bitop.
5173 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
5174 to named insn, correct maximum insn length.
5176 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
5179 * machmode.def: Define a complex mode for PARTIAL_INT.
5180 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
5182 * doc/rtl.texi: Document CSPImode.
5183 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
5185 (msp430_hard_regno_nregs_with_padding): Likewise.
5187 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
5190 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
5192 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
5194 PR rtl-optimization/84169
5195 * combine.c (try_combine): New variable split_i2i3. Set it to true if
5196 we generated a parallel as new i3 and we split that to new i2 and i3
5197 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
5198 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
5199 those to i2, not i1. Partially rewrite this scan code.
5201 2018-02-13 Jakub Jelinek <jakub@redhat.com>
5204 * stor-layout.c (place_field): For variable length fields, adjust
5205 offset_align afterwards not just based on the field's alignment,
5206 but also on the size.
5209 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
5210 of exps and logs in the use_exp2 case.
5212 2018-02-13 Jeff Law <law@redhat.com>
5214 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
5217 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
5220 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
5224 * common.opt (gas-loc-support, gas-locview-support): New.
5225 (ginline-points, ginternal-reset-location-views): New.
5226 * doc/invoke.texi: Document them. Use @itemx where intended.
5227 (gvariable-location-views): Adjust.
5228 * target.def (reset_location_view): New.
5229 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
5230 (TARGET_RESET_LOCATION_VIEW): New.
5231 * doc/tm.texi: Rebuilt.
5232 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
5233 (dwarf2out_default_as_locview_support): New.
5234 (output_asm_line_debug_info): Use option variables.
5235 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
5236 (output_loc_list): Likewise.
5237 (add_high_low_attributes): Check option variables.
5238 Don't output entry view attribute in strict mode.
5239 (gen_inlined_subroutine_die): Check option variables.
5240 (dwarf2out_inline_entry): Likewise.
5241 (init_sections_and_labels): Likewise.
5242 (dwarf2out_early_finish): Likewise.
5243 (maybe_reset_location_view): New, from...
5244 (dwarf2out_var_location): ... here. Call it.
5245 * debug.h (dwarf2out_default_as_loc_support): Declare.
5246 (dwarf2out_default_as_locview_support): Declare.
5247 * hooks.c (hook_int_rtx_insn_0): New.
5248 * hooks.h (hook_int_rtx_insn_0): Declare.
5249 * toplev.c (process_options): Take -gas-loc-support and
5250 -gas-locview-support from dwarf2out. Enable
5251 -gvariable-location-views by default only with locview
5252 assembler support. Enable -ginternal-reset-location-views by
5253 default only if the target defines the corresponding hook.
5254 Enable -ginline-points by default if location views are
5255 enabled; force it disabled if statement frontiers are
5257 * tree-inline.c (expand_call_inline): Check option variables.
5258 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
5260 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
5262 PR tree-optimization/84321
5263 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
5264 handling. Also check whether the anti-range contains any values
5265 that satisfy the mask; switch to a VR_RANGE if not.
5267 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
5270 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
5272 2018-02-13 Martin Jambor <mjambor@suse.cz>
5275 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
5276 of call statements, also set location of a load to a temporary.
5278 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
5280 * config/rl78/rl78.c (add_vector_labels): New function.
5281 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
5282 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
5283 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
5284 which checks that no arguments are passed.
5285 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
5286 * doc/extend.texi: Documentation for the new attribute.
5288 2018-02-13 Andreas Schwab <schwab@suse.de>
5290 * config/riscv/linux.h (CPP_SPEC): Define.
5292 2018-02-13 Jakub Jelinek <jakub@redhat.com>
5295 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
5296 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
5297 OPTION_MASK_ISA_AES as first argument to def_builtin_const
5298 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
5299 instead of OPTION_MASK_ISA_PCLMUL as first argument to
5300 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
5301 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
5302 temporarily for AES and PCLMUL builtins.
5304 PR tree-optimization/84339
5305 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
5306 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
5310 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
5311 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
5312 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
5314 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
5316 * omp-simd-clone.h: New file.
5317 * omp-simd-clone.c: Include omp-simd-clone.h.
5318 (expand_simd_clones): No longer static.
5319 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
5320 cgraph.h and omp-simd-clone.h.
5321 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
5322 (vect_recog_widen_shift_pattern): Formatting fix.
5323 (vect_pattern_recog_1): Don't check optab for calls.
5326 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
5327 operands[2] into a REG before using gen_lowpart on it.
5329 2018-02-12 Jeff Law <law@redhat.com>
5332 * config/sh/sh.c (find_barrier): Consider a sibling call
5335 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
5336 successfully back substituting a reg.
5338 2018-02-12 Richard Biener <rguenther@suse.de>
5340 PR tree-optimization/84037
5341 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
5342 parameter, move visited init to caller.
5343 (vect_slp_analyze_operations): Separate cost from validity
5344 check, initialize visited once for all instances.
5345 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
5347 * tree-vect-stmts.c (vect_model_simple_cost): Make early
5349 (vect_model_promotion_demotion_cost): Likewise.
5350 (vectorizable_bswap): Guard cost modeling with !slp_node
5351 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
5353 (vectorizable_call): Likewise.
5354 (vectorizable_conversion): Likewise.
5355 (vectorizable_assignment): Likewise.
5356 (vectorizable_shift): Likewise.
5357 (vectorizable_operation): Likewise.
5358 (vectorizable_store): Likewise.
5359 (vectorizable_load): Likewise.
5360 (vectorizable_condition): Likewise.
5361 (vectorizable_comparison): Likewise.
5363 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
5366 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
5367 (ASAN_MARK): Fix fnspec to account for return value, change pointer
5368 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
5370 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
5373 * params.def (inline-min-speedup): Increase from 8 to 15.
5374 (max-inline-insns-auto): Decrease from 40 to 30.
5375 * ipa-split.c (consider_split): Add some buffer for function to
5376 be considered inlining candidate.
5377 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
5380 2018-02-12 Richard Biener <rguenther@suse.de>
5382 PR tree-optimization/84037
5383 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
5384 matched stmts if we cannot swap the non-matched ones.
5386 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
5388 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
5389 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
5390 _mm_maskz_scalef_round_ss): New intrinsics.
5391 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
5392 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
5393 __builtin_ia32_scalefss_round): Remove.
5394 (__builtin_ia32_scalefsd_mask_round,
5395 __builtin_ia32_scalefss_mask_round): New intrinsics.
5396 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
5397 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
5398 ((match_operand:VF_128 2 "<round_nimm_predicate>"
5399 "<round_constraint>")): Changed to ...
5400 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
5401 "<round_scalar_constraint>")): ... this.
5402 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
5403 %0, %1, %2<round_op3>}"): Changed to ...
5404 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
5405 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
5406 %2<round_scalar_mask_op3>}"): ... this.
5407 * config/i386/subst.md (round_scalar_nimm_predicate): New.
5409 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
5411 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
5412 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
5413 (_mm_maskz_sqrt_round_ss): New intrinsics.
5414 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
5415 (__builtin_ia32_sqrtsd_mask_round)
5416 (__builtin_ia32_sqrtss_mask_round): New builtins.
5417 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
5418 (__builtin_ia32_sqrtss_round): Remove.
5419 (__builtin_ia32_sqrtsd_mask_round)
5420 (__builtin_ia32_sqrtss_mask_round): New builtins.
5421 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
5422 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
5423 ((match_operand:VF_128 1 "vector_operand"
5424 "xBm,<round_constraint>")): Changed to ...
5425 ((match_operand:VF_128 1 "vector_operand"
5426 "xBm,<round_scalar_constraint>")): ... this.
5427 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
5428 %0, %2, %<iptr>1<round_op3>}): Changed to ...
5429 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
5430 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
5431 %<iptr>1<round_scalar_mask_op3>}): ... this.
5432 ((set_attr "prefix" "<round_prefix>")): Changed to ...
5433 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
5435 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
5438 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
5439 Cast vec_cmpeq result to correct type.
5440 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
5441 Cast vec_cmpgt result to correct type.
5443 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
5445 * final.c (final_scan_insn_1): Renamed from...
5446 (final_scan_insn): ... this. New wrapper, to recover
5447 seen from the outermost call in recursive ones.
5448 * config/sparc/sparc.c (output_return): Drop seen from call.
5449 (output_sibcall): Likewise.
5450 * config/visium/visium.c (output_branch): Likewise.
5452 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
5454 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
5457 2018-02-10 Alan Modra <amodra@gmail.com>
5460 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
5461 Specify LR as an input.
5463 2018-02-10 Jakub Jelinek <jakub@redhat.com>
5466 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
5467 remove_member_access_dummy_vars): New functions.
5468 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
5469 lower_omp_1, execute_lower_omp): Use them.
5471 PR rtl-optimization/84308
5472 * shrink-wrap.c (spread_components): Release todo vector.
5474 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
5476 PR rtl-optimization/57193
5477 * ira-color.c (struct allocno_color_data): Add member
5478 conflict_allocno_hard_prefs.
5479 (update_conflict_allocno_hard_prefs): New.
5480 (bucket_allocno_compare_func): Add a preference based on
5481 conflict_allocno_hard_prefs.
5482 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
5483 (color_allocnos): Remove a dead code. Initiate
5484 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
5486 2018-02-09 Jakub Jelinek <jakub@redhat.com>
5489 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
5490 constraint from =wa to wa. Avoid a subreg on the output operand,
5491 instead use a pseudo and subreg it in a move.
5492 (p9_xxbrd_<mode>): Changed to ...
5493 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
5494 (p9_xxbrd_v2df): New expander.
5495 (p9_xxbrw_<mode>): Changed to ...
5496 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
5497 (p9_xxbrw_v4sf): New expander.
5499 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
5501 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
5503 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
5506 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
5507 multiply in 32-bit mode.
5508 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
5509 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
5512 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
5514 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
5515 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
5516 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
5517 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
5519 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
5522 * dwarf2out.c (is_trivial_indirect_ref): New function.
5523 (dwarf2out_late_global_decl): Do not generate a location
5524 attribute for variables that have a non-trivial DECL_VALUE_EXPR
5525 and that are not defined in the current unit.
5527 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
5529 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
5530 instead of a libcall for UNORDERED.
5532 2018-02-09 Tamar Christina <tamar.christina@arm.com>
5535 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
5536 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
5538 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5541 * config/s390/s390.c (s390_set_current_function): Invoke
5542 s390_indirect_branch_settings also if fndecl didn't change.
5544 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
5546 * config/rs6000/rs6000.md (blockage): Set length to zero.
5548 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
5550 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
5552 2018-02-09 Jakub Jelinek <jakub@redhat.com>
5555 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
5556 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
5560 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
5561 PARALLEL incoming that failed vt_get_decl_and_offset check.
5564 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
5565 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
5567 (get_variable_section): For decls in named .bss* sections pass true as
5568 second argument to bss_initializer_p.
5570 2018-02-09 Marek Polacek <polacek@redhat.com>
5571 Jakub Jelinek <jakub@redhat.com>
5574 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
5575 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
5576 Sync some changes from cxx_fold_indirect_ref.
5578 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
5580 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
5582 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
5583 (BLOCK_INLINE_ENTRY_LABEL): New.
5584 (dwarf2out_var_location): Disregard inline entry markers.
5585 (inline_entry_data): New struct.
5586 (inline_entry_data_hasher): New hashtable type.
5587 (inline_entry_data_hasher::hash): New.
5588 (inline_entry_data_hasher::equal): New.
5589 (inline_entry_data_table): New variable.
5590 (add_high_low_attributes): Add DW_AT_entry_pc and
5591 DW_AT_GNU_entry_view attributes if a pending entry is found
5592 in inline_entry_data_table. Add old entry_pc attribute only
5593 if debug nonbinding markers are disabled.
5594 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
5595 markers are enabled.
5596 (block_within_block_p, dwarf2out_inline_entry): New.
5597 (dwarf2out_finish): Check that no entries remained in
5598 inline_entry_data_table.
5599 * final.c (reemit_insn_block_notes): Handle inline entry notes.
5600 (final_scan_insn, notice_source_line): Likewise.
5601 (rest_of_clean_state): Skip inline entry markers.
5602 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
5604 * gimple.c (gimple_build_debug_inline_entry): New.
5605 * gimple.h (enum gimple_debug_subcode): Add
5606 GIMPLE_DEBUG_INLINE_ENTRY.
5607 (gimple_build_debug_inline_entry): Declare.
5608 (gimple_debug_inline_entry_p): New.
5609 (gimple_debug_nonbind_marker_p): Adjust.
5610 * insn-notes.def (INLINE_ENTRY): New.
5611 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
5612 inline entry marker notes.
5613 (print_insn): Likewise.
5614 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
5615 (INSN_DEBUG_MARKER_KIND): Likewise.
5616 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
5617 * tree-inline.c (expand_call_inline): Build and insert
5618 debug_inline_entry stmt.
5619 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
5620 inline entry blocks early, if nonbind markers are enabled.
5621 (dump_scope_block): Dump fragment info.
5622 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
5623 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
5624 (gimple_build_debug_inline_entry): New.
5625 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
5626 Enable/disable inline entry points too.
5627 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
5628 (DEBUG_INSN): Describe inline entry markers.
5630 * common.opt (gvariable-location-views): New.
5631 (gvariable-location-views=incompat5): New.
5632 * config.in: Rebuilt.
5633 * configure: Rebuilt.
5634 * configure.ac: Test assembler for view support.
5635 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
5636 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
5637 * dwarf2out.c (var_loc_view): New typedef.
5638 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
5639 (dwarf2out_locviews_in_attribute): New.
5640 (dwarf2out_locviews_in_loclist): New.
5641 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
5642 (enum dw_line_info_opcode): Add LI_adv_address.
5643 (struct dw_line_info_table): Add view.
5644 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
5645 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
5646 (zero_view_p): New variable.
5647 (ZERO_VIEW_P): New macro.
5648 (output_asm_line_debug_info): New.
5649 (struct var_loc_node): Add view.
5650 (add_AT_view_list, AT_loc_list): New.
5651 (add_var_loc_to_decl): Add view param. Test it against last.
5652 (new_loc_list): Add view params. Record them.
5653 (AT_loc_list_ptr): Handle loc and view lists.
5654 (view_list_to_loc_list_val_node): New.
5655 (print_dw_val): Handle dw_val_class_view_list.
5656 (size_of_die): Likewise.
5657 (value_format): Likewise.
5658 (loc_list_has_views): New.
5659 (gen_llsym): Set vl_symbol too.
5660 (maybe_gen_llsym, skip_loc_list_entry): New.
5661 (dwarf2out_maybe_output_loclist_view_pair): New.
5662 (output_loc_list): Output view list or entries too.
5663 (output_view_list_offset): New.
5664 (output_die): Handle dw_val_class_view_list.
5665 (output_dwarf_version): New.
5666 (output_compilation_unit_header): Use it.
5667 (output_skeleton_debug_sections): Likewise.
5668 (output_rnglists, output_line_info): Likewise.
5669 (output_pubnames, output_aranges): Update version comments.
5670 (output_one_line_info_table): Output view numbers in asm comments.
5671 (dw_loc_list): Determine current endview, pass it to new_loc_list.
5672 Call maybe_gen_llsym.
5673 (loc_list_from_tree_1): Adjust.
5674 (add_AT_location_description): Create view list attribute if
5675 needed, check it's absent otherwise.
5676 (convert_cfa_to_fb_loc_list): Adjust.
5677 (maybe_emit_file): Call output_asm_line_debug_info for test.
5678 (dwarf2out_var_location): Reset views as needed. Precompute
5679 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
5680 attribute. Set view.
5681 (new_line_info_table): Reset next view.
5682 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
5683 (dwarf2out_source_line): Likewise. Output view resets and labels to
5684 the assembler, or select appropriate line info opcodes.
5685 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
5686 (optimize_string_length): Catch it. Adjust.
5687 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
5688 dw_val_class_view_list, and remove it if no longer needed.
5689 (hash_loc_list): Hash view numbers.
5690 (loc_list_hasher::equal): Compare them.
5691 (optimize_location_lists): Check whether a view list symbol is
5692 needed, and whether the locview attribute is present, and
5693 whether they match. Remove the locview attribute if no longer
5695 (index_location_lists): Call skip_loc_list_entry for test.
5696 (dwarf2out_finish): Call output_asm_line_debug_info for test.
5697 Use output_dwarf_version.
5698 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
5699 (struct dw_val_node): Add val_view_list.
5700 * final.c (SEEN_NEXT_VIEW): New.
5701 (set_next_view_needed): New.
5702 (clear_next_view_needed): New.
5703 (maybe_output_next_view): New.
5704 (final_start_function): Rename to...
5705 (final_start_function_1): ... this. Take pointer to FIRST,
5706 add SEEN parameter. Emit param bindings in the initial view.
5707 (final_start_function): Reintroduce SEEN-less interface.
5708 (final): Rename to...
5709 (final_1): ... this. Take SEEN parameter. Output final pending
5710 next view at the end.
5711 (final): Reintroduce seen-less interface.
5712 (final_scan_insn): Output pending next view before switching
5713 sections or ending a block. Mark the next view as needed when
5714 outputting variable locations. Notify debug backend of section
5715 changes, and of location view changes.
5716 (rest_of_handle_final): Adjust.
5717 * toplev.c (process_options): Autodetect value for debug variable
5718 location views option. Warn on incompat5 without -gdwarf-5.
5719 * doc/invoke.texi (gvariable-location-views): New.
5720 (gvariable-location-views=incompat5): New.
5721 (gno-variable-location-views): New.
5723 2018-02-08 David Malcolm <dmalcolm@redhat.com>
5725 PR tree-optimization/84136
5726 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
5727 that the result of find_edge is non-NULL.
5729 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
5732 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
5733 storing integer register in SImode. Fix cost of 256 and 512
5734 byte aligned SSE register store.
5736 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
5738 * config/i386/i386.c (ix86_multiplication_cost): Fix
5739 multiplication cost for TARGET_AVX512DQ.
5741 2018-02-08 Marek Polacek <polacek@redhat.com>
5743 PR tree-optimization/84238
5744 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
5747 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5749 PR tree-optimization/84265
5750 * tree-vect-stmts.c (vectorizable_store): Don't treat
5751 VMAT_CONTIGUOUS accesses as grouped.
5752 (vectorizable_load): Likewise.
5754 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5756 PR tree-optimization/81635
5757 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
5758 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
5759 (test_round_for_mask): New functions.
5760 (wide_int_cc_tests): Call test_round_for_mask.
5761 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
5762 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
5763 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
5764 range returned by get_range_info.
5766 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
5769 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
5770 * symtab.c: Include builtins.h
5771 (symtab_node::output_to_lto_symbol_table_p): Move here
5772 from lto-streamer-out.c:output_symbol_p.
5773 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
5774 (output_symbol_p): Move all logic to symtab.c
5775 (produce_symtab): Update.
5777 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5779 * config/s390/s390-opts.h (enum indirect_branch): Define.
5780 * config/s390/s390-protos.h (s390_return_addr_from_memory)
5781 (s390_indirect_branch_via_thunk)
5782 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
5783 (enum s390_indirect_branch_type): Define.
5784 * config/s390/s390.c (struct s390_frame_layout, struct
5785 machine_function): Remove.
5786 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
5787 (indirect_branch_table_label_no, indirect_branch_table_name):
5789 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
5790 (enum s390_indirect_branch_option): Define.
5791 (s390_return_addr_from_memory): New function.
5792 (s390_handle_string_attribute): New function.
5793 (s390_attribute_table): Add new attribute handler.
5794 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
5795 (s390_indirect_branch_via_thunk): New function.
5796 (s390_indirect_branch_via_inline_thunk): New function.
5797 (s390_function_ok_for_sibcall): When jumping via thunk disallow
5798 sibling call optimization for non z10 compiles.
5799 (s390_emit_call): Force indirect branch target to be a single
5800 register. Add r1 clobber for non-z10 compiles.
5801 (s390_emit_epilogue): Emit return jump via return_use expander.
5802 (s390_reorg): Handle JUMP_INSNs as execute targets.
5803 (s390_option_override_internal): Perform validity checks for the
5804 new command line options.
5805 (s390_indirect_branch_attrvalue): New function.
5806 (s390_indirect_branch_settings): New function.
5807 (s390_set_current_function): Invoke s390_indirect_branch_settings.
5808 (s390_output_indirect_thunk_function): New function.
5809 (s390_code_end): Implement target hook.
5810 (s390_case_values_threshold): Implement target hook.
5811 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
5813 * config/s390/s390.h (struct s390_frame_layout)
5814 (struct machine_function): Move here from s390.c.
5815 (TARGET_INDIRECT_BRANCH_NOBP_RET)
5816 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
5817 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
5818 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
5819 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
5820 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
5821 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
5822 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
5823 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
5824 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
5825 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
5826 (mnemonic attribute): Add values which aren't recognized
5828 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
5829 pattern for branch conversion. Fix mnemonic attribute.
5830 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
5831 indirect branch via thunk if requested.
5832 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
5833 ("*indirect_jump"): Disable for branch conversion using out of
5835 ("indirect_jump_via_thunk<mode>_z10")
5836 ("indirect_jump_via_thunk<mode>")
5837 ("indirect_jump_via_inlinethunk<mode>_z10")
5838 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
5839 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
5840 ("casesi_jump_via_inlinethunk<mode>_z10")
5841 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
5842 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
5843 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
5844 ("*indirect2_jump"): Disable for branch conversion.
5845 ("casesi_jump"): Turn into expander and expand patterns for branch
5847 ("return_use"): New expander.
5848 ("*return"): Emit return via thunk and rename it to ...
5849 ("*return<mode>"): ... this one.
5850 * config/s390/s390.opt: Add new options and and enum for the
5853 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5855 * lra-constraints.c (match_reload): Unconditionally use
5856 gen_lowpart_SUBREG, rather than selecting between that
5857 and equivalent gen_rtx_SUBREG code.
5859 2018-02-08 Richard Biener <rguenther@suse.de>
5861 PR tree-optimization/84233
5862 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
5863 changed flag instead of boguously re-using phi_inserted.
5865 2018-02-08 Martin Jambor <mjambor@suse.cz>
5867 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
5868 static local variables.
5870 2018-02-08 Richard Biener <rguenther@suse.de>
5872 PR tree-optimization/84278
5873 * tree-vect-stmts.c (vectorizable_store): When looking for
5874 smaller vector types to perform grouped strided loads/stores
5875 make sure the mode is supported by the target.
5876 (vectorizable_load): Likewise.
5878 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
5880 * config/aarch64/aarch64.c (aarch64_components_for_bb):
5881 Increase LDP/STP opportunities by adding adjacent callee-saves.
5883 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
5885 PR rtl-optimization/84068
5886 PR rtl-optimization/83459
5887 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
5889 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
5891 PR tree-optimization/84224
5892 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
5893 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
5896 2018-02-07 Iain Sandoe <iain@codesourcery.com>
5899 * config/rs6000/altivec.md (*restore_world): Remove LR use.
5900 * config/rs6000/predicates.md (restore_world_operation): Adjust op
5901 count, remove one USE.
5903 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
5905 * doc/install.texi (Configuration): Document the
5906 --with-long-double-format={ibm,ieee} PowerPC configuration
5910 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
5911 Convert from define_expand to be define_insn_and_split. Rework
5912 float/double/_Float128 conversions to QI/HI/SImode to work with
5913 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
5914 conversions to QI/HImode types did a store and then a load to
5915 truncate the value. For conversions to VSX registers, don't split
5916 the insn, instead emit the code directly. Use the code iterator
5917 any_fix to combine signed and unsigned conversions.
5918 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
5919 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
5920 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
5921 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
5922 (fix_<mode>di2_hw): Likewise.
5923 (fixuns_<mode>di2_hw): Likewise.
5924 (fix_<mode>si2_hw): Likewise.
5925 (fixuns_<mode>si2_hw): Likewise.
5926 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
5927 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
5928 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
5929 fix<uns>_trunc<SFDF:mode>si2_p8.
5930 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
5932 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
5933 (fix<uns>_<mode>_mem): Likewise.
5934 (fctiw<u>z_<mode>_mem): Likewise.
5935 (fix<uns>_<mode>_mem): Likewise.
5936 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
5937 the register allocator from doing a direct move to the GPRs to do
5938 a store, and instead use the ISA 3.0 store byte/half-word from
5939 vector register instruction. For IEEE 128-bit floating point,
5940 also optimize stores of 32-bit ints.
5941 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
5943 2018-02-07 Alan Hayward <alan.hayward@arm.com>
5945 * genextract.c (push_pathstr_operand): New function to support
5947 (walk_rtx): Call push_pathstr_operand.
5948 (print_path): Support [a-zA-Z].
5950 2018-02-07 Richard Biener <rguenther@suse.de>
5952 PR tree-optimization/84037
5953 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
5954 (cse_and_gimplify_to_preheader): Declare.
5955 (vect_get_place_in_interleaving_chain): Likewise.
5956 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5958 (_loop_vec_info::~_loop_vec_info): Delete it.
5959 (cse_and_gimplify_to_preheader): New function.
5960 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
5961 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
5962 (vectorizable_load): Likewise. For grouped stores always base
5963 the IV on the first element.
5964 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
5965 condition before gimplifying.
5967 2018-02-07 Jakub Jelinek <jakub@redhat.com>
5969 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
5970 *DIV_EXPR and *MOD_EXPR.
5972 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
5975 * config/i386/i386.c (ix86_option_override_internal): Mask out
5976 the CF_SET bit when checking -fcf-protection.
5978 2018-02-07 Tom de Vries <tom@codesourcery.com>
5981 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
5984 2018-02-07 Richard Biener <rguenther@suse.de>
5986 PR tree-optimization/84204
5987 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
5990 PR tree-optimization/84205
5991 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
5992 special-case isl_ast_op_zdiv_r.
5994 PR tree-optimization/84223
5995 * graphite-scop-detection.c (gather_bbs::before_dom_children):
5996 Only add conditions from within the region.
5997 (gather_bbs::after_dom_children): Adjust.
5999 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
6002 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
6003 * config/avr/avr.md: Only post-reload split REG-REG moves if
6004 either register is GENERAL_REG_P.
6006 2018-02-07 Jakub Jelinek <jakub@redhat.com>
6008 PR tree-optimization/84235
6009 * tree-ssa-scopedtables.c
6010 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
6011 if the subtraction is performed in floating point type where NaNs are
6012 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
6013 build 1. Formatting fix.
6015 2018-02-06 Jakub Jelinek <jakub@redhat.com>
6018 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
6019 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
6020 and skip it regardless of bb boundaries. Use CALL_P macro,
6021 don't test INSN_P (insn) together with CALL_P or JUMP_P check
6022 unnecessarily, formatting fix.
6024 2018-02-06 Michael Collison <michael.collison@arm.com>
6026 * config/arm/thumb2.md:
6027 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
6028 (*thumb_mov_notscc): Ditto.
6030 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
6033 * config/rs6000/rs6000.md (su code attribute): Use "u" for
6034 unsigned_fix, not "s".
6036 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6038 * configure.ac (gcc_fn_eh_frame_ro): New function.
6039 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
6040 correct .eh_frame permissions.
