1 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
2 Shiva Chen <shiva0217@gmail.com>
4 * config/nds32/constants.md (unspec_volatile_element): Add
5 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
6 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
8 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
9 (make_pass_nds32_fp_as_gp): Declare.
10 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
12 (nds32_asm_function_end_prologue): Remove unused asm output.
13 (nds32_asm_function_begin_epilogue): Remove unused asm output.
14 (nds32_asm_file_start): Output necessary fp_as_gp information.
15 (nds32_option_override): Adjust register usage.
16 (nds32_expand_prologue): Consider fp_as_gp situation.
17 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
18 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
21 (simple_return): Ditto.
22 (omit_fp_begin): Output special directive for fp_as_gp.
23 (omit_fp_end): Output special directive for fp_as_gp.
24 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
25 mforbid-fp-as-gp): New options.
27 2018-06-01 Mark Wielaard <mark@klomp.org>
29 * dwarf2out.c (dwarf2out_finish): Remove generation of
32 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
34 * gimple-ssa-store-merging.c: Include gimple-fold.h.
35 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
36 (struct merged_store_group): Add bit_insertion field.
37 (dump_char_array): Use standard hexadecimal format.
38 (merged_store_group::merged_store_group): Set bit_insertion to false.
39 (merged_store_group::apply_stores): Use optimal buffer size. Deal
40 with BIT_INSERT_EXPR stores. Move up code updating the mask and
41 also print the mask in the dump file.
42 (pass_store_merging::gate): Minor tweak.
43 (imm_store_chain_info::coalesce_immediate): Fix wrong association
44 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
45 stores with INTEGER_CST stores.
46 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
47 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
48 and use it throughout. Generate bit insertion sequences if need be.
49 (pass_store_merging::process_store): Remove redundant condition.
50 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
52 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
54 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
55 the 128-bit floating point types. Fix function comment.
57 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
59 * config/aarch64/aarch64-simd.md
60 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
62 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
65 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
67 PR tree-optimization/85989
68 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
70 (backprop::intersect_uses): Check it when deciding whether this
71 is a backedge reference.
72 (backprop::process_block): Add each phi to m_visited_phis
73 after visiting it, then clear it at the end.
75 2018-06-01 Richard Biener <rguenther@suse.de>
77 * tree-vectorizer.h (vect_dr_stmt): New function.
78 (vect_get_load_cost): Adjust.
79 (vect_get_store_cost): Likewise.
80 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
81 Use vect_dr_stmt instead of DR_SMTT.
82 (vect_record_base_alignments): Likewise.
83 (vect_calculate_target_alignment): Likewise.
84 (vect_compute_data_ref_alignment): Likewise and make static.
85 (vect_update_misalignment_for_peel): Likewise.
86 (vect_verify_datarefs_alignment): Likewise.
87 (vector_alignment_reachable_p): Likewise.
88 (vect_get_data_access_cost): Likewise. Pass down
89 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
90 (vect_get_peeling_costs_all_drs): Likewise.
91 (vect_peeling_hash_get_lowest_cost): Likewise.
92 (vect_enhance_data_refs_alignment): Likewise.
93 (vect_find_same_alignment_drs): Likewise.
94 (vect_analyze_data_refs_alignment): Likewise.
95 (vect_analyze_group_access_1): Likewise.
96 (vect_analyze_group_access): Likewise.
97 (vect_analyze_data_ref_access): Likewise.
98 (vect_analyze_data_ref_accesses): Likewise.
99 (vect_vfa_segment_size): Likewise.
100 (vect_small_gap_p): Likewise.
101 (vectorizable_with_step_bound_p): Likewise.
102 (vect_prune_runtime_alias_test_list): Likewise.
103 (vect_analyze_data_refs): Likewise.
104 (vect_supportable_dr_alignment): Likewise.
105 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
106 (vect_gen_prolog_loop_niters): Likewise.
107 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
108 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
110 (vect_recog_mask_conversion_pattern): Likewise.
111 (vect_try_gather_scatter_pattern): Likewise.
112 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
113 to vect_get_store_cost.
114 (vect_get_store_cost): Get stmt_info instead of DR.
115 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
116 (vect_get_load_cost): Get stmt_info instead of DR.
118 2018-06-01 Richard Biener <rguenther@suse.de>
121 * gimple-fold.c (var_decl_component_p): Also allow offsetted
122 vars wrapped in MEM_REFs.
124 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
126 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
127 Fix subreg tests so that we only return a choice between
128 GENERAL_REGS and FP_REGS if the original classes included both.
130 2018-06-01 Richard Biener <rguenther@suse.de>
133 * tree-ssa-structalias.c (get_function_part_constraint):
134 Handle NULL fi->decl.
135 (find_func_aliases_for_call): Properly handle indirect
137 (find_func_clobbers): Likewise.
138 (ipa_pta_execute): Likewise.
139 (create_variable_info_for): For functions that are ifunc_resolver
140 resolve to a varinfo that contains the result of the resolver
142 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
145 2018-05-31 Michael Collison <michael.collison@arm.com>
147 * config/aarch64/aarch64.md:
148 (*fix_to_zero_extenddfdi2): New pattern.
149 * gcc.target/aarch64/fix_extend1.c: New testcase.
151 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
155 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
156 and BUILT_IN_STRNCMP_EQ.
157 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
159 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
160 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
161 (gimple_fold_builtin): Likewise.
162 * tree-ssa-strlen.c (compute_string_length): New function.
163 (determine_min_obsize): New function.
164 (handle_builtin_string_cmp): New function to handle calls to
165 string compare functions.
166 (strlen_optimize_stmt): Add handling to builtin string compare
168 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
169 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
170 * tree.c (build_common_builtin_nodes): Add new defines of
171 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
173 2018-05-31 Jakub Jelinek <jakub@redhat.com>
176 * bb-reorder.c (pass_partition_blocks::gate): Return false for
177 functions with naked attribute.
179 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
181 * config/i386/sse.md (avx_vec_concat<mode>):
182 Substitute concat_tg_mode mode attribute with xtg_mode.
183 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
184 (concat_tg_mode): Remove mode attribute.
186 2018-05-31 Martin Sebor <msebor@redhat.com>
189 * calls.c (alloc_max_size): Correct a logic error/typo.
190 Treat excessive arguments as infinite. Warn for invalid arguments.
191 * doc/invoke.texi (-Walloc-size-larger-than): Update.
193 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
196 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
197 and movx for Haswell.
199 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
200 Cesar Philippidis <cesar@codesourcery.com>
203 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
204 when emitting error on private/firstprivate reductions.
205 * omp-low.c (lower_omp_target): Avoid reference-type processing
206 on pointers for firstprivate clause.
208 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
210 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
213 * config/aarch64/aarch64-simd.md
214 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
215 (aarch64_ld1_x3_<mode>): Likewise
216 (aarch64_st1x2<VALLDIF:mode>): Likewise
217 (aarch64_st1_x2_<mode>): Likewise
218 (aarch64_st1x3<VALLDIF:mode>): Likewise
219 (aarch64_st1_x3_<mode>): Likewise
220 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
221 (vld1_s8_x3): Likewise.
222 (vld1_u16_x3): Likewise.
223 (vld1_s16_x3): Likewise.
224 (vld1_u32_x3): Likewise.
225 (vld1_s32_x3): Likewise.
226 (vld1_u64_x3): Likewise.
227 (vld1_s64_x3): Likewise.
228 (vld1_f16_x3): Likewise.
229 (vld1_f32_x3): Likewise.
230 (vld1_f64_x3): Likewise.
231 (vld1_p8_x3): Likewise.
232 (vld1_p16_x3): Likewise.
233 (vld1_p64_x3): Likewise.
234 (vld1q_u8_x3): Likewise.
235 (vld1q_s8_x3): Likewise.
236 (vld1q_u16_x3): Likewise.
237 (vld1q_s16_x3): Likewise.
238 (vld1q_u32_x3): Likewise.
239 (vld1q_s32_x3): Likewise.
240 (vld1q_u64_x3): Likewise.
241 (vld1q_s64_x3): Likewise.
242 (vld1q_f16_x3): Likewise.
243 (vld1q_f32_x3): Likewise.
244 (vld1q_f64_x3): Likewise.
245 (vld1q_p8_x3): Likewise.
246 (vld1q_p16_x3): Likewise.
247 (vld1q_p64_x3): Likewise.
248 (vst1_s64_x2): Likewise.
249 (vst1_u64_x2): Likewise.
250 (vst1_f64_x2): Likewise.
251 (vst1_s8_x2): Likewise.
252 (vst1_p8_x2): Likewise.
253 (vst1_s16_x2): Likewise.
254 (vst1_p16_x2): Likewise.
255 (vst1_s32_x2): Likewise.
256 (vst1_u8_x2): Likewise.
257 (vst1_u16_x2): Likewise.
258 (vst1_u32_x2): Likewise.
259 (vst1_f16_x2): Likewise.
260 (vst1_f32_x2): Likewise.
261 (vst1_p64_x2): Likewise.
262 (vst1q_s8_x2): Likewise.
263 (vst1q_p8_x2): Likewise.
264 (vst1q_s16_x2): Likewise.
265 (vst1q_p16_x2): Likewise.
266 (vst1q_s32_x2): Likewise.
267 (vst1q_s64_x2): Likewise.
268 (vst1q_u8_x2): Likewise.
269 (vst1q_u16_x2): Likewise.
270 (vst1q_u32_x2): Likewise.
271 (vst1q_u64_x2): Likewise.
272 (vst1q_f16_x2): Likewise.
273 (vst1q_f32_x2): Likewise.
274 (vst1q_f64_x2): Likewise.
275 (vst1q_p64_x2): Likewise.
276 (vst1_s64_x3): Likewise.
277 (vst1_u64_x3): Likewise.
278 (vst1_f64_x3): Likewise.
279 (vst1_s8_x3): Likewise.
280 (vst1_p8_x3): Likewise.
281 (vst1_s16_x3): Likewise.
282 (vst1_p16_x3): Likewise.
283 (vst1_s32_x3): Likewise.
284 (vst1_u8_x3): Likewise.
285 (vst1_u16_x3): Likewise.
286 (vst1_u32_x3): Likewise.
287 (vst1_f16_x3): Likewise.
288 (vst1_f32_x3): Likewise.
289 (vst1_p64_x3): Likewise.
290 (vst1q_s8_x3): Likewise.
291 (vst1q_p8_x3): Likewise.
292 (vst1q_s16_x3): Likewise.
293 (vst1q_p16_x3): Likewise.
294 (vst1q_s32_x3): Likewise.
295 (vst1q_s64_x3): Likewise.
296 (vst1q_u8_x3): Likewise.
297 (vst1q_u16_x3): Likewise.
298 (vst1q_u32_x3): Likewise.
299 (vst1q_u64_x3): Likewise.
300 (vst1q_f16_x3): Likewise.
301 (vst1q_f32_x3): Likewise.
302 (vst1q_f64_x3): Likewise.
303 (vst1q_p64_x3): Likewise.
305 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
307 * config/msp430/msp430.c (msp430_output_labelref): Prepend
308 user_label_prefix to name.
310 * tree-core.h: Update comment about the format of NAME string
311 passed to handler in attribute_spec.
313 * config/msp430/msp430.md: Remove erroneous subreg expression from
314 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
315 zero_extend{q,h}isi2.
317 2018-05-30 Borislav Petkov <bp@suse.de>
319 * doc/extend.texi: Document some architecture specific
320 constraints and sort entries.
322 2018-05-30 Martin Sebor <msebor@redhat.com>
325 * builtins.c (expand_builtin_stpcpy_1): New function.
326 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
327 only if the former succeeds.
329 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
331 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
334 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
336 * doc/invoke.texi (-flinker-output): Document
338 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
340 * passes.c (ipa_write_summaries): Only modify statements if body
342 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
343 incrementally linking.
344 (ipa_passes): Likewise.
345 * lto-cgraph.c (lto_output_node): When incrementally linking do not
346 pass down resolution info.
347 * common.opt (flag_incremental_link): Update info.
348 * gcc.c (plugin specs): Turn flinker-output=* to
349 -plugin-opt=-linker-output-known
350 * toplev.c (compile_file): Also cut compilation when doing incremental
352 * flag-types. (enum lto_partition_model): Add
353 LTO_LINKER_OUTPUT_NOLTOREL.
354 (invoke.texi): Add -flinker-output docs.
355 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
356 link same way as WPA; do not stream in dead initializers.
358 * dwarf2out.c (dwarf2out_die_ref_for_decl,
359 darf2out_register_external_decl): Support incremental link.
361 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
363 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
365 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
367 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
368 it down to simple_object_copy_lto_debug_sections.
369 (run_gcc): Determine incremental LTO link time and configure
370 lto1 into non-wpa mode, disable renaming of debug sections.
372 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
374 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
375 descriptions of various incorrectly documented functions.
377 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
380 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
381 address check not strict.
383 2018-05-30 Richard Biener <rguenther@suse.de>
385 PR tree-optimization/85964
386 * tracer.c (better_p): Drop initialized count check, we only
387 call the function with initialized counts now.
388 (find_best_successor): Do find a best edge if one
389 has uninitialized count.
390 (find_best_predecessor): Likewise. Do BB frequency check only
391 if count is initialized.
393 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
395 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
396 (aarch64_ldrstr_offset_compare): New.
397 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
398 load/store orderings.
399 (aarch64_gen_adjusted_ldpstp): Likewise.
401 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
403 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
404 Check for subset of GENERAL_REGS and FP_REGS.
405 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
408 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
410 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
411 and wi::to_poly_offset. Add the current offset and then check
412 whether the sum fits, rather than using an unchecked addition of
413 a checked term. Check for a shwi rather than a uhwi.
414 * expr.c (get_bit_range): Use tree_to_poly_uint64.
415 (store_constructor): Use poly_int_tree_p.
416 (expand_expr_real_1): Likewise.
417 * function.c (assign_temp): Likewise.
418 * fold-const.c (const_binop): Use poly_int_tree_p and
420 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
422 * ipa-icf-gimple.c (func_checker::compare_operand): Use
423 to_poly_offset for MEM offsets.
424 * ipa-icf.c (sem_variable::equals): Likewise.
425 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
426 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
427 wi::to_poly_offset for BIT_FIELD_REF offsets.
428 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
430 * var-tracking.c (emit_note_insn_var_location): Use
433 2018-05-29 Jim Wilson <jimw@sifive.com>
435 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
437 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
440 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
441 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
443 (sse4_1_round<mode>2): Use nonimmediate_operand
444 for operand 1 predicate.
446 2018-05-29 Martin Sebor <msebor@redhat.com>
447 Richard Biener <rguenther@suse.de>
450 * calls.c (get_size_range): Call determine_value_range instead
452 * tree-vrp.h (determine_value_range): Declared new function.
453 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
455 2018-05-29 Richard Biener <rguenther@suse.de>
457 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
458 sure to use non-pattern stmts for get_earlier_stmt arguments.
459 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
460 called on pattern stmts.
461 (get_later_stmt): Likewise.
463 2018-05-29 Martin Liska <mliska@suse.cz>
465 PR gcov-profile/85759
466 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
469 2018-05-29 Jakub Jelinek <jakub@redhat.com>
471 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
473 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
477 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
478 VEC_PACK_FLOAT_EXPR): New tree codes.
479 * tree-pretty-print.c (op_code_prio): Handle
480 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
481 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
482 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
483 * tree-inline.c (estimate_operator_cost): Likewise.
484 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
485 * fold-const.c (const_binop): Likewise.
486 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
487 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
488 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
489 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
490 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
491 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
492 * expr.c (expand_expr_real_2): Likewise.
493 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
494 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
495 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
497 * optabs.c (expand_widen_pattern_expr): For
498 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
499 sign from result type rather than operand's type.
500 (expand_binop_directly): For vec_packu_float_optab and
501 vec_packs_float_optab allow result type to be different from operand's
503 * optabs-tree.c (optab_for_tree_code): Handle
504 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
505 VEC_PACK_FLOAT_EXPR. Formatting fixes.
506 * tree-vect-generic.c (expand_vector_operations_1): Handle
507 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
509 * tree-vect-stmts.c (supportable_widening_operation): Handle
511 (supportable_narrowing_operation): Handle FLOAT_EXPR.
512 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
513 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
514 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
515 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
517 (vec_pack<floatprefix>_float_<mode>): New expander.
518 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
520 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
521 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
522 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
523 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
524 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
526 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
527 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
528 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
529 VEC_PACK_FLOAT_EXPR): Document.
531 2018-05-29 Richard Biener <rguenther@suse.de>
533 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
535 (stmt_vec_info_vec): Make pointer.
536 (init_stmt_vec_info_vec): Remove.
537 (free_stmt_vec_info_vec): Likewise.
538 (set_stmt_vec_info_vec): New function.
539 (free_stmt_vec_infos): Likewise.
540 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
541 (set_vinfo_for_stmt): Likewise.
542 (get_earlier_stmt): Likewise.
543 (get_later_stmt): Likewise.
544 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
545 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
546 (vec_info::~vec_info): Free stmt_vec_infos.
547 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
548 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
549 (pass_slp_vectorize::execute): Likewise.
550 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
551 (free_stmt_vec_info_vec): Likewise.
552 (set_stmt_vec_info_vec): New function.
553 (free_stmt_vec_infos): Likewise.
554 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
555 the global stmt_vec_info_vec.
556 * tree-parloops.c (gather_scalar_reductions): Use
557 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
560 2018-05-29 Richard Biener <rguenther@suse.de>
562 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
564 2018-05-29 Martin Liska <mliska@suse.cz>
565 David Malcolm <dmalcolm@redhat.com>
567 * vec.c (test_reverse): New.
568 (vec_c_tests): Add new test.
569 * vec.h (vl_ptr>::reverse): New function.
571 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
573 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
575 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
578 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
580 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
582 2018-05-28 Richard Biener <rguenther@suse.de>
584 PR tree-optimization/85933
585 * tree-vect-data-refs.c (vect_record_base_alignments): Only
586 look at stmts marked as vectorizable.