6041 * configure: Regenerate.
6043 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
6045 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
6048 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6050 * config/rs6000/rs6000.c (rs6000_option_override_internal):
6051 Display warning message for -mno-speculate-indirect-jumps.
6053 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
6055 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
6057 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
6059 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
6061 PR tree-optimization/84225
6062 * tree-eh.c (find_trapping_overflow): Only call
6063 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
6065 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6068 * config/i386/i386.c: Reimplement the check of possible options
6069 -mibt/-mshstk conbination. Change error messages.
6070 * doc/invoke.texi: Fix a typo: remove extra '='.
6072 2018-02-06 Marek Polacek <polacek@redhat.com>
6074 PR tree-optimization/84228
6075 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
6077 2018-02-06 Tamar Christina <tamar.christina@arm.com>
6080 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
6081 emitted arch directives.
6082 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
6083 __ARM_FEATURE_COPROC before changing architectures.
6085 2018-02-06 Richard Biener <rguenther@suse.de>
6087 * config/i386/i386.c (print_reg): Fix typo.
6088 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
6090 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
6092 * configure: Regenerate.
6094 2018-02-05 Martin Sebor <msebor@redhat.com>
6096 PR tree-optimization/83369
6097 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
6100 2018-02-05 Martin Liska <mliska@suse.cz>
6102 * doc/invoke.texi: Cherry-pick upstream r323995.
6104 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
6106 * ira.c (ira_init_register_move_cost): Adjust comment.
6108 2018-02-05 Martin Liska <mliska@suse.cz>
6110 PR gcov-profile/84137
6111 * doc/gcov.texi: Fix typo in documentation.
6113 2018-02-05 Martin Liska <mliska@suse.cz>
6115 PR gcov-profile/83879
6116 * doc/gcov.texi: Document necessity of --dynamic-list-data when
6117 using dlopen functionality.
6119 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
6121 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
6122 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
6123 _mm_maskz_range_ss, _mm_mask_range_round_ss,
6124 _mm_maskz_range_round_ss): New intrinsics.
6125 (__builtin_ia32_rangesd128_round)
6126 (__builtin_ia32_rangess128_round): Remove.
6127 (__builtin_ia32_rangesd128_mask_round,
6128 __builtin_ia32_rangess128_mask_round): New builtins.
6129 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
6130 __builtin_ia32_rangess128_round): Remove.
6131 (__builtin_ia32_rangesd128_mask_round,
6132 __builtin_ia32_rangess128_mask_round): New builtins.
6133 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
6134 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
6135 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
6136 "<round_saeonly_constraint>")): Changed to ...
6137 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
6138 "<round_saeonly_scalar_constraint>")): ... this.
6139 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
6140 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
6141 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
6142 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
6143 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
6145 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
6147 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
6149 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
6150 Remove all values except native, 8540 and 8548.
6152 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
6154 * config/i386/i386.c (ix86_output_function_return): Pass
6155 INVALID_REGNUM, instead of -1, as invalid register number to
6156 indirect_thunk_name and output_indirect_thunk.
6158 2018-02-02 Julia Koval <julia.koval@intel.com>
6160 * config.gcc: Add -march=icelake.
6161 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
6162 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
6163 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
6164 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
6165 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
6166 (processor_target_table): Add icelake.
6167 (ix86_option_override_internal): Handle new PTAs.
6168 (get_builtin_code_for_version): Handle icelake.
6169 (M_INTEL_COREI7_ICELAKE): New.
6170 (fold_builtin_cpu): Handle icelake.
6171 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
6172 * doc/invoke.texi: Add -march=icelake.
6174 2018-02-02 Julia Koval <julia.koval@intel.com>
6176 * config/i386/i386.c (ix86_option_override_internal): Change flags type
6177 to wide_int_bitmask.
6178 * wide-int-bitmask.h: New.
6180 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6183 * config/i386/i386.md: Replace Pmode with word_mode in
6184 builtin_setjmp_setup and builtin_longjmp to support x32.
6186 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
6190 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
6192 (rs6000_supported_cpu_names): New static variable.
6193 (linux_cpu_translation_table): Likewise.
6194 (elf_platform) <cpu>: Define new static variable and use it.
6195 Translate kernel AT_PLATFORM name to canonical name if needed.
6196 Error if platform name is unknown.
6198 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
6201 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
6203 2018-02-01 Jeff Law <law@redhat.com>
6206 * config/i386/i386.c (release_scratch_register_on_entry): Add new
6207 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
6208 the scratch if RELEASE_VIA_POP is false.
6209 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
6210 If we have to save a temporary register, decrement SIZE appropriately.
6211 Pass new arguments to release_scratch_register_on_entry.
6212 (ix86_adjust_stack_and_probe): Likewise.
6213 (ix86_emit_probe_stack_range): Pass new arguments to
6214 release_scratch_register_on_entry.
6216 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
6218 PR rtl-optimization/84157
6219 * combine.c (change_zero_ext): Use REG_P predicate in
6220 front of HARD_REGISTER_P predicate.
6222 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
6224 * config/avr/avr.c (avr_option_override): Move disabling of
6225 -fdelete-null-pointer-checks to...
6226 * common/config/avr/avr-common.c (avr_option_optimization_table):
6229 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6231 PR tree-optimization/81635
6232 * tree-data-ref.c (split_constant_offset_1): For types that
6233 wrap on overflow, try to use range info to prove that wrapping
6236 2018-02-01 Renlin Li <renlin.li@arm.com>
6239 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
6241 (aarch64_register_move_cost): Likewise.
6242 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
6244 (REG_CLASS_NAMES): Likewise.
6245 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
6246 TAILCALL_ADDR_REGS. Remove IP registers.
6247 * config/aarch64/aarch64.md (Ucs): Update register constraint.
6249 2018-02-01 Richard Biener <rguenther@suse.de>
6251 * domwalk.h (dom_walker::dom_walker): Add additional constructor
6252 for specifying RPO order and allow NULL for that.
6253 * domwalk.c (dom_walker::dom_walker): Likewise.
6254 (dom_walker::walk): Handle NULL RPO order.
6255 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
6257 (rewrite_update_dom_walker): Likewise.
6258 (mark_def_dom_walker): Likewise.
6260 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6262 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
6263 (aarch64_maybe_expand_sve_subreg_move): Declare.
6264 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
6265 * config/aarch64/predicates.md (aarch64_any_register_operand): New
6267 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
6268 that are semantically a reverse operation.
6269 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
6270 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
6271 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
6273 (aarch64_can_change_mode_class): For big-endian, forbid changes
6274 between two SVE modes if they have different element sizes.
6276 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6278 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
6279 the TImode handling for big-endian targets.
6281 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6283 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
6284 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
6286 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
6287 Remove BSWAP handing for big-endian targets and use the form of
6288 LD1RQ appropariate for the mode.
6290 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6292 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
6293 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
6296 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6299 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
6300 check for operands that need to go through aarch64_sve_reload_be.
6302 2018-02-01 Jakub Jelinek <jakub@redhat.com>
6304 PR tree-optimization/81661
6305 PR tree-optimization/84117
6306 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
6307 * tree-eh.c: Include gimplify.h.
6308 (find_trapping_overflow, replace_trapping_overflow,
6309 rewrite_to_non_trapping_overflow): New functions.
6310 * tree-vect-loop.c: Include tree-eh.h.
6311 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
6312 * tree-data-ref.c: Include tree-eh.h.
6313 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
6315 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
6317 PR rtl-optimization/84123
6318 * combine.c (change_zero_ext): Check if hard register satisfies
6319 can_change_dest_mode before calling gen_lowpart_SUBREG.
6321 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
6324 * ira.c (ira_init_register_move_cost): Remove assert.
6326 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
6328 PR rtl-optimization/84071
6329 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
6330 * doc/tm.texi: Regenerate.
6332 2018-01-31 Richard Biener <rguenther@suse.de>
6334 PR tree-optimization/84132
6335 * tree-data-ref.c (analyze_miv_subscript): Properly
6336 check whether evolution_function_is_affine_multivariate_p
6337 before calling gcd_of_steps_may_divide_p.
6339 2018-01-31 Julia Koval <julia.koval@intel.com>
6342 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
6343 * config/i386/i386.md (rdpid_rex64) New.
6344 (rdpid): Make 32bit only.
6346 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
6349 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
6350 an IDENTIFIER_NODE for FUNCTION_TYPE's.
6352 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
6355 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
6357 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
6359 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
6361 PR rtl-optimization/84071
6362 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
6363 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
6365 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
6367 * config/arc/arc.c (arc_handle_aux_attribute): New function.
6368 (arc_attribute_table): Add 'aux' attribute.
6369 (arc_in_small_data_p): Consider aux like variables.
6370 (arc_is_aux_reg_p): New function.
6371 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
6372 (arc_get_aux_arg): New function.
6373 (prepare_move_operands): Handle aux-register access.
6374 (arc_handle_aux_attribute): New function.
6375 * doc/extend.texi (ARC Variable attributes): Add subsection.
6377 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
6379 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
6380 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
6381 (arc_attribute_table): Add 'uncached' attribute.
6382 (arc_print_operand): Print '.di' flag for uncached memory
6384 (arc_in_small_data_p): Do not consider for small data the uncached
6386 (arc_is_uncached_mem_p): New function.
6387 * config/arc/predicates.md (compact_store_memory_operand): Check
6388 for uncached memory accesses.
6389 (nonvol_nonimm_operand): Likewise.
6390 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
6392 2018-01-31 Jakub Jelinek <jakub@redhat.com>
6395 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
6396 falign-loops=): Add Optimization flag.
6398 2018-01-30 Jeff Law <law@redhat.com>
6401 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
6402 INT_REGISTERS_SAVED. Check it prior to calling
6403 get_scratch_register_on_entry.
6404 (ix86_adjust_stack_and_probe): Similarly.
6405 (ix86_emit_probe_stack_range): Similarly.
6406 (ix86_expand_prologue): Corresponding changes.
6408 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6411 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
6412 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
6414 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
6417 * lra-constraints.c (curr_insn_transform): Process AND in the
6420 2018-01-30 Jakub Jelinek <jakub@redhat.com>
6422 PR rtl-optimization/83986
6423 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
6424 dependence against last_pending_memory_flush in addition to
6427 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
6429 PR tree-optimization/81611
6430 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
6433 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6436 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
6439 2018-01-30 Richard Biener <rguenther@suse.de>
6440 Jakub Jelinek <jakub@redhat.com>
6442 PR tree-optimization/84111
6443 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
6444 inner loops added during recursion, as they don't have up-to-date
6447 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
6450 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
6451 (can_inline_edge_by_limits_p): ... here.
6452 (can_early_inline_edge_p, check_callers,
6453 update_caller_keys, update_callee_keys, recursive_inlining,
6454 add_new_edges_to_heap, speculation_useful_p,
6455 inline_small_functions,
6456 inline_small_functions, flatten_function,
6457 inline_to_all_callers_1): Update.
6459 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
6461 * profile-count.c (profile_count::combine_with_ipa_count): Handle
6464 2018-01-30 Richard Biener <rguenther@suse.de>
6466 PR tree-optimization/83008
6467 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
6468 invariant and constant vector uses in stmts when they need
6471 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6474 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
6475 * configure: Regenerate.
6477 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
6479 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
6481 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
6482 Use gen_rtx_REG rather than gen_lowpart.
6484 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
6486 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
6487 rather than 0 when creating partial subregs.
6489 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
6491 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
6494 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
6497 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
6498 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
6499 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
6500 flags. This restores the settings used before the 2017-07-24.
6501 Turning off pre increment/decrement/modify allows IVOPTS to
6502 optimize DF/SF loops where the index is an int.
6504 2018-01-29 Richard Biener <rguenther@suse.de>
6505 Kelvin Nilsen <kelvin@gcc.gnu.org>
6508 * tree-vect-stmts.c (vectorizable_call): Don't call
6509 targetm.vectorize_builtin_md_vectorized_function if callee is
6512 2018-01-22 Carl Love <cel@us.ibm.com>
6514 * doc/extend.tex: Fix typo in second arg in
6515 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
6517 2018-01-29 Richard Biener <rguenther@suse.de>
6519 PR tree-optimization/84086
6520 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
6521 (flush_ssaname_freelist): When SSA names were released reset
6522 the SCEV hash table.
6524 2018-01-29 Richard Biener <rguenther@suse.de>
6526 PR tree-optimization/84057
6527 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
6528 removed paths when removing edges.
6530 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
6532 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
6533 -mfunction-return=@var{choice}.
6535 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
6538 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
6540 (layout::print_source_line): Likewise.
6541 (test_get_line_width_without_trailing_whitespace): Add test cases.
6543 2018-01-27 Jakub Jelinek <jakub@redhat.com>
6546 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
6549 2018-01-26 Jim Wilson <jimw@sifive.com>
6551 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
6553 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
6556 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6558 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
6559 and CMP + SUB-immediate -> SUBS.
6561 2018-01-26 Martin Sebor <msebor@redhat.com>
6563 PR tree-optimization/83896
6564 * tree-ssa-strlen.c (get_string_len): Rename...
6565 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
6566 Avoid assuming length is constant.
6567 (handle_char_store): Use HOST_WIDE_INT for string length.
6569 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
6572 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
6573 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
6575 2018-01-26 Richard Biener <rguenther@suse.de>
6577 PR rtl-optimization/84003
6578 * dse.c (record_store): Only record redundant stores when
6579 the earlier store aliases at least all accesses the later one does.
6581 2018-01-26 Jakub Jelinek <jakub@redhat.com>
6583 PR rtl-optimization/83985
6584 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
6585 REG_CFA_RESTORE insns.
6586 (delete_unmarked_insns): Don't ignore separate shrink wrapping
6587 REG_CFA_RESTORE insns here.
6590 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
6591 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
6593 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6595 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
6596 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
6597 (arc_init): Likewise.
6598 (arc_override_options): Likewise.
6599 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
6601 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
6603 * config/arc/arc.h (TARGET_DBNZ): Define.
6604 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
6605 properly set the tune attribute.
6606 (dbnz): Use TARGET_DBNZ guard.
6607 * config/arc/arc.opt (mtune): Add core3 option.
6609 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6611 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
6612 recognize new pic like addresses.
6613 (arc_delegitimize_address): Clean up.
6615 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6617 * config/arc/arc-arches.def: Option mrf16 valid for all
6619 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
6620 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
6621 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
6622 * config/arc/arc-tables.opt: Regenerate.
6623 * config/arc/arc.c (arc_conditional_register_usage): Handle
6624 reduced register file case.
6625 (arc_file_start): Set must have build attributes.
6626 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
6628 * config/arc/arc.opt (mrf16): Add new option.
6629 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
6630 * config/arc/genmultilib.awk: Handle new mrf16 option.
6631 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
6632 * config/arc/t-multilib: Regenerate.
6633 * doc/invoke.texi (ARC Options): Document mrf16 option.
6635 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6637 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
6638 * config/arc/arc.c (arc_handle_secure_attribute): New function.
6639 (arc_attribute_table): Add 'secure_call' attribute.
6640 (arc_print_operand): Print secure call operand.
6641 (arc_function_ok_for_sibcall): Don't optimize tail calls when
6643 (arc_is_secure_call_p): New function. * config/arc/arc.md
6644 (call_i): Add support for sjli instruction.
6645 (call_value_i): Likewise.
6646 * config/arc/constraints.md (Csc): New constraint.
6648 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6649 John Eric Martin <John.Martin@emmicro-us.com>
6651 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
6652 * config/arc/arc.c (_arc_jli_section): New struct.
6653 (arc_jli_section): New type.
6654 (rc_jli_sections): New static variable.
6655 (arc_handle_jli_attribute): New function.
6656 (arc_attribute_table): Add jli_always and jli_fixed attribute.
6657 (arc_file_end): New function.
6658 (TARGET_ASM_FILE_END): Define.
6659 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
6660 (arc_add_jli_section): New function.
6661 (jli_call_scan): Likewise.
6662 (arc_reorg): Call jli_call_scan.
6663 (arc_output_addsi): Remove 'S' from printing asm operand.
6664 (arc_is_jli_call_p): New function.
6665 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
6667 (movhi_insn): Likewise.
6668 (movsi_insn): Likewise.
6669 (movsi_set_cc_insn): Likewise.
6670 (loadqi_update): Likewise.
6671 (load_zeroextendqisi_update): Likewise.
6672 (load_signextendqisi_update): Likewise.
6673 (loadhi_update): Likewise.
6674 (load_zeroextendhisi_update): Likewise.
6675 (load_signextendhisi_update): Likewise.
6676 (loadsi_update): Likewise.
6677 (loadsf_update): Likewise.
6678 (movsicc_insn): Likewise.
6679 (bset_insn): Likewise.
6680 (bxor_insn): Likewise.
6681 (bclr_insn): Likewise.
6682 (bmsk_insn): Likewise.
6683 (bicsi3_insn): Likewise.
6684 (cmpsi_cc_c_insn): Likewise.
6685 (movsi_ne): Likewise.
6686 (movsi_cond_exec): Likewise.
6687 (clrsbsi2): Likewise.
6698 (call_i): Remove 'S' asm letter, add jli instruction.
6699 (call_value_i): Likewise.
6700 * config/arc/arc.op (mjli-always): New option.
6701 * config/arc/constraints.md (Cji): New constraint.
6702 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
6704 (subsf3_fpx): Likewise.
6705 (mulsf3_fpx): Likewise.
6706 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
6708 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
6710 * doc/invoke.texi (ARC): Document mjli-always option.
6712 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
6714 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
6715 avoid addition with 0 and use incw and decw where possible.
6717 2018-01-26 Richard Biener <rguenther@suse.de>
6719 PR tree-optimization/81082
6720 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
6721 association if it requires casting to unsigned.
6722 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
6723 from fold_plusminus_mult_expr to catch important cases late when
6724 range info is available.
6726 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6728 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
6729 * configure.ac (hidden_linkonce): New test.
6730 * configure: Regenerate.
6731 * config.in: Regenerate.
6733 2018-01-26 Julia Koval <julia.koval@intel.com>
6735 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
6736 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
6737 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
6738 _mm_mask_bitshuffle_epi64_mask): Fix type.
6739 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
6740 USI_FTYPE_V4DI_V4DI_USI): Remove.
6741 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
6742 __builtin_ia32_vpshufbitqmb256_mask,
6743 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
6744 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
6745 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
6747 2018-01-26 Alan Modra <amodra@gmail.com>
6750 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
6751 UNSPEC_VBPERMQ. Sort other unspecs.
6753 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
6755 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
6757 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
6760 * predict.c (drop_profile): Do not push/pop cfun; update also
6762 (handle_missing_profiles): Fix logic looking for zero profiles.
6764 2018-01-25 Jakub Jelinek <jakub@redhat.com>
6767 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
6768 on functions with #pragma omp declare simd or functions with simd
6770 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
6771 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
6772 Remove trailing \n from warning_at calls.
6774 2018-01-25 Tom de Vries <tom@codesourcery.com>
6777 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
6778 for neutered workers.
6780 2018-01-24 Joseph Myers <joseph@codesourcery.com>
6783 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
6784 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
6786 2018-01-24 Jeff Law <law@redhat.com>
6789 * i386.c (get_probe_interval): Move to earlier point.
6790 (ix86_compute_frame_layout): If -fstack-clash-protection and
6791 the frame is larger than the probe interval, then use pushes
6792 to save registers rather than reg->mem moves.
6793 (ix86_expand_prologue): Remove conditional for int_registers_saved
6796 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
6799 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
6800 min/max for never referenced object.
6802 2018-01-24 Jakub Jelinek <jakub@redhat.com>
6805 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
6807 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
6808 attributes from DECL_ATTRIBUTES (decl) without affecting
6809 DECL_ATTRIBUTES (current_function_decl).
6810 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
6811 functions with non-NULL DECL_ABSTRACT_ORIGIN.
6813 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
6815 PR tree-optimization/83979
6816 * fold-const.c (fold_comparison): Use constant_boolean_node
6817 instead of boolean_{true,false}_node.
6819 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
6821 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
6824 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6826 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
6827 Simplify the clause that sets the length attribute.
6828 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
6829 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
6830 clause that sets the length attribute.
6831 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
6833 2018-01-24 Tom de Vries <tom@codesourcery.com>
6836 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
6837 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
6838 Add strict parameter.
6839 (prevent_branch_around_nothing): Insert dummy insn between branch to
6840 label and label with no ptx insn inbetween.
6841 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
6843 2018-01-24 Tom de Vries <tom@codesourcery.com>
6846 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
6847 for neutered threads in warp.
6848 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
6850 2018-01-24 Richard Biener <rguenther@suse.de>
6852 PR tree-optimization/83176
6853 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
6856 2018-01-24 Richard Biener <rguenther@suse.de>
6858 PR tree-optimization/82819
6859 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
6860 code generating pluses that are no-ops in the target precision.
6862 2018-01-24 Richard Biener <rguenther@suse.de>
6865 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
6867 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
6869 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
6870 to merge probabilities.
6871 * predict.c (probably_never_executed): Also mark as cold functions
6872 with global 0 profile and guessed local profile.
6873 * profile-count.c (profile_probability::combine_with_count): New
6875 * profile-count.h (profile_probability::operator*,
6876 profile_probability::operator*=, profile_probability::operator/,
6877 profile_probability::operator/=): Reduce precision to adjusted
6878 and set value to guessed on contradictory divisions.
6879 (profile_probability::combine_with_freq): Remove.
6880 (profile_probability::combine_wiht_count): Declare.
6881 (profile_count::force_nonzero):: Set to adjusted.
6882 (profile_count::probability_in):: Set quality to adjusted.
6883 * tree-ssa-tail-merge.c (replace_block_by): Use
6886 2018-01-23 Andrew Waterman <andrew@sifive.com>
6887 Jim Wilson <jimw@sifive.com>
6889 * config/riscv/riscv.c (riscv_stack_boundary): New.
6890 (riscv_option_override): Set riscv_stack_boundary. Handle
6891 riscv_preferred_stack_boundary_arg.
6892 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
6893 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
6894 (STACK_BOUNDARY): Set to riscv_stack_boundary.
6895 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
6896 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
6897 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
6899 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
6902 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
6903 of struct ix86_frame.
6904 (ix86_expand_epilogue): Likewise. Add a local variable for
6905 the reg_save_offset field in struct ix86_frame.
6907 2018-01-23 Bin Cheng <bin.cheng@arm.com>
6909 PR tree-optimization/82604
6910 * tree-loop-distribution.c (enum partition_kind): New enum item
6911 PKIND_PARTIAL_MEMSET.
6912 (partition_builtin_p): Support above new enum item.
6913 (generate_code_for_partition): Ditto.
6914 (compute_access_range): Differentiate cases that equality can be
6915 proven at all loops, the innermost loops or no loops.
6916 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
6917 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
6918 (finalize_partitions, distribute_loop): Don't fuse partition of
6919 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
6920 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
6923 2018-01-23 Martin Liska <mliska@suse.cz>
6925 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
6926 order to ignore the predictor.
6927 (PRED_POLYMORPHIC_CALL): Likewise.
6928 (PRED_RECURSIVE_CALL): Likewise.
6930 2018-01-23 Martin Liska <mliska@suse.cz>
6932 * tree-profile.c (tree_profiling): Print function header to
6933 aware reader which function we are working on.
6934 * value-prof.c (gimple_find_values_to_profile): Do not print
6935 not interesting value histograms.
6937 2018-01-23 Martin Liska <mliska@suse.cz>
6939 * profile-count.h (enum profile_quality): Add
6940 profile_uninitialized as the first value. Do not number values
6941 as they are zero based.
6942 (profile_count::verify): Update sanity check.
6943 (profile_probability::verify): Likewise.
6945 2018-01-23 Nathan Sidwell <nathan@acm.org>
6947 * doc/invoke.texi (ffor-scope): Deprecate.
6949 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6951 PR tree-optimization/83510
6952 * domwalk.c (set_all_edges_as_executable): New function.
6953 (dom_walker::dom_walker): Convert bool param
6954 "skip_unreachable_blocks" to enum reachability. Move setup of
6955 edge flags to set_all_edges_as_executable and only do it when
6956 reachability is REACHABLE_BLOCKS.
6957 * domwalk.h (enum dom_walker::reachability): New enum.
6958 (dom_walker::dom_walker): Convert bool param
6959 "skip_unreachable_blocks" to enum reachability.
6960 (set_all_edges_as_executable): New decl.
6961 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
6962 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
6964 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
6965 but converting true to REACHABLE_BLOCKS.
6966 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
6968 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
6969 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
6970 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
6972 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
6973 if check_all_array_refs will be called.
6975 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6977 * tree.c (selftest::test_location_wrappers): Add more test
6980 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6982 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
6983 (selftest::test_bit_in_range): Likewise.
6985 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
6988 * doc/sourcebuild.texi (vect_float): Say that the selector
6989 only describes the situation when -funsafe-math-optimizations is on.
6990 (vect_float_strict): Document.
6992 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
6994 PR tree-optimization/83965
6995 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
6996 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
6997 instead of checking only for a reduction.
6998 (vect_recog_widen_sum_pattern): Likewise.
7000 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7002 * predict.c (probably_never_executed): Only use precise profile info.
7003 (compute_function_frequency): Skip after inlining hack since we now
7004 have quality checking.
7006 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7008 * profile-count.h (profile_probability::very_unlikely,
7009 profile_probability::unlikely, profile_probability::even): Set
7010 precision to guessed.
7012 2018-01-23 Richard Biener <rguenther@suse.de>
7014 PR tree-optimization/83963
7015 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
7016 Properly terminate dominator walk when crossing the exit edge not
7017 when visiting its source block.
7019 2018-01-23 Jakub Jelinek <jakub@redhat.com>
7022 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
7023 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
7025 2018-01-22 Jakub Jelinek <jakub@redhat.com>
7027 PR tree-optimization/83957
7028 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
7029 semicolon after for body surrounded by braces.
7031 PR tree-optimization/83081
7032 * profile-count.h (profile_probability::split): New method.
7033 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
7034 Use profile_probability::split.
7035 (do_compare_rtx_and_jump): Fix adjustment of probabilities
7036 when splitting a single conditional jump into 2.
7038 2018-01-22 David Malcolm <dmalcolm@redhat.com>
7040 PR tree-optimization/69452
7041 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
7044 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7046 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
7047 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
7048 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
7050 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7052 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
7053 * config/rl78/rl78.md (movdi): New define_expand.
7054 * config/rl78/rl78.c (rl78_split_movdi): New function.
7056 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
7059 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
7061 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
7062 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
7063 128-bit to produce an UNSPEC move to get the double word with the
7064 signbit and then a shift directly to do signbit.
7065 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
7066 implementation with a new version that just does either a direct
7067 move or a regular move. Move memory interface to separate insns.
7068 Move insns so they are next to the expander.
7069 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
7070 with signbit move. Split big and little endian case.
7071 (signbit<mode>2_dm_mem_le): Likewise.
7072 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
7073 (signbit<mode>2_dm2): Likewise.
7075 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7077 * config/rl78/rl78.md (anddi3): New define_expand.
7079 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7081 * config/rl78/rl78.md (umindi3): New define_expand.