588 2018-05-28 Richard Biener <rguenther@suse.de>
590 PR tree-optimization/85934
591 * tree-vect-generic.c (expand_vector_operations_1): Hoist
592 vector boolean check before scalar optimization.
594 2018-05-28 Jakub Jelinek <jakub@redhat.com>
596 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
599 2018-05-28 Mark Wielaard <mark@klomp.org>
601 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
602 if it is an expression containing a minus sign.
604 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
606 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
608 2018-05-27 Paul Koning <ni1d@arrl.net>
610 * config/pdp11/pdp11.md (truncsihi2): Remove.
612 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
613 Chung-Ju Wu <jasonwucj@gmail.com>
615 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
617 (unaligned_store_dw): Ditto.
618 * config/nds32/nds32-memory-manipulation.c
619 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
620 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
621 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
622 (emit_setmem_word_loop): Rename to ...
623 (emit_setmem_doubleword_loop): ... this.
624 (nds32_gen_dup_4_byte_to_word_value): New function.
625 (nds32_gen_dup_8_byte_to_double_word_value): New function.
626 (nds32_expand_setmem_loop): Refine implementation.
627 (nds32_expand_setmem_loop_v3m): Ditto.
628 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
631 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
633 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
635 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
637 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
638 (nds32_init_machine_status): Initialize machine->attr_naked_p and
639 machine->attr_no_prologue_p.
640 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
641 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
642 (nds32_expand_epilogue): Consider attr_naked_p.
643 (nds32_expand_epilogue_v3pop): Likewise.
644 (nds32_can_use_return_insn): Likewise.
645 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
646 attr_no_prologue_p fields.
647 * config/nds32/nds32.opt (mret-in-naked-func): New option.
649 2018-05-27 Jakub Jelinek <jakub@redhat.com>
652 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
655 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
657 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
659 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
661 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
663 (*<floatsuffix>floatv2div2sf2): Rename to ...
664 (*float<floatunssuffix>v2div2sf2): ... this.
665 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
666 (float<floatunssuffix>v2div2sf2_mask): ... this.
667 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
668 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
669 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
671 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
673 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
675 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
677 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
679 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
681 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
682 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
683 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
684 gen_ufix_truncv8dfv8si2.
685 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
686 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
687 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
688 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
689 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
690 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
691 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
692 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
694 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
698 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
700 2018-05-25 Jim Wilson <jimw@sifive.com>
702 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
703 * config/riscv/riscv.c (struct machine_function): Add
704 interrupt_handler_p and attribute_checked_p fields.
705 (riscv_attribute_table): Add interrupt.
706 (riscv_interrupt_type_p): New.
707 (riscv_save_reg_p): Save extra regs for interrupt handler.
708 (riscv_use_save_libcall): Return false for interrupt handler.
709 (riscv_first_stack_step): Add forward declaration.
710 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
711 for interrupt handler with large frame. Use it for saved reg list.
712 (riscv_expand_prologue): Move flag_stack_usage_info support to
713 eliminate duplication.
714 (riscv_expand_epilogue): Generate mret for interrupt handler.
715 (riscv_epilogue_uses): New.
716 (riscv_can_use_return_insn): Return false for interrupt handler.
717 (riscv_function_ok_for_sibcall): Likewise.
718 (riscv_set_current_function): Add interrupt handler support.
719 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
720 * config/riscv/riscv.md (UNSPECV_MRET): New.
722 (riscv_frflags, riscv_fsflags): Use tab after opcode.
724 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
726 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
728 PR tree-optimization/85712
729 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
730 this candidate has already been replaced in-situ by a copy.
732 2018-05-25 Jason Merrill <jason@redhat.com>
734 PR c++/80485 - inline function non-zero address.
735 * symtab.c (nonzero_address): Check DECL_COMDAT.
737 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
740 * config/alpha/alpha.md (ashlsi3): New insn pattern.
741 (*ashlsi_se): Rename from *ashldi_se. Define as sign
742 extension of SImode operation. Use const123_operand predicate.
744 (*saddl_se_1): Ditto.
746 (*ssubl_se_1): Ditto.
747 * config/alpha/predicates.md (const123_operand): New predicate.
748 * config/alpha/constraints.md (P): Use IN_RANGE.
750 2018-05-25 Richard Biener <rguenther@suse.de>
752 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
754 (ref_maybe_used_by_stmt_p): Likewise.
755 (stmt_may_clobber_ref_p): Likewise.
756 (stmt_may_clobber_ref_p_1): Likewise.
757 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
759 (ref_maybe_used_by_stmt_p): Likewise.
760 (stmt_may_clobber_ref_p): Likewise.
761 (stmt_may_clobber_ref_p_1): Likewise.
762 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
763 the alias oracle to disambiguate DRs with stmts DR analysis
765 (vect_analyze_data_refs): Do not give up on not analyzable
766 DRs for BB vectorization. Remove code truncating the dataref
769 2018-05-25 Jakub Jelinek <jakub@redhat.com>
772 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
773 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
774 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
776 2018-05-25 Richard Biener <rguenther@suse.de>
778 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
779 function, combining stmt data ref gathering and fatal analysis
781 (vect_analyze_data_refs): Remove now redudnant code and simplify.
782 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
783 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
784 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
785 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
787 2018-05-25 Bin Cheng <bin.cheng@arm.com>
789 PR tree-optimization/85720
790 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
791 SCC if all partitions are builtins.
792 (version_loop_by_alias_check): New parameter. Generate cancelable
793 runtime alias check if all partitions are builtins.
794 (distribute_loop): Update call to above function.
796 2018-05-25 Bin Cheng <bin.cheng@arm.com>
798 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
799 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
800 (parm_default_def_partition_arg): Ditto.
801 (set_parm_default_def_partition): Ditto.
802 (get_parm_default_def_partitions): Ditto and make it static.
803 (get_undefined_value_partitions): Ditto and make it static.
804 (remove_ssa_form): Refactor call to init_var_map here.
805 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
806 computation for loop region.
807 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
808 (register_default_def): Delete.
809 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
810 (parm_default_def_partition_arg): Ditto.
811 (set_parm_default_def_partition): Ditto.
812 (get_parm_default_def_partitions): Ditto and make it static.
813 (get_undefined_value_partitions): Ditto and make it static.
814 (coalesce_with_default, coalesce_with_default): Update comment.
815 (create_coalesce_list_for_region): New func factored out from
816 create_outofssa_var_map.
817 (populate_coalesce_list_for_outofssa): New func factored out from
818 create_outofssa_var_map and coalesce_ssa_name.
819 (create_outofssa_var_map): Delete.
820 (coalesce_ssa_name): Refactor to support live range computation.
821 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
822 (get_parm_default_def_partitions): Delete.
823 (get_undefined_value_partitions): Ditto.
824 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
825 computation for loop region.
826 (new_tree_live_info, loe_visit_block): Ditto.
827 (live_worklist, set_var_live_on_entry): Ditto.
828 (calculate_live_on_exit, verify_live_on_entry): Ditto.
829 * tree-ssa-live.h (struct _var_map): New fields.
830 (init_var_map): Change decl.
831 (region_contains_p): New.
833 2018-05-25 Bin Cheng <bin.cheng@arm.com>
835 * tree-ssa-live.h (live_merge_and_clear): Delete.
837 2018-05-25 Richard Biener <rguenther@suse.de>
840 * tree-dump.c (dequeue_and_dump): Remove access to removed
841 operand 2 of a SWITCH_EXPR.
843 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
845 * doc/sourcebuild.texi (vect_double_cond_arith): Include
846 multiplication and division.
847 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
848 (cond_udiv@var{m}, cond_umod@var{m}): Document.
849 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
850 (cond_udiv_optab, cond_umod_optab): New optabs.
851 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
852 (IFN_COND_RDIV): New internal functions.
853 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
854 TRUNC_MOD_EXPR and RDIV_EXPR.
855 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
856 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
858 (SVE_INT_BINARY): Include mult.
859 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
860 (optab, sve_int_op): Handle mult.
861 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
863 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
864 for SVE_INT_BINARY_SD.
866 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
868 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
869 (optab, sve_int_op): Handle div and udiv.
870 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
871 for SVE_INT_BINARY_SD.
872 (*<optab><mode>3): New insn for the same.
874 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
876 * tree-vect-patterns.c: Include predict.h.
877 (vect_recog_divmod_pattern): Restrict check for division support
878 to when optimizing for size.
880 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
882 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
883 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
884 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
885 (gimple_match_op::set_op): Likewise.
886 (gimple_resimplify4): Declare.
887 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
888 (expr::gen_transform): Likewise.
889 (decision_tree::gen): Generate a simplification routine for 4 operands.
890 * gimple-match-head.c (gimple_simplify): Add an overload for
891 4 operands. In the top-level function, handle up to 4 call
892 arguments and call gimple_resimplify4.
893 (gimple_resimplify4): New function.
894 (build_call_internal): Pass a fourth operand.
895 (maybe_push_to_seq): Likewise.
896 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
897 Fold VEC_COND_EXPRs of an operation and a default value into
898 an IFN_COND_* function if possible.
899 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
901 (SVE_COND_FP_BINARY): Include them.
902 (optab, sve_fp_op): Handle them.
903 (SVE_INT_BINARY_REV): New code iterator.
904 (SVE_COND_FP_BINARY_REV): New int iterator.
905 (commutative): New int attribute.
906 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
908 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
910 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
911 (*cond_<optab><mode>): New patterns for reversed operands.
913 2018-05-25 Richard Biener <rguenther@suse.de>
915 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
916 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
917 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
918 (STMT_VINFO_GROUPED_ACCESS): Adjust.
919 * tree-vect-data-refs.c (everywhere): Adjust users.
920 * tree-vect-loop.c (everywhere): Likewise.
921 * tree-vect-slp.c (everywhere): Likewise.
922 * tree-vect-stmts.c (everywhere): Likewise.
923 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
925 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
927 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
929 (gcc_cv_as_section_exclude): ... this.
930 Try Solaris as #exclude syntax.
931 * configure: Regenerate.
932 * config.in: Regenerate.
933 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
935 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
936 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
938 * varasm.c (default_elf_asm_named_section): Don't check if
939 HAVE_GAS_SECTION_EXCLUDE is defined.
941 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
943 * doc/md.texi: Update the documentation of the cond_* optabs
944 to mention the new final operand. Fix GET_MODE_NUNITS call.
945 Describe the scalar case too.
946 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
947 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
949 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
950 (get_conditional_internal_fn): Update comment.
951 * tree-vect-loop.c (vectorizable_reduction): Pass the original
952 accumulator value as a final argument to conditional functions.
953 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
954 a define_expand and add an "else" operand. Assert for now that
955 the else operand is equal to operand 2. Use SVE_INT_BINARY and
956 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
957 (*cond_<optab><mode>): New patterns.
958 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
959 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
960 (UNSPEC_COND_EOR): Delete.
961 (optab): Remove associated mappings.
962 (SVE_INT_BINARY): New code iterator.
963 (sve_int_op): Remove int attribute and add "minus" to the code
965 (SVE_COND_INT_OP): Delete.
966 (SVE_COND_FP_OP): Rename to...
967 (SVE_COND_FP_BINARY): ...this.
969 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
971 * optabs.c (can_reuse_operands_p): New function.
972 (maybe_legitimize_operands): Try to reuse the results for
975 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
977 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
978 Add {q} suffix to insn mnemonic.
980 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
982 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
983 (msp430_warn_func_return): New.
985 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
987 * fold-const.c (tree_nonzero_bits): New function.
988 * fold-const.h (tree_nonzero_bits): Likewise.
989 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
990 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
992 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
996 * varasm.c (assemble_alias): Check ifunc_resolver only on
999 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
1002 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
1003 when memory input operand is handled.
1005 2018-05-24 Luis Machado <luis.machado@linaro.org>
1007 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
1009 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
1011 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
1013 * match.pd: Delay FMA folds until after vectorization.
1015 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
1018 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
1019 address check not strict.
1021 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
1023 * gimple-match.h (gimple_match_op): New class.
1024 (mprts_hook): Replace parameters with a gimple_match_op *.
1025 (maybe_build_generic_op): Likewise.
1026 (gimple_simplified_result_is_gimple_val): Replace parameters with
1027 a const gimple_match_op *.
1028 (gimple_simplify): Replace code_helper * and tree * parameters with
1029 a gimple_match_op * parameter.
1030 (gimple_resimplify1): Replace code_helper *, tree and tree *
1031 parameters with a gimple_match_op * parameter.
1032 (gimple_resimplify2): Likewise.
1033 (gimple_resimplify3): Likewise.
1034 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
1035 parameters with a gimple_match_op * parameter.
1036 * gimple-match-head.c (gimple_simplify): Change prototypes of
1037 auto-generated functions to take a gimple_match_op * instead of
1038 separate code_helper * and tree * parameters. Make the same
1039 change in the top-level overload and update calls to the
1040 gimple_resimplify routines. Update calls to the auto-generated
1041 functions and to maybe_push_res_to_seq in the publicly-facing
1042 operation-specific gimple_simplify overloads.
1043 (gimple_match_op::MAX_NUM_OPS): Define.
1044 (gimple_resimplify1): Replace rcode and ops with a single res_op
1045 parameter. Update call to gimple_simplify.
1046 (gimple_resimplify2): Likewise.
1047 (gimple_resimplify3): Likewise.
1048 (mprts_hook): Replace parameters with a gimple_match_op *.
1049 (maybe_build_generic_op): Likewise.
1050 (build_call_internal): Replace type, nargs and ops with
1051 a gimple_match_op *.
1052 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
1053 with a single gimple_match_op *. Update calls to mprts_hook,
1054 build_call_internal and gimple_simplified_result_is_gimple_val.
1055 Factor out code that is common to the tree_code and combined_fn cases.
1056 * genmatch.c (expr::gen_transform): Replace tem_code and
1057 tem_ops with a gimple_match_op called tem_op. Update calls
1058 to the gimple_resimplify functions and maybe_push_res_to_seq.
1059 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
1060 res_ops. Update call to the gimple_resimplify functions.
1061 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
1062 (decision_tree::gen): Make the functions take a gimple_match_op *
1063 called res_op instead of separate res_code and res_ops parameters.
1064 Update call accordingly.
1065 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
1066 and ops with a single res_op parameter. Update calls to
1067 maybe_build_generic_op and maybe_push_res_to_seq.
1068 (fold_stmt_1): Update calls to gimple_simplify and
1069 replace_stmt_with_simplification.
1070 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
1071 and gimple_simplified_result_is_gimple_val.
1072 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
1074 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
1075 with a gimple_match_op *.
1076 (vn_nary_build_or_lookup): Likewise. Update call to
1077 vn_nary_build_or_lookup_1.
1078 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
1079 gimple_match_op *. Update calls to the gimple_resimplify routines
1080 and to gimple_simplified_result_is_gimple_val.
1081 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
1082 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
1083 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
1084 (visit_nary_op): Likewise.
1085 (visit_reference_op_load): Likewise.
1087 2018-05-23 Luis Machado <luis.machado@linaro.org>
1089 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
1090 modifier for printing the step amount.
1092 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
1095 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
1098 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
1100 * doc/sourcebuild.texi (Endianness): New subsubsection.
1102 2018-05-23 Luis Machado <luis.machado@linaro.org>
1104 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1105 <prefetch_dynamic_strides>: New const bool field.
1106 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1107 prefetch_dynamic_strides.
1108 (exynosm1_prefetch_tune): Likewise.
1109 (thunderxt88_prefetch_tune): Likewise.
1110 (thunderx_prefetch_tune): Likewise.
1111 (thunderx2t99_prefetch_tune): Likewise.
1112 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
1114 (aarch64_override_options_internal): Update to set
1115 PARAM_PREFETCH_DYNAMIC_STRIDES.
1116 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1117 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1118 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1119 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1120 prefetch-dynamic-strides setting.
1122 2018-05-23 Luis Machado <luis.machado@linaro.org>
1124 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1125 <minimum_stride>: New const int field.
1126 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1127 minimum_stride field defaulting to -1.
1128 (exynosm1_prefetch_tune): Likewise.
1129 (thunderxt88_prefetch_tune): Likewise.
1130 (thunderx_prefetch_tune): Likewise.
1131 (thunderx2t99_prefetch_tune): Likewise.
1132 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
1133 <default_opt_level>: Set to 3.
1134 (aarch64_override_options_internal): Update to set
1135 PARAM_PREFETCH_MINIMUM_STRIDE.
1136 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1137 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1138 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1139 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1140 stride is constant and is below the minimum stride threshold.
1142 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1144 * config/arm/arm-cpus.in (mode26): Delete.
1145 (armv4): Delete mode26 reference.
1146 * config/arm/arm.c (arm_configure_build_target): Delete use of
1149 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
1151 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
1153 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
1154 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
1156 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
1157 (floatunsdidf2): Ditto.
1159 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
1161 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
1162 (fixuns_trunc<mode>si2_avx512f): Ditto.
1163 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
1164 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
1165 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
1167 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
1169 PR rtl-optimization/79985
1170 * df-scan.c (df_insn_refs_collect): Remove special case for
1171 global registers and asm statements.
1173 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
1175 * extend.texi (Global Register Variables): Rewrite the bullet list.
1176 Note that the register is available for allocation. Note that access
1177 via inline asm must use constraints. Add note about async-signal
1178 handlers. Remove paragraph about automagic register selection.
1180 2018-05-23 Richard Biener <rguenther@suse.de>
1182 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
1183 of fixed offset from memset VN.
1185 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
1187 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
1189 (alloc_cand_and_find_basis): Initialize first_interp field.
1190 (slsr_process_mul): Modify first_interp field.
1191 (slsr_process_add): Likewise.
1192 (slsr_process_cast): Modify first_interp field for each new
1194 (slsr_process_copy): Likewise.
1195 (dump_candidate): Dump first_interp field.
1196 (replace_mult_candidate): Process all interpretations, not just
1198 (replace_rhs_if_not_dup): Likewise.
1199 (replace_one_candidate): Likewise.
1201 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
1203 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
1205 (aarch64_needs_frame_chain): New function.
1206 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
1208 2018-05-23 Sudakshina Das <sudi.das@arm.com>
1211 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
1212 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
1213 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
1214 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
1215 as true for strict-align.