7083 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7085 * config/rl78/rl78.md (smindi3): New define_expand.
7087 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7089 * config/rl78/rl78.md (smaxdi3): New define_expand.
7091 2018-01-22 Carl Love <cel@us.ibm.com>
7093 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
7094 LVX_V1TI): Add macro expansion.
7095 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
7096 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
7097 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
7098 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
7099 Change check to determine if the instruction is a byte reversing
7100 entry. Fix typo in comment.
7101 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
7102 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
7103 Add def_builtin calls for new builtins.
7104 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
7105 Add define_insn expansion.
7107 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7109 * config/rl78/rl78.md (umaxdi3): New define_expand.
7111 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7113 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
7114 for non-QImode registers.
7116 2018-01-22 Richard Biener <rguenther@suse.de>
7118 PR tree-optimization/83963
7119 * graphite-scop-detection.c (scop_detection::get_sese): Delay
7120 including the loop exit block.
7121 (scop_detection::merge_sese): Likewise.
7122 (scop_detection::add_scop): Do it here instead.
7124 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7126 * doc/sourcebuild.texi (arm_softfloat): Document.
7128 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
7131 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
7132 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
7133 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
7135 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7136 David Edelsohn <dje.gcc@gmail.com>
7139 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7140 Change "crset eq" to "crset 2".
7141 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7142 (*call_indirect_aix<mode>_nospec): Likewise.
7143 (*call_value_indirect_aix<mode>_nospec): Likewise.
7144 (*call_indirect_elfv2<mode>_nospec): Likewise.
7145 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
7146 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
7147 change assembly output from . to $.
7148 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7149 (indirect_jump<mode>_nospec): Change assembly output from . to $.
7150 (*tablejump<mode>_internal1_nospec): Likewise.
7152 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
7155 * config/sh/sh_optimize_sett_clrt.cc:
7156 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
7158 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
7160 PR tree-optimization/83940
7161 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
7162 offset_dt to vect_constant_def rather than vect_unknown_def_type.
7163 (vect_check_load_store_mask): Add a mask_dt_out parameter and
7164 use it to pass back the definition type.
7165 (vect_check_store_rhs): Likewise rhs_dt_out.
7166 (vect_build_gather_load_calls): Add a mask_dt argument and use
7167 it instead of a call to vect_is_simple_use.
7168 (vectorizable_store): Update calls to vect_check_load_store_mask
7169 and vect_check_store_rhs. Use the dt returned by the latter instead
7170 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
7171 instead of calls to vect_is_simple_use. Pass the scalar rather
7172 than the vector operand to vect_is_simple_use when handling
7173 second and subsequent copies of an rhs value.
7174 (vectorizable_load): Update calls to vect_check_load_store_mask
7175 and vect_build_gather_load_calls. Use the cached mask_dt and
7176 gs_info.offset_dt instead of calls to vect_is_simple_use.
7178 2018-01-20 Jakub Jelinek <jakub@redhat.com>
7181 * tree-emutls.c: Include gimplify.h.
7182 (lower_emutls_2): New function.
7183 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
7184 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
7185 it before further processing.
7188 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
7189 UINTVAL (trueop1) instead of INTVAL (op1).
7191 2018-01-19 Jakub Jelinek <jakub@redhat.com>
7195 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
7196 INCOMING_FRAME_SP_OFFSET if not defined.
7197 (scan_trace): Add ENTRY argument. If true and
7198 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
7199 emit a note to adjust the CFA offset.
7200 (create_cfi_notes): Adjust scan_trace callers.
7201 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
7202 INCOMING_FRAME_SP_OFFSET in the CIE.
7203 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
7204 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
7206 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
7207 * doc/tm.texi: Regenerated.
7209 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7211 PR rtl-optimization/83147
7212 * lra-constraints.c (remove_inheritance_pseudos): Use
7213 lra_substitute_pseudo_within_insn.
7215 2018-01-19 Tom de Vries <tom@codesourcery.com>
7216 Cesar Philippidis <cesar@codesourcery.com>
7219 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
7221 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
7224 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
7225 spaces for function labels.
7227 2018-01-19 Martin Liska <mliska@suse.cz>
7229 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
7230 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
7231 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
7232 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
7233 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
7234 (PRED_CONST_RETURN): Change from 69 to 65.
7235 (PRED_NULL_RETURN): Change from 91 to 71.
7236 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
7237 (PRED_LOOP_GUARD): Change from 66 to 73.
7239 2018-01-19 Martin Liska <mliska@suse.cz>
7241 * predict.c (predict_insn_def): Add new assert.
7242 (struct branch_predictor): Change type to signed integer.
7243 (test_prediction_value_range): Amend test to cover
7245 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
7246 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
7247 (PRED_LOOP_ITERATIONS_MAX): Likewise.
7248 (PRED_LOOP_IV_COMPARE): Likewise.
7249 * predict.h (PROB_UNINITIALIZED): Define new constant.
7251 2018-01-19 Martin Liska <mliska@suse.cz>
7253 * predict.c (dump_prediction): Add new format for
7254 analyze_brprob.py script which is enabled with -details
7256 * profile-count.h (precise_p): New function.
7258 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
7260 PR tree-optimization/83922
7261 * tree-vect-loop.c (vect_verify_full_masking): Return false if
7262 there are no statements that need masking.
7263 (vect_active_double_reduction_p): New function.
7264 (vect_analyze_loop_operations): Use it when handling phis that
7265 are not in the loop header.
7267 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
7269 PR tree-optimization/83914
7270 * tree-vect-loop.c (vectorizable_induction): Don't convert
7271 init_expr or apply the peeling adjustment for inductions
7272 that are nested within the vectorized loop.
7274 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7276 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
7279 2018-01-18 Jakub Jelinek <jakub@redhat.com>
7283 * function.h (gimplify_parameters): Add gimple_seq * argument.
7284 * function.c: Include gimple.h and options.h.
7285 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
7286 for the added local temporaries if needed.
7287 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
7288 if there are any parameter cleanups, wrap whole body into a
7289 try/finally with the cleanups.
7291 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
7294 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
7295 Use GET_MODE_CLASS for scalar floating point.
7297 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
7301 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
7302 Fix call of call_cgraph_insertion_hooks.
7304 2018-01-18 Martin Sebor <msebor@redhat.com>
7306 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
7308 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
7311 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
7314 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
7317 * common.opt: (-ffile-prefix-map): New option.
7318 * opts.c (common_handle_option): Defer it.
7319 * opts-global.c (handle_common_deferred_options): Handle it.
7320 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
7321 * file-prefix-map.h: New file.
7322 (remap_debug_filename, add_debug_prefix_map): ...here.
7323 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
7324 * final.c (debug_prefix_map, add_debug_prefix_map
7325 remap_debug_filename): Move to...
7326 * file-prefix-map.c: New file.
7327 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
7328 generalize, get rid of alloca(), use strrchr() instead of strchr().
7329 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
7330 Implement in terms of add_prefix_map().
7331 (remap_macro_filename, remap_debug_filename): Implement in term of
7333 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
7334 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
7335 * dbxout.c: Include file-prefix-map.h.
7336 * varasm.c: Likewise.
7337 * vmsdbgout.c: Likewise.
7338 * xcoffout.c: Likewise.
7339 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
7340 * doc/cppopts.texi (-fmacro-prefix-map): Document.
7341 * doc/invoke.texi (-ffile-prefix-map): Document.
7342 (-fdebug-prefix-map): Update description.
7344 2018-01-18 Martin Liska <mliska@suse.cz>
7346 * config/i386/i386.c (indirect_thunk_name): Document that also
7348 (output_indirect_thunk): Document why both instructions
7349 (pause and lfence) are generated.
7351 2018-01-18 Richard Biener <rguenther@suse.de>
7353 PR tree-optimization/83887
7354 * graphite-scop-detection.c
7355 (scop_detection::get_nearest_dom_with_single_entry): Remove.
7356 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
7357 (scop_detection::merge_sese): Re-implement with a flood-fill
7358 algorithm that properly finds a SESE region if it exists.
7360 2018-01-18 Jakub Jelinek <jakub@redhat.com>
7363 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
7364 pointer_diff optimizations use view_convert instead of convert.
7366 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7368 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7369 Generate different code for -mno-speculate-indirect-jumps.
7370 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7371 (*call_indirect_aix<mode>): Disable for
7372 -mno-speculate-indirect-jumps.
7373 (*call_indirect_aix<mode>_nospec): New define_insn.
7374 (*call_value_indirect_aix<mode>): Disable for
7375 -mno-speculate-indirect-jumps.
7376 (*call_value_indirect_aix<mode>_nospec): New define_insn.
7377 (*sibcall_nonlocal_sysv<mode>): Generate different code for
7378 -mno-speculate-indirect-jumps.
7379 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7381 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
7383 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
7384 long double type, set the flags for noting the default long double
7385 type, even if we don't pass or return a long double type.
7387 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
7390 * ipa-inline.c (flatten_function): Do not overwrite final inlining
7393 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
7395 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
7396 support for merge[hl].
7397 (fold_mergehl_helper): New helper function.
7398 (tree-vector-builder.h): New #include for tree_vector_builder usage.
7399 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
7400 (altivec_vmrglw_direct): Add xxmrglw insn.
7402 2018-01-17 Andrew Waterman <andrew@sifive.com>
7404 * config/riscv/riscv.c (riscv_conditional_register_usage): If
7405 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
7407 2018-01-17 David Malcolm <dmalcolm@redhat.com>
7410 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
7411 call the lto_location_cache before reading the
7412 DECL_SOURCE_LOCATION of the types.
7414 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
7415 Richard Sandiford <richard.sandiford@linaro.org>
7417 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
7418 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
7419 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
7420 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
7421 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
7423 * config/aarch64/constraints.md (aarch64_movti_operand):
7425 * config/aarch64/predicates.md (Uti): Add new constraint.
7427 2018-01-17 Carl Love <cel@us.ibm.com>
7429 * config/rs6000/vsx.md (define_expand xl_len_r,
7430 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
7431 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
7433 (define_expand, define_insn): Move the shift left from the
7434 define_insn to the define_expand for lxvl and stxvl instructions.
7435 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
7436 and XL_LEN_R definitions to PURE.
7438 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
7440 * config/i386/i386.c (indirect_thunk_name): Declare regno
7441 as unsigned int. Compare regno with INVALID_REGNUM.
7442 (output_indirect_thunk): Ditto.
7443 (output_indirect_thunk_function): Ditto.
7444 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
7445 in the call to output_indirect_thunk_function.
7447 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
7450 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
7451 rather than the size of inner_type to determine the stack slot size
7452 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
7454 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
7457 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
7460 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
7462 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
7463 endian Linux systems to optionally enable multilibs for selecting
7464 the long double type if the user configured an explicit type.
7465 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
7466 have no long double multilibs if not defined.
7467 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
7468 warn if the user used -mabi={ieee,ibm}longdouble and we built
7469 multilibs for long double.
7470 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
7471 appropriate multilib option.
7472 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
7474 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
7475 for building long double multilibs.
7476 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
7478 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
7480 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
7483 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
7485 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
7488 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
7491 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
7494 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
7496 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
7497 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
7499 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
7501 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
7502 different rtl trees depending on TARGET_64BIT.
7503 (rs6000_gen_lvx): Likewise.
7505 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
7507 * config/visium/visium.md (nop): Tweak comment.
7508 (hazard_nop): Likewise.
7510 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7512 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
7513 -mspeculate-indirect-jumps.
7514 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
7515 for -mno-speculate-indirect-jumps.
7516 (*call_indirect_elfv2<mode>_nospec): New define_insn.
7517 (*call_value_indirect_elfv2<mode>): Disable for
7518 -mno-speculate-indirect-jumps.
7519 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
7520 (indirect_jump): Emit different RTL for
7521 -mno-speculate-indirect-jumps.
7522 (*indirect_jump<mode>): Disable for
7523 -mno-speculate-indirect-jumps.
7524 (*indirect_jump<mode>_nospec): New define_insn.
7525 (tablejump): Emit different RTL for
7526 -mno-speculate-indirect-jumps.
7527 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
7528 (tablejumpsi_nospec): New define_expand.
7529 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
7530 (tablejumpdi_nospec): New define_expand.
7531 (*tablejump<mode>_internal1): Disable for
7532 -mno-speculate-indirect-jumps.
7533 (*tablejump<mode>_internal1_nospec): New define_insn.
7534 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
7537 2018-01-16 Artyom Skrobov tyomitch@gmail.com
7539 * caller-save.c (insert_save): Drop unnecessary parameter. All
7542 2018-01-16 Jakub Jelinek <jakub@redhat.com>
7543 Richard Biener <rguenth@suse.de>
7546 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
7547 return early, inline manually is_gimple_sizepos. Make sure if we
7548 call gimplify_expr we don't end up with a gimple constant.
7549 * tree.c (variably_modified_type_p): Don't return true for
7550 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
7551 * gimplify.h (is_gimple_sizepos): Remove.
7553 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7555 PR tree-optimization/83857
7556 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
7557 vectorizable_live_operation for pure SLP statements.
7558 (vectorizable_live_operation): Handle PHIs.
7560 2018-01-16 Richard Biener <rguenther@suse.de>
7562 PR tree-optimization/83867
7563 * tree-vect-stmts.c (vect_transform_stmt): Precompute
7564 nested_in_vect_loop_p since the scalar stmt may get invalidated.
7566 2018-01-16 Jakub Jelinek <jakub@redhat.com>
7569 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
7570 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
7571 If off is not INTEGER_CST, issue a may not be aligned warning
7572 rather than isn't aligned. Use isn%'t rather than isn't.
7573 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
7575 <case MULT_EXPR>: Improve the case when bottom and one of the
7576 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
7577 operand, in that case check if the other operand is multiple of
7578 bottom divided by the INTEGER_CST operand.
7580 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7583 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
7584 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
7585 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
7586 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
7587 * config/pa/pa.c (pa_function_arg_advance): Likewise.
7588 (pa_function_arg, pa_arg_partial_bytes): Likewise.
7589 (pa_function_arg_size): New function.
7591 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7593 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
7594 in a separate statement.
7596 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7598 PR tree-optimization/83847
7599 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
7600 group gathers and scatters.
7602 2018-01-16 Jakub Jelinek <jakub@redhat.com>
7604 PR rtl-optimization/86620
7605 * params.def (max-sched-ready-insns): Bump minimum value to 1.
7607 PR rtl-optimization/83213
7608 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
7609 to last if both are JUMP_INSNs.
7611 PR tree-optimization/83843
7612 * gimple-ssa-store-merging.c
7613 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
7614 store_immediate_info for bswap/nop orig_stores.
7616 2018-01-15 Andrew Waterman <andrew@sifive.com>
7618 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
7620 <UDIV>: Increase cost if !TARGET_DIV.
7622 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
7624 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
7625 (define_attr "cr_logical_3op"): New.
7626 (cceq_ior_compare): Adjust.
7627 (cceq_ior_compare_complement): Adjust.
7628 (*cceq_rev_compare): Adjust.
7629 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
7630 (is_cracked_insn): Adjust.
7631 (insn_must_be_first_in_group): Adjust.
7632 * config/rs6000/40x.md: Adjust.
7633 * config/rs6000/440.md: Adjust.
7634 * config/rs6000/476.md: Adjust.
7635 * config/rs6000/601.md: Adjust.
7636 * config/rs6000/603.md: Adjust.
7637 * config/rs6000/6xx.md: Adjust.
7638 * config/rs6000/7450.md: Adjust.
7639 * config/rs6000/7xx.md: Adjust.
7640 * config/rs6000/8540.md: Adjust.
7641 * config/rs6000/cell.md: Adjust.
7642 * config/rs6000/e300c2c3.md: Adjust.
7643 * config/rs6000/e500mc.md: Adjust.
7644 * config/rs6000/e500mc64.md: Adjust.
7645 * config/rs6000/e5500.md: Adjust.
7646 * config/rs6000/e6500.md: Adjust.
7647 * config/rs6000/mpc.md: Adjust.
7648 * config/rs6000/power4.md: Adjust.
7649 * config/rs6000/power5.md: Adjust.
7650 * config/rs6000/power6.md: Adjust.
7651 * config/rs6000/power7.md: Adjust.
7652 * config/rs6000/power8.md: Adjust.
7653 * config/rs6000/power9.md: Adjust.
7654 * config/rs6000/rs64.md: Adjust.
7655 * config/rs6000/titan.md: Adjust.
7657 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7659 * config/i386/predicates.md (indirect_branch_operand): Rewrite
7660 ix86_indirect_branch_register logic.
7662 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7664 * config/i386/constraints.md (Bs): Update
7665 ix86_indirect_branch_register check. Don't check
7666 ix86_indirect_branch_register with GOT_memory_operand.
7668 * config/i386/predicates.md (GOT_memory_operand): Don't check
7669 ix86_indirect_branch_register here.
7670 (GOT32_symbol_operand): Likewise.
7672 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7674 * config/i386/predicates.md (constant_call_address_operand):
7675 Rewrite ix86_indirect_branch_register logic.
7676 (sibcall_insn_operand): Likewise.
7678 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7680 * config/i386/constraints.md (Bs): Replace
7681 ix86_indirect_branch_thunk_register with
7682 ix86_indirect_branch_register.
7684 * config/i386/i386.md (indirect_jump): Likewise.
7685 (tablejump): Likewise.
7686 (*sibcall_memory): Likewise.
7687 (*sibcall_value_memory): Likewise.
7688 Peepholes of indirect call and jump via memory: Likewise.
7689 * config/i386/i386.opt: Likewise.
7690 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7691 (GOT_memory_operand): Likewise.
7692 (call_insn_operand): Likewise.
7693 (sibcall_insn_operand): Likewise.
7694 (GOT32_symbol_operand): Likewise.
7696 2018-01-15 Jakub Jelinek <jakub@redhat.com>
7699 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
7700 type rather than type addr's type points to.
7701 (expand_omp_atomic_mutex): Likewise.
7702 (expand_omp_atomic): Likewise.
7704 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7707 * config/i386/i386.c (output_indirect_thunk_function): Use
7708 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
7709 for __x86_return_thunk.
7711 2018-01-15 Richard Biener <rguenther@suse.de>
7714 * expmed.c (extract_bit_field_1): Fix typo.
7716 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7719 * config/arm/iterators.md (VF): New mode iterator.
7720 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
7721 Remove integer-related logic from pattern.
7722 (neon_vabd<mode>_3): Likewise.
7724 2018-01-15 Jakub Jelinek <jakub@redhat.com>
7727 * common.opt (fstrict-overflow): No longer an alias.
7728 (fwrapv-pointer): New option.
7729 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
7730 also for pointer types based on flag_wrapv_pointer.
7731 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
7732 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
7733 opts->x_flag_wrapv got set.
7734 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
7735 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
7736 POINTER_TYPE_OVERFLOW_UNDEFINED.
7737 * match.pd: Likewise in address comparison pattern.
7738 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
7740 2018-01-15 Richard Biener <rguenther@suse.de>
7743 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
7744 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
7745 Reset type names to their identifier if their TYPE_DECL doesn't
7746 have linkage (and thus is used for ODR and devirt).
7747 (save_debug_info_for_decl): Remove.
7748 (save_debug_info_for_type): Likewise.
7749 (add_tree_to_fld_list): Adjust.
7750 * tree-pretty-print.c (dump_generic_node): Make dumping of
7751 type names more robust.
7753 2018-01-15 Richard Biener <rguenther@suse.de>
7755 * BASE-VER: Bump to 8.0.1.
7757 2018-01-14 Martin Sebor <msebor@redhat.com>
7760 * builtins.c (check_access): Avoid warning when the no-warning bit
7763 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
7765 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
7766 * ira-color (allocno_hard_regs_compare): Likewise.
7768 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
7771 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
7772 Use .pushsection/.popsection.
7774 2018-01-14 Martin Sebor <msebor@redhat.com>
7777 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
7779 2018-01-14 Jakub Jelinek <jakub@redhat.com>
7781 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
7782 entry from extra_headers.
7783 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
7784 extra_headers, make the list bitwise identical to the i?86-*-* one.
7786 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7788 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
7789 -mcmodel=large with -mindirect-branch=thunk,
7790 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
7791 -mfunction-return=thunk-extern.
7792 * doc/invoke.texi: Document -mcmodel=large is incompatible with
7793 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
7794 -mfunction-return=thunk and -mfunction-return=thunk-extern.
7796 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7798 * config/i386/i386.c (print_reg): Print the name of the full
7799 integer register without '%'.
7800 (ix86_print_operand): Handle 'V'.
7801 * doc/extend.texi: Document 'V' modifier.
7803 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7805 * config/i386/constraints.md (Bs): Disallow memory operand for
7806 -mindirect-branch-register.
7808 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7809 (GOT_memory_operand): Likewise.
7810 (call_insn_operand): Likewise.
7811 (sibcall_insn_operand): Likewise.
7812 (GOT32_symbol_operand): Likewise.
7813 * config/i386/i386.md (indirect_jump): Call convert_memory_address
7814 for -mindirect-branch-register.
7815 (tablejump): Likewise.
7816 (*sibcall_memory): Likewise.
7817 (*sibcall_value_memory): Likewise.
7818 Disallow peepholes of indirect call and jump via memory for
7819 -mindirect-branch-register.
7820 (*call_pop): Replace m with Bw.
7821 (*call_value_pop): Likewise.
7822 (*sibcall_pop_memory): Replace m with Bs.
7823 * config/i386/i386.opt (mindirect-branch-register): New option.
7824 * doc/invoke.texi: Document -mindirect-branch-register option.
7826 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7828 * config/i386/i386-protos.h (ix86_output_function_return): New.
7829 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
7830 set function_return_type.
7831 (indirect_thunk_name): Add ret_p to indicate thunk for function
7833 (output_indirect_thunk_function): Pass false to
7834 indirect_thunk_name.
7835 (ix86_output_indirect_branch_via_reg): Likewise.
7836 (ix86_output_indirect_branch_via_push): Likewise.
7837 (output_indirect_thunk_function): Create alias for function
7838 return thunk if regno < 0.
7839 (ix86_output_function_return): New function.
7840 (ix86_handle_fndecl_attribute): Handle function_return.
7841 (ix86_attribute_table): Add function_return.
7842 * config/i386/i386.h (machine_function): Add
7843 function_return_type.
7844 * config/i386/i386.md (simple_return_internal): Use
7845 ix86_output_function_return.
7846 (simple_return_internal_long): Likewise.
7847 * config/i386/i386.opt (mfunction-return=): New option.
7848 (indirect_branch): Mention -mfunction-return=.
7849 * doc/extend.texi: Document function_return function attribute.
7850 * doc/invoke.texi: Document -mfunction-return= option.
7852 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7854 * config/i386/i386-opts.h (indirect_branch): New.
7855 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
7856 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
7857 with local indirect jump when converting indirect call and jump.
7858 (ix86_set_indirect_branch_type): New.
7859 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
7860 (indirectlabelno): New.
7861 (indirect_thunk_needed): Likewise.
7862 (indirect_thunk_bnd_needed): Likewise.
7863 (indirect_thunks_used): Likewise.
7864 (indirect_thunks_bnd_used): Likewise.
7865 (INDIRECT_LABEL): Likewise.
7866 (indirect_thunk_name): Likewise.
7867 (output_indirect_thunk): Likewise.
7868 (output_indirect_thunk_function): Likewise.
7869 (ix86_output_indirect_branch_via_reg): Likewise.
7870 (ix86_output_indirect_branch_via_push): Likewise.
7871 (ix86_output_indirect_branch): Likewise.
7872 (ix86_output_indirect_jmp): Likewise.
7873 (ix86_code_end): Call output_indirect_thunk_function if needed.
7874 (ix86_output_call_insn): Call ix86_output_indirect_branch if
7876 (ix86_handle_fndecl_attribute): Handle indirect_branch.
7877 (ix86_attribute_table): Add indirect_branch.
7878 * config/i386/i386.h (machine_function): Add indirect_branch_type
7879 and has_local_indirect_jump.
7880 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
7882 (tablejump): Likewise.
7883 (*indirect_jump): Use ix86_output_indirect_jmp.
7884 (*tablejump_1): Likewise.
7885 (simple_return_indirect_internal): Likewise.
7886 * config/i386/i386.opt (mindirect-branch=): New option.
7887 (indirect_branch): New.
7890 (thunk-inline): Likewise.
7891 (thunk-extern): Likewise.
7892 * doc/extend.texi: Document indirect_branch function attribute.
7893 * doc/invoke.texi: Document -mindirect-branch= option.
7895 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
7898 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
7900 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
7902 * ipa-inline.c (want_inline_small_function_p): Return false if
7903 inlining has already failed with CIF_FINAL_ERROR.
7904 (update_caller_keys): Call want_inline_small_function_p before
7906 (update_callee_keys): Likewise.
7908 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
7910 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
7912 (rs6000_quadword_masked_address_p): Likewise.
7913 (quad_aligned_load_p): Likewise.
7914 (quad_aligned_store_p): Likewise.
7915 (const_load_sequence_p): Add comment to describe the outer-most loop.
7916 (mimic_memory_attributes_and_flags): New function.
7917 (rs6000_gen_stvx): Likewise.
7918 (replace_swapped_aligned_store): Likewise.
7919 (rs6000_gen_lvx): Likewise.
7920 (replace_swapped_aligned_load): Likewise.
7921 (replace_swapped_load_constant): Capitalize argument name in
7922 comment describing this function.
7923 (rs6000_analyze_swaps): Add a third pass to search for vector loads
7924 and stores that access quad-word aligned addresses and replace
7925 with stvx or lvx instructions when appropriate.
7926 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
7927 New function prototype.
7928 (rs6000_quadword_masked_address_p): Likewise.
7929 (rs6000_gen_lvx): Likewise.
7930 (rs6000_gen_stvx): Likewise.
7931 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
7932 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
7933 when memory address is aligned.
7934 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
7935 this split to select lvx instruction when memory address is aligned.
7936 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
7937 instruction when memory address is aligned.
7938 (*vsx_le_perm_load_v16qi): Likewise.
7939 (four unnamed splitters): Modify to select the stvx instruction
7940 when memory is aligned.
7942 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
7944 * predict.c (determine_unlikely_bbs): Handle correctly BBs
7945 which appears in the queue multiple times.
7947 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7948 Alan Hayward <alan.hayward@arm.com>
7949 David Sherwood <david.sherwood@arm.com>
7951 * tree-vectorizer.h (vec_lower_bound): New structure.
7952 (_loop_vec_info): Add check_nonzero and lower_bounds.
7953 (LOOP_VINFO_CHECK_NONZERO): New macro.
7954 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
7955 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
7956 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
7957 fields. Make seg_len the distance travelled, not including the
7959 (dr_direction_indicator): Declare.
7960 (dr_zero_step_indicator): Likewise.
7961 (dr_known_forward_stride_p): Likewise.
7962 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
7964 (runtime_alias_check_p): Allow runtime alias checks with
7966 (operator ==): Compare access_size and align.
7967 (prune_runtime_alias_test_list): Rework for new distinction between
7968 the access_size and seg_len.
7969 (create_intersect_range_checks_index): Likewise. Cope with polynomial
7971 (get_segment_min_max): New function.
7972 (create_intersect_range_checks): Use it.
7973 (dr_step_indicator): New function.
7974 (dr_direction_indicator): Likewise.
7975 (dr_zero_step_indicator): Likewise.
7976 (dr_known_forward_stride_p): Likewise.
7977 * tree-loop-distribution.c (data_ref_segment_size): Return
7978 DR_STEP * (niters - 1).
7979 (compute_alias_check_pairs): Update call to the dr_with_seg_len
7981 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
7982 (vect_preserves_scalar_order_p): New function, split out from...
7983 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
7984 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
7985 (vect_vfa_access_size): New function.
7986 (vect_vfa_align): Likewise.
7987 (vect_compile_time_alias): Take access_size_a and access_b arguments.
7988 (dump_lower_bound): New function.
7989 (vect_check_lower_bound): Likewise.
7990 (vect_small_gap_p): Likewise.
7991 (vectorizable_with_step_bound_p): Likewise.
7992 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
7993 depencies if the vectorization factor is 1. Convert the checks
7994 for nonzero steps into checks on the bounds of DR_STEP. Try using
7995 a bunds check for variable steps if the minimum required step is
7996 relatively small. Update calls to the dr_with_seg_len
7997 constructor and to vect_compile_time_alias.
7998 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
8000 (vect_loop_versioning): Call it.
8001 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
8003 (vect_estimate_min_profitable_iters): Account for any bounds checks.
8005 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8006 Alan Hayward <alan.hayward@arm.com>
8007 David Sherwood <david.sherwood@arm.com>
8009 * doc/sourcebuild.texi (vect_scatter_store): Document.
8010 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
8012 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
8014 * genopinit.c (main): Add supports_vec_scatter_store and
8015 supports_vec_scatter_store_cached to target_optabs.
8016 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
8017 IFN_MASK_SCATTER_STORE.
8018 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
8020 * internal-fn.h (internal_store_fn_p): Declare.
8021 (internal_fn_stored_value_index): Likewise.
8022 * internal-fn.c (scatter_store_direct): New macro.
8023 (expand_scatter_store_optab_fn): New function.
8024 (direct_scatter_store_optab_supported_p): New macro.
8025 (internal_store_fn_p): New function.
8026 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
8027 IFN_MASK_SCATTER_STORE.
8028 (internal_fn_mask_index): Likewise.
8029 (internal_fn_stored_value_index): New function.
8030 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
8032 * optabs-query.h (supports_vec_scatter_store_p): Declare.
8033 * optabs-query.c (supports_vec_scatter_store_p): New function.
8034 * tree-vectorizer.h (vect_get_store_rhs): Declare.
8035 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
8036 true for scatter stores.
8037 (vect_gather_scatter_fn_p): Handle scatter stores too.
8038 (vect_check_gather_scatter): Consider using scatter stores if
8039 supports_vec_scatter_store_p.
8040 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
8042 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
8043 internal_fn_stored_value_index.
8044 (check_load_store_masking): Handle scatter stores too.
8045 (vect_get_store_rhs): Make public.
8046 (vectorizable_call): Use internal_store_fn_p.
8047 (vectorizable_store): Handle scatter store internal functions.
8048 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
8049 when deciding whether the end of the group has been reached.
8050 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
8051 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
8052 (mask_scatter_store<mode>): New insns.
8054 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8055 Alan Hayward <alan.hayward@arm.com>
8056 David Sherwood <david.sherwood@arm.com>
8058 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
8059 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
8060 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
8062 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
8063 Use vect_truncate_gather_scatter_offset if we can't treat the
8064 operation as a normal gather load or scatter store.
8065 (get_group_load_store_type): Take the gather_scatter_info
8066 as argument. Try using a gather load or scatter store for
8067 single-element groups.
8068 (get_load_store_type): Update calls to get_group_load_store_type
8069 and vect_use_strided_gather_scatters_p.
8071 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8072 Alan Hayward <alan.hayward@arm.com>
8073 David Sherwood <david.sherwood@arm.com>
8075 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
8076 optional tree argument.
8077 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
8079 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
8080 but continue to use the current value as a fallback.
8081 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
8082 to compare the updates.
8083 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
8084 (get_load_store_type): Use it when handling a strided access.
8085 (vect_get_strided_load_store_ops): New function.
8086 (vect_get_data_ptr_increment): Likewise.
8087 (vectorizable_load): Handle strided gather loads. Always pass
8088 a step to vect_create_data_ref_ptr and bump_vector_ptr.
8090 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8091 Alan Hayward <alan.hayward@arm.com>
8092 David Sherwood <david.sherwood@arm.com>
8094 * doc/md.texi (gather_load@var{m}): Document.
8095 (mask_gather_load@var{m}): Likewise.
8096 * genopinit.c (main): Add supports_vec_gather_load and
8097 supports_vec_gather_load_cached to target_optabs.
8098 * optabs-tree.c (init_tree_optimization_optabs): Use
8099 ggc_cleared_alloc to allocate target_optabs.
8100 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
8101 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
8103 * internal-fn.h (internal_load_fn_p): Declare.
8104 (internal_gather_scatter_fn_p): Likewise.
8105 (internal_fn_mask_index): Likewise.
8106 (internal_gather_scatter_fn_supported_p): Likewise.
8107 * internal-fn.c (gather_load_direct): New macro.
8108 (expand_gather_load_optab_fn): New function.
8109 (direct_gather_load_optab_supported_p): New macro.
8110 (direct_internal_fn_optab): New function.
8111 (internal_load_fn_p): Likewise.
8112 (internal_gather_scatter_fn_p): Likewise.
8113 (internal_fn_mask_index): Likewise.
8114 (internal_gather_scatter_fn_supported_p): Likewise.
8115 * optabs-query.c (supports_at_least_one_mode_p): New function.
8116 (supports_vec_gather_load_p): Likewise.
8117 * optabs-query.h (supports_vec_gather_load_p): Declare.
8118 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
8119 and memory_type field.
8120 (NUM_PATTERNS): Bump to 15.
8121 * tree-vect-data-refs.c: Include internal-fn.h.
8122 (vect_gather_scatter_fn_p): New function.
8123 (vect_describe_gather_scatter_call): Likewise.
8124 (vect_check_gather_scatter): Try using internal functions for
8125 gather loads. Recognize existing calls to a gather load function.
8126 (vect_analyze_data_refs): Consider using gather loads if
8127 supports_vec_gather_load_p.
8128 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
8129 (vect_get_gather_scatter_offset_type): Likewise.
8130 (vect_convert_mask_for_vectype): Likewise.
8131 (vect_add_conversion_to_patterm): Likewise.
8132 (vect_try_gather_scatter_pattern): Likewise.
8133 (vect_recog_gather_scatter_pattern): New pattern recognizer.
8134 (vect_vect_recog_func_ptrs): Add it.
8135 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
8136 internal_fn_mask_index and internal_gather_scatter_fn_p.
8137 (check_load_store_masking): Take the gather_scatter_info as an
8138 argument and handle gather loads.
8139 (vect_get_gather_scatter_ops): New function.
8140 (vectorizable_call): Check internal_load_fn_p.
8141 (vectorizable_load): Likewise. Handle gather load internal
8143 (vectorizable_store): Update call to check_load_store_masking.
8144 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
8145 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
8146 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
8147 (aarch64_gather_scale_operand_d): New predicates.
8148 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
8149 (mask_gather_load<mode>): New insns.
8151 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8152 Alan Hayward <alan.hayward@arm.com>
8153 David Sherwood <david.sherwood@arm.com>
8155 * optabs.def (fold_left_plus_optab): New optab.
8156 * doc/md.texi (fold_left_plus_@var{m}): Document.
8157 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
8158 * internal-fn.c (fold_left_direct): Define.
8159 (expand_fold_left_optab_fn): Likewise.
8160 (direct_fold_left_optab_supported_p): Likewise.
8161 * fold-const-call.c (fold_const_fold_left): New function.
8162 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
8163 * tree-parloops.c (valid_reduction_p): New function.
8164 (gather_scalar_reductions): Use it.
8165 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
8166 (vect_finish_replace_stmt): Declare.
8167 * tree-vect-loop.c (fold_left_reduction_fn): New function.
8168 (needs_fold_left_reduction_p): New function, split out from...
8169 (vect_is_simple_reduction): ...here. Accept reductions that
8170 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
8171 (vect_force_simple_reduction): Also store the reduction type in
8172 the assignment's STMT_VINFO_REDUC_TYPE.
8173 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
8174 (merge_with_identity): New function.
8175 (vect_expand_fold_left): Likewise.
8176 (vectorize_fold_left_reduction): Likewise.
8177 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
8178 scalar phi in place for it. Check for target support and reject
8179 cases that would reassociate the operation. Defer the transform
8180 phase to vectorize_fold_left_reduction.
8181 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
8182 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
8183 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
8185 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8187 * tree-if-conv.c (predicate_mem_writes): Remove redundant
8188 call to ifc_temp_var.
8190 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8191 Alan Hayward <alan.hayward@arm.com>
8192 David Sherwood <david.sherwood@arm.com>
8194 * target.def (legitimize_address_displacement): Take the original
8195 offset as a poly_int.
8196 * targhooks.h (default_legitimize_address_displacement): Update
8198 * targhooks.c (default_legitimize_address_displacement): Likewise.
8199 * doc/tm.texi: Regenerate.
8200 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
8201 as an argument, moving assert of ad->disp == ad->disp_term to...
8202 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
8203 Try calling targetm.legitimize_address_displacement before expanding
8204 the address rather than afterwards, and adjust for the new interface.
8205 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
8206 Match the new hook interface. Handle SVE addresses.
8207 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
8210 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8212 * Makefile.in (OBJS): Add early-remat.o.
8213 * target.def (select_early_remat_modes): New hook.
8214 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
8215 * doc/tm.texi: Regenerate.
8216 * targhooks.h (default_select_early_remat_modes): Declare.
8217 * targhooks.c (default_select_early_remat_modes): New function.
8218 * timevar.def (TV_EARLY_REMAT): New timevar.
8219 * passes.def (pass_early_remat): New pass.
8220 * tree-pass.h (make_pass_early_remat): Declare.
8221 * early-remat.c: New file.
8222 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
8224 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
8226 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8227 Alan Hayward <alan.hayward@arm.com>
8228 David Sherwood <david.sherwood@arm.com>
8230 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
8231 vfm1 with a bound_epilog parameter.
8232 (vect_do_peeling): Update calls accordingly, and move the prologue
8233 call earlier in the function. Treat the base bound_epilog as 0 for
8234 fully-masked loops and retain vf - 1 for other loops. Add 1 to
8235 this base when peeling for gaps.
8236 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
8237 with fully-masked loops.
8238 (vect_estimate_min_profitable_iters): Handle the single peeled
8239 iteration in that case.
8241 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8242 Alan Hayward <alan.hayward@arm.com>
8243 David Sherwood <david.sherwood@arm.com>
8245 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
8246 single-element interleaving even if the size is not a power of 2.
8247 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
8248 accesses for single-element interleaving if the group size is
8251 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8252 Alan Hayward <alan.hayward@arm.com>
8253 David Sherwood <david.sherwood@arm.com>
8255 * doc/md.texi (fold_extract_last_@var{m}): Document.
8256 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
8257 * optabs.def (fold_extract_last_optab): New optab.
8258 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
8259 * internal-fn.c (fold_extract_direct): New macro.
8260 (expand_fold_extract_optab_fn): Likewise.
8261 (direct_fold_extract_optab_supported_p): Likewise.
8262 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
8263 * tree-vect-loop.c (vect_model_reduction_cost): Handle
8264 EXTRACT_LAST_REDUCTION.
8265 (get_initial_def_for_reduction): Do not create an initial vector
8266 for EXTRACT_LAST_REDUCTION reductions.
8267 (vectorizable_reduction): Leave the scalar phi in place for
8268 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
8269 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
8270 epilogue code for EXTRACT_LAST_REDUCTION and defer the
8271 transform phase to vectorizable_condition.
8272 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
8274 (vect_finish_stmt_generation): ...here.
8275 (vect_finish_replace_stmt): New function.
8276 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
8277 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
8279 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
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 * doc/md.texi (extract_last_@var{m}): Document.
8286 * optabs.def (extract_last_optab): New optab.
8287 * internal-fn.def (EXTRACT_LAST): New internal function.
8288 * internal-fn.c (cond_unary_direct): New macro.
8289 (expand_cond_unary_optab_fn): Likewise.
8290 (direct_cond_unary_optab_supported_p): Likewise.
8291 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
8292 loops using EXTRACT_LAST.
8293 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
8294 (extract_last_<mode>): ...this optab.
8295 (vec_extract<mode><Vel>): Update accordingly.
8297 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8298 Alan Hayward <alan.hayward@arm.com>
8299 David Sherwood <david.sherwood@arm.com>
8301 * target.def (empty_mask_is_expensive): New hook.
8302 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
8303 * doc/tm.texi: Regenerate.
8304 * targhooks.h (default_empty_mask_is_expensive): Declare.
8305 * targhooks.c (default_empty_mask_is_expensive): New function.
8306 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
8307 if the target says that empty masks are expensive.
8308 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
8310 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
8312 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8313 Alan Hayward <alan.hayward@arm.com>
8314 David Sherwood <david.sherwood@arm.com>
8316 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
8317 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
8318 (vect_use_loop_mask_for_alignment_p): New function.
8319 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
8320 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
8321 niters_skip argument. Make sure that the first niters_skip elements
8322 of the first iteration are inactive.
8323 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
8324 Update call to vect_set_loop_masks_directly.
8325 (get_misalign_in_elems): New function, split out from...
8326 (vect_gen_prolog_loop_niters): ...here.
8327 (vect_update_init_of_dr): Take a code argument that specifies whether
8328 the adjustment should be added or subtracted.
8329 (vect_update_init_of_drs): Likewise.
8330 (vect_prepare_for_masked_peels): New function.
8331 (vect_do_peeling): Skip prologue peeling if we're using a mask
8332 instead. Update call to vect_update_inits_of_drs.
8333 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
8335 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
8336 alignment. Do not include the number of peeled iterations in
8337 the minimum threshold in that case.
8338 (vectorizable_induction): Adjust the start value down by
8339 LOOP_VINFO_MASK_SKIP_NITERS iterations.
8340 (vect_transform_loop): Call vect_prepare_for_masked_peels.
8341 Take the number of skipped iterations into account when calculating
8343 * tree-vect-stmts.c (vect_gen_while_not): New function.
8345 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8346 Alan Hayward <alan.hayward@arm.com>
8347 David Sherwood <david.sherwood@arm.com>
8349 * doc/sourcebuild.texi (vect_fully_masked): Document.
8350 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
8352 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
8354 (vect_analyze_loop_2): ...here. Don't check the vectorization
8355 factor against the number of loop iterations if the loop is
8358 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8359 Alan Hayward <alan.hayward@arm.com>
8360 David Sherwood <david.sherwood@arm.com>
8362 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
8363 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
8364 (dump_groups): Update accordingly.
8365 (iv_use::mem_type): New member variable.
8366 (address_p): New function.
8367 (record_use): Add a mem_type argument and initialize the new
8369 (record_group_use): Add a mem_type argument. Use address_p.
8370 Remove obsolete null checks of base_object. Update call to record_use.
8371 (find_interesting_uses_op): Update call to record_group_use.
8372 (find_interesting_uses_cond): Likewise.
8373 (find_interesting_uses_address): Likewise.
8374 (get_mem_type_for_internal_fn): New function.
8375 (find_address_like_use): Likewise.
8376 (find_interesting_uses_stmt): Try find_address_like_use before
8377 calling find_interesting_uses_op.
8378 (addr_offset_valid_p): Use the iv mem_type field as the type
8379 of the addressed memory.
8380 (add_autoinc_candidates): Likewise.
8381 (get_address_cost): Likewise.
8382 (split_small_address_groups_p): Use address_p.
8383 (split_address_groups): Likewise.
8384 (add_iv_candidate_for_use): Likewise.
8385 (autoinc_possible_for_pair): Likewise.
8386 (rewrite_groups): Likewise.
8387 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
8388 (determine_group_iv_cost): Update after split of USE_ADDRESS.
8389 (get_alias_ptr_type_for_ptr_address): New function.
8390 (rewrite_use_address): Rewrite address uses in calls that were
8391 identified by find_address_like_use.
8393 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8394 Alan Hayward <alan.hayward@arm.com>
8395 David Sherwood <david.sherwood@arm.com>
8397 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
8399 * gimple-expr.h (is_gimple_addressable: Likewise.
8400 * gimple-expr.c (is_gimple_address): Likewise.
8401 * internal-fn.c (expand_call_mem_ref): New function.
8402 (expand_mask_load_optab_fn): Use it.
8403 (expand_mask_store_optab_fn): Likewise.
8405 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8406 Alan Hayward <alan.hayward@arm.com>
8407 David Sherwood <david.sherwood@arm.com>
8409 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
8410 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
8411 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
8412 (cond_umax@var{mode}): Document.
8413 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
8414 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
8415 (cond_umin_optab, cond_umax_optab): New optabs.
8416 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
8417 (COND_IOR, COND_XOR): New internal functions.
8418 * internal-fn.h (get_conditional_internal_fn): Declare.
8419 * internal-fn.c (cond_binary_direct): New macro.
8420 (expand_cond_binary_optab_fn): Likewise.
8421 (direct_cond_binary_optab_supported_p): Likewise.
8422 (get_conditional_internal_fn): New function.
8423 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
8424 Cope with reduction statements that are vectorized as calls rather
8426 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
8427 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
8428 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
8429 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
8430 (UNSPEC_COND_EOR): New unspecs.
8431 (optab): Add mappings for them.
8432 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
8433 (sve_int_op, sve_fp_op): New int attributes.
8435 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8436 Alan Hayward <alan.hayward@arm.com>
8437 David Sherwood <david.sherwood@arm.com>
8439 * optabs.def (while_ult_optab): New optab.
8440 * doc/md.texi (while_ult@var{m}@var{n}): Document.
8441 * internal-fn.def (WHILE_ULT): New internal function.
8442 * internal-fn.h (direct_internal_fn_supported_p): New override
8443 that takes two types as argument.
8444 * internal-fn.c (while_direct): New macro.
8445 (expand_while_optab_fn): New function.
8446 (convert_optab_supported_p): Likewise.
8447 (direct_while_optab_supported_p): New macro.
8448 * wide-int.h (wi::udiv_ceil): New function.
8449 * tree-vectorizer.h (rgroup_masks): New structure.
8450 (vec_loop_masks): New typedef.
8451 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
8453 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
8454 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
8455 (vect_max_vf): New function.
8456 (slpeel_make_loop_iterate_ntimes): Delete.
8457 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
8458 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
8459 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
8460 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
8461 internal-fn.h, stor-layout.h and optabs-query.h.
8462 (vect_set_loop_mask): New function.
8463 (add_preheader_seq): Likewise.
8464 (add_header_seq): Likewise.
8465 (interleave_supported_p): Likewise.
8466 (vect_maybe_permute_loop_masks): Likewise.
8467 (vect_set_loop_masks_directly): Likewise.
8468 (vect_set_loop_condition_masked): Likewise.
8469 (vect_set_loop_condition_unmasked): New function, split out from
8470 slpeel_make_loop_iterate_ntimes.
8471 (slpeel_make_loop_iterate_ntimes): Rename to..
8472 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
8473 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
8474 (vect_do_peeling): Update call accordingly.
8475 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
8477 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
8478 mask_compare_type, can_fully_mask_p and fully_masked_p.
8479 (release_vec_loop_masks): New function.
8480 (_loop_vec_info): Use it to free the loop masks.
8481 (can_produce_all_loop_masks_p): New function.
8482 (vect_get_max_nscalars_per_iter): Likewise.
8483 (vect_verify_full_masking): Likewise.
8484 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
8485 retries, and free the mask rgroups before retrying. Check loop-wide
8486 reasons for disallowing fully-masked loops. Make the final decision
8487 about whether use a fully-masked loop or not.
8488 (vect_estimate_min_profitable_iters): Do not assume that peeling
8489 for the number of iterations will be needed for fully-masked loops.
8490 (vectorizable_reduction): Disable fully-masked loops.
8491 (vectorizable_live_operation): Likewise.
8492 (vect_halve_mask_nunits): New function.
8493 (vect_double_mask_nunits): Likewise.
8494 (vect_record_loop_mask): Likewise.
8495 (vect_get_loop_mask): Likewise.
8496 (vect_transform_loop): Handle the case in which the final loop
8497 iteration might handle a partial vector. Call vect_set_loop_condition
8498 instead of slpeel_make_loop_iterate_ntimes.
8499 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
8500 (check_load_store_masking): New function.
8501 (prepare_load_store_mask): Likewise.
8502 (vectorizable_store): Handle fully-masked loops.
8503 (vectorizable_load): Likewise.
8504 (supportable_widening_operation): Use vect_halve_mask_nunits for
8506 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
8507 (vect_gen_while): New function.
8508 * config/aarch64/aarch64.md (umax<mode>3): New expander.
8509 (aarch64_uqdec<mode>): New insn.
8511 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8512 Alan Hayward <alan.hayward@arm.com>
8513 David Sherwood <david.sherwood@arm.com>
8515 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
8516 (reduc_xor_scal_optab): New optabs.
8517 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
8518 (reduc_xor_scal_@var{m}): Document.
8519 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
8520 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
8522 * fold-const-call.c (fold_const_call): Handle them.
8523 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
8524 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
8525 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
8526 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
8527 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
8528 (UNSPEC_XORV): New unspecs.
8529 (optab): Add entries for them.
8530 (BITWISEV): New int iterator.
8531 (bit_reduc_op): New int attributes.
8533 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8534 Alan Hayward <alan.hayward@arm.com>
8535 David Sherwood <david.sherwood@arm.com>
8537 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
8538 * internal-fn.def (VEC_SHL_INSERT): New internal function.
8539 * optabs.def (vec_shl_insert_optab): New optab.
8540 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
8541 (duplicate_and_interleave): Likewise.
8542 * tree-vect-loop.c: Include internal-fn.h.
8543 (neutral_op_for_slp_reduction): New function, split out from
8544 get_initial_defs_for_reduction.
8545 (get_initial_def_for_reduction): Handle option 2 for variable-length
8546 vectors by loading the neutral value into a vector and then shifting
8547 the initial value into element 0.
8548 (get_initial_defs_for_reduction): Replace the code argument with
8549 the neutral value calculated by neutral_op_for_slp_reduction.
8550 Use gimple_build_vector for constant-length vectors.
8551 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
8552 but the first group_size elements have a neutral value.
8553 Use duplicate_and_interleave otherwise.
8554 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
8555 Update call to get_initial_defs_for_reduction. Handle SLP
8556 reductions for variable-length vectors by creating one vector
8557 result for each scalar result, with the elements associated
8558 with other scalar results stubbed out with the neutral value.
8559 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
8560 Require IFN_VEC_SHL_INSERT for double reductions on
8561 variable-length vectors, or SLP reductions that have
8562 a neutral value. Require can_duplicate_and_interleave_p
8563 support for variable-length unchained SLP reductions if there
8564 is no neutral value, such as for MIN/MAX reductions. Also require
8565 the number of vector elements to be a multiple of the number of
8566 SLP statements when doing variable-length unchained SLP reductions.
8567 Update call to vect_create_epilog_for_reduction.
8568 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
8569 and remove initial values.
8570 (duplicate_and_interleave): Make public.
8571 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
8572 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
8574 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8575 Alan Hayward <alan.hayward@arm.com>
8576 David Sherwood <david.sherwood@arm.com>
8578 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
8579 (can_duplicate_and_interleave_p): New function.
8580 (vect_get_and_check_slp_defs): Take the vector of statements
8581 rather than just the current one. Remove excess parentheses.
8582 Restriction rejectinon of vect_constant_def and vect_external_def
8583 for variable-length vectors to boolean types, or types for which
8584 can_duplicate_and_interleave_p is false.
8585 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
8586 (duplicate_and_interleave): New function.
8587 (vect_get_constant_vectors): Use gimple_build_vector for
8588 constant-length vectors and suitable variable-length constant
8589 vectors. Use duplicate_and_interleave for other variable-length
8590 vectors. Don't defer the update when inserting new statements.
8592 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8593 Alan Hayward <alan.hayward@arm.com>
8594 David Sherwood <david.sherwood@arm.com>
8596 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
8597 min_profitable_iters doesn't go negative.
8599 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8600 Alan Hayward <alan.hayward@arm.com>
8601 David Sherwood <david.sherwood@arm.com>
8603 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
8604 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
8605 * optabs.def (vec_mask_load_lanes_optab): New optab.
8606 (vec_mask_store_lanes_optab): Likewise.
8607 * internal-fn.def (MASK_LOAD_LANES): New internal function.
8608 (MASK_STORE_LANES): Likewise.
8609 * internal-fn.c (mask_load_lanes_direct): New macro.
8610 (mask_store_lanes_direct): Likewise.
8611 (expand_mask_load_optab_fn): Handle masked operations.
8612 (expand_mask_load_lanes_optab_fn): New macro.
8613 (expand_mask_store_optab_fn): Handle masked operations.
8614 (expand_mask_store_lanes_optab_fn): New macro.
8615 (direct_mask_load_lanes_optab_supported_p): Likewise.
8616 (direct_mask_store_lanes_optab_supported_p): Likewise.
8617 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
8619 (vect_load_lanes_supported): Likewise.
8620 * tree-vect-data-refs.c (strip_conversion): New function.
8621 (can_group_stmts_p): Likewise.
8622 (vect_analyze_data_ref_accesses): Use it instead of checking
8623 for a pair of assignments.
8624 (vect_store_lanes_supported): Take a masked_p parameter.
8625 (vect_load_lanes_supported): Likewise.
8626 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
8627 vect_store_lanes_supported and vect_load_lanes_supported.
8628 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
8629 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
8630 parameter. Don't allow gaps for masked accesses.
8631 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
8632 and vect_load_lanes_supported.
8633 (get_load_store_type): Take a masked_p parameter and update
8634 call to get_group_load_store_type.
8635 (vectorizable_store): Update call to get_load_store_type.
8636 Handle IFN_MASK_STORE_LANES.
8637 (vectorizable_load): Update call to get_load_store_type.
8638 Handle IFN_MASK_LOAD_LANES.
8640 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8641 Alan Hayward <alan.hayward@arm.com>
8642 David Sherwood <david.sherwood@arm.com>
8644 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
8646 * config/aarch64/aarch64-protos.h
8647 (aarch64_sve_struct_memory_operand_p): Declare.
8648 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
8649 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
8650 (VPRED, vpred): Handle SVE structure modes.
8651 * config/aarch64/constraints.md (Utx): New constraint.
8652 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
8653 (aarch64_sve_struct_nonimmediate_operand): New predicates.
8654 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
8655 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
8656 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
8657 structure modes. Split into pieces after RA.
8658 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
8659 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
8661 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
8662 SVE structure modes.
8663 (aarch64_classify_address): Likewise.
8664 (sizetochar): Move earlier in file.
8665 (aarch64_print_operand): Handle SVE register lists.