1216 (aarch64_can_inline_p): Perform checks even when callee has no
1217 attributes to check for strict alignment.
1218 * doc/extend.texi (AArch64 Function Attributes): Document
1220 * doc/invoke.texi: (AArch64 Options): Likewise.
1222 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
1224 PR tree-optimization/85853
1225 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
1226 the handling of the root of the node to...
1227 (vect_slp_analyze_node_operations_1): ...this new function,
1228 and run the whole thing with the child nodes' def types
1229 set according to their SLP node's def type.
1231 2018-05-23 Richard Biener <rguenther@suse.de>
1234 * tree-data-ref.c (create_runtime_alias_checks): Defer
1235 and ignore overflow warnings.
1237 2018-05-23 Yury Gribov <tetra2005@gmail.com>
1239 PR tree-optimization/85822
1240 * tree-vrp.c (is_masked_range_test): Fix handling of negative
1243 2018-05-23 Richard Biener <rguenther@suse.de>
1245 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
1246 memset constants via native_interpret_expr.
1248 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
1251 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
1253 (cgraph_node::create_alias): Likewise.
1254 (cgraph_node::get_availability): Check ifunc_resolver instead
1255 of looking up ifunc attribute.
1256 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
1257 * varasm.c (do_assemble_alias): Likewise.
1258 (assemble_alias): Likewise.
1259 (default_binds_local_p_3): Likewise.
1260 * cgraph.h (cgraph_node): Add ifunc_resolver.
1261 (cgraph_node::only_called_directly_or_aliased_p): Return false
1263 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
1265 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
1266 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
1267 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
1268 instead of looking up ifunc attribute.
1270 2018-05-22 Luis Machado <luis.machado@linaro.org>
1272 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
1274 2018-05-22 Martin Sebor <msebor@redhat.com>
1277 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
1278 only when expasion succeeds.
1279 (expand_builtin_strcmp): Same.
1280 (expand_builtin_strncmp): Same.
1282 2018-05-22 Martin Sebor <msebor@redhat.com>
1284 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
1286 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
1287 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1289 * config/aarch64/aarch64-ldpstp.md: Replace uses of
1290 aarch64_mem_pair_operand with memory_operand and delete operand swapping
1292 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
1293 Add check for legitimate_address.
1294 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
1295 (aarch64_swap_ldrstr_operands): New.
1296 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
1299 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
1300 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1302 * config/aarch64/aarch64.md: New patterns to generate stp
1304 (store_pair_sw, store_pair_dw): New patterns to generate stp for
1305 single words and double words.
1306 (load_pair_sw, load_pair_dw): Likewise.
1307 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
1309 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
1311 * config/aarch64/aarch64-ldpstp.md: Modify peephole
1312 for different mode ldpstp and add peephole for merged zero stores.
1314 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
1316 (aarch64_gen_store_pair): Rename calls to match new patterns.
1317 (aarch64_gen_load_pair): Rename calls to match new patterns.
1318 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
1319 (load_pair<DREG:mode><DREG2:mode>): ... This.
1320 (store_pair<mode>): Rename to...
1321 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
1322 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
1324 (V_INT_EQUIV): Handle SImode.
1325 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
1328 2018-05-22 Martin Sebor <msebor@redhat.com>
1331 * calls.c (maybe_warn_nonstring_arg): Use string length to set
1332 or ajust the presumed bound on an operation to avoid unnecessary
1335 2018-05-22 Martin Sebor <msebor@redhat.com>
1337 PR tree-optimization/85826
1338 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
1339 assuming that a DECL necesarily has a constant size.
1341 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
1344 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
1346 2018-05-22 Richard Biener <rguenther@suse.de>
1348 PR tree-optimization/85834
1349 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
1350 non-constant and non-zero memset arguments.
1352 2018-05-22 Martin Liska <mliska@suse.cz>
1355 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
1357 2018-05-22 Richard Biener <rguenther@suse.de>
1359 PR tree-optimization/85863
1360 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
1361 comparisons when vectype is specified.
1362 (vectorizable_condition): Do not specify vectype for
1363 vect_is_simple_cond when SLP vectorizing.
1365 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
1368 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
1369 define __ibm128 as long double.
1370 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
1371 as a distinct type when IEEE 128-bit support is enabled.
1372 (init_float128_ieee): Fix up conversions between IFmode and IEEE
1373 128-bit types to use the correct functions.
1374 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
1375 convert between 128-bit floating point types that have different
1376 modes but the same representation, instead of using gen_lowpart to
1378 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
1380 (IFKF_reg): New attributes to give the register constraints for
1382 (extend<mode>tf2_internal): New insns to mark an explicit
1383 conversion between 128-bit floating point types that have a
1384 different mode but share the same representation.
1386 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
1388 PR tree-optimization/85814
1389 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
1390 a null return from get_strinfo when unsharing the next
1391 strinfo in the chain.
1393 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
1396 * varasm.c (weak_finish): Clean up weak_decls.
1398 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1400 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
1401 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
1402 UNSPEC_UADALP values.
1403 * config/aarch64/iterators.md (ABAL): New int iterator.
1406 (sur): Add mappings for the above.
1407 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
1409 (aarch64_<sur>abal<mode>_4): Likewise.
1410 (aarch64_<sur>adalp<mode>_3): Likewise.
1411 (<sur>sadv16qi): New define_expand.
1413 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
1415 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
1416 (*movdf_internal): Ditto.
1417 (*rcpsf2_sse): Ditto.
1418 (*rsqrtsf2_sse): Ditto.
1419 (*sqrt<mode>2_sse): Ditto.
1421 2018-05-21 Tamar Christina <tamar.christina@arm.com>
1423 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
1425 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
1426 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
1427 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
1428 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
1430 * config/aarch64/arm_neon.h: Likewise.
1431 * config/aarch64/iterators.md (VQ_I): New.
1433 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
1435 * config.gcc: Add arc/t-multilib-linux to tmake_file for
1437 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
1440 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
1442 * config/nds32/constraints.md (S): New constraint.
1443 * config/nds32/nds32.md (call_internal): Use constraint S.
1444 (call_value_internal): Likewise.
1445 (sibcall_internal): Likewise.
1446 (sibcall_value_internal): Likewise.
1448 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
1449 Chung-Ju Wu <jasonwucj@gmail.com>
1451 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
1454 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
1455 Chung-Ju Wu <jasonwucj@gmail.com>
1457 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
1458 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
1459 (nds32_rtx_costs_impl): Simplify.
1460 (nds32_address_cost_impl): Simplify.
1461 (nds32_init_rtx_costs): New function.
1462 (nds32_rtx_costs_speed_prefer): Likewise.
1463 (nds32_rtx_costs_size_prefer): Likewise.
1464 (nds32_address_cost_speed_prefer): Likewise.
1465 (nds32_address_cost_speed_fwprop): Likewise.
1466 (nds32_address_cost_size_prefer): Likewise.
1467 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
1468 * config/nds32/nds32.c (nds32_option_override): Use
1469 nds32_init_rtx_costs function.
1471 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
1473 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
1474 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
1475 (TARGET_PIPELINE_N8): Likewise.
1476 (TARGET_PIPELINE_N10): Likewise.
1477 (TARGET_PIPELINE_N13): Likewise.
1478 (TARGET_PIPELINE_GRAYWOLF): Likewise.
1480 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
1482 * config/nds32/nds32-fpu.md: Update copyright year.
1484 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1486 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
1488 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1490 * config/nds32/nds32.c
1491 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
1492 * config/nds32/nds32.opt (minline-asm-r15): New option.
1494 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1496 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
1498 * config/nds32/nds32.md (abssi2): New pattern.
1500 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
1502 * config/i386/i386.md (rex64namesuffix): New mode attribute.
1503 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
1504 Merge insn pattern from sse_cvtsi2ss<round_name> and
1505 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
1506 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
1507 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
1508 using SWI48 mode iterator.
1509 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
1510 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
1511 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
1512 pattern from sse_cvttss2si<round_saeonly_name>
1513 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
1514 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
1515 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
1516 using SWI48 mode iterator.
1517 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
1518 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
1519 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
1520 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
1521 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
1522 using SWI48 mode iterator.
1523 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
1524 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
1525 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
1526 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
1527 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
1528 SWI48 mode iterator.
1529 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
1530 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
1531 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
1532 pattern from sse_cvttsd2si<round_saeonly_name>
1533 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
1535 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1537 * config/nds32/nds32-md-auxiliary.c
1538 (nds32_valid_smw_lwm_base_p): Refine.
1539 (nds32_output_smw_single_word): Refine.
1540 (nds32_output_smw_double_word): New.
1541 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
1543 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1545 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
1546 (nds32_output_stack_pop): Refine.
1547 (nds32_expand_unaligned_load): Refine.
1548 (nds32_expand_unaligned_store): Refine.
1550 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1551 Chung-Ju Wu <jasonwucj@gmail.com>
1553 * config/nds32/constants.md: Add TP_REGNUM constant.
1554 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
1555 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
1557 * config/nds32/nds32-doubleword.md: Consider flag_pic.
1558 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
1559 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
1560 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
1561 and PIC code generation.
1562 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
1564 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
1566 * config/nds32/nds32.md: Support TLS and PIC.
1567 * config/nds32/nds32.c: Support TLS and PIC.
1568 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
1569 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
1572 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1574 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
1575 mode with E_ prefix.
1577 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1578 Chung-Ju Wu <jasonwucj@gmail.com>
1580 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
1581 * config/nds32/nds32-md-auxiliary.c
1582 (symbolic_reference_mentioned_p): New.
1583 (nds32_legitimize_ict_address): New.
1584 (nds32_expand_ict_move): New.
1585 (nds32_indirect_call_referenced_p): New.
1586 (nds32_symbol_binds_local_p): Delete.
1587 (nds32_long_call_p): Modify.
1588 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
1589 * config/nds32/nds32-protos.h
1590 (symbolic_reference_mentioned_p): Declare.
1591 (nds32_legitimize_ict_address): Declare.
1592 (nds32_expand_ict_move): Declare.
1593 (nds32_indirect_call_referenced_p): Declare.
1594 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
1595 (nds32_relax_group): Use nds32_ict_const_p as condition.
1596 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
1597 (nds32_asm_file_start): Output ict_model directive in asm code.
1598 (nds32_legitimate_address_p): Consider indirect call.
1599 (nds32_print_operand): Consider indirect call.
1600 (nds32_print_operand_address): Consider indirect call.
1601 (nds32_insert_attributes): Handle "indirect_call" attribute.
1602 (TARGET_LEGITIMATE_ADDRESS_P): Define.
1603 (TARGET_LEGITIMATE_CONSTANT_P): Define.
1604 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
1605 (TARGET_DELEGITIMIZE_ADDRESS): Define.
1606 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
1607 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
1608 (TARGET_ICT_MODEL_SMALL): Define.
1609 (TARGET_ICT_MODEL_LARGE): Define.
1610 * config/nds32/nds32.md (movsi): Consider ict model.
1611 (call, call_value): Consider ict model.
1612 (sibcall, sibcall_value): Consider ict model.
1613 * config/nds32/nds32.opt (mict-model): New option.
1614 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
1617 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
1618 Monk Chiang <sh.chiang04@gmail.com>
1619 Jim Wilson <jimw@sifive.com>
1621 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
1622 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
1623 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
1624 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
1625 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
1626 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
1627 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
1628 compute save_libcall_adjustment properly.
1629 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
1630 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
1631 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
1632 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
1633 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
1634 (ABI_SPEC): Handle mabi=ilp32e.
1635 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
1636 (RVE): Add RVE mask.
1637 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
1638 <-march>: Add rv32e as an example.
1640 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
1643 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
1644 (intra_create_variable_infos): Handle C++ constructors.
1646 2018-05-18 Martin Liska <mliska@suse.cz>
1648 * passes.def: Remove a redundant pass.
1650 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
1653 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
1655 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1657 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
1659 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
1661 (armv2, armv2a, armv3, armv3m): Delete architectures.
1662 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
1663 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
1664 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
1666 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
1667 (*mulsidi3adddi): Likewise.
1668 (mulsidi3): Likewise.
1669 (*mulsidi3_nov6): Likewise.
1670 (umulsidi3): Likewise.
1671 (umulsidi3_nov6): Likewise.
1672 (umaddsidi4): Likewise.
1673 (*umulsidi3adddi): Likewise.
1674 (smulsi3_highpart): Likewise.
1675 (*smulsi3_highpart_nov6): Likewise.
1676 (umulsi3_highpart): Likewise.
1677 (*umulsi3_highpart_nov6): Likewise.
1678 * config/arm/arm.h (arm_arch3m): Delete.
1679 * config/arm/arm.c (arm_arch3m): Delete.
1680 (arm_option_override_internal): Update armv3-related comment.
1681 (arm_configure_build_target): Delete use of isa_bit_mode32.
1682 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
1683 (arm_rtx_costs_internal): Delete check of arm_arch3m.
1684 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
1686 (mulusa3): Likewise.
1687 * config/arm/arm-protos.h (arm_arch3m): Delete.
1688 * config/arm/arm-tables.opt: Regenerate.
1689 * config/arm/arm-tune.md: Likewise.
1690 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
1691 deleted architectures.
1693 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1695 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
1696 (armv5t, armv5te): New features.
1697 (ARMv5, ARMv5e): Delete fgroups.
1698 (ARMv5t, ARMv5te): Adjust for above changes.
1700 (armv5, armv5e): Delete arches.
1701 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
1703 (*call_reg_arm): Likewise.
1704 (*call_value_reg_armv5): Likewise.
1705 (*call_value_reg_arm): Likewise.
1706 (*call_symbol): Likewise.
1707 (*call_value_symbol): Likewise.
1708 (*sibcall_insn): Likewise.
1709 (*sibcall_value_insn): Likewise.
1711 (prefetch): Likewise.
1712 (define_split and define_peephole2 dependent on arm_arch5):
1714 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
1716 (TARGET_ARM_QBIT): Likewise.
1717 (TARGET_DSP_MULTIPLY): Likewise.
1718 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
1719 (arm_arch5, arm_arch5e): Delete.
1720 (arm_arch5t, arm_arch5te): Declare.
1721 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
1722 (arm_arch5t): Declare.
1723 (arm_option_reconfigure_globals): Update for the above.
1724 (arm_options_perform_arch_sanity_checks): Update comment, replace
1725 use of arm_arch5 with arm_arch5t.
1726 (use_return_insn): Likewise.
1727 (arm_emit_call_insn): Likewise.
1728 (output_return_instruction): Likewise.
1729 (arm_final_prescan_insn): Likewise.
1730 (arm_coproc_builtin_available): Likewise.
1731 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
1732 arm_arch5e with arm_arch5t and arm_arch5te.
1733 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
1734 (arm_arch5t, arm_arch5te): Declare.
1735 * config/arm/arm-tables.opt: Regenerate.
1736 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
1737 * config/arm/t-multilib: Likewise.
1738 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
1739 instead of arm_arch5.
1740 (*call_reg_thumb1): Likewise.
1741 (*call_value_reg_thumb1_v5): Likewise.
1742 (*call_value_reg_thumb1): Likewise.
1743 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
1745 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
1747 2018-05-18 Martin Liska <mliska@suse.cz>
1749 PR gcov-profile/84846
1750 * doc/gcov.texi: Document -t option of gcov tool.
1752 2018-05-18 Martin Liska <mliska@suse.cz>
1754 PR gcov-profile/84846
1755 * gcov.c (print_usage): Add new -t option.
1756 (process_args): Handle the option.
1757 (generate_results): Use stdout as output when requested by
1760 2018-05-18 Martin Liska <mliska@suse.cz>
1762 PR gcov-profile/84846
1763 * coverage.c (coverage_init): Write PWD to .gcno file.
1764 * doc/gcov.texi: Document how working directory is printed.
1765 * gcov-dump.c (dump_gcov_file): Print PWD.
1766 * gcov.c (output_intermediate_file): Likewise.
1767 (read_graph_file): Read PWD string.
1768 (output_lines): Print PWD.
1770 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1773 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
1774 for retval and return false if all args to phi are zero.
1776 2018-05-18 Richard Biener <rguenther@suse.de>
1778 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
1780 (evrp_dom_walker::before_dom_children): Call it.
1782 2018-05-18 Richard Biener <rguenther@suse.de>
1784 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
1785 results when processing array refs with variable index.
1787 2018-05-18 Toon Moene <toon@moene.org>
1789 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
1790 directly after that of -floop-interchange. Indicate that both
1791 options are enabled by default when specifying -O3.
1793 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1795 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
1796 iterator. Delete separate integer-mode vec_set<mode> expander.
1797 (aarch64_simd_vec_setv2di): Delete.
1798 (vec_setv2di): Delete.
1799 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
1800 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
1801 the "w, r" alternative.
1803 2018-05-18 Martin Liska <mliska@suse.cz>
1805 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
1806 * tree-pass.h (make_pass_lower_switch_O0): New function.
1807 * tree-switch-conversion.c (node_has_low_bound): Remove.
1808 (node_has_high_bound): Likewise.
1809 (node_is_bounded): Likewise.
1810 (class pass_lower_switch): Make it a template type and create
1812 (pass_lower_switch::execute): Add template argument.
1813 (make_pass_lower_switch): New function.
1814 (make_pass_lower_switch_O0): New function.
1815 (do_jump_if_equal): Remove.
1816 (emit_case_nodes): Simplify to just handle all 3 cases and leave
1817 all the hard work to tree optimization passes.
1819 2018-05-18 Martin Liska <mliska@suse.cz>
1821 * dbgcnt.c (limit_low): Renamed from limit.
1822 (limit_high): New variable.
1823 (dbg_cnt_is_enabled): Check for upper limit.
1824 (dbg_cnt): Adjust dumping.
1825 (dbg_cnt_set_limit_by_index): Add new argument for high
1827 (dbg_cnt_set_limit_by_name): Likewise.
1828 (dbg_cnt_process_single_pair): Parse new format.
1829 (dbg_cnt_process_opt): Use strtok.