8666 (aarch64_array_mode): New function.
8667 (aarch64_sve_struct_memory_operand_p): Likewise.
8668 (TARGET_ARRAY_MODE): Redefine.
8670 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8671 Alan Hayward <alan.hayward@arm.com>
8672 David Sherwood <david.sherwood@arm.com>
8674 * target.def (array_mode): New target hook.
8675 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
8676 * doc/tm.texi: Regenerate.
8677 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
8678 * hooks.c (hook_optmode_mode_uhwi_none): New function.
8679 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
8681 * stor-layout.c (mode_for_array): Likewise. Support polynomial
8684 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8685 Alan Hayward <alan.hayward@arm.com>
8686 David Sherwood <david.sherwood@arm.com>
8688 * fold-const.c (fold_binary_loc): Check the argument types
8689 rather than the result type when testing for a vector operation.
8691 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8693 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
8694 * doc/tm.texi: Regenerate.
8696 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8697 Alan Hayward <alan.hayward@arm.com>
8698 David Sherwood <david.sherwood@arm.com>
8700 * doc/invoke.texi (-msve-vector-bits=): Document new option.
8701 (sve): Document new AArch64 extension.
8702 * doc/md.texi (w): Extend the description of the AArch64
8703 constraint to include SVE vectors.
8704 (Upl, Upa): Document new AArch64 predicate constraints.
8705 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
8707 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
8708 (msve-vector-bits=): New option.
8709 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
8710 SVE when these are disabled.
8711 (sve): New extension.
8712 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
8713 modes. Adjust their number of units based on aarch64_sve_vg.
8714 (MAX_BITSIZE_MODE_ANY_MODE): Define.
8715 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
8716 aarch64_addr_query_type.
8717 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
8718 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
8719 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
8720 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
8721 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
8722 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
8723 (aarch64_simd_imm_zero_p): Delete.
8724 (aarch64_check_zero_based_sve_index_immediate): Declare.
8725 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
8726 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
8727 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
8728 (aarch64_sve_float_mul_immediate_p): Likewise.
8729 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
8731 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
8732 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
8733 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
8734 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
8735 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
8736 (aarch64_regmode_natural_size): Likewise.
8737 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
8738 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
8740 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
8741 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
8742 for VG and the SVE predicate registers.
8743 (V_ALIASES): Add a "z"-prefixed alias.
8744 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
8745 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
8746 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
8747 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
8748 (REG_CLASS_NAMES): Add entries for them.
8749 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
8750 and the predicate registers.
8751 (aarch64_sve_vg): Declare.
8752 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
8753 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
8754 (REGMODE_NATURAL_SIZE): Define.
8755 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
8757 * config/aarch64/aarch64.c: Include cfgrtl.h.
8758 (simd_immediate_info): Add a constructor for series vectors,
8759 and an associated step field.
8760 (aarch64_sve_vg): New variable.
8761 (aarch64_dbx_register_number): Handle VG and the predicate registers.
8762 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
8763 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
8764 (VEC_ANY_DATA, VEC_STRUCT): New constants.
8765 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
8766 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
8767 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
8768 (aarch64_get_mask_mode): New functions.
8769 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
8770 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
8771 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
8772 predicate modes and predicate registers. Explicitly restrict
8773 GPRs to modes of 16 bytes or smaller. Only allow FP registers
8774 to store a vector mode if it is recognized by
8775 aarch64_classify_vector_mode.
8776 (aarch64_regmode_natural_size): New function.
8777 (aarch64_hard_regno_caller_save_mode): Return the original mode
8779 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
8780 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
8781 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
8782 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
8784 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
8785 does not overlap dest if the function is frame-related. Handle
8787 (aarch64_split_add_offset): New function.
8788 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
8789 them aarch64_add_offset.
8790 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
8791 and update call to aarch64_sub_sp.
8792 (aarch64_add_cfa_expression): New function.
8793 (aarch64_expand_prologue): Pass extra temporary registers to the
8794 functions above. Handle the case in which we need to emit new
8795 DW_CFA_expressions for registers that were originally saved
8796 relative to the stack pointer, but now have to be expressed
8797 relative to the frame pointer.
8798 (aarch64_output_mi_thunk): Pass extra temporary registers to the
8800 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
8801 IP0 and IP1 values for SVE frames.
8802 (aarch64_expand_vec_series): New function.
8803 (aarch64_expand_sve_widened_duplicate): Likewise.
8804 (aarch64_expand_sve_const_vector): Likewise.
8805 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
8806 Handle SVE constants. Use emit_move_insn to move a force_const_mem
8807 into the register, rather than emitting a SET directly.
8808 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
8809 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
8810 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
8811 (offset_9bit_signed_scaled_p): New functions.
8812 (aarch64_replicate_bitmask_imm): New function.
8813 (aarch64_bitmask_imm): Use it.
8814 (aarch64_cannot_force_const_mem): Reject expressions involving
8815 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
8816 (aarch64_classify_index): Handle SVE indices, by requiring
8817 a plain register index with a scale that matches the element size.
8818 (aarch64_classify_address): Handle SVE addresses. Assert that
8819 the mode of the address is VOIDmode or an integer mode.
8820 Update call to aarch64_classify_symbol.
8821 (aarch64_classify_symbolic_expression): Update call to
8822 aarch64_classify_symbol.
8823 (aarch64_const_vec_all_in_range_p): New function.
8824 (aarch64_print_vector_float_operand): Likewise.
8825 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
8826 "vN" for FP registers with SVE modes. Handle (const ...) vectors
8827 and the FP immediates 1.0 and 0.5.
8828 (aarch64_print_address_internal): Handle SVE addresses.
8829 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
8830 (aarch64_regno_regclass): Handle predicate registers.
8831 (aarch64_secondary_reload): Handle big-endian reloads of SVE
8833 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
8834 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
8835 (aarch64_convert_sve_vector_bits): New function.
8836 (aarch64_override_options): Use it to handle -msve-vector-bits=.
8837 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
8839 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
8840 Handle SVE vector and predicate modes. Accept VL-based constants
8841 that need only one temporary register, and VL offsets that require
8842 no temporary registers.
8843 (aarch64_conditional_register_usage): Mark the predicate registers
8844 as fixed if SVE isn't available.
8845 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
8846 Return true for SVE vector and predicate modes.
8847 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
8848 rather than an unsigned int. Handle SVE modes.
8849 (aarch64_preferred_simd_mode): Update call accordingly. Handle
8851 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
8853 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
8854 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
8855 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
8856 (aarch64_sve_float_mul_immediate_p): New functions.
8857 (aarch64_sve_valid_immediate): New function.
8858 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
8859 Explicitly reject structure modes. Check for INDEX constants.
8860 Handle PTRUE and PFALSE constants.
8861 (aarch64_check_zero_based_sve_index_immediate): New function.
8862 (aarch64_simd_imm_zero_p): Delete.
8863 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
8864 vector modes. Accept constants in the range of CNT[BHWD].
8865 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
8866 ask for an Advanced SIMD mode.
8867 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
8868 (aarch64_simd_vector_alignment): Handle SVE predicates.
8869 (aarch64_vectorize_preferred_vector_alignment): New function.
8870 (aarch64_simd_vector_alignment_reachable): Use it instead of
8872 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
8873 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
8875 (MAX_VECT_LEN): Delete.
8876 (expand_vec_perm_d): Add a vec_flags field.
8877 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
8878 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
8879 (aarch64_evpc_ext): Don't apply a big-endian lane correction
8881 (aarch64_evpc_rev): Rename to...
8882 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
8883 (aarch64_evpc_rev_global): New function.
8884 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
8885 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
8887 (aarch64_evpc_sve_tbl): New function.
8888 (aarch64_expand_vec_perm_const_1): Update after rename of
8889 aarch64_evpc_rev. Handle SVE permutes too, trying
8890 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
8891 than aarch64_evpc_tbl.
8892 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
8893 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
8894 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
8895 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
8896 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
8897 (aarch64_expand_sve_vcond): New functions.
8898 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
8899 of aarch64_vector_mode_p.
8900 (aarch64_dwarf_poly_indeterminate_value): New function.
8901 (aarch64_compute_pressure_classes): Likewise.
8902 (aarch64_can_change_mode_class): Likewise.
8903 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
8904 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
8905 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
8906 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
8907 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
8908 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
8909 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
8910 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
8912 (Dn, Dl, Dr): Accept const as well as const_vector.
8913 (Dz): Likewise. Compare against CONST0_RTX.
8914 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
8915 of "vector" where appropriate.
8916 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
8917 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
8918 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
8919 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
8920 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
8921 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
8922 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
8923 (v_int_equiv): Extend to SVE modes.
8924 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
8926 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
8927 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
8928 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
8929 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
8930 (SVE_COND_FP_CMP): New int iterators.
8931 (perm_hilo): Handle the new unpack unspecs.
8932 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
8934 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
8935 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
8936 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
8937 (aarch64_equality_operator, aarch64_constant_vector_operand)
8938 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
8939 (aarch64_sve_nonimmediate_operand): Likewise.
8940 (aarch64_sve_general_operand): Likewise.
8941 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
8942 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
8943 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
8944 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
8945 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
8946 (aarch64_sve_float_arith_immediate): Likewise.
8947 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
8948 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
8949 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
8950 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
8951 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
8952 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
8953 (aarch64_sve_float_arith_operand): Likewise.
8954 (aarch64_sve_float_arith_with_sub_operand): Likewise.
8955 (aarch64_sve_float_mul_operand): Likewise.
8956 (aarch64_sve_vec_perm_operand): Likewise.
8957 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
8958 (aarch64_mov_operand): Accept const_poly_int and const_vector.
8959 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
8960 as well as const_vector.
8961 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
8962 in file. Use CONST0_RTX and CONSTM1_RTX.
8963 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
8964 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
8965 Use aarch64_simd_imm_zero.
8966 * config/aarch64/aarch64-sve.md: New file.
8967 * config/aarch64/aarch64.md: Include it.
8968 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
8969 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
8970 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
8971 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
8972 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
8973 (sve): New attribute.
8974 (enabled): Disable instructions with the sve attribute unless
8976 (movqi, movhi): Pass CONST_POLY_INT operaneds through
8977 aarch64_expand_mov_immediate.
8978 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
8979 CNT[BHSD] immediates.
8980 (movti): Split CONST_POLY_INT moves into two halves.
8981 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
8982 Split additions that need a temporary here if the destination
8983 is the stack pointer.
8984 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
8985 (*add<mode>3_poly_1): New instruction.
8986 (set_clobber_cc): New expander.
8988 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8990 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
8991 parameter and use it instead of GET_MODE_SIZE (innermode). Use
8992 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
8993 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
8994 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
8995 Change innermode from fixed_mode_size to machine_mode.
8996 (simplify_subreg): Update call accordingly. Handle a constant-sized
8997 subreg of a variable-length CONST_VECTOR.
8999 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9000 Alan Hayward <alan.hayward@arm.com>
9001 David Sherwood <david.sherwood@arm.com>
9003 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
9004 (add_offset_to_base): New function, split out from...
9005 (create_mem_ref): ...here. When handling a scale other than 1,
9006 check first whether the address is valid without the offset.
9007 Add it into the base if so, leaving the index and scale as-is.
9009 2018-01-12 Jakub Jelinek <jakub@redhat.com>
9012 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
9013 fold_for_warn before checking if arg2 is INTEGER_CST.
9015 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
9017 * config/rs6000/predicates.md (load_multiple_operation): Delete.
9018 (store_multiple_operation): Delete.
9019 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
9020 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
9021 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
9022 guarded by TARGET_STRING.
9023 (rs6000_output_load_multiple): Delete.
9024 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
9025 OPTION_MASK_STRING / TARGET_STRING handling.
9026 (print_operand) <'N', 'O'>: Add comment that these are unused now.
9027 (const rs6000_opt_masks) <"string">: Change mask to 0.
9028 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
9029 (MASK_STRING): Delete.
9030 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
9032 (load_multiple): Delete.
9039 (store_multiple): Delete.
9046 (movmemsi_8reg): Delete.
9047 (corresponding unnamed define_insn): Delete.
9048 (movmemsi_6reg): Delete.
9049 (corresponding unnamed define_insn): Delete.
9050 (movmemsi_4reg): Delete.
9051 (corresponding unnamed define_insn): Delete.
9052 (movmemsi_2reg): Delete.
9053 (corresponding unnamed define_insn): Delete.
9054 (movmemsi_1reg): Delete.
9055 (corresponding unnamed define_insn): Delete.
9056 * config/rs6000/rs6000.opt (mno-string): New.
9057 (mstring): Replace by deprecation warning stub.
9058 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
9060 2018-01-12 Jakub Jelinek <jakub@redhat.com>
9062 * regrename.c (regrename_do_replace): If replacing the same
9063 reg multiple times, try to reuse last created gen_raw_REG.
9066 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
9067 main to workaround a bug in GDB.
9069 2018-01-12 Tom de Vries <tom@codesourcery.com>
9072 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
9074 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
9076 PR rtl-optimization/80481
9077 * ira-color.c (get_cap_member): New function.
9078 (allocnos_conflict_by_live_ranges_p): Use it.
9079 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
9080 (setup_slot_coalesced_allocno_live_ranges): Ditto.
9082 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
9085 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
9086 (*saddl_se_1): Ditto.
9088 (*ssubl_se_1): Ditto.
9090 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9092 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
9093 rather than wi::to_widest for DR_INITs.
9094 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
9095 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
9096 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
9098 (vect_analyze_group_access_1): Note that here.
9100 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9102 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
9103 polynomial type sizes.
9105 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9107 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
9108 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
9109 (gimple_add_tmp_var): Likewise.
9111 2018-01-12 Martin Liska <mliska@suse.cz>
9113 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
9114 (gimple_alloc_sizes): Likewise.
9115 (dump_gimple_statistics): Use PRIu64 in printf format.
9116 * gimple.h: Change uint64_t to int.
9118 2018-01-12 Martin Liska <mliska@suse.cz>
9120 * tree-core.h: Use uint64_t instead of int.
9121 * tree.c (tree_node_counts): Likewise.
9122 (tree_node_sizes): Likewise.
9123 (dump_tree_statistics): Use PRIu64 in printf format.
9125 2018-01-12 Martin Liska <mliska@suse.cz>
9127 * Makefile.in: As qsort_chk is implemented in vec.c, add
9128 vec.o to linkage of gencfn-macros.
9129 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
9130 passing the info to record_node_allocation_statistics.
9131 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
9133 * ggc-common.c (struct ggc_usage): Add operator== and use
9134 it in operator< and compare function.
9135 * mem-stats.h (struct mem_usage): Likewise.
9136 * vec.c (struct vec_usage): Remove operator< and compare
9137 function. Can be simply inherited.
9139 2018-01-12 Martin Jambor <mjambor@suse.cz>
9142 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
9143 * tree-ssa-math-opts.c: Include domwalk.h.
9144 (convert_mult_to_fma_1): New function.
9145 (fma_transformation_info): New type.
9146 (fma_deferring_state): Likewise.
9147 (cancel_fma_deferring): New function.
9148 (result_of_phi): Likewise.
9149 (last_fma_candidate_feeds_initial_phi): Likewise.
9150 (convert_mult_to_fma): Added deferring logic, split actual
9151 transformation to convert_mult_to_fma_1.
9152 (math_opts_dom_walker): New type.
9153 (math_opts_dom_walker::after_dom_children): New method, body moved
9154 here from pass_optimize_widening_mul::execute, added deferring logic
9156 (pass_optimize_widening_mul::execute): Moved most of code to
9157 math_opts_dom_walker::after_dom_children.
9158 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
9159 * config/i386/i386.c (ix86_option_override_internal): Added
9160 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
9162 2018-01-12 Richard Biener <rguenther@suse.de>
9165 * dwarf2out.c (gen_variable_die): Do not reset old_die for
9166 inline instance vars.
9168 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
9171 * config/rx/rx.c (rx_is_restricted_memory_address):
9174 2018-01-12 Richard Biener <rguenther@suse.de>
9176 PR tree-optimization/80846
9177 * target.def (split_reduction): New target hook.
9178 * targhooks.c (default_split_reduction): New function.
9179 * targhooks.h (default_split_reduction): Declare.
9180 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
9181 target requests first reduce vectors by combining low and high
9183 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
9184 (get_vectype_for_scalar_type_and_size): Export.
9185 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
9186 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
9187 * doc/tm.texi: Regenerate.
9188 * config/i386/i386.c (ix86_split_reduction): Implement
9189 TARGET_VECTORIZE_SPLIT_REDUCTION.
9191 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
9194 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
9195 in PIC mode except for TARGET_VXWORKS_RTP.
9196 * config/sparc/sparc.c: Include cfgrtl.h.
9197 (TARGET_INIT_PIC_REG): Define.
9198 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
9199 (sparc_pic_register_p): New predicate.
9200 (sparc_legitimate_address_p): Use it.
9201 (sparc_legitimize_pic_address): Likewise.
9202 (sparc_delegitimize_address): Likewise.
9203 (sparc_mode_dependent_address_p): Likewise.
9204 (gen_load_pcrel_sym): Remove 4th parameter.
9205 (load_got_register): Adjust call to above. Remove obsolete stuff.
9206 (sparc_expand_prologue): Do not call load_got_register here.
9207 (sparc_flat_expand_prologue): Likewise.
9208 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
9209 (sparc_use_pseudo_pic_reg): New function.
9210 (sparc_init_pic_reg): Likewise.
9211 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
9212 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
9214 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
9216 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
9217 Add item for branch_cost.
9219 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
9221 PR rtl-optimization/83565
9222 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
9223 not extend the result to a larger mode for rotate operations.
9224 (num_sign_bit_copies1): Likewise.
9226 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9229 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
9231 Use values-Xc.o for -pedantic.
9232 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
9234 2018-01-12 Martin Liska <mliska@suse.cz>
9237 * ipa-devirt.c (final_warning_record::grow_type_warnings):
9239 (possible_polymorphic_call_targets): Use it.
9240 (ipa_devirt): Likewise.
9242 2018-01-12 Martin Liska <mliska@suse.cz>
9244 * profile-count.h (enum profile_quality): Use 0 as invalid
9245 enum value of profile_quality.
9247 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
9249 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
9250 -mext-string options.
9252 2018-01-12 Richard Biener <rguenther@suse.de>
9254 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
9255 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
9256 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
9258 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
9260 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
9262 * configure.ac (--with-long-double-format): Add support for the
9263 configuration option to change the default long double format on
9265 * config.gcc (powerpc*-linux*-*): Likewise.
9266 * configure: Regenerate.
9267 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
9268 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
9269 used without modification.
9271 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9273 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
9274 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
9275 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
9276 MISC_BUILTIN_SPEC_BARRIER.
9277 (rs6000_init_builtins): Likewise.
9278 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
9280 (speculation_barrier): New define_insn.
9281 * doc/extend.texi: Document __builtin_speculation_barrier.
9283 2018-01-11 Jakub Jelinek <jakub@redhat.com>
9286 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
9287 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
9288 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
9290 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
9291 integral modes instead of "ss" and "sd".
9292 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
9293 vectors with 32-bit and 64-bit elements.
9294 (vecdupssescalarmodesuffix): New mode attribute.
9295 (vec_dup<mode>): Use it.
9297 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
9300 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
9301 frame if argument is passed on stack.
9303 2018-01-11 Jakub Jelinek <jakub@redhat.com>
9306 * ree.c (combine_reaching_defs): Optimize also
9307 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
9308 reg2=any_extend(exp); reg1=reg2;, formatting fix.
9310 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
9313 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
9315 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
9318 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
9319 after they are computed.
9321 2018-01-11 Bin Cheng <bin.cheng@arm.com>
9323 PR tree-optimization/83695
9324 * gimple-loop-linterchange.cc
9325 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
9326 reset cached scev information after interchange.
9327 (pass_linterchange::execute): Remove call to scev_reset_htab.
9329 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9331 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
9332 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
9333 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
9334 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
9335 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
9336 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
9337 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
9338 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
9339 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
9340 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
9341 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
9342 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
9343 (V_lane_reg): Likewise.
9344 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
9346 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
9347 (vfmal_lane_low<mode>_intrinsic,
9348 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
9349 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
9350 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
9351 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
9352 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
9353 vfmsl_lane_high<mode>_intrinsic): New define_insns.
9355 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9357 * config/arm/arm-cpus.in (fp16fml): New feature.
9358 (ALL_SIMD): Add fp16fml.
9359 (armv8.2-a): Add fp16fml as an option.
9360 (armv8.3-a): Likewise.
9361 (armv8.4-a): Add fp16fml as part of fp16.
9362 * config/arm/arm.h (TARGET_FP16FML): Define.
9363 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
9365 * config/arm/arm-modes.def (V2HF): Define.
9366 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
9367 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
9368 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
9369 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
9370 vfmsl_low, vfmsl_high): New set of builtins.
9371 * config/arm/iterators.md (PLUSMINUS): New code iterator.
9372 (vfml_op): New code attribute.
9373 (VFMLHALVES): New int iterator.
9374 (VFML, VFMLSEL): New mode attributes.
9375 (V_reg): Define mapping for V2HF.
9376 (V_hi, V_lo): New mode attributes.
9377 (VF_constraint): Likewise.
9378 (vfml_half, vfml_half_selector): New int attributes.
9379 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
9381 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
9382 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
9384 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
9385 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
9386 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
9387 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
9389 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
9390 Document new effective target and option set.
9392 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9394 * config/arm/arm-cpus.in (armv8_4): New feature.
9395 (ARMv8_4a): New fgroup.
9396 (armv8.4-a): New arch.
9397 * config/arm/arm-tables.opt: Regenerate.
9398 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
9399 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
9400 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
9401 Add matching rules for -march=armv8.4-a and extensions.
9402 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
9404 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
9407 * config/rx/rx.md (BW): New mode attribute.
9408 (sync_lock_test_and_setsi): Add mode suffix to insn output.
9410 2018-01-11 Richard Biener <rguenther@suse.de>
9412 PR tree-optimization/83435
9413 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
9414 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
9415 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
9417 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9418 Alan Hayward <alan.hayward@arm.com>
9419 David Sherwood <david.sherwood@arm.com>
9421 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
9423 (aarch64_classify_address): Initialize it. Track polynomial offsets.
9424 (aarch64_print_address_internal): Use it to check for a zero offset.
9426 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9427 Alan Hayward <alan.hayward@arm.com>
9428 David Sherwood <david.sherwood@arm.com>
9430 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
9431 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
9432 Return a poly_int64 rather than a HOST_WIDE_INT.
9433 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
9434 rather than a HOST_WIDE_INT.
9435 * config/aarch64/aarch64.h (aarch64_frame): Protect with
9436 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
9437 hard_fp_offset, frame_size, initial_adjust, callee_offset and
9438 final_offset from HOST_WIDE_INT to poly_int64.
9439 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
9440 to_constant when getting the number of units in an Advanced SIMD
9442 (aarch64_builtin_vectorized_function): Check for a constant number
9444 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
9446 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
9447 attribute instead of GET_MODE_NUNITS.
9448 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
9449 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
9450 GET_MODE_SIZE for fixed-size registers.
9451 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
9452 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
9453 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
9454 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
9455 (aarch64_print_operand, aarch64_print_address_internal)
9456 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
9457 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
9458 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
9459 Handle polynomial GET_MODE_SIZE.
9460 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
9461 wider than SImode without modification.
9462 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
9463 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
9464 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
9465 passing and returning SVE modes.
9466 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
9467 rather than GEN_INT.
9468 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
9469 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
9470 (aarch64_allocate_and_probe_stack_space): Likewise.
9471 (aarch64_layout_frame): Cope with polynomial offsets.
9472 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
9473 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
9475 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
9476 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
9477 poly_int64 rather than a HOST_WIDE_INT.
9478 (aarch64_get_separate_components, aarch64_process_components)
9479 (aarch64_expand_prologue, aarch64_expand_epilogue)
9480 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
9481 (aarch64_anchor_offset): New function, split out from...
9482 (aarch64_legitimize_address): ...here.
9483 (aarch64_builtin_vectorization_cost): Handle polynomial
9484 TYPE_VECTOR_SUBPARTS.
9485 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
9487 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
9488 number of elements from the PARALLEL rather than the mode.
9489 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
9490 rather than GET_MODE_BITSIZE.
9491 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
9492 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
9493 (aarch64_expand_vec_perm_const_1): Handle polynomial
9494 d->perm.length () and d->perm elements.
9495 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
9496 Apply to_constant to d->perm elements.
9497 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
9498 polynomial CONST_VECTOR_NUNITS.
9499 (aarch64_move_pointer): Take amount as a poly_int64 rather
9501 (aarch64_progress_pointer): Avoid temporary variable.
9502 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
9503 the mode attribute instead of GET_MODE.
9505 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9506 Alan Hayward <alan.hayward@arm.com>
9507 David Sherwood <david.sherwood@arm.com>
9509 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
9510 x exists before using it.
9511 (aarch64_add_constant_internal): Rename to...
9512 (aarch64_add_offset_1): ...this. Replace regnum with separate
9513 src and dest rtxes. Handle the case in which they're different,
9514 including when the offset is zero. Replace scratchreg with an rtx.
9515 Use 2 additions if there is no spare register into which we can
9516 move a 16-bit constant.
9517 (aarch64_add_constant): Delete.
9518 (aarch64_add_offset): Replace reg with separate src and dest
9519 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
9520 Use aarch64_add_offset_1.
9521 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
9522 an rtx rather than an int. Take the delta as a poly_int64
9523 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
9524 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
9525 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
9526 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
9527 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
9529 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
9530 aarch64_add_constant.
9532 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9534 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
9535 Use scalar_float_mode.
9537 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9539 * config/aarch64/aarch64-simd.md
9540 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
9541 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
9542 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
9543 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
9544 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
9545 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
9546 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
9547 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
9548 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
9549 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
9551 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9554 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
9555 targ_options->x_arm_arch_string is non NULL.
9557 2018-01-11 Tamar Christina <tamar.christina@arm.com>
9559 * config/aarch64/aarch64.h
9560 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
9562 2018-01-11 Sudakshina Das <sudi.das@arm.com>
9565 * expmed.c (emit_store_flag_force): Swap if const op0
9566 and change VOIDmode to mode of op0.
9568 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9570 PR rtl-optimization/83761
9571 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
9572 than bytes to mode_for_size.
9574 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9577 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
9578 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
9581 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9584 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
9585 when in layout mode.
9586 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
9587 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
9590 2018-01-10 Michael Collison <michael.collison@arm.com>
9592 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
9593 * config/aarch64/aarch64-option-extension.def: Add
9594 AARCH64_OPT_EXTENSION of 'fp16fml'.
9595 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9596 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
9597 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
9598 * config/aarch64/constraints.md (Ui7): New constraint.
9599 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
9600 (VFMLA_SEL_W): Ditto.
9603 (VFMLA16_LOW): New int iterator.
9604 (VFMLA16_HIGH): Ditto.
9605 (UNSPEC_FMLAL): New unspec.
9606 (UNSPEC_FMLSL): Ditto.
9607 (UNSPEC_FMLAL2): Ditto.