1830 (dbg_cnt_list_all_counters): Remove 'value' and add
1832 * doc/invoke.texi: Document changes.
1834 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
1836 * doc/sourcebuild.texi (scalar_all_fma): Document.
1837 * tree.def (FMA_EXPR): Delete.
1838 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
1839 * internal-fn.c (ternary_direct): New macro.
1840 (expand_ternary_optab_fn): Likewise.
1841 (direct_ternary_optab_supported_p): Likewise.
1842 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
1843 * builtins.c (fold_builtin_fma): Delete.
1844 (fold_builtin_3): Don't call it.
1845 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
1846 * expr.c (expand_expr_real_2): Likewise.
1847 * fold-const.c (operand_equal_p): Likewise.
1848 (fold_ternary_loc): Likewise.
1849 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
1850 * gimple.c (DEFTREECODE): Likewise.
1851 * gimplify.c (gimplify_expr): Likewise.
1852 * optabs-tree.c (optab_for_tree_code): Likewise.
1853 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
1854 * tree-eh.c (operation_could_trap_p): Likewise.
1855 (stmt_could_throw_1_p): Likewise.
1856 * tree-inline.c (estimate_operator_cost): Likewise.
1857 * tree-pretty-print.c (dump_generic_node): Likewise.
1858 (op_code_prio): Likewise.
1859 * tree-ssa-loop-im.c (stmt_cost): Likewise.
1860 * tree-ssa-operands.c (get_expr_operands): Likewise.
1861 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
1862 * fold-const-call.h (fold_fma): Delete.
1863 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
1864 CFN_FNMA and CFN_FNMS.
1866 * genmatch.c (combined_fn): New enum.
1867 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
1868 (commutative_op): New function.
1869 (commutate): Use it. Handle more than 2 operands.
1870 (dt_operand::gen_gimple_expr): Use commutative_op.
1871 (parser::parse_expr): Allow :c to be used with non-binary
1872 operators if the commutative operand is known.
1873 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
1874 CFN_FMS, CFN_FNMA and CFN_FNMS.
1875 (backprop::process_assign_use): Remove FMA_EXPR handling.
1876 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
1877 (gen_hsa_fma): New function.
1878 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
1879 IFN_FNMA and IFN_FNMS.
1880 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
1881 * gimple-fold.h (follow_all_ssa_edges): Declare.
1882 * gimple-fold.c (follow_all_ssa_edges): New function.
1883 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
1884 gimple_build interface and use follow_all_ssa_edges to fold the result.
1885 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
1886 instead of checking for optabs directly.
1887 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
1888 rather than FMA_EXPRs.
1889 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
1890 call to IFN_FMA instead of an FMA_EXPR.
1892 2018-05-17 Jim Wilson <jimw@sifive.com>
1894 * expr.c (do_tablejump): When converting index to Pmode, if we have a
1895 sign extended promoted subreg, and the range does not have the sign bit
1896 set, then do a sign extend.
1898 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
1899 test, check for sign extended subreg and/or constant operands, and
1900 do a sign extend in that case.
1902 2018-05-17 Steve Ellcey <sellcey@cavium.com>
1904 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
1905 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
1907 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
1908 Change logics_shift_reg to logics_shift_imm.
1909 (thunderx2t99_fp_loadpair_basic): Delete.
1910 (thunderx2t99_fp_storepair_basic): Delete.
1911 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
1912 (thunderx2t99_asimd_polynomial): Delete.
1913 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
1914 and neon_fp_mul_d_scalar_q.
1915 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
1916 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
1917 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
1918 (thunderx2t99_asimd_lut): Add missing tbl types.
1919 (thunderx2t99_asimd_ext): Delete.
1920 (thunderx2t99_asimd_load1_1_mult): Delete.
1921 (thunderx2t99_asimd_load1_2_mult): Delete.
1922 (thunderx2t99_asimd_load1_ldp): New.
1923 (thunderx2t99_asimd_load1): New.
1924 (thunderx2t99_asimd_load2): Add missing *load2* types.
1925 (thunderx2t99_asimd_load3): New.
1926 (thunderx2t99_asimd_load4): New.
1927 (thunderx2t99_asimd_store1_1_mult): Delete.
1928 (thunderx2t99_asimd_store1_2_mult): Delete.
1929 (thunderx2t99_asimd_store2_mult): Delete.
1930 (thunderx2t99_asimd_store2_onelane): Delete.
1931 (thunderx2t99_asimd_store_stp): New.
1932 (thunderx2t99_asimd_store1): New.
1933 (thunderx2t99_asimd_store2): New.
1934 (thunderx2t99_asimd_store3): New.
1935 (thunderx2t99_asimd_store4): New.
1937 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
1939 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
1940 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
1942 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
1943 Segher Boessenkool <segher@kernel.crashing.org>
1946 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
1949 2018-05-17 Richard Biener <rguenther@suse.de>
1951 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
1952 for pruning loop and prune defs feeding only already visited PHIs.
1954 2018-05-17 Richard Biener <rguenther@suse.de>
1956 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
1958 2018-05-17 Bin Cheng <bin.cheng@arm.com>
1959 Richard Biener <rguenther@suse.de>
1961 PR tree-optimization/85793
1962 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
1963 for VMAT_ELEMENTWISE.
1965 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
1967 * internal-fn.h (lookup_internal_fn): Declare
1968 * internal-fn.c (lookup_internal_fn): New function.
1969 * gimple.c (gimple_build_call_from_tree): Handle calls to
1971 * gimple-pretty-print.c (dump_gimple_call): Print "." before
1972 internal function names.
1973 * tree-pretty-print.c (dump_generic_node): Likewise.
1974 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
1976 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
1978 * gimple-fold.h (gimple_build): Make the function forms take
1979 combined_fn rather than built_in_function.
1980 (gimple_simplify): Likewise.
1981 * gimple-match-head.c (gimple_simplify): Likewise.
1982 * gimple-fold.c (gimple_build): Likewise.
1983 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
1984 rather than gimple_build_call_internal.
1985 (get_initial_defs_for_reduction): Likewise.
1986 (vect_create_epilog_for_reduction): Likewise.
1987 (vectorizable_live_operation): Likewise.
1989 2018-05-17 Martin Liska <mliska@suse.cz>
1991 * gimple-ssa-sprintf.c (format_directive): Do not use
1992 space in between 'G_' and '('.
1994 2018-05-17 Jakub Jelinek <jakub@redhat.com>
1997 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
1998 even if the mask is not all ones.
2001 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
2003 (ix86_gimple_fold_builtin): Likewise.
2006 * config/i386/i386.c: Include tree-vector-builder.h.
2007 (ix86_vector_shift_count): New function.
2008 (ix86_fold_builtin): Fold shift builtins by scalar count.
2009 (ix86_gimple_fold_builtin): Likewise.
2011 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
2012 _mm512_setzero): New intrinsics.
2014 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
2015 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2017 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
2018 code generation for cases where splatting a value is not useful.
2019 * simplify-rtx.c (simplify_ternary_operation): Simplify
2020 vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
2021 mode to a vec_concat.
2023 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
2025 * config.gcc: Support "goldmont-plus".
2026 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
2028 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2029 PROCESSOR_GOLDMONT_PLUS.
2030 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
2031 (processor_target_table): Add "goldmont-plus".
2032 (PTA_GOLDMONT_PLUS): Define.
2033 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
2034 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
2035 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
2036 (fold_builtin_cpu): Add "goldmont-plus".
2037 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
2038 (ix86_option_override_internal): Add "goldmont-plus".
2039 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
2040 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
2041 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
2042 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
2044 2018-05-17 Richard Biener <rguenther@suse.de>
2046 PR tree-optimization/85757
2047 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
2048 remove defs that only feed that PHI from further processing.
2050 2018-05-16 Jim Wilson <jimw@sifive.com>
2052 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
2054 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
2056 2018-05-16 Mark Wielaard <mark@klomp.org>
2058 * dwarf2out.c (count_index_strings): New function.
2059 (output_indirect_strings): Call count_index_strings and generate
2060 header for dwarf_version >= 5.
2062 2018-05-16 Mark Wielaard <mark@klomp.org>
2064 * dwarf2out.c (dwarf_FORM): New function.
2065 (set_indirect_string): Use dwarf_FORM.
2066 (reset_indirect_string): Likewise.
2067 (size_of_die): Likewise.
2068 (value_format): Likewise.
2069 (output_die): Likewise.
2070 (add_skeleton_AT_string): Likewise.
2071 (output_macinfo_op): Likewise.
2072 (index_string): Likewise.
2073 (output_index_string_offset): Likewise.
2074 (output_index_string): Likewise.
2075 (count_index_strings): Likewise.
2077 2018-05-16 Carl Love <cel@us.ibm.com>
2079 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
2080 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
2082 2018-05-16 Martin Jambor <mjambor@suse.cz>
2084 * ipa-prop.c (ipa_free_all_edge_args): Remove.
2085 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
2087 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
2089 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
2090 (fnma<mode>4): Likewise.
2091 (fms<mode>4): Likewise.
2092 (fnms<mode>4): Likewise.
2093 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
2094 (aarch64_fnma<mode>4): Likewise.
2095 (aarch64_fms<mode>4): Likewise.
2096 (aarch64_fnms<mode>4): Likewise.
2097 (aarch64_fnmadd<mode>4): Likewise.
2099 2018-05-16 Jason Merrill <jason@redhat.com>
2101 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
2103 2018-05-16 Richard Biener <rguenther@suse.de>
2105 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
2106 (dump_stmt_cost): Declare.
2107 (add_stmt_cost): Dump cost we add.
2108 (add_stmt_costs): New function.
2109 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
2111 (vect_analyze_stmt): Adjust prototype.
2112 (vectorizable_condition): Likewise.
2113 (vectorizable_live_operation): Likewise.
2114 (vectorizable_reduction): Likewise.
2115 (vectorizable_induction): Likewise.
2116 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
2117 cost vector to pass to vectorizable_ and record afterwards.
2118 (vect_model_reduction_cost): Take cost vector argument and adjust.
2119 (vect_model_induction_cost): Likewise.
2120 (vectorizable_reduction): Likewise.
2121 (vectorizable_induction): Likewise.
2122 (vectorizable_live_operation): Likewise.
2123 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
2124 SLP_TREE_NUMBER_OF_VEC_STMTS.
2125 (vect_analyze_slp_cost_1): Remove.
2126 (vect_analyze_slp_cost): Likewise.
2127 (vect_slp_analyze_node_operations): Take visited args and
2128 a target cost vector. Avoid processing already visited stmt sets.
2129 (vect_slp_analyze_operations): Use a local cost vector to gather
2130 costs and register those of non-discarded instances.
2131 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
2132 (vect_schedule_slp_instance): Remove copying of
2133 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
2135 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
2136 adding cost. Record cost entry location.
2137 (vect_prologue_cost_for_slp_op): Function to compute cost of
2138 a constant or invariant generated for SLP vect in the prologue,
2139 split out from vect_analyze_slp_cost_1.
2140 (vect_model_simple_cost): Make static. Adjust for SLP costing.
2141 (vect_model_promotion_demotion_cost): Likewise.
2142 (vect_model_store_cost): Likewise, make static.
2143 (vect_model_load_cost): Likewise.
2144 (vectorizable_bswap): Add cost vector arg and adjust.
2145 (vectorizable_call): Likewise.
2146 (vectorizable_simd_clone_call): Likewise.
2147 (vectorizable_conversion): Likewise.
2148 (vectorizable_assignment): Likewise.
2149 (vectorizable_shift): Likewise.
2150 (vectorizable_operation): Likewise.
2151 (vectorizable_store): Likewise.
2152 (vectorizable_load): Likewise.
2153 (vectorizable_condition): Likewise.
2154 (vectorizable_comparison): Likewise.
2155 (can_vectorize_live_stmts): Likewise.
2156 (vect_analyze_stmt): Likewise.
2157 (vect_transform_stmt): Adjust calls to vectorizable_*.
2158 * tree-vectorizer.c: Include gimple-pretty-print.h.
2159 (dump_stmt_cost): New function.
2161 2018-05-16 Richard Biener <rguenther@suse.de>
2163 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
2164 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
2165 * tree-ssa-dse.c: Include tree-ssa-loop.h.
2166 (check_name): New callback.
2167 (dse_classify_store): Track cycles via a visited bitmap of PHI
2168 defs and simplify handling of in-loop and across loop dead stores
2169 and properly fail for loop-variant refs. Handle byte-tracking with
2170 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
2173 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
2175 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
2176 (vect_get_mask_type_for_stmt): Likewise.
2177 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
2179 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
2180 to determine the statement's vector type and the vector type that
2181 should be used for calculating nunits. Deal with cases in which
2182 the type has to be deferred.
2183 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
2184 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
2185 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
2186 (vect_determine_vf_for_stmt): New functions, split out from...
2187 (vect_determine_vectorization_factor): ...here.
2188 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
2189 (vect_get_mask_type_for_stmt): New functions, split out from
2190 vect_determine_vectorization_factor.
2192 2018-05-16 Richard Biener <rguenther@suse.de>
2194 * tree-cfg.c (verify_gimple_assign_ternary): Properly
2195 verify the [VEC_]COND_EXPR embedded comparison.
2197 2018-05-15 Martin Sebor <msebor@redhat.com>
2199 PR tree-optimization/85753
2200 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
2201 RECORD_TYPE in addition to ARRAY_TYPE.
2203 2018-05-15 Martin Sebor <msebor@redhat.com>
2206 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
2208 2018-05-15 Richard Biener <rguenther@suse.de>
2210 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
2211 add by_clobber_p one. Change algorithm to collect all defs
2212 representing uses we need to walk and try reducing them to
2213 a single one before failing.
2214 (dse_dom_walker::dse_optimize_stmt): Adjust.
2216 2018-05-13 Mark Wielaard <mark@klomp.org>
2218 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
2219 (size_of_loc_descr): Likewise.
2220 (output_loc_operands): Likewise.
2221 (output_loc_operands_raw): Likewise.
2222 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
2223 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
2224 (hash_loc_operands): Likewise.
2225 (compare_loc_operands): Likewise.
2227 2018-05-14 Mark Wielaard <mark@klomp.org>
2229 * dwarf2out.c (count_index_addrs): New function.
2230 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
2232 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2234 PR tree-optimization/83648
2235 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
2236 return value as malloc candidate.
2238 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2241 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
2242 as true in call to suggest_attribute.
2244 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
2246 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
2247 -mreadonly-in-sdata.
2249 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2251 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
2253 (aarch64_crypto_aesd_fused): Likewise.
2255 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
2257 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
2258 (movsi_aarch64): Likewise.
2259 (load_pairsi): Likewise.
2260 (load_pairdi): Likewise.
2261 (store_pairsi): Likewise.
2262 (store_pairdi): Likewise.
2263 (load_pairsf): Likewise.
2264 (load_pairdf): Likewise.
2265 (store_pairsf): Likewise.
2266 (store_pairdf): Likewise.
2267 (zero_extend): Likewise.
2268 (trunc): Swap alternatives.
2269 (fcvt_target): Add '?' to prefer w over r.
2271 2018-05-14 Jakub Jelinek <jakub@redhat.com>
2274 * config/i386/i386.md: Disallow non-commutative arithmetics in
2275 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
2276 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
2277 in the peephole2 before it.
2279 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
2281 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
2282 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
2283 (ix86_handle_option): Handle -mcldemote.
2284 * config.gcc: New header.
2285 * config/i386/cldemoteintrin.h: New file.
2286 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
2287 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
2289 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2290 OPTION_MASK_ISA_CLDEMOTE.
2291 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
2292 (ix86_valid_target_attribute_inner_p): Ditto.
2293 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
2294 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
2295 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
2296 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
2297 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
2299 * config/i386/i386.opt: Add -mcldemote.
2300 * config/i386/x86intrin.h: New header.
2301 * doc/invoke.texi: Add -mcldemote.
2303 2018-05-14 Richard Biener <rguenther@suse.de>
2305 * doc/match-and-simplify.texi: Adjust :s documentation.
2307 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
2309 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
2310 intended memcpy size.
2311 (REORDER_45): Likewise.
2313 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
2315 * sort.cc: New file.
2316 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
2317 * vec.c (qsort_chk): Use gcc_qsort.
2318 * Makefile.in (OBJS-libcommon): Add sort.o.
2319 (build/sort.o): New target. Use it...
2320 (BUILD_RTL): ... here, and...
2321 (build/gencfn-macros): ... here, and...
2322 (build/genmatch): ... here.
2324 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
2325 Chung-Ju Wu <jasonwucj@gmail.com>
2327 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
2328 * config/nds32/nds32-graywolf.md: New file.
2329 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
2330 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
2332 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
2333 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
2334 * config/nds32/nds32.md (pipeline_model): Add graywolf.
2335 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
2336 * config/nds32/pipelines.md: Include n15 settings.
2338 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
2339 Chung-Ju Wu <jasonwucj@gmail.com>
2341 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
2342 * config/nds32/nds32-n13.md: New file.
2343 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
2344 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
2346 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
2347 * config/nds32/nds32.md (pipeline_model): Add n13.
2348 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
2349 * config/nds32/pipelines.md: Include n13 settings.
2351 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
2352 Chung-Ju Wu <jasonwucj@gmail.com>
2354 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
2355 * config/nds32/nds32-n10.md: New file.
2356 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
2357 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
2359 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
2360 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
2361 * config/nds32/nds32.md (pipeline_model): Add n10.
2362 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
2363 * config/nds32/pipelines.md: Include n10 settings.
2365 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
2366 Kito Cheng <kito.cheng@gmail.com>
2367 Chung-Ju Wu <jasonwucj@gmail.com>
2369 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
2370 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2371 Add enum values for DSP extension instructions.
2372 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
2374 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
2375 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
2377 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
2378 * config/nds32/nds32-dspext.md: New file for DSP implementation.
2379 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
2380 * config/nds32/nds32-intrinsic.md: Likewise.
2381 * config/nds32/nds32_intrinsic.h: Likewise.
2382 * config/nds32/nds32-md-auxiliary.c: Likewise.
2383 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
2384 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
2385 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
2386 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
2387 * config/nds32/nds32-protos.h: New declarations for DSP extension.