9608 (UNSPEC_FMLSL2): Ditto.
9609 (f16mac): New code attribute.
9610 * config/aarch64/aarch64-simd-builtins.def
9611 (aarch64_fmlal_lowv2sf): Ditto.
9612 (aarch64_fmlsl_lowv2sf): Ditto.
9613 (aarch64_fmlalq_lowv4sf): Ditto.
9614 (aarch64_fmlslq_lowv4sf): Ditto.
9615 (aarch64_fmlal_highv2sf): Ditto.
9616 (aarch64_fmlsl_highv2sf): Ditto.
9617 (aarch64_fmlalq_highv4sf): Ditto.
9618 (aarch64_fmlslq_highv4sf): Ditto.
9619 (aarch64_fmlal_lane_lowv2sf): Ditto.
9620 (aarch64_fmlsl_lane_lowv2sf): Ditto.
9621 (aarch64_fmlal_laneq_lowv2sf): Ditto.
9622 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
9623 (aarch64_fmlalq_lane_lowv4sf): Ditto.
9624 (aarch64_fmlsl_lane_lowv4sf): Ditto.
9625 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
9626 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
9627 (aarch64_fmlal_lane_highv2sf): Ditto.
9628 (aarch64_fmlsl_lane_highv2sf): Ditto.
9629 (aarch64_fmlal_laneq_highv2sf): Ditto.
9630 (aarch64_fmlsl_laneq_highv2sf): Ditto.
9631 (aarch64_fmlalq_lane_highv4sf): Ditto.
9632 (aarch64_fmlsl_lane_highv4sf): Ditto.
9633 (aarch64_fmlalq_laneq_highv4sf): Ditto.
9634 (aarch64_fmlsl_laneq_highv4sf): Ditto.
9635 * config/aarch64/aarch64-simd.md:
9636 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
9637 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
9638 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
9639 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
9640 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
9641 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
9642 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
9643 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
9644 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
9645 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
9646 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
9647 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
9648 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
9649 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
9650 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
9651 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
9652 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
9653 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
9654 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
9655 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
9656 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
9657 (vfmlsl_low_u32): Ditto.
9658 (vfmlalq_low_u32): Ditto.
9659 (vfmlslq_low_u32): Ditto.
9660 (vfmlal_high_u32): Ditto.
9661 (vfmlsl_high_u32): Ditto.
9662 (vfmlalq_high_u32): Ditto.
9663 (vfmlslq_high_u32): Ditto.
9664 (vfmlal_lane_low_u32): Ditto.
9665 (vfmlsl_lane_low_u32): Ditto.
9666 (vfmlal_laneq_low_u32): Ditto.
9667 (vfmlsl_laneq_low_u32): Ditto.
9668 (vfmlalq_lane_low_u32): Ditto.
9669 (vfmlslq_lane_low_u32): Ditto.
9670 (vfmlalq_laneq_low_u32): Ditto.
9671 (vfmlslq_laneq_low_u32): Ditto.
9672 (vfmlal_lane_high_u32): Ditto.
9673 (vfmlsl_lane_high_u32): Ditto.
9674 (vfmlal_laneq_high_u32): Ditto.
9675 (vfmlsl_laneq_high_u32): Ditto.
9676 (vfmlalq_lane_high_u32): Ditto.
9677 (vfmlslq_lane_high_u32): Ditto.
9678 (vfmlalq_laneq_high_u32): Ditto.
9679 (vfmlslq_laneq_high_u32): Ditto.
9680 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
9681 (AARCH64_FL_FOR_ARCH8_4): New.
9682 (AARCH64_ISA_F16FML): New ISA flag.
9683 (TARGET_F16FML): New feature flag for fp16fml.
9684 (doc/invoke.texi): Document new fp16fml option.
9686 2018-01-10 Michael Collison <michael.collison@arm.com>
9688 * config/aarch64/aarch64-builtins.c:
9689 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
9690 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9691 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
9692 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
9693 (AARCH64_ISA_SHA3): New ISA flag.
9694 (TARGET_SHA3): New feature flag for sha3.
9695 * config/aarch64/iterators.md (sha512_op): New int attribute.
9696 (CRYPTO_SHA512): New int iterator.
9697 (UNSPEC_SHA512H): New unspec.
9698 (UNSPEC_SHA512H2): Ditto.
9699 (UNSPEC_SHA512SU0): Ditto.
9700 (UNSPEC_SHA512SU1): Ditto.
9701 * config/aarch64/aarch64-simd-builtins.def
9702 (aarch64_crypto_sha512hqv2di): New builtin.
9703 (aarch64_crypto_sha512h2qv2di): Ditto.
9704 (aarch64_crypto_sha512su0qv2di): Ditto.
9705 (aarch64_crypto_sha512su1qv2di): Ditto.
9706 (aarch64_eor3qv8hi): Ditto.
9707 (aarch64_rax1qv2di): Ditto.
9708 (aarch64_xarqv2di): Ditto.
9709 (aarch64_bcaxqv8hi): Ditto.
9710 * config/aarch64/aarch64-simd.md:
9711 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
9712 (aarch64_crypto_sha512su0qv2di): Ditto.
9713 (aarch64_crypto_sha512su1qv2di): Ditto.
9714 (aarch64_eor3qv8hi): Ditto.
9715 (aarch64_rax1qv2di): Ditto.
9716 (aarch64_xarqv2di): Ditto.
9717 (aarch64_bcaxqv8hi): Ditto.
9718 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
9719 (vsha512h2q_u64): Ditto.
9720 (vsha512su0q_u64): Ditto.
9721 (vsha512su1q_u64): Ditto.
9722 (veor3q_u16): Ditto.
9723 (vrax1q_u64): Ditto.
9725 (vbcaxq_u16): Ditto.
9726 * config/arm/types.md (crypto_sha512): New type attribute.
9727 (crypto_sha3): Ditto.
9728 (doc/invoke.texi): Document new sha3 option.
9730 2018-01-10 Michael Collison <michael.collison@arm.com>
9732 * config/aarch64/aarch64-builtins.c:
9733 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
9734 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9735 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
9736 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
9737 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
9738 (AARCH64_ISA_SM4): New ISA flag.
9739 (TARGET_SM4): New feature flag for sm4.
9740 * config/aarch64/aarch64-simd-builtins.def
9741 (aarch64_sm3ss1qv4si): Ditto.
9742 (aarch64_sm3tt1aq4si): Ditto.
9743 (aarch64_sm3tt1bq4si): Ditto.
9744 (aarch64_sm3tt2aq4si): Ditto.
9745 (aarch64_sm3tt2bq4si): Ditto.
9746 (aarch64_sm3partw1qv4si): Ditto.
9747 (aarch64_sm3partw2qv4si): Ditto.
9748 (aarch64_sm4eqv4si): Ditto.
9749 (aarch64_sm4ekeyqv4si): Ditto.
9750 * config/aarch64/aarch64-simd.md:
9751 (aarch64_sm3ss1qv4si): Ditto.
9752 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
9753 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
9754 (aarch64_sm4eqv4si): Ditto.
9755 (aarch64_sm4ekeyqv4si): Ditto.
9756 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
9757 (sm3part_op): Ditto.
9758 (CRYPTO_SM3TT): Ditto.
9759 (CRYPTO_SM3PART): Ditto.
9760 (UNSPEC_SM3SS1): New unspec.
9761 (UNSPEC_SM3TT1A): Ditto.
9762 (UNSPEC_SM3TT1B): Ditto.
9763 (UNSPEC_SM3TT2A): Ditto.
9764 (UNSPEC_SM3TT2B): Ditto.
9765 (UNSPEC_SM3PARTW1): Ditto.
9766 (UNSPEC_SM3PARTW2): Ditto.
9767 (UNSPEC_SM4E): Ditto.
9768 (UNSPEC_SM4EKEY): Ditto.
9769 * config/aarch64/constraints.md (Ui2): New constraint.
9770 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
9771 * config/arm/types.md (crypto_sm3): New type attribute.
9772 (crypto_sm4): Ditto.
9773 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
9774 (vsm3tt1aq_u32): Ditto.
9775 (vsm3tt1bq_u32): Ditto.
9776 (vsm3tt2aq_u32): Ditto.
9777 (vsm3tt2bq_u32): Ditto.
9778 (vsm3partw1q_u32): Ditto.
9779 (vsm3partw2q_u32): Ditto.
9780 (vsm4eq_u32): Ditto.
9781 (vsm4ekeyq_u32): Ditto.
9782 (doc/invoke.texi): Document new sm4 option.
9784 2018-01-10 Michael Collison <michael.collison@arm.com>
9786 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
9787 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
9788 (AARCH64_FL_FOR_ARCH8_4): New.
9789 (AARCH64_FL_V8_4): New flag.
9790 (doc/invoke.texi): Document new armv8.4-a option.
9792 2018-01-10 Michael Collison <michael.collison@arm.com>
9794 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9795 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
9796 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
9797 * config/aarch64/aarch64-option-extension.def: Add
9798 AARCH64_OPT_EXTENSION of 'sha2'.
9799 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
9800 (crypto): Disable sha2 and aes if crypto disabled.
9801 (crypto): Enable aes and sha2 if enabled.
9802 (simd): Disable sha2 and aes if simd disabled.
9803 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
9805 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
9806 (TARGET_SHA2): New feature flag for sha2.
9807 (TARGET_AES): New feature flag for aes.
9808 * config/aarch64/aarch64-simd.md:
9809 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
9810 conditional on TARGET_AES.
9811 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
9812 (aarch64_crypto_sha1hsi): Make pattern conditional
9814 (aarch64_crypto_sha1hv4si): Ditto.
9815 (aarch64_be_crypto_sha1hv4si): Ditto.
9816 (aarch64_crypto_sha1su1v4si): Ditto.
9817 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
9818 (aarch64_crypto_sha1su0v4si): Ditto.
9819 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
9820 (aarch64_crypto_sha256su0v4si): Ditto.
9821 (aarch64_crypto_sha256su1v4si): Ditto.
9822 (doc/invoke.texi): Document new aes and sha2 options.
9824 2018-01-10 Martin Sebor <msebor@redhat.com>
9826 PR tree-optimization/83781
9827 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
9830 2018-01-11 Martin Sebor <msebor@gmail.com>
9831 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9833 PR tree-optimization/83501
9834 PR tree-optimization/81703
9836 * tree-ssa-strlen.c (get_string_cst): Rename...
9837 (get_string_len): ...to this. Handle global constants.
9838 (handle_char_store): Adjust.
9840 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
9841 Jim Wilson <jimw@sifive.com>
9843 * config/riscv/riscv-protos.h (riscv_output_return): New.
9844 * config/riscv/riscv.c (struct machine_function): New naked_p field.
9845 (riscv_attribute_table, riscv_output_return),
9846 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
9847 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
9848 (riscv_compute_frame_info): Only compute frame->mask if not a naked
9850 (riscv_expand_prologue): Add early return for naked function.
9851 (riscv_expand_epilogue): Likewise.
9852 (riscv_function_ok_for_sibcall): Return false for naked function.
9853 (riscv_set_current_function): New.
9854 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
9855 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
9856 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
9857 * doc/extend.texi (RISC-V Function Attributes): New.
9859 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
9861 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
9862 check for 128-bit long double before checking TCmode.
9863 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
9864 128-bit long doubles before checking TFmode or TCmode.
9865 (FLOAT128_IBM_P): Likewise.
9867 2018-01-10 Martin Sebor <msebor@redhat.com>
9869 PR tree-optimization/83671
9870 * builtins.c (c_strlen): Unconditionally return zero for the empty
9872 Use -Warray-bounds for warnings.
9873 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
9874 for non-constant array indices with COMPONENT_REF, arrays of
9875 arrays, and pointers to arrays.
9876 (gimple_fold_builtin_strlen): Determine and set length range for
9877 non-constant character arrays.
9879 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
9882 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
9885 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
9887 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
9889 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
9892 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
9893 VECTOR_MEM_ALTIVEC_OR_VSX_P.
9894 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
9895 indexed_or_indirect_operand predicate.
9896 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
9897 (*vsx_le_perm_load_v8hi): Likewise.
9898 (*vsx_le_perm_load_v16qi): Likewise.
9899 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
9900 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
9901 (*vsx_le_perm_store_v8hi): Likewise.
9902 (*vsx_le_perm_store_v16qi): Likewise.
9903 (eight unnamed splitters): Likewise.
9905 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
9907 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
9908 * config/rs6000/emmintrin.h: Likewise.
9909 * config/rs6000/mmintrin.h: Likewise.
9910 * config/rs6000/xmmintrin.h: Likewise.
9912 2018-01-10 David Malcolm <dmalcolm@redhat.com>
9915 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
9917 * tree.c (tree_nop_conversion): Return true for location wrapper
9919 (maybe_wrap_with_location): New function.
9920 (selftest::check_strip_nops): New function.
9921 (selftest::test_location_wrappers): New function.
9922 (selftest::tree_c_tests): Call it.
9923 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
9924 (maybe_wrap_with_location): New decl.
9925 (EXPR_LOCATION_WRAPPER_P): New macro.
9926 (location_wrapper_p): New inline function.
9927 (tree_strip_any_location_wrapper): New inline function.
9929 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
9932 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
9933 stack_realign_offset for the largest alignment of stack slot
9935 (ix86_find_max_used_stack_alignment): New function.
9936 (ix86_finalize_stack_frame_flags): Use it. Set
9937 max_used_stack_alignment if we don't realign stack.
9938 * config/i386/i386.h (machine_function): Add
9939 max_used_stack_alignment.
9941 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
9943 * config/arm/arm.opt (-mbranch-cost): New option.
9944 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
9947 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
9950 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
9951 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
9953 2018-01-10 Richard Biener <rguenther@suse.de>
9956 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
9957 early out so it also covers the case where we have a non-NULL
9960 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
9962 PR tree-optimization/83753
9963 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
9964 for non-strided grouped accesses if the number of elements is 1.
9966 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9969 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
9970 * i386.h (TARGET_USE_GATHER): Define.
9971 * x86-tune.def (X86_TUNE_USE_GATHER): New.
9973 2018-01-10 Martin Liska <mliska@suse.cz>
9976 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
9977 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
9979 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
9980 CLEANUP_NO_PARTITIONING is not set.
9982 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
9984 * doc/rtl.texi: Remove documentation of (const ...) wrappers
9985 for vectors, as a partial revert of r254296.
9986 * rtl.h (const_vec_p): Delete.
9987 (const_vec_duplicate_p): Don't test for vector CONSTs.
9988 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
9989 * expmed.c (make_tree): Likewise.
9992 * common.md (E, F): Use CONSTANT_P instead of checking for
9994 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
9995 checking for CONST_VECTOR.
9997 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10000 * predict.c (force_edge_cold): Handle in more sane way edges
10001 with no prediction.
10003 2018-01-09 Carl Love <cel@us.ibm.com>
10005 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
10007 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
10008 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
10009 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
10010 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
10011 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
10012 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
10013 * config/rs6000/rs6000-protos.h: Add extern defition for
10014 rs6000_generate_float2_double_code.
10015 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
10017 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
10018 (float2_v2df): Add define_expand.
10020 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
10023 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
10024 op_mode in the force_to_mode call.
10026 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10028 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
10029 instead of checking each element individually.
10030 (aarch64_evpc_uzp): Likewise.
10031 (aarch64_evpc_zip): Likewise.
10032 (aarch64_evpc_ext): Likewise.
10033 (aarch64_evpc_rev): Likewise.
10034 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
10035 instead of checking each element individually. Return true without
10037 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
10038 whether all selected elements come from the same input, instead of
10039 checking each element individually. Remove calls to gen_rtx_REG,
10040 start_sequence and end_sequence and instead assert that no rtl is
10043 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10045 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
10046 order of HIGH and CONST checks.
10048 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10050 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
10051 if the destination isn't an SSA_NAME.
10053 2018-01-09 Richard Biener <rguenther@suse.de>
10055 PR tree-optimization/83668
10056 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
10058 (canonicalize_loop_form): ... here, renamed from ...
10059 (canonicalize_loop_closed_ssa_form): ... this and amended to
10060 swap successor edges for loop exit blocks to make us use
10061 the RPO order we need for initial schedule generation.
10063 2018-01-09 Joseph Myers <joseph@codesourcery.com>
10065 PR tree-optimization/64811
10066 * match.pd: When optimizing comparisons with Inf, avoid
10067 introducing or losing exceptions from comparisons with NaN.
10069 2018-01-09 Martin Liska <mliska@suse.cz>
10072 * asan.c (shadow_mem_size): Add gcc_assert.
10074 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
10076 Don't save registers in main().
10079 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
10080 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
10081 * config/avr/avr.c (avr_set_current_function): Don't error if
10082 naked, OS_task or OS_main are specified at the same time.
10083 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
10085 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
10087 * common/config/avr/avr-common.c (avr_option_optimization_table):
10088 Switch on -mmain-is-OS_task for optimizing compilations.
10090 2018-01-09 Richard Biener <rguenther@suse.de>
10092 PR tree-optimization/83572
10093 * graphite.c: Include cfganal.h.
10094 (graphite_transform_loops): Connect infinite loops to exit
10095 and remove fake edges at the end.
10097 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10099 * ipa-inline.c (edge_badness): Revert accidental checkin.
10101 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10104 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
10105 symbols; not inline clones.
10107 2018-01-09 Jakub Jelinek <jakub@redhat.com>
10110 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
10111 hard registers. Formatting fixes.
10113 PR preprocessor/83722
10114 * gcc.c (try_generate_repro): Pass
10115 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
10116 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
10119 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
10120 Kito Cheng <kito.cheng@gmail.com>
10122 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
10123 (riscv_leaf_function_p): Delete.
10124 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
10126 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10128 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
10130 (do_ifelse): New function.
10131 (do_isel): New function.
10132 (do_sub3): New function.
10133 (do_add3): New function.
10134 (do_load_mask_compare): New function.
10135 (do_overlap_load_compare): New function.
10136 (expand_compare_loop): New function.
10137 (expand_block_compare): Call expand_compare_loop() when appropriate.
10138 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
10139 option description.
10140 (-mblock-compare-inline-loop-limit): New option.
10142 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10145 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
10146 Reverse order of second and third operands in first alternative.
10147 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
10148 of first and second elements in UNSPEC_VPERMR vector.
10149 (altivec_expand_vec_perm_le): Likewise.
10151 2018-01-08 Jeff Law <law@redhat.com>
10153 PR rtl-optimizatin/81308
10154 * tree-switch-conversion.c (cfg_altered): New file scoped static.
10155 (process_switch): If group_case_labels makes a change, then set
10157 (pass_convert_switch::execute): If a switch is converted, then
10158 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
10160 PR rtl-optimization/81308
10161 * recog.c (split_all_insns): Conditionally cleanup the CFG after
10164 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
10166 PR target/83663 - Revert r255946
10167 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
10168 generation for cases where splatting a value is not useful.
10169 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
10170 across a vec_duplicate and a paradoxical subreg forming a vector
10171 mode to a vec_concat.
10173 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10175 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
10176 -march=armv8.3-a variants.
10177 * config/arm/t-multilib: Likewise.
10178 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
10180 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10182 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
10184 (cceq_ior_compare_complement): Give it a name so I can use it, and
10185 change boolean_or_operator predicate to boolean_operator so it can
10186 be used to generate a crand.
10187 (eqne): New code iterator.
10188 (bd/bd_neg): New code_attrs.
10189 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
10190 a single define_insn.
10191 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
10192 decrement (bdnzt/bdnzf/bdzt/bdzf).
10193 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
10194 with the new names of the branch decrement patterns, and added the
10195 names of the branch decrement conditional patterns.
10197 2018-01-08 Richard Biener <rguenther@suse.de>
10199 PR tree-optimization/83563
10200 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
10203 2018-01-08 Richard Biener <rguenther@suse.de>
10205 PR middle-end/83713
10206 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
10208 2018-01-08 Richard Biener <rguenther@suse.de>
10210 PR tree-optimization/83685
10211 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
10212 references to abnormals.
10214 2018-01-08 Richard Biener <rguenther@suse.de>
10217 * dwarf2out.c (output_indirect_strings): Handle empty
10218 skeleton_debug_str_hash.
10219 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
10221 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
10223 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
10224 (emit_store_direct): Likewise.
10225 (arc_trampoline_adjust_address): Likewise.
10226 (arc_asm_trampoline_template): New function.
10227 (arc_initialize_trampoline): Use asm_trampoline_template.
10228 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
10229 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
10230 * config/arc/arc.md (flush_icache): Delete pattern.
10232 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
10234 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
10235 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
10238 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
10241 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
10242 by not USED_FOR_TARGET.
10243 (make_pass_resolve_sw_modes): Likewise.
10245 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
10247 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
10250 2018-01-08 Richard Biener <rguenther@suse.de>
10252 PR middle-end/83580
10253 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
10255 2018-01-08 Richard Biener <rguenther@suse.de>
10257 PR middle-end/83517
10258 * match.pd ((t * 2) / 2) -> t): Add missing :c.
10260 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
10262 PR middle-end/81897
10263 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
10264 basic blocks with a small number of successors.
10265 (convert_control_dep_chain_into_preds): Improve handling of
10267 (dump_predicates): Split apart into...
10268 (dump_pred_chain): ...here...
10269 (dump_pred_info): ...and here.
10270 (can_one_predicate_be_invalidated_p): Add debugging printfs.
10271 (can_chain_union_be_invalidated_p): Improve check for invalidation
10273 (uninit_uses_cannot_happen): Avoid unnecessary if
10274 convert_control_dep_chain_into_preds yielded nothing.
10276 2018-01-06 Martin Sebor <msebor@redhat.com>
10278 PR tree-optimization/83640
10279 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
10280 subtracting negative offset from size.
10281 (builtin_access::overlap): Adjust offset bounds of the access to fall
10282 within the size of the object if possible.
10284 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
10286 PR rtl-optimization/83699
10287 * expmed.c (extract_bit_field_1): Restrict the vector usage of
10288 extract_bit_field_as_subreg to cases in which the extracted
10289 value is also a vector.
10291 * lra-constraints.c (process_alt_operands): Test for the equivalence
10292 substitutions when detecting a possible reload cycle.
10294 2018-01-06 Jakub Jelinek <jakub@redhat.com>
10297 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
10298 by default if flag_selective_schedling{,2}. Formatting fixes.
10300 PR rtl-optimization/83682
10301 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
10302 if it has non-VECTOR_MODE element mode.
10303 (vec_duplicate_p): Likewise.
10305 PR middle-end/83694
10306 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
10307 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
10309 2018-01-05 Jakub Jelinek <jakub@redhat.com>
10312 * config/i386/i386-builtin.def
10313 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
10314 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
10315 Require also OPTION_MASK_ISA_AVX512F in addition to
10316 OPTION_MASK_ISA_GFNI.
10317 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
10318 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
10319 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
10320 to OPTION_MASK_ISA_GFNI.
10321 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
10322 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
10323 OPTION_MASK_ISA_AVX512BW.
10324 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
10325 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
10326 addition to OPTION_MASK_ISA_GFNI.
10327 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
10328 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
10329 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
10330 to OPTION_MASK_ISA_GFNI.
10331 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
10332 a requirement for all ISAs rather than any of them with a few
10334 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
10336 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
10337 bitmasks to be enabled with 3 exceptions, instead of requiring any
10338 enabled ISA with lots of exceptions.
10339 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
10340 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
10341 Change avx512bw in isa attribute to avx512f.
10342 * config/i386/sgxintrin.h: Add license boilerplate.
10343 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
10344 to __AVX512F__ and __AVX512VL to __AVX512VL__.
10345 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
10346 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
10348 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
10349 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
10350 temporarily sse2 rather than sse if not enabled already.
10353 * config/i386/sse.md (VI248_VLBW): Rename to ...
10354 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
10355 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
10356 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
10357 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
10358 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
10359 mode iterator instead of VI248_VLBW.
10361 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
10363 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
10364 (record_modified): Skip clobbers; add debug output.
10365 (param_change_prob): Use sreal frequencies.
10367 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
10369 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
10370 punt for user-aligned variables.
10372 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
10374 * tree-chrec.c (chrec_contains_symbols): Return true for
10377 2018-01-05 Sudakshina Das <sudi.das@arm.com>
10380 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
10381 of (x|y) == x for BICS pattern.
10383 2018-01-05 Jakub Jelinek <jakub@redhat.com>
10385 PR tree-optimization/83605
10386 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
10387 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
10390 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
10392 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
10393 * config/epiphany/rtems.h: New file.
10395 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10396 Uros Bizjak <ubizjak@gmail.com>
10399 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
10400 QIreg_operand instead of register_operand predicate.
10401 * config/i386/i386.c (ix86_rop_should_change_byte_p,
10402 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
10403 comments instead of -fmitigate[-_]rop.
10405 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10408 * cgraphunit.c (symbol_table::compile): Switch to text_section
10409 before calling assembly_start debug hook.
10410 * run-rtl-passes.c (run_rtl_passes): Likewise.
10413 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10415 * tree-vrp.c (extract_range_from_binary_expr_1): Check
10416 range_int_cst_p rather than !symbolic_range_p before calling
10417 extract_range_from_multiplicative_op_1.
10419 2018-01-04 Jeff Law <law@redhat.com>
10421 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
10422 redundant test in assertion.
10424 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10426 * doc/rtl.texi: Document machine_mode wrapper classes.
10428 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10430 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
10431 using tree_to_uhwi.
10433 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10435 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
10436 the VEC_PERM_EXPR fold to fail.
10438 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10441 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
10442 to switched_sections.
10444 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10447 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
10448 test for d.testing.
10450 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
10453 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
10454 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
10456 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10459 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
10460 is BLKmode and bitpos not zero or mode change is needed.
10462 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10465 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
10468 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
10471 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
10472 instead of MULT rtx. Update all corresponding splitters.
10473 (*saddl_se): Ditto.
10474 (*ssub<modesuffix>): Ditto.
10475 (*ssubl_se): Ditto.
10476 (*cmp_sadd_di): Update split patterns.
10477 (*cmp_sadd_si): Ditto.
10478 (*cmp_sadd_sidi): Ditto.
10479 (*cmp_ssub_di): Ditto.
10480 (*cmp_ssub_si): Ditto.
10481 (*cmp_ssub_sidi): Ditto.
10482 * config/alpha/predicates.md (const23_operand): New predicate.
10483 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
10484 Look for ASHIFT, not MULT inner operand.
10485 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
10487 2018-01-04 Martin Liska <mliska@suse.cz>
10489 PR gcov-profile/83669
10490 * gcov.c (output_intermediate_file): Add version to intermediate
10492 * doc/gcov.texi: Document new field 'version' in intermediate
10493 file format. Fix location of '-k' option of gcov command.
10495 2018-01-04 Martin Liska <mliska@suse.cz>
10498 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
10500 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10502 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
10504 2018-01-03 Martin Sebor <msebor@redhat.com>
10506 PR tree-optimization/83655
10507 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
10508 checking calls with invalid arguments.
10510 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10512 * tree-vect-stmts.c (vect_get_store_rhs): New function.
10513 (vectorizable_mask_load_store): Delete.
10514 (vectorizable_call): Return false for masked loads and stores.
10515 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
10516 instead of gimple_assign_rhs1.
10517 (vectorizable_load): Handle IFN_MASK_LOAD.
10518 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
10520 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10522 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
10524 (vectorizable_mask_load_store): ...here.
10525 (vectorizable_load): ...and here.
10527 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10529 * tree-vect-stmts.c (vect_build_all_ones_mask)
10530 (vect_build_zero_merge_argument): New functions, split out from...
10531 (vectorizable_load): ...here.
10533 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10535 * tree-vect-stmts.c (vect_check_store_rhs): New function,
10537 (vectorizable_mask_load_store): ...here.
10538 (vectorizable_store): ...and here.
10540 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10542 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
10544 (vectorizable_mask_load_store): ...here.
10546 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10548 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
10549 (vect_model_store_cost): Take a vec_load_store_type instead of a
10551 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
10552 (vect_model_store_cost): Take a vec_load_store_type instead of a
10554 (vectorizable_mask_load_store): Update accordingly.
10555 (vectorizable_store): Likewise.
10556 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
10558 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10560 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
10561 IFN_MASK_LOAD calls here rather than...
10562 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
10564 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10565 Alan Hayward <alan.hayward@arm.com>
10566 David Sherwood <david.sherwood@arm.com>
10568 * expmed.c (extract_bit_field_1): For vector extracts,
10569 fall back to extract_bit_field_as_subreg if vec_extract
10572 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10573 Alan Hayward <alan.hayward@arm.com>
10574 David Sherwood <david.sherwood@arm.com>
10576 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
10577 they are variable or constant sized.
10578 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
10579 slots for constant-sized data.
10581 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10582 Alan Hayward <alan.hayward@arm.com>
10583 David Sherwood <david.sherwood@arm.com>
10585 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
10586 handling COND_EXPRs with boolean comparisons, try to find a better
10587 basis for the mask type than the boolean itself.
10589 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10591 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
10592 is calculated and how it can be overridden.
10593 * genmodes.c (max_bitsize_mode_any_mode): New variable.
10594 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
10596 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
10599 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10600 Alan Hayward <alan.hayward@arm.com>
10601 David Sherwood <david.sherwood@arm.com>
10603 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
10604 Remove the mode argument.
10605 (aarch64_simd_valid_immediate): Remove the mode and inverse
10607 * config/aarch64/iterators.md (bitsize): New iterator.
10608 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
10609 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
10610 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
10611 aarch64_simd_valid_immediate.
10612 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
10613 (aarch64_reg_or_bic_imm): Likewise.
10614 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
10615 with an insn_type enum and msl with a modifier_type enum.
10616 Replace element_width with a scalar_mode. Change the shift
10617 to unsigned int. Add constructors for scalar_float_mode and
10618 scalar_int_mode elements.
10619 (aarch64_vect_float_const_representable_p): Delete.
10620 (aarch64_can_const_movi_rtx_p)
10621 (aarch64_simd_scalar_immediate_valid_for_move)
10622 (aarch64_simd_make_constant): Update call to
10623 aarch64_simd_valid_immediate.
10624 (aarch64_advsimd_valid_immediate_hs): New function.
10625 (aarch64_advsimd_valid_immediate): Likewise.
10626 (aarch64_simd_valid_immediate): Remove mode and inverse
10627 arguments. Rewrite to use the above. Use const_vec_duplicate_p
10628 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
10629 and aarch64_float_const_representable_p on the result.
10630 (aarch64_output_simd_mov_immediate): Remove mode argument.
10631 Update call to aarch64_simd_valid_immediate and use of
10632 simd_immediate_info.
10633 (aarch64_output_scalar_simd_mov_immediate): Update call
10636 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10637 Alan Hayward <alan.hayward@arm.com>
10638 David Sherwood <david.sherwood@arm.com>
10640 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
10641 (mode_nunits): Likewise CONST_MODE_NUNITS.
10642 * machmode.def (ADJUST_NUNITS): Document.
10643 * genmodes.c (mode_data::need_nunits_adj): New field.
10644 (blank_mode): Update accordingly.
10645 (adj_nunits): New variable.
10646 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
10648 (emit_mode_size_inline): Set need_bytesize_adj for all modes
10649 listed in adj_nunits.
10650 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
10651 listed in adj_nunits. Don't emit case statements for such modes.
10652 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
10653 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
10654 nothing if adj_nunits is nonnull.
10655 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
10656 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
10657 (emit_mode_fbit): Update use of print_maybe_const_decl.
10658 (emit_move_size): Likewise. Treat the array as non-const
10660 (emit_mode_adjustments): Handle adj_nunits.
10662 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10664 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
10665 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
10666 (VECTOR_MODES): Use it.
10667 (make_vector_modes): Take the prefix as an argument.
10669 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10670 Alan Hayward <alan.hayward@arm.com>
10671 David Sherwood <david.sherwood@arm.com>
10673 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
10674 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
10675 for MODE_VECTOR_BOOL.
10676 * machmode.def (VECTOR_BOOL_MODE): Document.
10677 * genmodes.c (VECTOR_BOOL_MODE): New macro.
10678 (make_vector_bool_mode): New function.
10679 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
10681 * lto-streamer-in.c (lto_input_mode_table): Likewise.
10682 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
10684 * stor-layout.c (int_mode_for_mode): Likewise.
10685 * tree.c (build_vector_type_for_mode): Likewise.
10686 * varasm.c (output_constant_pool_2): Likewise.
10687 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
10688 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
10689 for MODE_VECTOR_BOOL.
10690 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
10691 of mode class checks.
10692 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
10693 instead of a list of mode class checks.
10694 (expand_vector_scalar_condition): Likewise.
10695 (type_for_widest_vector_mode): Handle BImode as an inner mode.
10697 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10698 Alan Hayward <alan.hayward@arm.com>
10699 David Sherwood <david.sherwood@arm.com>
10701 * machmode.h (mode_size): Change from unsigned short to
10703 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
10704 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
10705 or if measurement_type is not polynomial.
10706 (fixed_size_mode::includes_p): Check for constant-sized modes.
10707 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
10708 return a poly_uint16 rather than an unsigned short.
10709 (emit_mode_size): Change the type of mode_size from unsigned short
10710 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
10711 (emit_mode_adjustments): Cope with polynomial vector sizes.
10712 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10714 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10716 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
10717 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
10718 * caller-save.c (setup_save_areas): Likewise.
10719 (replace_reg_with_saved_mem): Likewise.
10720 * calls.c (emit_library_call_value_1): Likewise.
10721 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
10722 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
10723 (gen_lowpart_for_combine): Likewise.
10724 * convert.c (convert_to_integer_1): Likewise.
10725 * cse.c (equiv_constant, cse_insn): Likewise.
10726 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
10727 (cselib_subst_to_values): Likewise.
10728 * dce.c (word_dce_process_block): Likewise.
10729 * df-problems.c (df_word_lr_mark_ref): Likewise.
10730 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
10731 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
10732 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
10733 (rtl_for_decl_location): Likewise.
10734 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
10735 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
10736 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
10737 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
10738 (expand_expr_real_1): Likewise.
10739 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
10740 (pad_below): Likewise.
10741 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
10742 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
10743 * ira.c (get_subreg_tracking_sizes): Likewise.
10744 * ira-build.c (ira_create_allocno_objects): Likewise.
10745 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
10746 (ira_sort_regnos_for_alter_reg): Likewise.
10747 * ira-costs.c (record_operand_costs): Likewise.
10748 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
10749 (resolve_simple_move): Likewise.
10750 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
10751 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
10752 (lra_constraints): Likewise.
10753 (CONST_POOL_OK_P): Reject variable-sized modes.
10754 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
10755 (add_pseudo_to_slot, lra_spill): Likewise.
10756 * omp-low.c (omp_clause_aligned_alignment): Likewise.
10757 * optabs-query.c (get_best_extraction_insn): Likewise.
10758 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
10759 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
10760 (expand_mult_highpart, valid_multiword_target_p): Likewise.
10761 * recog.c (offsettable_address_addr_space_p): Likewise.
10762 * regcprop.c (maybe_mode_change): Likewise.
10763 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
10764 * regrename.c (build_def_use): Likewise.
10765 * regstat.c (dump_reg_info): Likewise.
10766 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
10767 (find_reloads, find_reloads_subreg_address): Likewise.
10768 * reload1.c (eliminate_regs_1): Likewise.
10769 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
10770 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
10771 (simplify_binary_operation_1, simplify_subreg): Likewise.
10772 * targhooks.c (default_function_arg_padding): Likewise.
10773 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
10774 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
10775 (verify_gimple_assign_ternary): Likewise.
10776 * tree-inline.c (estimate_move_cost): Likewise.
10777 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10778 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
10779 (get_address_cost_ainc): Likewise.
10780 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
10781 (vect_supportable_dr_alignment): Likewise.
10782 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
10783 (vectorizable_reduction): Likewise.
10784 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
10785 (vectorizable_operation, vectorizable_load): Likewise.
10786 * tree.c (build_same_sized_truth_vector_type): Likewise.
10787 * valtrack.c (cleanup_auto_inc_dec): Likewise.
10788 * var-tracking.c (emit_note_insn_var_location): Likewise.
10789 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
10790 (ADDR_VEC_ALIGN): Likewise.
10792 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10793 Alan Hayward <alan.hayward@arm.com>
10794 David Sherwood <david.sherwood@arm.com>
10796 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
10798 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
10799 or if measurement_type is polynomial.
10800 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
10801 * combine.c (make_extraction): Likewise.
10802 * dse.c (find_shift_sequence): Likewise.
10803 * dwarf2out.c (mem_loc_descriptor): Likewise.
10804 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
10805 (extract_bit_field, extract_low_bits): Likewise.
10806 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
10807 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
10808 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
10809 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
10810 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
10811 * reload.c (find_reloads): Likewise.
10812 * reload1.c (alter_reg): Likewise.
10813 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
10814 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
10815 * tree-if-conv.c (predicate_mem_writes): Likewise.
10816 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
10817 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
10818 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
10819 * valtrack.c (dead_debug_insert_temp): Likewise.
10820 * varasm.c (mergeable_constant_section): Likewise.
10821 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
10823 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10824 Alan Hayward <alan.hayward@arm.com>
10825 David Sherwood <david.sherwood@arm.com>
10827 * expr.c (expand_assignment): Cope with polynomial mode sizes
10828 when assigning to a CONCAT.
10830 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10831 Alan Hayward <alan.hayward@arm.com>
10832 David Sherwood <david.sherwood@arm.com>
10834 * machmode.h (mode_precision): Change from unsigned short to
10836 (mode_to_precision): Return a poly_uint16 rather than an unsigned
10838 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
10839 or if measurement_type is not polynomial.
10840 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
10841 in which the mode is already known to be a scalar_int_mode.
10842 * genmodes.c (emit_mode_precision): Change the type of mode_precision
10843 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
10845 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10846 for GET_MODE_PRECISION.
10847 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10848 for GET_MODE_PRECISION.
10849 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
10851 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
10852 (expand_field_assignment, make_extraction): Likewise.
10853 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
10854 (get_last_value): Likewise.
10855 * convert.c (convert_to_integer_1): Likewise.
10856 * cse.c (cse_insn): Likewise.
10857 * expr.c (expand_expr_real_1): Likewise.
10858 * lra-constraints.c (simplify_operand_subreg): Likewise.
10859 * optabs-query.c (can_atomic_load_p): Likewise.
10860 * optabs.c (expand_atomic_load): Likewise.
10861 (expand_atomic_store): Likewise.
10862 * ree.c (combine_reaching_defs): Likewise.
10863 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
10864 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
10865 * tree.h (type_has_mode_precision_p): Likewise.
10866 * ubsan.c (instrument_si_overflow): Likewise.
10868 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10869 Alan Hayward <alan.hayward@arm.com>
10870 David Sherwood <david.sherwood@arm.com>
10872 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
10873 polynomial numbers of units.
10874 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
10875 (valid_vector_subparts_p): New function.
10876 (build_vector_type): Remove temporary shim and take the number
10877 of units as a poly_uint64 rather than an int.
10878 (build_opaque_vector_type): Take the number of units as a
10879 poly_uint64 rather than an int.
10880 * tree.c (build_vector_from_ctor): Handle polynomial
10881 TYPE_VECTOR_SUBPARTS.
10882 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
10883 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
10884 (build_vector_from_val): If the number of units is variable,
10885 use build_vec_duplicate_cst for constant operands and
10886 VEC_DUPLICATE_EXPR otherwise.
10887 (make_vector_type): Remove temporary is_constant ().
10888 (build_vector_type, build_opaque_vector_type): Take the number of
10889 units as a poly_uint64 rather than an int.
10890 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
10892 * cfgexpand.c (expand_debug_expr): Likewise.
10893 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
10894 (store_constructor, expand_expr_real_1): Likewise.
10895 (const_scalar_mask_from_tree): Likewise.
10896 * fold-const-call.c (fold_const_reduction): Likewise.
10897 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
10898 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
10899 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
10900 (fold_relational_const): Likewise.
10901 (native_interpret_vector): Likewise. Change the size from an
10902 int to an unsigned int.
10903 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
10904 TYPE_VECTOR_SUBPARTS.
10905 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
10906 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
10907 duplicating a non-constant operand into a variable-length vector.
10908 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
10909 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
10910 * ipa-icf.c (sem_variable::equals): Likewise.
10911 * match.pd: Likewise.
10912 * omp-simd-clone.c (simd_clone_subparts): Likewise.
10913 * print-tree.c (print_node): Likewise.
10914 * stor-layout.c (layout_type): Likewise.
10915 * targhooks.c (default_builtin_vectorization_cost): Likewise.
10916 * tree-cfg.c (verify_gimple_comparison): Likewise.
10917 (verify_gimple_assign_binary): Likewise.
10918 (verify_gimple_assign_ternary): Likewise.
10919 (verify_gimple_assign_single): Likewise.
10920 * tree-pretty-print.c (dump_generic_node): Likewise.
10921 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10922 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
10923 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
10924 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
10925 (vect_shift_permute_load_chain): Likewise.
10926 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
10927 (expand_vector_condition, optimize_vector_constructor): Likewise.
10928 (lower_vec_perm, get_compute_type): Likewise.
10929 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
10930 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
10931 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
10932 (vect_recog_mask_conversion_pattern): Likewise.
10933 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
10934 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
10935 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10936 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
10937 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
10938 (vectorizable_shift, vectorizable_operation, vectorizable_store)
10939 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
10940 (supportable_widening_operation): Likewise.
10941 (supportable_narrowing_operation): Likewise.
10942 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
10944 * varasm.c (output_constant): Likewise.
10946 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10947 Alan Hayward <alan.hayward@arm.com>
10948 David Sherwood <david.sherwood@arm.com>
10950 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
10951 so that both the length == 3 and length != 3 cases set up their
10952 own permute vectors. Add comments explaining why we know the
10953 number of elements is constant.
10954 (vect_permute_load_chain): Likewise.
10956 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10957 Alan Hayward <alan.hayward@arm.com>
10958 David Sherwood <david.sherwood@arm.com>
10960 * machmode.h (mode_nunits): Change from unsigned char to
10962 (ONLY_FIXED_SIZE_MODES): New macro.
10963 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
10964 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
10965 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
10967 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
10968 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
10969 or if measurement_type is not polynomial.
10970 * genmodes.c (ZERO_COEFFS): New macro.
10971 (emit_mode_nunits_inline): Make mode_nunits_inline return a
10973 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
10974 Use ZERO_COEFFS when emitting initializers.
10975 * data-streamer.h (bp_pack_poly_value): New function.
10976 (bp_unpack_poly_value): Likewise.
10977 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10978 for GET_MODE_NUNITS.
10979 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10980 for GET_MODE_NUNITS.
10981 * tree.c (make_vector_type): Remove temporary shim and make
10982 the real function take the number of units as a poly_uint64
10983 rather than an int.
10984 (build_vector_type_for_mode): Handle polynomial nunits.
10985 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
10986 * emit-rtl.c (const_vec_series_p_1): Likewise.
10987 (gen_rtx_CONST_VECTOR): Likewise.
10988 * fold-const.c (test_vec_duplicate_folding): Likewise.
10989 * genrecog.c (validate_pattern): Likewise.
10990 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
10991 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
10992 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
10993 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
10994 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
10995 * rtlanal.c (subreg_get_info): Likewise.
10996 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
10997 (vect_grouped_load_supported): Likewise.
10998 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
10999 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
11000 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
11001 (simplify_const_unary_operation, simplify_binary_operation_1)
11002 (simplify_const_binary_operation, simplify_ternary_operation)
11003 (test_vector_ops_duplicate, test_vector_ops): Likewise.
11004 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
11005 instead of CONST_VECTOR_NUNITS.
11006 * varasm.c (output_constant_pool_2): Likewise.
11007 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
11008 explicit-encoded elements in the XVEC for variable-length vectors.
11010 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11012 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
11014 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11015 Alan Hayward <alan.hayward@arm.com>
11016 David Sherwood <david.sherwood@arm.com>
11018 * coretypes.h (fixed_size_mode): Declare.
11019 (fixed_size_mode_pod): New typedef.
11020 * builtins.h (target_builtins::x_apply_args_mode)
11021 (target_builtins::x_apply_result_mode): Change type to
11022 fixed_size_mode_pod.
11023 * builtins.c (apply_args_size, apply_result_size, result_vector)
11024 (expand_builtin_apply_args_1, expand_builtin_apply)
11025 (expand_builtin_return): Update accordingly.
11027 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11029 * cse.c (hash_rtx_cb): Hash only the encoded elements.
11030 * cselib.c (cselib_hash_rtx): Likewise.
11031 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
11032 CONST_VECTOR encoding.
11034 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11035 Jeff Law <law@redhat.com>
11038 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
11039 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
11040 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
11041 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
11044 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
11045 explicitly probe *sp in a noreturn function if there were any callee
11046 register saves or frame pointer is needed.
11048 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11051 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
11052 BLKmode for ternary, binary or unary expressions.
11055 * var-tracking.c (delete_vta_debug_insn): New inline function.
11056 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
11057 insns from get_insns () to NULL instead of each bb separately.
11058 Use delete_vta_debug_insn. No longer static.
11059 (vt_debug_insns_local, variable_tracking_main_1): Adjust
11060 delete_vta_debug_insns callers.
11061 * rtl.h (delete_vta_debug_insns): Declare.
11062 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
11063 instead of variable_tracking_main.
11065 2018-01-03 Martin Sebor <msebor@redhat.com>
11067 PR tree-optimization/83603
11068 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
11069 arguments past the endof the argument list in functions declared
11070 without a prototype.
11071 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
11072 Avoid checking when arguments are null.
11074 2018-01-03 Martin Sebor <msebor@redhat.com>
11077 * doc/extend.texi (attribute const): Fix a typo.
11078 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
11079 issuing -Wsuggest-attribute for void functions.
11081 2018-01-03 Martin Sebor <msebor@redhat.com>
11083 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
11084 offset_int::from instead of wide_int::to_shwi.
11085 (maybe_diag_overlap): Remove assertion.
11086 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
11087 * gimple-ssa-sprintf.c (format_directive): Same.
11088 (parse_directive): Same.
11089 (sprintf_dom_walker::compute_format_length): Same.
11090 (try_substitute_return_value): Same.
11092 2018-01-03 Jeff Law <law@redhat.com>
11094 PR middle-end/83654
11095 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
11096 non-constant residual for zero at runtime and avoid probing in
11097 that case. Reorganize code for trailing problem to mirror handling
11100 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11102 PR tree-optimization/83501
11103 * tree-ssa-strlen.c (get_string_cst): New.
11104 (handle_char_store): Call get_string_cst.
11106 2018-01-03 Martin Liska <mliska@suse.cz>
11108 PR tree-optimization/83593
11109 * tree-ssa-strlen.c: Include tree-cfg.h.
11110 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
11111 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
11112 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
11114 (strlen_dom_walker::before_dom_children): Call
11115 gimple_purge_dead_eh_edges. Dump tranformation with details
11117 (strlen_dom_walker::before_dom_children): Update call by adding
11118 new argument cleanup_eh.
11119 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
11121 2018-01-03 Martin Liska <mliska@suse.cz>
11124 * cif-code.def (VARIADIC_THUNK): New enum value.
11125 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
11128 2018-01-03 Jan Beulich <jbeulich@suse.com>
11130 * sse.md (mov<mode>_internal): Tighten condition for when to use
11131 vmovdqu<ssescalarsize> for TI and OI modes.
11133 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11135 Update copyright years.
11137 2018-01-03 Martin Liska <mliska@suse.cz>
11140 * ipa-visibility.c (function_and_variable_visibility): Skip
11141 functions with noipa attribure.
11143 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11145 * gcc.c (process_command): Update copyright notice dates.
11146 * gcov-dump.c (print_version): Ditto.
11147 * gcov.c (print_version): Ditto.
11148 * gcov-tool.c (print_version): Ditto.
11149 * gengtype.c (create_file): Ditto.
11150 * doc/cpp.texi: Bump @copying's copyright year.
11151 * doc/cppinternals.texi: Ditto.
11152 * doc/gcc.texi: Ditto.
11153 * doc/gccint.texi: Ditto.
11154 * doc/gcov.texi: Ditto.
11155 * doc/install.texi: Ditto.
11156 * doc/invoke.texi: Ditto.
11158 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11160 * vector-builder.h (vector_builder::m_full_nelts): Change from
11161 unsigned int to poly_uint64.
11162 (vector_builder::full_nelts): Update prototype accordingly.
11163 (vector_builder::new_vector): Likewise.
11164 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
11165 (vector_builder::operator ==): Likewise.
11166 (vector_builder::finalize): Likewise.
11167 * int-vector-builder.h (int_vector_builder::int_vector_builder):
11168 Take the number of elements as a poly_uint64 rather than an
11170 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
11171 from unsigned int to poly_uint64.
11172 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
11173 (vec_perm_indices::new_vector): Likewise.
11174 (vec_perm_indices::length): Likewise.
11175 (vec_perm_indices::nelts_per_input): Likewise.
11176 (vec_perm_indices::input_nelts): Likewise.
11177 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
11178 number of elements per input as a poly_uint64 rather than an
11179 unsigned int. Use the original encoding for variable-length
11180 vectors, rather than clamping each individual element.
11181 For the second and subsequent elements in each pattern,
11182 clamp the step and base before clamping their sum.
11183 (vec_perm_indices::series_p): Handle polynomial element counts.
11184 (vec_perm_indices::all_in_range_p): Likewise.
11185 (vec_perm_indices_to_tree): Likewise.
11186 (vec_perm_indices_to_rtx): Likewise.
11187 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
11188 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
11189 (tree_vector_builder::new_binary_operation): Handle polynomial
11190 element counts. Return false if we need to know the number
11191 of elements at compile time.
11192 * fold-const.c (fold_vec_perm): Punt if the number of elements
11193 isn't known at compile time.
11195 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11197 * vec-perm-indices.h (vec_perm_builder): Change element type
11198 from HOST_WIDE_INT to poly_int64.
11199 (vec_perm_indices::element_type): Update accordingly.
11200 (vec_perm_indices::clamp): Handle polynomial element_types.
11201 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
11202 (vec_perm_indices::all_in_range_p): Likewise.
11203 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
11205 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
11206 polynomial vec_perm_indices element types.
11207 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
11208 * fold-const.c (fold_vec_perm): Likewise.
11209 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
11210 * tree-vect-generic.c (lower_vec_perm): Likewise.
11211 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
11212 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
11213 element type to HOST_WIDE_INT.
11215 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11216 Alan Hayward <alan.hayward@arm.com>
11217 David Sherwood <david.sherwood@arm.com>
11219 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
11220 rather than an int. Use plus_constant.
11221 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
11222 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
11224 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11225 Alan Hayward <alan.hayward@arm.com>
11226 David Sherwood <david.sherwood@arm.com>
11228 * calls.c (emit_call_1, expand_call): Change struct_value_size from
11229 a HOST_WIDE_INT to a poly_int64.
11231 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11232 Alan Hayward <alan.hayward@arm.com>
11233 David Sherwood <david.sherwood@arm.com>
11235 * calls.c (load_register_parameters): Cope with polynomial
11236 mode sizes. Require a constant size for BLKmode parameters
11237 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
11238 forces a parameter to be padded at the lsb end in order to
11239 fill a complete number of words, require the parameter size
11240 to be ordered wrt UNITS_PER_WORD.
11242 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11243 Alan Hayward <alan.hayward@arm.com>
11244 David Sherwood <david.sherwood@arm.com>
11246 * reload1.c (spill_stack_slot_width): Change element type
11247 from unsigned int to poly_uint64_pod.
11248 (alter_reg): Treat mode sizes as polynomial.
11250 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11251 Alan Hayward <alan.hayward@arm.com>
11252 David Sherwood <david.sherwood@arm.com>
11254 * reload.c (complex_word_subreg_p): New function.
11255 (reload_inner_reg_of_subreg, push_reload): Use it.
11257 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11258 Alan Hayward <alan.hayward@arm.com>
11259 David Sherwood <david.sherwood@arm.com>
11261 * lra-constraints.c (process_alt_operands): Reject matched
11262 operands whose sizes aren't ordered.
11263 (match_reload): Refer to this check here.
11265 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11266 Alan Hayward <alan.hayward@arm.com>
11267 David Sherwood <david.sherwood@arm.com>
11269 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
11270 that the mode size is in the set {1, 2, 4, 8, 16}.
11272 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11273 Alan Hayward <alan.hayward@arm.com>
11274 David Sherwood <david.sherwood@arm.com>
11276 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
11277 Use plus_constant instead of gen_rtx_PLUS.
11279 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11280 Alan Hayward <alan.hayward@arm.com>
11281 David Sherwood <david.sherwood@arm.com>
11283 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
11284 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
11285 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
11286 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
11287 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
11288 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
11289 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
11290 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
11291 * config/i386/i386.c (ix86_push_rounding): ...this new function.
11292 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
11294 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
11295 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
11296 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
11297 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
11298 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
11299 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
11300 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
11301 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
11302 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
11303 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
11305 * expr.c (emit_move_resolve_push): Treat the input and result
11306 of PUSH_ROUNDING as a poly_int64.
11307 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
11308 (emit_push_insn): Likewise.
11309 * lra-eliminations.c (mark_not_eliminable): Likewise.
11310 * recog.c (push_operand): Likewise.
11311 * reload1.c (elimination_effects): Likewise.
11312 * rtlanal.c (nonzero_bits1): Likewise.
11313 * calls.c (store_one_arg): Likewise. Require the padding to be
11314 known at compile time.
11316 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11317 Alan Hayward <alan.hayward@arm.com>
11318 David Sherwood <david.sherwood@arm.com>
11320 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
11321 Use plus_constant instead of gen_rtx_PLUS.
11323 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11324 Alan Hayward <alan.hayward@arm.com>
11325 David Sherwood <david.sherwood@arm.com>
11327 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
11328 rather than an int.
11330 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11331 Alan Hayward <alan.hayward@arm.com>
11332 David Sherwood <david.sherwood@arm.com>
11334 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
11335 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
11336 via stack temporaries. Treat the mode size as polynomial too.