2388 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
2389 TYPE_DMAC in switch statement.
2390 * config/nds32/nds32.c: New checking and implementation for DSP
2391 extension instructions.
2392 * config/nds32/nds32.h: Likewise.
2393 * config/nds32/nds32.md: Likewise.
2394 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
2395 * config/nds32/predicates.md: Implement new predicates for DSP
2398 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
2400 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
2401 Reformat alternatives and attributes so it is easier to identify
2402 which constraints/attributes go with which instruction.
2403 (mov<mode>_hardfloat32, FMOVE64): Likewise.
2404 (mov<mode>_softfloat32, FMOVE64): Likewise.
2405 (mov<mode>_hardfloat64, FMOVE64): Likewise.
2406 (mov<mode>_softfloat64, FMOVE64): Likewise.
2408 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
2410 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2412 (Basic PowerPC Built-in Functions): The new name of the
2413 subsection previously known as "PowerPC Built-in Functions".
2414 (Basic PowerPC Built-in Functions Available on all Configurations):
2416 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
2417 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
2418 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
2419 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
2421 2018-05-11 Martin Jambor <mjambor@suse.cz>
2424 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
2427 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
2430 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
2432 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
2434 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
2435 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
2436 (ix86_handle_option): Handle -mwaitpkg.
2437 * config.gcc: New header.
2438 * config/i386/cpuid.h (bit_WAITPKG): New bit.
2439 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
2440 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
2442 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2443 OPTION_MASK_ISA_WAITPKG.
2444 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
2445 (ix86_option_override_internal): Add PTA_WAITPKG.
2446 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
2447 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
2448 IX86_BUILTIN_TPAUSE.
2449 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
2450 __builtin_ia32_umwait and __builtin_ia32_tpause.
2451 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
2452 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
2453 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
2454 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
2455 UNSPECV_TPAUSE): New.
2456 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
2457 * config/i386/i386.opt: Add -mwaitpkg.
2458 * config/i386/waitpkgintrin.h: New file.
2459 * config/i386/x86intrin.h: New header.
2460 * doc/invoke.texi: Add -mwaitpkg.
2462 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
2465 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
2467 (cortex-m0): Use armv6s-m isa.
2468 (cortex-m0plus): Likewise.
2469 (cortex-m1): Likewise.
2470 (cortex-m0.small-multiply): Likewise.
2471 (cortex-m0plus.small-multiply): Likewise.
2472 (cortex-m1.small-multiply): Likewise.
2474 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
2475 Jakub Jelinek <jakub@redhat.com>
2477 PR tree-optimization/85692
2478 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
2479 source permute as well.
2481 2018-05-11 Martin Liska <mliska@suse.cz>
2484 * doc/extend.texi: Document LLVM style format for no_sanitize
2487 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
2489 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
2490 mode_supports_vsx_dform_quad to mode_supports_dq_form.
2491 (mode_supports_vsx_dform_quad): Likewise.
2492 (mode_supports_vmx_dform): Move these functions to be next to the
2493 other mode_supports functions.
2494 (mode_supports_dq_form): Likewise.
2495 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
2496 mode_supports_dq_form.
2497 (reg_offset_addressing_ok_p): Likewise.
2498 (offsettable_ok_by_alignment): Likewise.
2499 (rs6000_legitimate_offset_address_p): Likewise.
2500 (legitimate_lo_sum_address_p): Likewise.
2501 (rs6000_legitimize_address): Likewise.
2502 (rs6000_legitimize_reload_address): Likewise.
2503 (rs6000_secondary_reload_inner): Likewise.
2504 (rs6000_preferred_reload_class): Likewise.
2505 (rs6000_output_move_128bit): Likewise.
2507 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
2509 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
2510 Generate SImode target register for null target.
2511 <case IX86_BUILTIN_XGETBV>: Ditto.
2512 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
2513 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
2515 2018-05-10 Carl Love <cel@us.ibm.com>
2517 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
2518 dcbtt and dcbtstt if operands[2] is 0.
2520 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
2523 * config/i386/sse.md (usadv64qi): New expander.
2525 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
2527 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
2528 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
2529 -maltivec=be support.
2530 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
2531 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
2532 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
2533 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
2534 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
2535 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
2536 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
2537 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
2538 altivec_vsumsws): Adjust.
2539 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
2540 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
2541 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
2543 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
2544 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
2545 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
2546 (altivec_lve<VI_char>x): Delete expand.
2547 (*altivec_lve<VI_char>x_internal): Rename to...
2548 (altivec_lve<VI_char>x): ... this.
2549 (altivec_lvxl_<mode>): Delete expand.
2550 (*altivec_lvxl_<mode>_internal): Rename to ...
2551 (altivec_lvxl_<mode>): ... this.
2552 (altivec_stvxl_<mode>): Delete expand.
2553 (*altivec_stvxl_<mode>_internal): Rename to ...
2554 (altivec_stvxl_<mode>): ... this.
2555 (altivec_stve<VI_char>x): Delete expand.
2556 (*altivec_stve<VI_char>x_internal): Rename to ...
2557 (altivec_stve<VI_char>x): ... this.
2558 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
2559 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
2560 reduc_plus_scal_<mode>): Adjust.
2561 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
2563 (rs6000_cpu_cpp_builtins): Adjust.
2564 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
2565 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
2566 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
2567 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
2568 -maltivec=be support.
2569 (rs6000_split_vec_extract_var): Adjust.
2570 (rs6000_split_v4si_init): Adjust.
2571 (swap_selector_for_mode): Delete.
2572 (altivec_expand_lvx_be, altivec_expand_stvx_be,
2573 altivec_expand_stvex_be): Delete.
2574 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
2575 -maltivec=be support.
2576 (rs6000_gimple_fold_builtin): Ditto.
2577 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
2579 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
2580 (TARGET_DIRECT_MOVE_64BIT): Adjust.
2581 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
2582 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
2583 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
2584 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
2585 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
2586 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
2587 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
2588 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
2589 anonymous split): Adjust.
2590 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
2591 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
2593 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
2595 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
2596 when --with-gxx-include-dir is also specified.
2597 * configure: Regenerate.
2599 2018-05-09 Jim Wilson <jimw@sifive.com>
2602 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
2603 * config/riscv/t-withmultilib: New.
2604 * config/riscv/withmultilib.h: New.
2605 * doc/install.texi: Document RISC-V --with-multilib-list support.
2607 2018-05-09 Richard Biener <rguenther@suse.de>
2609 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
2611 (vect_bb_vectorization_profitable_p): Adjust. Compute
2612 actual scalar cost using the cost vector and the add_stmt_cost
2615 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2617 PR rtl-optimization/85645
2618 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
2619 in the REG_CFA_REGISTER note for LR, don't leave it empty.
2621 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2623 PR rtl-optimization/85645
2624 * shrink-wrap.c (spread_components): Return a boolean saying if
2625 anything was changed.
2626 (try_shrink_wrapping_separate): Iterate spread_components until
2627 nothing changes anymore.
2629 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2631 PR rtl-optimization/85645
2632 * regrename.c (build_def_use): Also kill the chains that include the
2633 destination of a REG_CFA_REGISTER note.
2635 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2637 PR rtl-optimization/85645
2638 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
2639 insn that has a REG_CFA_REGISTER note.
2641 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
2643 * cfgexpand.c (expand_clobber): New function.
2644 (expand_gimple_stmt_1): Use it.
2645 * tree-vect-stmts.c (vect_clobber_variable): New function,
2647 (vectorizable_simd_clone_call): ...here.
2648 (vectorizable_store): Emit a clobber either side of an
2649 IFN_STORE_LANES sequence.
2650 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
2652 2018-05-09 Tom de Vries <tom@codesourcery.com>
2655 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
2656 (define_insn "trap_if_false"): Add exit after trap.
2658 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
2660 PR rtl-optimization/85638
2661 * bb-reorder.c: Include common/common-target.h.
2662 (create_forwarder_block): New function extracted from...
2663 (fix_up_crossing_landing_pad): ...here. Rename into...
2664 (dw2_fix_up_crossing_landing_pad): ...this.
2665 (sjlj_fix_up_crossing_landing_pad): New function.
2666 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
2667 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
2668 from both partitions and exit the loop after one iteration.
2670 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
2673 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2675 (Basic PowerPC Built-in Functions): The new name of the
2676 subsection previously known as "PowerPC Built-in Functions".
2677 (Basic PowerPC Built-in Functions Available on all Configurations):
2679 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
2681 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
2683 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
2685 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
2688 2018-05-08 Jim Wilson <jimw@sifive.com>
2690 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
2691 (LD_EMUL_SUFFIX): New.
2692 (LINK_SPEC): Use it.
2694 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
2696 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2698 (Basic PowerPC Built-in Functions): The new name of the
2699 subsection previously known as "PowerPC Built-in Functions".
2700 (Basic PowerPC Built-in Functions Available on all Configurations):
2702 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
2704 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
2706 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
2708 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
2711 2018-05-08 Jakub Jelinek <jakub@redhat.com>
2714 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
2715 after cmpelim optimization.
2717 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
2719 * config.gcc: Support "goldmont".
2720 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
2721 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2723 * config/i386/i386.c (m_GOLDMONT): Define.
2724 (processor_target_table): Add "goldmont".
2725 (PTA_GOLDMONT): Define.
2726 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
2727 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
2728 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
2729 (fold_builtin_cpu): Add "goldmont".
2730 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
2731 (ix86_option_override_internal): Add "goldmont".
2732 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
2733 (processor_type): Add PROCESSOR_GOLDMONT.
2734 * config/i386/i386.md: Add CPU "glm".
2735 * config/i386/glm.md: New file.
2736 * config/i386/x86-tune.def: Add m_GOLDMONT.
2737 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
2739 2018-05-08 Jakub Jelinek <jakub@redhat.com>
2742 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
2744 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
2745 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
2746 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
2749 * config/i386/i386.c (ix86_fold_builtin): Handle
2750 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
2753 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
2754 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
2756 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
2759 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
2760 (check_arch): Likewise.
2761 (check_fpu): Return the result rather than printing it.
2762 (end arch): Fix operator precedence.
2763 (end cpu): Likewise.
2764 (END): Print the result from check_fpu.
2766 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
2767 Alan Hayward <alan.hayward@arm.com>
2768 David Sherwood <david.sherwood@arm.com>
2770 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
2771 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
2772 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
2773 (*fcmuo<mode>_and): New patterns.
2775 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
2777 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
2778 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
2779 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
2780 (cmp_op, sve_imm_con): New code attributes.
2781 (SVE_COND_INT_CMP, imm_con): Delete.
2782 (cmp_op): Remove above unspecs from int attribute.
2783 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
2785 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
2786 comparison-specific unspecs.
2787 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
2788 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
2789 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
2790 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
2791 (*vec_fcm<cmp_op><mode>): Rename to...
2792 (*fcm<cmp_op><mode>): ...this and adjust likewise.
2793 (*vec_fcmuo<mode>): Rename to...
2794 (*fcmuo<mode>): ...this and adjust likewise.
2795 (*pred_fcm<cmp_op><mode>): New pattern.
2796 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
2797 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
2799 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
2801 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
2802 (aarch64_emit_sve_predicated_cond): New function.
2803 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
2804 (aarch64_emit_unspec_cond_or): Replace with...
2805 (aarch64_emit_sve_or_conds): ...this new function. Use
2806 aarch64_emit_sve_ptrue_op for the individual comparisons and
2807 aarch64_emit_binop to OR them together.
2808 (aarch64_emit_inverted_unspec_cond): Replace with...
2809 (aarch64_emit_sve_inverted_cond): ...this new function. Use
2810 aarch64_emit_sve_ptrue_op for the comparison and
2811 aarch64_emit_unop to invert the result.
2812 (aarch64_expand_sve_vec_cmp_float): Update after the above
2813 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
2815 2018-05-07 Nathan Sidwell <nathan@acm.org>
2817 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
2818 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
2819 (Backwards Compatibility): Likewise.
2821 2018-05-07 Luis Machado <luis.machado@linaro.org>
2825 2018-05-07 Luis Machado <luis.machado@linaro.org>
2827 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2828 <prefetch_dynamic_strides>: New const bool field.
2829 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2830 prefetch_dynamic_strides.
2831 (exynosm1_prefetch_tune): Likewise.
2832 (thunderxt88_prefetch_tune): Likewise.
2833 (thunderx_prefetch_tune): Likewise.
2834 (thunderx2t99_prefetch_tune): Likewise.
2835 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
2836 (aarch64_override_options_internal): Update to set
2837 PARAM_PREFETCH_DYNAMIC_STRIDES.
2838 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
2839 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
2840 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
2841 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
2842 prefetch-dynamic-strides setting.
2844 2018-05-07 Luis Machado <luis.machado@linaro.org>
2846 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2847 <minimum_stride>: New const int field.
2848 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2849 minimum_stride field.
2850 (exynosm1_prefetch_tune): Likewise.
2851 (thunderxt88_prefetch_tune): Likewise.
2852 (thunderx_prefetch_tune): Likewise.
2853 (thunderx2t99_prefetch_tune): Likewise.
2854 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
2855 (aarch64_override_options_internal): Update to set
2856 PARAM_PREFETCH_MINIMUM_STRIDE.
2857 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
2858 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
2859 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
2860 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
2861 stride is constant and is below the minimum stride threshold.
2863 2018-05-07 Luis Machado <luis.machado@linaro.org>
2865 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
2868 2018-05-07 Luis Machado <luis.machado@linaro.org>
2870 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2871 <prefetch_dynamic_strides>: New const bool field.
2872 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2873 prefetch_dynamic_strides.
2874 (exynosm1_prefetch_tune): Likewise.
2875 (thunderxt88_prefetch_tune): Likewise.
2876 (thunderx_prefetch_tune): Likewise.
2877 (thunderx2t99_prefetch_tune): Likewise.
2878 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
2879 (aarch64_override_options_internal): Update to set
2880 PARAM_PREFETCH_DYNAMIC_STRIDES.
2881 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
2882 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
2883 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
2884 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
2885 prefetch-dynamic-strides setting.
2887 2018-05-07 Luis Machado <luis.machado@linaro.org>
2889 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2890 <minimum_stride>: New const int field.
2891 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2892 minimum_stride field.
2893 (exynosm1_prefetch_tune): Likewise.
2894 (thunderxt88_prefetch_tune): Likewise.
2895 (thunderx_prefetch_tune): Likewise.
2896 (thunderx2t99_prefetch_tune): Likewise.
2897 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
2898 (aarch64_override_options_internal): Update to set
2899 PARAM_PREFETCH_MINIMUM_STRIDE.
2900 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
2901 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
2902 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
2903 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
2904 stride is constant and is below the minimum stride threshold.
2906 2018-05-06 Jakub Jelinek <jakub@redhat.com>
2909 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
2910 the type is addressable. Don't force op into register if it has
2913 2018-05-05 Roland McGrath <mcgrathr@google.com>
2916 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
2917 any section for which we don't know a specific type it should have,
2918 regardless of name. Previously this was done only for the exact
2919 names ".init_array", ".fini_array", and ".preinit_array".
2920 (default_elf_asm_named_section): Add comment about
2921 relationship with default_section_type_flags and SECTION_NOTYPE.
2922 (get_section): Don't consider it a type conflict if one side has
2923 SECTION_NOTYPE and the other doesn't, as long as neither has the
2924 SECTION_BSS et al used in the default_section_type_flags logic.
2926 2018-05-05 Tom de Vries <tom@codesourcery.com>
2929 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
2930 (workaround_barsyncs): New function.
2931 (nvptx_reorg): Use workaround_barsyncs.
2932 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
2933 (define_expand "nvptx_membar_cta"): New define_expand.
2934 (define_insn "*nvptx_membar_cta"): New insn.
2936 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
2938 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
2939 To improve optimization opportunities.
2940 * builtin-types.def: The new needed builtin types for the above.
2942 2018-05-04 Richard Biener <rguenther@suse.de>
2944 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
2945 * gimple-ssa-store-merging.c
2946 (imm_store_chain_info::output_merged_store): Remove redundant create,
2947 release split_store vector contents on failure.
2948 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
2949 scalar stmt vector on cache hit.
2951 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
2953 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
2955 * config.gcc (powerpc-xilinx-eabi*): Remove.
2956 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
2958 (fusion_addis_mem_combo_load): Ditto.
2959 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
2961 (rs6000_cpu_cpp_builtins): Ditto.
2962 * config/rs6000/rs6000-linux.c
2963 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
2964 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
2965 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
2967 (rs6000_setup_reg_addr_masks): Ditto.
2968 (rs6000_init_hard_regno_mode_ok): Ditto.
2969 (rs6000_option_override_internal): Ditto.
2970 (legitimate_lo_sum_address_p): Ditto.
2971 (rs6000_legitimize_address): Ditto.
2972 (rs6000_legitimize_reload_address): Ditto.
2973 (rs6000_legitimate_address_p): Ditto.
2974 (abi_v4_pass_in_fpr): Ditto.
2975 (setup_incoming_varargs): Ditto.
2976 (rs6000_gimplify_va_arg): Ditto.
2977 (rs6000_split_multireg_move): Ditto.
2978 (rs6000_savres_strategy): Ditto.
2979 (rs6000_emit_prologue_components): Ditto.
2980 (rs6000_emit_epilogue_components): Ditto.
2981 (rs6000_emit_prologue): Ditto.
2982 (rs6000_emit_epilogue): Ditto.
2983 (rs6000_elf_file_end): Ditto.
2984 (rs6000_function_value): Ditto.
2985 (rs6000_libcall_value): Ditto.
2986 * config/rs6000/rs6000.h: Ditto.
2987 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
2988 (TARGET_MINMAX): ... this. New.
2989 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
2990 * config/rs6000/rs6000.md: Remove Xilinx FP support.
2991 (*movsi_internal1_single): Delete.
2992 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
2993 mfpu=, mxilinx-fpu): Delete.
2994 * config/rs6000/singlefp.h: Delete.
2995 * config/rs6000/sysv4.h: Remove Xilinx FP support.
2996 * config/rs6000/t-rs6000: Ditto.
2997 * config/rs6000/t-xilinx: Delete.
2998 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
2999 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
3000 (VStype_simple): Delete.
3001 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
3002 * config/rs6000/xfpu.h: Delete.
3003 * config/rs6000/xfpu.md: Delete.
3004 * config/rs6000/xilinx.h: Delete.
3005 * config/rs6000/xilinx.opt: Delete.
3006 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
3007 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
3009 2018-05-04 Tom de Vries <tom@codesourcery.com>
3012 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
3015 2018-05-04 Richard Biener <rguenther@suse.de>
3018 * tree-complex.c (update_complex_assignment): We are always in SSA form.
3019 (expand_complex_div_wide): Likewise.
3020 (expand_complex_operations_1): Likewise.
3021 (expand_complex_libcall): Preserve EH info of the original stmt.
3022 (tree_lower_complex): Handle removed blocks.
3023 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
3024 on complex multiplication and division libcall builtins.
3026 2018-05-04 Richard Biener <rguenther@suse.de>
3029 * fold-const.c (negate_expr_p): Restrict negation of operand
3030 zero of a division to when we know that can happen without
3032 (fold_negate_expr_1): Likewise.
3034 2018-05-04 Jakub Jelinek <jakub@redhat.com>
3037 * real.h (real_nextafter): Declare.
3038 * real.c (real_nextafter): New function.
3039 * fold-const-call.c (fold_const_nextafter): New function.
3040 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
3041 CASE_CFN_NEXTTOWARD.
3042 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
3043 even when arg1_mode is different from arg0_mode.
3045 2018-05-03 Nathan Sidwell <nathan@acm.org>
3047 * doc/extend.texi (Deprecated Features): Remove
3049 (Backwards Compatibility): Likewise.
3050 * doc/invoke.texi (C++ Language Options): Likewise.
3051 (C++ Dialect Options): Likewise.
3053 2018-05-03 Jakub Jelinek <jakub@redhat.com>
3056 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
3057 _mm512_mask_mullox_epi64): New intrinsics.
3059 2018-05-03 Tom de Vries <tom@codesourcery.com>
3062 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
3063 dump files): Add offload-tree.
3065 2018-05-03 Richard Biener <rguenther@suse.de>
3067 PR tree-optimization/85615
3068 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
3069 to loops not nested in BBs loop father to avoid creating multi-entry
3072 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3074 PR tree-optimization/70291
3075 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
3076 arguments. Change return type to tree. Emit libcall as a new
3077 statement rather than replacing existing one when inplace_p is true.
3078 (expand_complex_multiplication_components): New function.
3079 (expand_complex_multiplication): Expand floating-point complex
3080 multiplication using the above.
3081 (expand_complex_division): Rename inner_type parameter to type.
3082 Update expand_complex_libcall call-site.
3083 (expand_complex_operations_1): Update expand_complex_multiplication
3084 and expand_complex_division call-sites.
3086 2018-05-02 Jakub Jelinek <jakub@redhat.com>
3089 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
3090 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
3091 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
3092 the highest significant bit of the shift count mask is clear. In
3093 check whether and[sq]i3 is needed verify that all significant bits
3094 of the shift count other than the highest are set.
3096 2018-05-02 Tom de Vries <tom@codesourcery.com>
3099 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
3100 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
3101 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
3102 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
3103 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
3104 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
3105 __builtin_goacc_parlevel_size.
3107 2018-05-02 Richard Biener <rguenther@suse.de>
3109 PR tree-optimization/85597
3110 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
3111 do not use split vect_get_vec_defs call but call vect_get_slp_defs
3114 2018-05-02 Tom de Vries <tom@codesourcery.com>
3117 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
3118 dump files): Add ltrans-tree.
3120 2018-05-02 Tom de Vries <tom@codesourcery.com>
3123 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
3124 dump files): Add wpa-ipa.
3126 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
3128 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
3129 powerpc*-*-linux*paired* target.
3130 * config/rs6000/750cl.h: Delete.
3131 * config/rs6000/paired.h: Delete.
3132 * config/rs6000/paired.md: Delete.
3133 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
3135 * config/rs6000/rs6000-builtin.def: Remove paired float support.
3136 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
3137 comment. Remove paired float support.
3138 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
3139 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
3141 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
3142 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
3144 * config/rs6000/rs6000.c: Remove paired float support.
3145 (paired_expand_vector_init, paired_expand_vector_move,
3146 paired_emit_vector_compare, paired_emit_vector_cond_expr,
3147 (paired_expand_lv_builtin, paired_expand_stv_builtin,
3148 paired_expand_builtin, paired_expand_predicate_builtin,
3149 paired_init_builtins): Delete.
3150 * config/rs6000/rs6000.h: Remove paired float support.
3151 * config/rs6000/rs6000.md: Remove paired float support.
3152 (move_from_CR_ov_bit): Delete.
3153 * config/rs6000/rs6000.opt (mpaired): Delete.
3154 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
3155 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
3157 2018-05-02 Richard Biener <rguenther@suse.de>
3160 * gimplify.c (gimplify_save_expr): When in SSA form allow
3161 SAVE_EXPRs to compute to SSA vars.
3163 2018-05-02 Jakub Jelinek <jakub@redhat.com>
3166 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
3167 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
3168 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
3169 clobber operands[2], instead use a new pseudo. Formatting fixes.
3171 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
3173 PR tree-optimization/85586
3174 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
3175 exit early for statements in the same group if the accesses are
3178 2018-05-02 Tom de Vries <tom@codesourcery.com>
3181 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
3184 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
3186 PR tree-optimization/85143
3187 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
3189 2018-05-01 Tom de Vries <tom@codesourcery.com>
3192 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
3193 not found" error message.
3195 2018-05-01 Tom de Vries <tom@codesourcery.com>
3198 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
3199 * vec.c (test_ordered_remove_if): New function.
3200 (vec_c_tests): Call test_ordered_remove_if.
3201 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
3202 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
3203 * tree-vect-patterns.c (vect_pattern_recog_1): Use
3204 VEC_ORDERED_REMOVE_IF.
3206 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3208 PR tree-optimization/82665
3209 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
3210 pointer subtraction where arguments come from a memchr call.
3212 2018-05-01 Jakub Jelinek <jakub@redhat.com>
3214 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
3215 --push-state --as-needed and --pop-state instead of --as-needed and
3216 --no-as-needed if ld supports it.
3217 * configure: Regenerated.
3220 * doc/install.texi2html: Replace _002d with - and _002a with * in
3221 generated html files using sed.
3223 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3226 * gcc-rich-location.c (blank_line_before_p): New function.
3227 (use_new_line): New function.
3228 (gcc_rich_location::add_fixit_insert_formatted): New function.
3229 * gcc-rich-location.h
3230 (gcc_rich_location::add_fixit_insert_formatted): New function.
3232 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3234 * selftest.c (assert_streq): Rename "expected" and "actual" to
3235 "val1" and "val2". Extend NULL-handling to cover both inputs
3236 symmetrically, while still requiring both to be non-NULL for a pass.
3237 * selftest.h (assert_streq): Rename "expected" and "actual" to
3239 (ASSERT_EQ): Likewise.
3240 (ASSERT_EQ_AT): Likewise.
3241 (ASSERT_KNOWN_EQ): Likewise.
3242 (ASSERT_KNOWN_EQ_AT): Likewise.
3243 (ASSERT_NE): Likewise.
3244 (ASSERT_MAYBE_NE): Likewise.
3245 (ASSERT_MAYBE_NE_AT): Likewise.
3246 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
3247 the assertion to pass.
3248 (ASSERT_STREQ_AT): Likewise.
3250 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
3252 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
3253 interaction with -pie.
3255 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3257 * selftest.h: Fix alphabetization of per-source-file selftest
3260 2018-04-30 Jason Merrill <jason@redhat.com>
3262 PR c++/61982 - dead stores to destroyed objects.
3263 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
3266 2018-04-30 Jason Merrill <jason@redhat.com>
3268 * tree.c (build_clobber): New.
3269 * tree.h: Declare it.
3270 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
3272 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3274 * diagnostic-show-locus.c (layout::layout): Update for
3275 location_get_source_line returning a char_span.
3276 (struct char_span): Move to input.h.
3277 (struct correction): Update for fields in char_span becoming
3279 (struct source_line): Update for location_get_source_line
3280 returning a char_span.
3281 (layout::print_line): Likewise.
3282 * edit-context.c (edited_file::print_content): Likewise.
3283 (edited_file::print_diff_hunk): Likewise.
3284 (edited_file::print_run_of_changed_lines): Likewise.
3285 (edited_file::get_num_lines): Likewise.
3286 (edited_line::edited_line): Likewise.
3287 * final.c (asm_show_source): Likewise.
3288 * input.c (location_get_source_line): Convert return type
3289 from const char * to char_span, losing the final "line_len"
3291 (dump_location_info): Update for the above.
3292 (get_substring_ranges_for_loc): Likewise. Use a char_span
3293 when handling the literal within the line.
3294 (test_reading_source_line): Update for location_get_source_line
3295 returning a char_span.
3296 * input.h (class char_span): Move here from
3297 diagnostic-show-locus.c, converting from a struct to a class.
3298 Make data members private.
3299 (char_span::operator bool): New.
3300 (char_span::length): New.
3301 (char_span::get_buffer): New.
3302 (char_span::operator[]): New.
3303 (char_span::subspan): Make const.
3304 (char_span::xstrdup): New.
3305 (location_get_source_line): Convert return type from const char *
3306 to char_span, losing the final "line_size" param.
3308 2018-04-30 Jan Hubicka <jh@suse.cz>
3310 * lto-wrapper.c (ltrans_priorities): New static var.
3311 (cmp_priority): New.
3312 (run_gcc): Read priorities and if doing parallel build order
3313 the Makefile by them.
3315 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3317 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
3319 2018-04-30 Richard Biener <rguenther@suse.de>
3321 * tree-cfg.c (verify_address): Remove base argument, add
3322 flag whether to check TREE_ADDRESSABLE and do that.
3323 (verify_expr): Remove.
3324 (verify_types_in_gimple_reference): Add pieces from verify_expr.
3325 (verify_gimple_assign_single): Likewise.
3326 (verify_gimple_switch): Likewise.
3327 (verify_expr_location_1): Dereference tp once. Add (disabled)
3328 piece from verify_expr.
3329 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
3331 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3333 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
3335 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3337 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
3338 (small_data_pattern): Likewise.
3339 (arc_rewrite_small_data): Likewise.
3340 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
3341 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
3342 (get_symbol_alignment): New function.
3343 (legitimate_small_data_address_p): Likewise.
3344 (legitimate_scaled_address): Update, call
3345 legitimate_small_data_address_p.
3346 (output_sdata): New static variable.
3347 (arc_print_operand): Update how we handle small data operands.
3348 (arc_print_operand_address): Likewise.
3349 (arc_legitimate_address_p): Update, use
3350 legitimate_small_data_address_p.
3351 (arc_rewrite_small_data_p): Remove.
3352 (arc_rewrite_small_data_1): Likewise.
3353 (arc_rewrite_small_data): Likewise.
3354 (small_data_pattern): Likewise.
3355 (compact_sda_memory_operand): Update to use
3356 legitimate_small_data_address_p and get_symbol_alignment.
3357 (prepare_move_operands): Don't rewite sdata pattern.
3358 (prepare_extend_operands): Remove.
3359 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
3361 (zero_extendqisi2): Likewise.
3362 (zero_extendhisi2): Likewise.
3363 (extendqihi2): Likewise.
3364 (extendqisi2): Likewise.
3365 (extendhisi2): Likewise.
3369 * config/arc/constraints.md (Usd): Change it to memory constraint.
3371 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3373 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
3374 as source of std instructions.
3375 * config/arc/arc.md (movsi_insn): Update pattern predicate to
3376 allow 6-bit constants as source for store instructions.
3377 (movdi_insn): Update instruction pattern to allow 6-bit constants
3378 as source for store instructions.
3380 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
3382 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
3384 2018-04-30 Nathan Sidwell <nathan@acm.org>
3385 Sandra Loosemore <sandra@codesourcery.com>
3387 * dumpfile.c (dump_open): Allow '-' for stdout.
3388 * doc/invoke.texi (Developer Options): Document dump filename
3389 determination early. Document stdin/stdout selection.
3391 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
3393 Microblaze Target: PIC data text relative
3395 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
3396 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
3398 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
3399 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
3400 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
3401 New addressing mode for data-text relative position indepenedent code.
3402 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
3403 'ADDRESS_SYMBOLIC_TXT_REL'.
3404 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
3405 (microblaze_legitimate_pic_operand): Exclude function calls from
3406 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
3407 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
3409 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
3410 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
3411 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
3412 for 'address + offset'.
3413 (microblaze_expand_prologue): Add new function prologue call for
3415 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
3416 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
3417 table in case of TARGET_PIC_DATA_TEXT_REL.
3418 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
3419 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
3420 Add new macros 'UNSPEC_TEXT',
3421 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
3422 + exclude function calls from 'UNSPEC_PLT' in case of data text
3424 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
3425 new target hook for generating address diff vector tables in case of
3427 * doc/tm.texi : Regenerate.
3428 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
3429 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
3430 of addr diff vector generation.
3431 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
3432 target hook definition.
3433 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
3434 Add default function for generate_pic_addr_diff_vec -> flag_pic.
3435 * doc/invoke.texi (Add new pic option): Add new microblaze pic
3436 option for data text relative.
3438 2018-04-30 Richard Biener <rguenther@suse.de>
3440 * tree-chrec.h (evolution_function_is_constant_p): Remove
3442 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
3444 2018-04-30 Richard Biener <rguenther@suse.de>
3447 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
3449 2018-04-30 Richard Biener <rguenther@suse.de>
3451 PR tree-optimization/28364
3452 PR tree-optimization/85275
3453 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
3454 copying first exit test.
3456 2018-04-28 Mark Wielaard <mark@klomp.org>
3458 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
3460 (dwarf_AT): Handle DW_AT_addr_base.
3461 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
3463 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
3466 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
3467 (*ashl<dwi>3_doubleword_mask_1): Ditto.
3468 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
3469 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
3471 2018-04-28 Richard Biener <rguenther@suse.de>
3473 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
3474 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
3475 to reflect use. Only add interesting stmts.
3477 2018-04-27 Martin Jambor <mjambor@suse.cz>
3480 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
3481 the jump function allows for passing through aggregate values.
3483 2018-04-27 David Malcolm <dmalcolm@redhat.com>
3485 * input.h (in_system_header_at): Convert from macro to inline
3487 (from_macro_expansion_at): Likewise.
3488 (from_macro_definition_at): Likewise.
3490 2018-04-27 Jeff Law <law@redhat.com>
3492 * config.gcc: Mark tile* targets as deprecated/obsolete.
3494 2018-04-27 Richard Biener <rguenther@suse.de>
3496 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
3499 2018-04-27 Richard Biener <rguenther@suse.de>
3501 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
3503 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
3505 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
3506 with Yd constraint. Set "preferred_for_speed" attribute from
3507 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
3509 (*movdi_internal): Ditto.
3510 (movti_interunit splitters): Remove
3511 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
3512 (movdi_interunit splitters): Ditto.
3513 * config/i386/constraints.md (Ye): Remove.
3514 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
3516 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3519 * config/aarch64/constraints.md (Usg): Limit to 31.
3522 2018-04-27 Jakub Jelinek <jakub@redhat.com>
3524 PR tree-optimization/85529
3525 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
3526 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
3527 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
3528 zero extension or masking of the MSB bit.
3529 (optimize_range_tests): Add FIRST_BB argument, pass it through
3530 to optimize_range_tests_var_bound.
3531 (maybe_optimize_range_tests, reassociate_bb): Adjust
3532 optimize_range_tests callers.
3534 2018-04-26 Richard Biener <rguenther@suse.de>
3535 Jakub Jelinek <jakub@redhat.com>
3537 * cgraph.h (symbol_table): Just declare debug method here.
3538 * symtab.c (symbol_table::debug): Define.
3540 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
3542 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
3544 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
3546 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
3547 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
3548 (*movdi_internal): Substitute Yi and Yj constraint with x
3549 and Ym and Yn constraint with y constraint. Update "isa"
3550 attribute and set "preferred_for_speed" attribute from
3551 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
3552 (*movsi_internal): Ditto.
3553 (*movdf_internal): Ditto.
3554 (*movsf_internal): Ditto.
3555 (*zero_extendsidi2): Ditto.
3556 * config/i386/sse.md (vec_set<mode>_0): Ditto.
3557 (sse2_loadld): Ditto.
3558 (*vec_extract<ssevecmodelower>_0): Ditto.
3559 (*vec_extractv4si_0_zext_sse4): Ditto.
3560 (vec_concatv2di): Ditto.
3561 (*vec_dup<mode>): Ditto.
3562 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
3563 * config/i386/constraints.md (Yi): Remove.
3568 2018-04-26 Nathan Sidwell <nathan@acm.org>
3570 * dumpfile.c (dump_open): New.
3571 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
3572 (dump_finish): Detect stdio/stderr by value not name.
3574 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
3576 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
3578 2018-04-26 Tom de Vries <tom@codesourcery.com>
3581 * config/nvptx/nvptx.c (verify_neutering_jumps)
3582 (verify_neutering_labels): New function
3583 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
3585 2018-04-26 Tom de Vries <tom@codesourcery.com>
3588 * config/nvptx/nvptx.c (needs_neutering_p): New function.
3589 (nvptx_single): Use needs_neutering_p to skip over insns that do not
3592 2018-04-26 Richard Biener <rguenther@suse.de>
3593 Tom de Vries <tom@codesourcery.com>
3596 * lto-streamer-out.c (output_function): Fixup loops if required to match
3597 discovery done in the reader.
3599 2018-04-26 Richard Biener <rguenther@suse.de>
3601 PR tree-optimization/85116
3602 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
3603 have a loop exit from the single latch predecessor. Remove
3604 case of header with just condition.
3605 (ch_base::copy_headers): Exclude infinite loops from any
3607 (pass_ch::execute): Record exits.