11338 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11339 Alan Hayward <alan.hayward@arm.com>
11340 David Sherwood <david.sherwood@arm.com>
11342 * expr.c (expand_expr_real_2): When handling conversions involving
11343 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
11344 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
11345 as a poly_uint64 too.
11347 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11348 Alan Hayward <alan.hayward@arm.com>
11349 David Sherwood <david.sherwood@arm.com>
11351 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
11353 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11354 Alan Hayward <alan.hayward@arm.com>
11355 David Sherwood <david.sherwood@arm.com>
11357 * combine.c (can_change_dest_mode): Handle polynomial
11358 REGMODE_NATURAL_SIZE.
11359 * expmed.c (store_bit_field_1): Likewise.
11360 * expr.c (store_constructor): Likewise.
11361 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
11362 and polynomial REGMODE_NATURAL_SIZE.
11363 (gen_lowpart_common): Likewise.
11364 * reginfo.c (record_subregs_of_mode): Likewise.
11365 * rtlanal.c (read_modify_subreg_p): Likewise.
11367 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11368 Alan Hayward <alan.hayward@arm.com>
11369 David Sherwood <david.sherwood@arm.com>
11371 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
11372 numbers of elements.
11374 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11375 Alan Hayward <alan.hayward@arm.com>
11376 David Sherwood <david.sherwood@arm.com>
11378 * match.pd: Cope with polynomial numbers of vector elements.
11380 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11381 Alan Hayward <alan.hayward@arm.com>
11382 David Sherwood <david.sherwood@arm.com>
11384 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
11385 in a POINTER_PLUS_EXPR.
11387 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11388 Alan Hayward <alan.hayward@arm.com>
11389 David Sherwood <david.sherwood@arm.com>
11391 * omp-simd-clone.c (simd_clone_subparts): New function.
11392 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
11393 (ipa_simd_modify_function_body): Likewise.
11395 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11396 Alan Hayward <alan.hayward@arm.com>
11397 David Sherwood <david.sherwood@arm.com>
11399 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
11400 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
11401 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
11402 (expand_vector_condition, vector_element): Likewise.
11403 (subparts_gt): New function.
11404 (get_compute_type): Use subparts_gt.
11405 (count_type_subparts): Delete.
11406 (expand_vector_operations_1): Use subparts_gt instead of
11407 count_type_subparts.
11409 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11410 Alan Hayward <alan.hayward@arm.com>
11411 David Sherwood <david.sherwood@arm.com>
11413 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
11414 (vect_compile_time_alias): ...this new function. Do the calculation
11415 on poly_ints rather than trees.
11416 (vect_prune_runtime_alias_test_list): Update call accordingly.
11418 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11419 Alan Hayward <alan.hayward@arm.com>
11420 David Sherwood <david.sherwood@arm.com>
11422 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
11424 (vect_schedule_slp_instance): Likewise.
11426 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11427 Alan Hayward <alan.hayward@arm.com>
11428 David Sherwood <david.sherwood@arm.com>
11430 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
11431 constant and extern definitions for variable-length vectors.
11432 (vect_get_constant_vectors): Note that the number of units
11433 is known to be constant.
11435 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11436 Alan Hayward <alan.hayward@arm.com>
11437 David Sherwood <david.sherwood@arm.com>
11439 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
11440 of units as polynomial. Choose between WIDE and NARROW based
11443 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11444 Alan Hayward <alan.hayward@arm.com>
11445 David Sherwood <david.sherwood@arm.com>
11447 * tree-vect-stmts.c (simd_clone_subparts): New function.
11448 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
11450 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11451 Alan Hayward <alan.hayward@arm.com>
11452 David Sherwood <david.sherwood@arm.com>
11454 * tree-vect-stmts.c (vectorizable_call): Treat the number of
11455 vectors as polynomial. Use build_index_vector for
11456 IFN_GOMP_SIMD_LANE.
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 * tree-vect-stmts.c (get_load_store_type): Treat the number of
11463 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
11464 for variable-length vectors.
11465 (vectorizable_mask_load_store): Treat the number of units as
11466 polynomial, asserting that it is constant if the condition has
11467 already been enforced.
11468 (vectorizable_store, vectorizable_load): Likewise.
11470 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11471 Alan Hayward <alan.hayward@arm.com>
11472 David Sherwood <david.sherwood@arm.com>
11474 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
11475 of units as polynomial. Punt if we can't tell at compile time
11476 which vector contains the final result.
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 * tree-vect-loop.c (vectorizable_induction): Treat the number
11483 of units as polynomial. Punt on SLP inductions. Use an integer
11484 VEC_SERIES_EXPR for variable-length integer reductions. Use a
11485 cast of such a series for variable-length floating-point
11488 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11489 Alan Hayward <alan.hayward@arm.com>
11490 David Sherwood <david.sherwood@arm.com>
11492 * tree.h (build_index_vector): Declare.
11493 * tree.c (build_index_vector): New function.
11494 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
11495 of units as polynomial, forcibly converting it to a constant if
11496 vectorizable_reduction has already enforced the condition.
11497 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
11498 to create a {1,2,3,...} vector.
11499 (vectorizable_reduction): Treat the number of units as polynomial.
11500 Choose vectype_in based on the largest scalar element size rather
11501 than the smallest number of units. Enforce the restrictions
11504 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11505 Alan Hayward <alan.hayward@arm.com>
11506 David Sherwood <david.sherwood@arm.com>
11508 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
11509 number of units as polynomial.
11511 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11512 Alan Hayward <alan.hayward@arm.com>
11513 David Sherwood <david.sherwood@arm.com>
11515 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
11516 * target.def (autovectorize_vector_sizes): Return the vector sizes
11517 by pointer, using vector_sizes rather than a bitmask.
11518 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
11519 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
11520 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
11522 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
11523 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
11524 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
11525 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
11526 * omp-general.c (omp_max_vf): Likewise.
11527 * omp-low.c (omp_clause_aligned_alignment): Likewise.
11528 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
11529 * tree-vect-loop.c (vect_analyze_loop): Likewise.
11530 * tree-vect-slp.c (vect_slp_bb): Likewise.
11531 * doc/tm.texi: Regenerate.
11532 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
11534 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
11535 the vector size as a poly_uint64 rather than an unsigned int.
11536 (current_vector_size): Change from an unsigned int to a poly_uint64.
11537 (get_vectype_for_scalar_type): Update accordingly.
11538 * tree.h (build_truth_vector_type): Take the size and number of
11539 units as a poly_uint64 rather than an unsigned int.
11540 (build_vector_type): Add a temporary overload that takes
11541 the number of units as a poly_uint64 rather than an unsigned int.
11542 * tree.c (make_vector_type): Likewise.
11543 (build_truth_vector_type): Take the number of units as a poly_uint64
11544 rather than an unsigned int.
11546 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11547 Alan Hayward <alan.hayward@arm.com>
11548 David Sherwood <david.sherwood@arm.com>
11550 * target.def (get_mask_mode): Take the number of units and length
11551 as poly_uint64s rather than unsigned ints.
11552 * targhooks.h (default_get_mask_mode): Update accordingly.
11553 * targhooks.c (default_get_mask_mode): Likewise.
11554 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
11555 * doc/tm.texi: Regenerate.
11557 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11558 Alan Hayward <alan.hayward@arm.com>
11559 David Sherwood <david.sherwood@arm.com>
11561 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
11562 * omp-general.c (omp_max_vf): Likewise.
11563 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
11564 (expand_omp_simd): Handle polynomial safelen.
11565 * omp-low.c (omplow_simd_context): Add a default constructor.
11566 (omplow_simd_context::max_vf): Change from int to poly_uint64.
11567 (lower_rec_simd_input_clauses): Update accordingly.
11568 (lower_rec_input_clauses): Likewise.
11570 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11571 Alan Hayward <alan.hayward@arm.com>
11572 David Sherwood <david.sherwood@arm.com>
11574 * tree-vectorizer.h (vect_nunits_for_cost): New function.
11575 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
11576 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
11577 (vect_analyze_slp_cost): Likewise.
11578 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
11579 (vect_model_load_cost): Likewise.
11581 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11582 Alan Hayward <alan.hayward@arm.com>
11583 David Sherwood <david.sherwood@arm.com>
11585 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
11586 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
11587 from an unsigned int * to a poly_uint64_pod *.
11588 (calculate_unrolling_factor): New function.
11589 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
11591 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11592 Alan Hayward <alan.hayward@arm.com>
11593 David Sherwood <david.sherwood@arm.com>
11595 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
11596 from an unsigned int to a poly_uint64.
11597 (_loop_vec_info::slp_unrolling_factor): Likewise.
11598 (_loop_vec_info::vectorization_factor): Change from an int
11600 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
11601 (vect_get_num_vectors): New function.
11602 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
11603 (vect_get_num_copies): Use vect_get_num_vectors.
11604 (vect_analyze_data_ref_dependences): Change max_vf from an int *
11605 to an unsigned int *.
11606 (vect_analyze_data_refs): Change min_vf from an int * to a
11608 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
11609 than an unsigned HOST_WIDE_INT.
11610 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
11611 (vect_analyze_data_ref_dependence): Change max_vf from an int *
11612 to an unsigned int *.
11613 (vect_analyze_data_ref_dependences): Likewise.
11614 (vect_compute_data_ref_alignment): Handle polynomial vf.
11615 (vect_enhance_data_refs_alignment): Likewise.
11616 (vect_prune_runtime_alias_test_list): Likewise.
11617 (vect_shift_permute_load_chain): Likewise.
11618 (vect_supportable_dr_alignment): Likewise.
11619 (dependence_distance_ge_vf): Take the vectorization factor as a
11620 poly_uint64 rather than an unsigned HOST_WIDE_INT.
11621 (vect_analyze_data_refs): Change min_vf from an int * to a
11623 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
11624 vfm1 as a poly_uint64 rather than an int. Make the same change
11625 for the returned bound_scalar.
11626 (vect_gen_vector_loop_niters): Handle polynomial vf.
11627 (vect_do_peeling): Likewise. Update call to
11628 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
11629 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
11631 * tree-vect-loop.c (vect_determine_vectorization_factor)
11632 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
11633 (vect_get_known_peeling_cost): Likewise.
11634 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
11635 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
11636 (vect_transform_loop): Likewise. Use the lowest possible VF when
11637 updating the upper bounds of the loop.
11638 (vect_min_worthwhile_factor): Make static. Return an unsigned int
11639 rather than an int.
11640 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
11641 polynomial unroll factors.
11642 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
11643 (vect_make_slp_decision): Likewise.
11644 (vect_supported_load_permutation_p): Likewise, and polynomial
11646 (vect_analyze_slp_cost): Handle polynomial vf.
11647 (vect_slp_analyze_node_operations): Likewise.
11648 (vect_slp_analyze_bb_1): Likewise.
11649 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
11650 than an unsigned HOST_WIDE_INT.
11651 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
11652 (vectorizable_load): Handle polynomial vf.
11653 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
11655 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
11657 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11658 Alan Hayward <alan.hayward@arm.com>
11659 David Sherwood <david.sherwood@arm.com>
11661 * match.pd: Handle bit operations involving three constants
11662 and try to fold one pair.
11664 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11666 * tree-vect-loop-manip.c: Include gimple-fold.h.
11667 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
11668 niters_maybe_zero parameters. Handle other cases besides a step of 1.
11669 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
11670 Add a path that uses a step of VF instead of 1, but disable it
11672 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
11673 and niters_no_overflow parameters. Update calls to
11674 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
11675 Create a new SSA name if the latter choses to use a ste other
11676 than zero, and return it via niters_vector_mult_vf_var.
11677 * tree-vect-loop.c (vect_transform_loop): Update calls to
11678 vect_do_peeling, vect_gen_vector_loop_niters and
11679 slpeel_make_loop_iterate_ntimes.
11680 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
11681 (vect_gen_vector_loop_niters): Update declarations after above changes.
11683 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
11685 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
11686 128-bit round to integer instructions.
11687 (ceil<mode>2): Likewise.
11688 (btrunc<mode>2): Likewise.
11689 (round<mode>2): Likewise.
11691 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11693 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
11694 unaligned VSX load/store on P8/P9.
11695 (expand_block_clear): Allow the use of unaligned VSX
11696 load/store on P8/P9.
11698 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11700 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
11702 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
11703 swap associated with both a load and a store.
11705 2018-01-02 Andrew Waterman <andrew@sifive.com>
11707 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
11708 * config/riscv/riscv.md (clear_cache): Use it.
11710 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
11712 * web.c: Remove out-of-date comment.
11714 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11716 * expr.c (fixup_args_size_notes): Check that any existing
11717 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
11718 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
11719 (emit_single_push_insn): ...here.
11721 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11723 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
11724 (const_vector_encoded_nelts): New function.
11725 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
11726 (const_vector_int_elt, const_vector_elt): Declare.
11727 * emit-rtl.c (const_vector_int_elt_1): New function.
11728 (const_vector_elt): Likewise.
11729 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
11730 of CONST_VECTOR_ELT.
11732 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11734 * expr.c: Include rtx-vector-builder.h.
11735 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
11736 directly on the tree encoding.
11737 (const_vector_from_tree): Likewise.
11738 * optabs.c: Include rtx-vector-builder.h.
11739 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
11740 sequence of "u" values.
11741 * vec-perm-indices.c: Include rtx-vector-builder.h.
11742 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
11743 directly on the vec_perm_indices encoding.
11745 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11747 * doc/rtl.texi (const_vector): Describe new encoding scheme.
11748 * Makefile.in (OBJS): Add rtx-vector-builder.o.
11749 * rtx-vector-builder.h: New file.
11750 * rtx-vector-builder.c: Likewise.
11751 * rtl.h (rtx_def::u2): Add a const_vector field.
11752 (CONST_VECTOR_NPATTERNS): New macro.
11753 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
11754 (CONST_VECTOR_DUPLICATE_P): Likewise.
11755 (CONST_VECTOR_STEPPED_P): Likewise.
11756 (CONST_VECTOR_ENCODED_ELT): Likewise.
11757 (const_vec_duplicate_p): Check for a duplicated vector encoding.
11758 (unwrap_const_vec_duplicate): Likewise.
11759 (const_vec_series_p): Check for a non-duplicated vector encoding.
11760 Say that the function only returns true for integer vectors.
11761 * emit-rtl.c: Include rtx-vector-builder.h.
11762 (gen_const_vec_duplicate_1): Delete.
11763 (gen_const_vector): Call gen_const_vec_duplicate instead of
11764 gen_const_vec_duplicate_1.
11765 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
11766 (gen_const_vec_duplicate): Use rtx_vector_builder.
11767 (gen_const_vec_series): Likewise.
11768 (gen_rtx_CONST_VECTOR): Likewise.
11769 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
11770 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
11771 Build a new vector rather than modifying a CONST_VECTOR in-place.
11772 (handle_special_swappables): Update call accordingly.
11773 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
11774 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
11775 Build a new vector rather than modifying a CONST_VECTOR in-place.
11776 (handle_special_swappables): Update call accordingly.
11778 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11780 * simplify-rtx.c (simplify_const_binary_operation): Use
11781 CONST_VECTOR_ELT instead of XVECEXP.
11783 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11785 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
11786 the selector elements to be different from the data elements
11787 if the selector is a VECTOR_CST.
11788 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
11789 ssizetype for the selector.
11791 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11793 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
11794 before testing each element individually.
11795 * tree-vect-generic.c (lower_vec_perm): Likewise.
11797 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11799 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
11800 * selftest-run-tests.c (selftest::run_tests): Call it.
11801 * vector-builder.h (vector_builder::operator ==): New function.
11802 (vector_builder::operator !=): Likewise.
11803 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
11804 (vec_perm_indices::all_from_input_p): New function.
11805 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
11806 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
11807 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
11808 instead of reading the VECTOR_CST directly. Detect whether both
11809 vector inputs are the same before constructing the vec_perm_indices,
11810 and update the number of inputs argument accordingly. Use the
11811 utility functions added above. Only construct sel2 if we need to.
11813 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11815 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
11816 the broadcast of the low byte.
11817 (expand_mult_highpart): Use an explicit encoding for the permutes.
11818 * optabs-query.c (can_mult_highpart_p): Likewise.
11819 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
11820 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11821 (vectorizable_bswap): Likewise.
11822 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
11823 explicit encoding for the power-of-2 permutes.
11824 (vect_permute_store_chain): Likewise.
11825 (vect_grouped_load_supported): Likewise.
11826 (vect_permute_load_chain): Likewise.
11828 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11830 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
11831 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
11832 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
11833 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
11834 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
11835 (vect_gen_perm_mask_any): Likewise.
11837 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11839 * int-vector-builder.h: New file.
11840 * vec-perm-indices.h: Include int-vector-builder.h.
11841 (vec_perm_indices): Redefine as an int_vector_builder.
11842 (auto_vec_perm_indices): Delete.
11843 (vec_perm_builder): Redefine as a stand-alone class.
11844 (vec_perm_indices::vec_perm_indices): New function.
11845 (vec_perm_indices::clamp): Likewise.
11846 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
11847 (vec_perm_indices::new_vector): New function.
11848 (vec_perm_indices::new_expanded_vector): Update for new
11849 vec_perm_indices class.
11850 (vec_perm_indices::rotate_inputs): New function.
11851 (vec_perm_indices::all_in_range_p): Operate directly on the
11852 encoded form, without computing elided elements.
11853 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
11854 encoding. Update for new vec_perm_indices class.
11855 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
11856 the given vec_perm_builder.
11857 (expand_vec_perm_var): Update vec_perm_builder constructor.
11858 (expand_mult_highpart): Use vec_perm_builder instead of
11859 auto_vec_perm_indices.
11860 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
11861 vec_perm_indices instead of auto_vec_perm_indices. Use a single
11862 or double series encoding as appropriate.
11863 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
11864 vec_perm_indices instead of auto_vec_perm_indices.
11865 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11866 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11867 (vect_permute_store_chain): Likewise.
11868 (vect_grouped_load_supported): Likewise.
11869 (vect_permute_load_chain): Likewise.
11870 (vect_shift_permute_load_chain): Likewise.
11871 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
11872 (vect_transform_slp_perm_load): Likewise.
11873 (vect_schedule_slp_instance): Likewise.
11874 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11875 (vectorizable_mask_load_store): Likewise.
11876 (vectorizable_bswap): Likewise.
11877 (vectorizable_store): Likewise.
11878 (vectorizable_load): Likewise.
11879 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
11880 vec_perm_indices instead of auto_vec_perm_indices. Use
11881 tree_to_vec_perm_builder to read the vector from a tree.
11882 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
11883 vec_perm_builder instead of a vec_perm_indices.
11884 (have_whole_vector_shift): Use vec_perm_builder and
11885 vec_perm_indices instead of auto_vec_perm_indices. Leave the
11886 truncation to calc_vec_perm_mask_for_shift.
11887 (vect_create_epilog_for_reduction): Likewise.
11888 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
11889 from auto_vec_perm_indices to vec_perm_indices.
11890 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
11891 instead of changing individual elements.
11892 (aarch64_vectorize_vec_perm_const): Use new_vector to install
11893 the vector in d.perm.
11894 * config/arm/arm.c (expand_vec_perm_d::perm): Change
11895 from auto_vec_perm_indices to vec_perm_indices.
11896 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
11897 instead of changing individual elements.
11898 (arm_vectorize_vec_perm_const): Use new_vector to install
11899 the vector in d.perm.
11900 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
11901 Update vec_perm_builder constructor.
11902 (rs6000_expand_interleave): Likewise.
11903 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
11904 (rs6000_expand_interleave): Likewise.
11906 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11908 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
11909 to qimode could truncate the indices.
11910 * optabs.c (expand_vec_perm_var): Likewise.
11912 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11914 * Makefile.in (OBJS): Add vec-perm-indices.o.
11915 * vec-perm-indices.h: New file.
11916 * vec-perm-indices.c: Likewise.
11917 * target.h (vec_perm_indices): Replace with a forward class
11919 (auto_vec_perm_indices): Move to vec-perm-indices.h.
11920 * optabs.h: Include vec-perm-indices.h.
11921 (expand_vec_perm): Delete.
11922 (selector_fits_mode_p, expand_vec_perm_var): Declare.
11923 (expand_vec_perm_const): Declare.
11924 * target.def (vec_perm_const_ok): Replace with...
11925 (vec_perm_const): ...this new hook.
11926 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
11927 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
11928 * doc/tm.texi: Regenerate.
11929 * optabs.def (vec_perm_const): Delete.
11930 * doc/md.texi (vec_perm_const): Likewise.
11931 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
11932 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
11933 expand_vec_perm for constant permutation vectors. Assert that
11934 the mode of variable permutation vectors is the integer equivalent
11935 of the mode that is being permuted.
11936 * optabs-query.h (selector_fits_mode_p): Declare.
11937 * optabs-query.c: Include vec-perm-indices.h.
11938 (selector_fits_mode_p): New function.
11939 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
11940 is defined, instead of checking whether the vec_perm_const_optab
11941 exists. Use targetm.vectorize.vec_perm_const instead of
11942 targetm.vectorize.vec_perm_const_ok. Check whether the indices
11943 fit in the vector mode before using a variable permute.
11944 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
11945 vec_perm_indices instead of an rtx.
11946 (expand_vec_perm): Replace with...
11947 (expand_vec_perm_const): ...this new function. Take the selector
11948 as a vec_perm_indices rather than an rtx. Also take the mode of
11949 the selector. Update call to shift_amt_for_vec_perm_mask.
11950 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
11951 Use vec_perm_indices::new_expanded_vector to expand the original
11952 selector into bytes. Check whether the indices fit in the vector
11953 mode before using a variable permute.
11954 (expand_vec_perm_var): Make global.
11955 (expand_mult_highpart): Use expand_vec_perm_const.
11956 * fold-const.c: Includes vec-perm-indices.h.
11957 * tree-ssa-forwprop.c: Likewise.
11958 * tree-vect-data-refs.c: Likewise.
11959 * tree-vect-generic.c: Likewise.
11960 * tree-vect-loop.c: Likewise.
11961 * tree-vect-slp.c: Likewise.
11962 * tree-vect-stmts.c: Likewise.
11963 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
11965 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
11966 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
11967 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
11968 (aarch64_vectorize_vec_perm_const): ...this new function.
11969 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11970 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11971 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
11972 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
11973 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11974 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11975 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
11977 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
11978 check for NEON modes.
11979 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
11980 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
11981 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
11982 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
11984 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
11985 the old VEC_PERM_CONST conditions.
11986 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
11987 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
11988 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
11989 (ia64_vectorize_vec_perm_const_ok): Merge into...
11990 (ia64_vectorize_vec_perm_const): ...this new function.
11991 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
11992 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
11993 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
11994 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
11995 * config/mips/mips.c (mips_expand_vec_perm_const)
11996 (mips_vectorize_vec_perm_const_ok): Merge into...
11997 (mips_vectorize_vec_perm_const): ...this new function.
11998 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
11999 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
12000 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
12001 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
12002 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
12003 (rs6000_expand_vec_perm_const): Delete.
12004 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
12006 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12007 (altivec_expand_vec_perm_const_le): Take each operand individually.
12008 Operate on constant selectors rather than rtxes.
12009 (altivec_expand_vec_perm_const): Likewise. Update call to
12010 altivec_expand_vec_perm_const_le.
12011 (rs6000_expand_vec_perm_const): Delete.
12012 (rs6000_vectorize_vec_perm_const_ok): Delete.
12013 (rs6000_vectorize_vec_perm_const): New function.
12014 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
12015 an element count and rtx array.
12016 (rs6000_expand_extract_even): Update call accordingly.
12017 (rs6000_expand_interleave): Likewise.
12018 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
12019 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
12020 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
12021 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
12022 (rs6000_expand_vec_perm_const): Delete.
12023 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12024 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12025 (altivec_expand_vec_perm_const_le): Take each operand individually.
12026 Operate on constant selectors rather than rtxes.
12027 (altivec_expand_vec_perm_const): Likewise. Update call to
12028 altivec_expand_vec_perm_const_le.
12029 (rs6000_expand_vec_perm_const): Delete.
12030 (rs6000_vectorize_vec_perm_const_ok): Delete.
12031 (rs6000_vectorize_vec_perm_const): New function. Remove stray
12032 reference to the SPE evmerge intructions.
12033 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
12034 an element count and rtx array.
12035 (rs6000_expand_extract_even): Update call accordingly.
12036 (rs6000_expand_interleave): Likewise.
12037 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
12038 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
12040 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12042 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12044 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
12045 vector mode and that that mode matches the mode of the data
12047 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
12048 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
12049 directly using expand_vec_perm_1 when forcing selectors into
12051 (expand_vec_perm_var): New function, split out from expand_vec_perm.
12053 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12055 * optabs-query.h (can_vec_perm_p): Delete.
12056 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
12057 * optabs-query.c (can_vec_perm_p): Split into...
12058 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
12059 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
12060 particular selector is valid.
12061 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12062 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12063 (vect_grouped_load_supported): Likewise.
12064 (vect_shift_permute_load_chain): Likewise.
12065 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
12066 (vect_transform_slp_perm_load): Likewise.
12067 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12068 (vectorizable_bswap): Likewise.
12069 (vect_gen_perm_mask_checked): Likewise.
12070 * fold-const.c (fold_ternary_loc): Likewise. Don't take
12071 implementations of variable permutation vectors into account
12072 when deciding which selector to use.
12073 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
12074 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
12075 with a false third argument.
12076 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
12077 to test whether the constant selector is valid and can_vec_perm_var_p
12078 to test whether a variable selector is valid.
12080 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12082 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
12083 * optabs-query.c (can_vec_perm_p): Likewise.
12084 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
12085 instead of vec_perm_indices.
12086 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
12087 (vect_gen_perm_mask_checked): Likewise,
12088 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
12089 (vect_gen_perm_mask_checked): Likewise,
12091 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12093 * optabs-query.h (qimode_for_vec_perm): Declare.
12094 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
12095 (qimode_for_vec_perm): ...this new function.
12096 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
12098 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12100 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
12101 does not have a conditional at the top.
12103 2018-01-02 Richard Biener <rguenther@suse.de>
12105 * ipa-inline.c (big_speedup_p): Fix expression.
12107 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
12110 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
12113 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
12117 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
12118 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
12119 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
12120 cond_taken_branch_cost 3->4.
12122 2018-01-01 Jakub Jelinek <jakub@redhat.com>
12124 PR tree-optimization/83581
12125 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
12126 TODO_cleanup_cfg if any changes have been made.
12128 PR middle-end/83608
12129 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
12130 convert_modes if target mode has the right side, but different mode
12133 PR middle-end/83609
12134 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
12135 last argument when extracting from CONCAT. If either from_real or
12136 from_imag is NULL, use expansion through memory. If result is not
12137 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
12138 the parts directly to inner mode, if even that fails, use expansion
12141 PR middle-end/83623
12142 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
12143 check for bswap in mode rather than HImode and use that in expand_unop
12146 Copyright (C) 2018 Free Software Foundation, Inc.
12148 Copying and distribution of this file, with or without modification,
12149 are permitted in any medium without royalty provided the copyright
12150 notice and this notice are preserved.