3609 2018-04-26 Richard Biener <rguenther@suse.de>
3611 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
3612 prologue cost vector and pass it to vect_get_load_cost.
3613 (vect_get_peeling_costs_all_drs): Likewise.
3614 (vect_peeling_hash_get_lowest_cost): Likewise.
3615 (vect_enhance_data_refs_alignment): Likewise.
3617 2018-04-26 Richard Biener <rguenther@suse.de>
3620 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
3621 checking of integer<->pointer conversions.
3622 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
3623 sign-/zero-extending pointer types.
3624 (expand_omp_for_static_chunk): Likewise.
3626 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
3627 Jean Lee <xiaoyur347@gmail.com>
3629 * config/mips/mips.c (mips_asan_shadow_offset): New function.
3630 (TARGET_ASAN_SHADOW_OFFSET): Define.
3631 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
3632 true for -fsanitize=address.
3634 2018-04-25 Mark Wielaard <mark@klomp.org>
3636 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
3639 2018-04-25 Jakub Jelinek <jakub@redhat.com>
3641 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
3642 than "alu", remove explicit "memory" and "imm_disp" attributes.
3643 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
3646 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
3647 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
3648 gen_lowpart_no_emit.
3650 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
3653 * config/i386/i386.c (ix86_expand_builtin): Change memory
3654 operand to XI, extend p0 to Pmode.
3655 * config/i386/i386.md: Change unspec volatile and operand
3656 1 mode to XI, change operand 0 mode to P.
3658 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3660 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
3661 GET_MODE_MASK before any checking.
3662 (nds32_can_use_bset_p): Likewise.
3663 (nds32_can_use_btgl_p): Likewise.
3665 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3667 * config/nds32/nds32-doubleword.md: New define_split pattern for
3668 illegal register number.
3670 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3672 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
3674 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3676 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
3678 2018-04-25 Richard Biener <rguenther@suse.de>
3680 * lto-streamer.h (LTO_major_version): Bump to 8.
3682 2018-04-25 Jakub Jelinek <jakub@redhat.com>
3684 * BASE-VER: Set to 9.0.0.
3686 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
3688 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
3689 in __abskf2 and __powikf2.
3691 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3694 * config/aarch64/constraints.md (Usg, Usj): New constraints.
3695 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
3696 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
3697 Use the above on operand 2. Reindent.
3698 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
3700 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
3703 * common/config/i386/i386-common.c (ix86_handle_option): Don't
3705 * config/i386/i386.opt (mcet): Removed.
3706 * doc/install.texi: Remove -mcet documentation.
3707 * doc/invoke.texi: Likewise.
3709 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
3712 * doc/install.texi: Remove -mcet from bootstrap-cet.
3714 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3717 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
3718 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
3722 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
3723 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
3724 containing a CONST_VECTOR.
3726 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
3728 * doc/install.texi: Update newlib dependency for nvptx.
3730 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3733 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
3734 instead of INTVAL when shifting x left.
3736 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
3738 PR tree-optimization/85478
3739 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
3740 vect_grouped_store_supported for single element vectors.
3742 2018-04-24 Richard Biener <rguenther@suse.de>
3745 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
3746 load cost increase to the case of non-constant step.
3748 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3751 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
3752 destination if any_malformed_asm.
3754 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
3757 * expr.c (store_field): In the bitfield case, if the value comes from
3758 a function call and is returned in registers by means of a PARALLEL,
3759 do not change the mode of the temporary unless BLKmode and VOIDmode.
3761 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
3763 PR rtl-optimization/85423
3764 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
3765 dependencies to debug insns when the previous insn is non-debug.
3767 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
3769 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
3770 enums into a single definition.
3771 (fls): Fix predicates and printing.
3774 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
3776 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
3777 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
3778 and short u6 immediate.
3779 (check_if_valid_sleep_operand): Remove.
3780 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
3782 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3784 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
3785 flag_always_save_lp condition.
3786 * config/nds32/nds32.opt (malways-save-lp): New option.
3788 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3790 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
3791 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
3792 * config/nds32/nds32.h
3793 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
3794 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
3796 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3798 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
3799 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
3801 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3802 Chung-Ju Wu <jasonwucj@gmail.com>
3804 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
3806 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
3807 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
3809 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3811 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
3813 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3815 * config/nds32/nds32-protos.h (nds32_data_alignment,
3816 nds32_local_alignment): Declare.
3817 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
3818 nds32_local_alignment): New functions.
3819 (TARGET_CONSTANT_ALIGNMENT): Define.
3820 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
3822 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3824 * config/nds32/nds32.c
3825 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
3826 (TARGET_MODES_TIEABLE_P): Likewise.
3828 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3830 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
3833 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
3834 Chung-Ju Wu <jasonwucj@gmail.com>
3836 * config/nds32/constants.md (unspec_volatile_element): Add enum values
3837 for unaligned access.
3838 * config/nds32/nds32-intrinsic.c: Implementation of expanding
3840 * config/nds32/nds32-intrinsic.md: Likewise.
3841 * config/nds32/nds32_intrinsic.h: Likewise.
3842 * config/nds32/nds32.h (nds32_builtins): Likewise.
3843 * config/nds32/nds32.opt (munaligned-access): New option.
3844 * config/nds32/nds32.c (nds32_asm_file_start): Display
3845 flag_unaligned_access status.
3847 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
3849 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
3850 -mno-relax is present.
3851 * config/riscv/linux.h (LINK_SPEC): Ditto.
3853 2018-04-20 Martin Sebor <msebor@redhat.com>
3856 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
3858 (gimple_fold_builtin_stxcpy_chk): Same.
3859 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
3861 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
3864 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
3865 __powikf2 when long double is IEEE 128-bit.
3867 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
3869 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
3870 step to make sure stack always aligned.
3872 2018-04-20 Carl Love <cel@us.ibm.com>
3875 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
3876 size check for arg0.
3878 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
3879 Tom de Vries <tom@codesourcery.com>
3882 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
3883 Emit insns for calls too.
3884 (nvptx_find_par): Always look for worker-level predecessor insn.
3885 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
3887 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
3888 (nvptx_process_pars): Propagate frames for calls.
3890 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
3893 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
3895 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
3896 (ix86_handle_option): Don't handle OPT_mibt.
3897 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
3899 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
3901 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
3902 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
3903 (ix86_target_macros): Define __CET__ with flag_cf_protection
3904 for -fcf-protection.
3905 * config/i386/i386.c (isa2_opts): Remove -mibt.
3906 * config/i386/i386.h (TARGET_IBT): Removed.
3907 (TARGET_IBT_P): Likewise.
3908 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
3909 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
3910 * config/i386/i386.opt (mcet): Update help message.
3913 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
3914 -mcet as an alias for -mshstk.
3916 2018-04-20 Richard Biener <rguenther@suse.de>
3919 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
3920 complexity by forcing a single use of the multiply operand.
3922 2018-04-20 Martin Jambor <mjambor@suse.cz>
3925 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
3926 recursion dependency to only apply to non-clones.
3928 2018-04-20 Martin Jambor <mjambor@suse.cz>
3931 * ipa-cp.c (create_specialized_node): Check that clones of
3932 self-recursive edges exist during IPA-CP.
3934 2018-04-19 Toon Moene <toon@moene.org>
3936 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
3939 2018-04-19 Jakub Jelinek <jakub@redhat.com>
3941 PR tree-optimization/85467
3942 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
3943 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
3944 VECTOR_CST element to type.
3946 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
3949 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
3950 * config/i386/i386.md (builtin_setjmp_setup): Removed.
3951 (builtin_longjmp): Likewise.
3952 (save_stack_nonlocal): New pattern.
3953 (restore_stack_nonlocal): Likewise.
3955 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
3958 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
3959 Replace ASM_OUTPUT_LABEL with fprintf.
3961 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
3964 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
3965 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
3966 * config/i386/i386-c.c (ix86_target_macros_internal): Also
3967 define __IBT__ and __SHSTK__ for -fcf-protection.
3968 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
3970 (ix86_trampoline_init): Likewise.
3971 (x86_output_mi_thunk): Likewise.
3972 (ix86_notrack_prefixed_insn_p): Likewise.
3973 (ix86_option_override_internal): Don't disallow -fcf-protection.
3974 * config/i386/i386.md (rdssp<mode>): Also enable for
3976 (incssp<mode>): Likewise.
3977 (nop_endbr): Likewise.
3978 * config/i386/i386.opt (mcet): Change help message to built-in
3982 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
3983 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
3984 enable CET built-in functions.
3986 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
3988 * common/config/i386/i386-common.c
3989 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
3990 OPTION_MASK_ISA_MOVDIRI_UNSET,
3991 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
3992 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
3993 * config.gcc (movdirintrin.h): New header.
3994 * config/i386/cpuid.h (bit_MOVDIRI,
3995 bit_MOVDIR64B): New bits.
3996 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
3998 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
3999 (VOID, PVOID, PCVOID)): New function types.
4000 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
4001 __builtin_ia32_directstoreu_u64,
4002 __builtin_ia32_movdir64b): New builtins.
4003 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
4004 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
4006 (ix86_valid_target_attribute_inner_p): Ditto.
4007 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
4008 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
4009 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
4010 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
4011 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
4012 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
4013 (movdiri<mode>, movdir64b_<mode>): New.
4014 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
4015 * config/i386/immintrin.h: Include movdirintrin.h.
4016 * config/i386/movdirintrin.h: New file.
4017 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
4019 2018-04-19 Richard Biener <rguenther@suse.de>
4022 * cfg.c (clear_bb_flags): When loop state says we have
4023 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
4025 2018-04-19 Richard Biener <rguenther@suse.de>
4027 PR tree-optimization/84737
4028 * tree-vect-data-refs.c (vect_copy_ref_info): New function
4029 copying restrict info.
4030 (vect_setup_realignment): Use it.
4031 * tree-vectorizer.h (vect_copy_ref_info): Declare.
4032 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
4033 the first DR to all generated stores.
4034 (vectorizable_load): Likewise for loads.
4036 2018-04-19 Jakub Jelinek <jakub@redhat.com>
4038 PR tree-optimization/85446
4039 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
4040 the integral and pointer types to have the same precision.
4042 * doc/install.texi: Document --disable-cet being the default and
4045 2018-04-18 Martin Liska <mliska@suse.cz>
4047 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
4050 2018-04-18 Martin Liska <mliska@suse.cz>
4053 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
4056 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
4057 arguments if they are comparable.
4059 2018-04-18 Martin Liska <mliska@suse.cz>
4062 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
4065 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
4068 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
4071 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
4072 ENDBR after calling __morestack.
4074 2018-04-18 David Malcolm <dmalcolm@redhat.com>
4077 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
4078 by using gcc_base_ver to generate a gcc_driver_version, and use
4079 it when generating GCC_DRIVER_NAME.
4080 * configure: Regenerate.
4082 2018-04-18 Jakub Jelinek <jakub@redhat.com>
4085 * config.gcc: Obsolete powerpc*-*-*spe*.
4087 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4090 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
4091 (stabstr_D): Change type of unum from unsigned int to
4092 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
4095 2018-04-17 Jim Wilson <jimw@sifive.com>
4098 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
4099 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
4100 Set arg_pointer_offset after using pretend_args_size.
4102 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4104 PR rtl-optimization/85431
4105 * dse.c (record_store): Ignore zero width stores.
4108 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
4109 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
4110 __builtin_stack_restore rather than after it.
4111 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
4112 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
4113 argument instead of virtual_dynamic_stack_rtx.
4115 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
4117 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
4119 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
4120 Add note to error message to explain internal mapping of overloaded
4121 built-in function name to non-overloaded built-in function name.
4122 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
4125 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
4128 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
4129 where the inputs overlap with the output.
4131 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4133 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
4134 (=v, v) alternative and explicit "memory" attribute.
4135 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
4136 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4138 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
4139 "sselog1" type instead of "sselog".
4140 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
4141 "sselog". Remove explicit "memory" attribute.
4142 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
4143 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4145 (vec_extract_hi_v32hi): Merge all alternatives into one, use
4146 "sselog1" type instead of "sselog". Remove explicit "memory"
4148 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
4149 use "sselog1" type instead of "sselog". Remove explicit "memory"
4151 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
4152 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4154 (vec_extract_hi_v64qi): Merge all alternatives into one, use
4155 "sselog1" type instead of "sselog". Remove explicit "memory"
4157 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
4158 use "sselog1" type instead of "sselog". Remove explicit "memory"
4162 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
4165 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
4168 2018-04-17 Martin Jambor <mjambor@suse.cz>
4171 * ipa-cp.c (create_specialized_node): Call
4172 expand_all_artificial_thunks if necessary.
4174 2018-04-17 Martin Liska <mliska@suse.cz>
4177 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
4178 in message, remote space in between '_G' and '('.
4180 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4183 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
4184 avx512f_vmcmp<mode>3<round_saeonly_name>,
4185 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
4186 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
4187 avx512f_rndscale<mode><round_saeonly_name>,
4188 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
4189 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
4190 Use %<iptr>2 instead of %2 for -masm=intel.
4191 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
4192 avx512f_vcvttss2usi<round_saeonly_name>,
4193 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
4195 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
4196 avx512f_vcvttsd2usi<round_saeonly_name>,
4197 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
4198 Use %q1 instead of %1 for -masm=intel.
4199 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
4200 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
4201 of %3 for -masm=intel.
4202 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
4204 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
4206 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
4208 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
4209 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
4211 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
4213 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
4214 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
4215 %g1 and one with %0 and %1.
4216 (avx512er_vmrcp28<mode><round_saeonly_name>,
4217 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
4219 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
4220 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
4221 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
4222 of %0 and %{%4%} for -masm=intel.
4223 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
4224 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
4225 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
4226 order of %0 and %{%5%}%{z%} for -masm=intel.
4228 2018-04-17 Jan Hubicka <jh@suse.cz>
4231 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
4233 2018-04-17 Martin Liska <mliska@suse.cz>
4236 * multiple_target.c (create_dispatcher_calls): Set apostrophes
4237 for target_clone error message. Make default implementation
4238 clone to be a local declaration.
4239 (separate_attrs): Add new argument and check for an empty
4241 (expand_target_clones): Handle it.
4242 (ipa_target_clone): Make redirection just for target_clones
4245 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
4246 Tom de Vries <tom@codesourcery.com>
4249 * omp-expand.c (expand_oacc_for): Add dummy false branch for
4250 tiled basic blocks without omp continue statements.
4252 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
4255 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
4256 vec_extract expression as having side effects to make sure it gets
4259 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
4262 * config/i386/i386.c (get_builtin_code_for_version): Check
4265 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
4268 * gcc/config.gcc: Support "skylake".
4269 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
4271 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
4272 (processor_target_table): Add "skylake".
4273 (ix86_option_override_internal): Add "skylake".
4274 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
4275 PROCESSOR_CANNONLAKE.
4276 (get_builtin_code_for_version): Fix priority for
4277 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
4278 PROCESSOR_SKYLAKE-AVX512.
4279 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
4280 (processor_type): Add PROCESSOR_SKYLAKE.
4282 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
4283 Jason Merrill <jason@redhat.com>
4286 * convert.c (convert_to_integer_1): Use direct recursion for
4287 enumeral types and types with a precision less than the number
4288 of bits in their mode.
4290 2018-04-16 Julia Koval <julia.koval@intel.com>
4293 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
4294 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
4296 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
4299 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
4300 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
4301 and -mno-direct-move.
4303 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
4306 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
4307 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
4308 Ensure negative shifts result in {0}.
4310 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
4312 PR rtl-optimization/79916
4313 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
4314 regs (if any) to define how to gnerate SD moves when LRA is in
4317 2018-04-13 Jakub Jelinek <jakub@redhat.com>
4319 PR rtl-optimization/85393
4320 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
4321 * except.c (expand_dw2_landing_pad_for_region): Make static.
4322 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
4323 a label and unconditional jump to old_bb, rather than
4324 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
4327 PR rtl-optimization/85376
4328 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
4329 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
4330 instead of a specific value.
4332 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
4333 Bin Cheng <bin.cheng@arm.com>
4335 PR tree-optimization/82965
4336 PR tree-optimization/83991
4337 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
4338 by_profile_only parameter.
4339 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
4340 information if the loop was predicted to iterate too many times.
4341 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
4343 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
4346 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
4349 2018-04-13 Martin Liska <mliska@suse.cz>
4350 Jakub Jelinek <jakub@redhat.com>
4353 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
4354 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
4355 * builtins.c (expand_builtin_memory_copy_args): Use
4356 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
4357 handle dest_addr == pc_rtx.
4359 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
4362 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
4363 asked to not generate direct moves.
4364 (fix_trunc<mode>si2_stfiwx): Similar.
4365 (fix_trunc<mode>si2_internal): Similar.
4367 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4370 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
4371 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
4372 lookup if dest in some wider mode is known to be const0_rtx and
4373 if so, record permanent equivalence for it to be ZERO_EXTEND of
4374 the narrower mode destination.
4376 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
4378 * lto-streamer-out.c (output_function): Revert 259346.
4379 * omp-expand.c (expand_oacc_for): Likewise.
4381 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
4383 PR rtl-optimization/85354
4384 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
4385 * sel-sched.c (sel_global_init): ... here.
4387 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
4390 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
4391 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
4392 mode for PE-COFF targets.
4393 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
4394 (i386_pe_asm_lto_end): Likewise.
4395 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
4396 (TARGET_ASM_LTO_END): Likewise.
4397 * config/i386/winnt.c (saved_debug_info_level): New static variable.
4398 (i386_pe_asm_lto_start): New function.
4399 (i386_pe_asm_lto_end): Likewise.
4401 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
4402 Richard Biener <rguenther@suse.de>
4405 * lto-streamer-out.c (output_function): Fix CFG loop state before
4407 * omp-expand.c (expand_oacc_for): Handle calls to internal
4408 functions like regular functions.
4410 2018-04-12 Richard Biener <rguenther@suse.de>
4413 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
4414 for the early LTO debug to properly generate references to it
4415 during DIE emission. Do not re-use that for the skeleton for
4417 (dwarf2out_early_finish): Likewise.
4419 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4422 * config/i386/sse.md
4423 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
4424 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
4425 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
4426 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
4427 and output is a reg, avoid creating invalid lowpart subreg, but
4428 instead split into a 512-bit move. Don't split if not AVX512VL,
4429 input is xmm16+ reg and output is a mem.
4430 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
4431 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
4432 xmm16+ reg and output is a mem.
4434 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4436 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
4437 also for flag_dwarf2_cfi_asm.
4439 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4441 PR rtl-optimization/85342
4442 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
4443 a bool scalar var inside of the loop instead. Don't try to update
4444 recog_data.operand after failed apply_change_group.
4446 2018-04-12 Tom de Vries <tom@codesourcery.com>
4449 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
4450 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
4451 array with flexible array member as array without given dimension.
4452 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
4453 argument for undefined param to true.
4455 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
4458 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
4459 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
4460 from PowerPC section.
4461 * config/rs6000/sysv4.opt (mcall-): Improve help text.
4462 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
4463 help text that is too long.
4464 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
4465 help text that is too long.
4466 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
4467 help text that is too long.
4469 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
4471 * config/alpha/alpha.md (stack_probe_internal): Rename
4472 from "probe_stack". Update all callers.
4474 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4476 PR rtl-optimization/84566
4477 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
4478 sched_macro_fuse_insns.
4480 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4483 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
4484 (compute_block_dependences): ... from here.
4486 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4488 PR tree-optimization/85331
4489 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
4490 from int to HOST_WIDE_INT.
4492 2018-04-11 Martin Jambor <mjambor@suse.cz>
4495 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
4496 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
4497 not the same as the source val.
4498 (cgraph_edge_brings_value_p): New parameter.
4499 (gather_edges_for_value): Pass destination value to
4500 cgraph_edge_brings_value_p.
4501 (perhaps_add_new_callers): Likewise.
4502 (get_info_about_necessary_edges): Likewise and exclude values brought
4503 only by self-recursive edges.
4504 (create_specialized_node): Redirect only clones of self-calling edges.
4505 (+self_recursive_pass_through_p): New function.
4506 (find_more_scalar_values_for_callers_subset): Use it.
4507 (find_aggregate_values_for_callers_subset): Likewise.
4508 (known_aggs_to_agg_replacement_list): Removed.
4509 (decide_whether_version_node): Re-calculate known constants for all
4510 remaining context clones.
4512 2018-04-11 Richard Biener <rguenther@suse.de>
4515 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
4516 from early DWARF output.
4517 (dwarf2out_early_finish): Output line info unconditionally into
4518 early DWARF and add reference to it.
4520 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4523 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
4524 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
4525 other than V2DFmode using iptr mode attribute.
4526 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
4528 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4530 PR rtl-optimization/84659
4531 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
4533 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4536 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
4538 (output_loc_list): Pass address of a dummy size variable even in the
4539 locview handling loop.
4540 (index_location_lists): Add comment on why skip_loc_list_entry can't
4543 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
4546 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
4549 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
4552 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
4553 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
4554 and -mstring-compare-inline-limit.
4556 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4559 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
4560 for stack clash protection in a register whenever we need it to be in
4563 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4565 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
4566 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
4568 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4571 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
4573 (mlong-double-): Ditto.
4574 * config/rs6000/sysv4.opt (msdata=): Ditto.
4575 (mtls-size=): Ditto.
4577 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
4579 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
4580 erroneous entries for
4581 "vector int vec_ldl (int, long int *)", and
4582 "vector unsigned int vec_ldl (int, unsigned long int *)".
4583 Add comments and entries for
4584 "vector bool char vec_ldl (int, bool char *)",
4585 "vector bool short vec_ldl (int, bool short *)",
4586 "vector bool int vec_ldl (int, bool int *)",
4587 "vector bool long long vec_ldl (int, bool long long *)",
4588 "vector pixel vec_ldl (int, pixel *)",
4589 "vector long long vec_ldl (int, long long *)",
4590 "vector unsigned long long vec_ldl (int, unsigned long long *)".
4591 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
4592 type tree bool_long_long_type_node and correct definition of
4593 bool_V2DI_type_node to make reference to this new type tree.
4594 (rs6000_mangle_type): Replace erroneous reference to
4595 bool_long_type_node with bool_long_long_type_node.
4596 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
4597 comments to emphasize sign distinctions for char and int types and
4598 replace RS6000_BTI_bool_long constant with
4599 RS6000_BTI_bool_long_long constant. Also add comment to restrict
4600 use of RS6000_BTI_pixel.
4601 (bool_long_type_node): Remove this macro definition.
4602 (bool_long_long_type_node): New macro definition
4604 2018-04-10 Jakub Jelinek <jakub@redhat.com>
4606 PR rtl-optimization/85300
4607 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
4608 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
4609 simplify_unary_operation fails.
4611 2018-04-10 Martin Liska <mliska@suse.cz>
4613 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
4614 cgraph_edge and ipa_ref.
4616 2018-04-10 Jakub Jelinek <jakub@redhat.com>
4620 * config/i386/sse.md
4621 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
4622 computation of the VEC_MERGE selector from mask.
4623 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
4624 Fix decoding of the VEC_MERGE selector into mask.
4626 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
4628 PR tree-optimization/85286
4629 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
4631 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
4633 * final.c (final_1): Set insn_last_address as well as
4634 insn_current_address.
4636 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4639 * explow.c (emit_stack_probe): Call validize_mem on memory location
4640 before passing it to gen_probe_stack. Create address operand and
4641 legitimize it for the probe_stack_address case.
4643 2018-04-09 Jan Hubicka <jh@suse.cz>
4646 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
4647 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
4648 * tree.c (free_lang_data_in_type): Fix handling of binfos;
4650 (free_lang_data): Rebuild type inheritance graph.
4652 2018-04-09 Martin Sebor <msebor@redhat.com>
4654 * invoke.texi (-finline-small-functions): Mention other optimization
4656 (-findirect-inlining, -fpartial-inlining): Same.
4657 (-finline-functions-called-once): Same.
4658 (-freorder-blocks-and-partition): Same.
4660 2018-04-09 Jan Hubicka <jh@suse.cz>
4663 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
4664 jumps; choose last target that matches the criteria (i.e.
4665 no partition changes for non-crossing jumps).
4666 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
4667 support for redirecting crossing jumps to non-crossing.
4669 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
4671 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
4672 also for naked functions.
4674 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
4676 * config/arc/arc.md (add_shift): New pattern.
4677 (add_shift2): Likewise.
4678 (sub_shift): Likewise.
4679 (sub_shift_cmp0_noout): Likewise.
4680 (compare_si_ashiftsi): Likewise.
4681 (xbfu_cmp0_noout): New combine pattern.
4682 (xbfu_cmp0"): Likewise.
4683 (movsi_set_cc_insn): Place the predicable variant first.
4684 (commutative_binary_cmp0_noout): Remove clobber.
4685 (commutative_binary_cmp0): New pattern.
4686 (noncommutative_binary_cmp0): Likewise.
4687 (noncommutative_binary_cmp0_noout): Likewise.
4688 (noncommutative_binary_comparison_result_used): Removed.
4689 (rsub_cmp0): New pattern.
4690 (rsub_cmp0_noout): Likewise.
4691 (extzvsi): Changed, keep only meaningful variants.
4692 (SQH, SEZ): New iterators.
4693 (SQH_postfix): New mode attribute.
4694 (SEZ_prefix): New code attribute.
4695 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
4696 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
4697 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
4699 (noncommutative_operator): Check the availability of barrel
4702 2018-04-09 Richard Biener <rguenther@suse.de>
4704 PR tree-optimization/85284
4705 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
4706 Only use the niter constraining form of simple_iv when the exit
4709 2018-04-09 Tom de Vries <tom@codesourcery.com>
4712 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
4713 (define_expand "*memory_barrier"): New define_expand.
4714 (define_insn "memory_barrier"): New insn.
4716 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4718 PR rtl-optimization/80463
4719 PR rtl-optimization/83972
4720 PR rtl-optimization/83480
4722 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
4723 correct producer for the insn.
4724 (tidy_control_flow): Fixup seqnos in case of debug insns.
4726 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4728 PR rtl-optimization/83913
4730 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
4731 different sched-times when merging exprs.
4733 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4735 PR rtl-optimization/83962
4737 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
4738 tidy_fallthru_edge and tidy_control_flow.
4740 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4742 PR rtl-optimization/83530
4744 * sel-sched.c (force_next_insn): New global variable.
4745 (remove_insn_for_debug): When force_next_insn is true, also leave only
4746 next insn in the ready list.
4747 (sel_sched_region): When the region wasn't scheduled, make another pass
4748 over it with force_next_insn set to 1.
4750 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
4752 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
4754 * config/nds32/constants.md (unspec_volatile_element): Add enum values
4755 for interrupt control.
4756 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
4757 functions for interrupt control.
4758 * config/nds32/nds32-intrinsic.md: Likewise.
4759 * config/nds32/nds32_intrinsic.h: Likewise.
4760 * config/nds32/nds32.h (nds32_builtins): Likewise.
4762 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
4764 * config/nds32/nds32.c (nds32_init_machine_status,
4765 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
4766 strict_aligned_p field.
4767 (nds32_expand_to_rtl_hook): New function.
4768 (TARGET_EXPAND_TO_RTL_HOOK): Define.
4769 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
4771 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4772 Chung-Ju Wu <jasonwucj@gmail.com>
4774 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
4775 * config/nds32/nds32-n7.md: New file.
4776 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
4777 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
4779 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
4780 * config/nds32/nds32.md (pipeline_model): Add n7.
4781 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
4782 * config/nds32/pipelines.md: Include n7 settings.
4784 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4785 Chung-Ju Wu <jasonwucj@gmail.com>
4787 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
4788 * config/nds32/nds32-e8.md: New file.
4789 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
4790 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
4792 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
4793 * config/nds32/nds32.md (pipeline_model): Add e8.
4794 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
4795 * config/nds32/pipelines.md: Include e8 settings.
4797 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4798 Chung-Ju Wu <jasonwucj@gmail.com>
4800 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
4801 * config/nds32/nds32-n8.md: New file.
4802 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
4803 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
4805 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
4806 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
4807 * config/nds32/nds32.md (pipeline_model): Add n8.
4808 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
4809 * config/nds32/pipelines.md: Include n8 settings.
4811 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4812 Chung-Ju Wu <jasonwucj@gmail.com>
4814 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
4815 * config/nds32/nds32-n9-2r1w.md: New file.
4816 * config/nds32/nds32-n9-3r2w.md: New file.
4817 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
4818 nds32_register_ports): New or modify for cpu n9.
4819 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
4821 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
4822 * config/nds32/nds32-utils.c: New file.
4823 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
4824 TARGET_MUL_SLOW): Define.
4825 * config/nds32/nds32.md (pipeline_model): New attribute.
4826 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
4827 New options that support cpu n9.
4828 * config/nds32/pipelines.md: Include n9 settings.
4829 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
4831 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
4833 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
4834 information if necessary.
4835 (output_cond_branch_compare_zero): Likewise.
4836 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
4837 (nds32_target_alignment): Refine for alignment.
4838 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
4839 (FUNCTION_BOUNDARY): Modify.
4840 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
4842 * config/nds32/nds32.opt (malways-align, malign-functions): New.
4844 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
4846 * config/nds32/constants.md (unspec_volatile_element): Add values for
4847 TLB operation and data prefetch.
4848 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
4849 functions for TLB operation and data prefetch.
4850 * config/nds32/nds32-intrinsic.md: Likewise.
4851 * config/nds32/nds32_intrinsic.h: Likewise.
4852 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
4853 (nds32_print_operand): Likewise.
4854 * config/nds32/nds32.h (nds32_builtins): Likewise.
4856 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
4857 Andrew Pinski <pinsika@gcc.gnu.org>
4860 * match.pd: Use constant_boolean_node of correct type instead of
4861 boolean_true_node or boolean_false_node for simplifying
4862 pointer comparisons to zero.
4864 2018-04-07 Jakub Jelinek <jakub@redhat.com>
4866 PR tree-optimization/80021
4867 * tree.c (verify_type_variant): Make error call in verify_variant_match
4868 translatable and remove final full stop.
4870 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4872 * config/nds32/constants.md (unspec_volatile_element): Add
4873 UNSPEC_VOLATILE_EH_RETURN.
4874 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
4875 nds32_output_stack_pop): Support dwarf exception handling process.
4876 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
4877 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
4878 exception handling process.
4879 (nds32_compute_stack_frame): Likewise.
4880 (nds32_return_addr_rtx): Likewise.
4881 (nds32_initial_elimination_offset): Likewise.
4882 (nds32_expand_prologue): Likewise.
4883 (nds32_expand_epilogue): Likewise.
4884 (nds32_dynamic_chain_address): New function.
4885 * config/nds32/nds32.h (machine_function): Add fields for dwarf
4887 (DYNAMIC_CHAIN_ADDRESS): Define.
4888 (EH_RETURN_DATA_REGNO): Define.
4889 (EH_RETURN_STACKADJ_RTX): Define.
4890 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
4891 patterns for dwarf exception handling.
4893 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4895 * config/nds32/nds32.h: Clean up obsolete macros.
4897 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4899 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4900 Add enum values for particular instructions.
4901 * config/nds32/nds32-intrinsic.c: Implementation of expanding
4902 particular intrinsic functions.
4903 * config/nds32/nds32-intrinsic.md: Likewise.
4904 * config/nds32/nds32_intrinsic.h: Likewise.
4905 * config/nds32/nds32.h (nds32_builtins): Likewise.
4906 * config/nds32/nds32.md (type): Add pbsad and pbsada.
4907 (btst, ave): New patterns for particular instructions.
4909 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4911 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4912 Add enum values for atomic load/store and memory sync.
4913 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
4915 * config/nds32/nds32-intrinsic.md: Likewise.
4916 * config/nds32/nds32_intrinsic.h: Likewise.
4917 * config/nds32/nds32.h (nds32_builtins): Likewise.
4919 2018-04-07 Jakub Jelinek <jakub@redhat.com>
4921 PR tree-optimization/85257
4922 * fold-const.c (native_encode_vector): If not all elts could fit
4923 and off is -1, return 0 rather than offset.
4924 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
4925 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
4926 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
4927 adjust buffer in native_interpret_expr call.
4929 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4931 * config/nds32/constants.md (unspec_volatile_element): Add cache
4932 control enum values.
4933 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
4934 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
4935 * config/nds32/nds32.c (nds32_cctl_names): New.
4936 (nds32_print_operand): Handle cache control register names.
4937 * config/nds32/nds32.h (nds32_builtins): New enum values.
4938 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
4940 * config/nds32/nds32.md (type): Add mmu.
4941 * config/nds32/pipelines.md (simple_insn): Add mmu.
4943 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4945 * config/nds32/nds32.md (type): Remove call.
4946 * config/nds32/pipelines.md (simple_insn): Likewise.
4948 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4950 * config/nds32/constants.md (unspec_volatile_element): Add
4951 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
4952 UNSPEC_VOLATILE_FMFCFG.
4953 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
4954 description for fmfcfg and fmfcsr.
4955 (bdesc_1arg): Add fmtcsr.
4956 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
4957 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
4958 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
4959 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
4960 unspec_fmfcfg): New patterns.
4961 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
4962 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
4963 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
4964 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
4965 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
4966 __nds32__fmfcfg): Define.
4968 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4970 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
4971 intrinsic register names.
4972 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
4973 intrinsic register enum values and macros.
4975 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4977 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
4978 for load/store addressing form.
4979 (nds32_print_operand_address): Likewise.
4981 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
4984 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
4985 based on LABEL_REF. Remove useless assertion.
4986 (pic_address_needs_scratch): Fix formatting.
4987 (sparc_legitimize_pic_address): Minor tweaks.
4988 (sparc_delegitimize_address): Adjust assertion accordingly.
4989 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
4990 into symbolic_operand.
4991 (movsi_high_pic_label_ref): Likewise.
4992 (movsi_lo_sum_pic_label_ref): Likewise.
4993 (movdi_pic_label_ref): Likewise.
4994 (movdi_high_pic_label_ref): Likewise.
4995 (movdi_lo_sum_pic_label_ref): Likewise.
4997 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
4999 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
5000 custom LIB_SPEC setup.
5002 2018-04-06 Ruslan Bukin <br@bsdpad.com>
5003 Kito Cheng <kito.cheng@gmail.com>
5005 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
5006 * config/riscv/freebsd.h: New.
5008 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
5010 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
5011 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
5014 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
5015 Kito Cheng <kito.cheng@gmail.com>
5017 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
5018 nds32_output_call, nds32_symbol_binds_local_p): New functions.
5019 * config/nds32/nds32-protos.h (nds32_output_call,
5020 nds32_output_return): Declare.
5021 * config/nds32/nds32.md: Refine all the call and return patterns.
5023 2018-04-06 Jakub Jelinek <jakub@redhat.com>
5026 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
5027 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
5029 PR rtl-optimization/84872
5030 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
5031 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
5034 2018-04-06 Tamar Christina <tamar.christina@arm.com>
5036 * expr.c (copy_blkmode_to_reg): Revert 254862.
5037 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
5039 2018-04-06 Richard Biener <rguenther@suse.de>
5042 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
5043 after seeing a component reference with an adjacent field. Treat
5044 refs to arrays at struct end of external decls similar to
5045 refs to unconstrained commons.
5047 2018-04-06 Jakub Jelinek <jakub@redhat.com>
5050 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
5051 look through SAVE_EXPRs with non-side-effects argument. Adjust
5053 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
5056 2018-04-06 Richard Biener <rguenther@suse.de>
5059 * alias.c (find_base_term): New wrapper around find_base_term
5060 unwinding CSELIB_VAL_PTR changes.
5061 (find_base_term): Do not restore CSELIB_VAL_PTR during the
5064 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5066 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
5068 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
5069 constant definitions.
5070 ("nop"): lr 0,0 -> nopr r0
5071 ("nop_lr0", "nop_lr1"): New insn definitions.
5073 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
5075 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
5076 NDS32_V3PUSH_AVAILABLE_P macro.
5078 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
5079 Chung-Ju Wu <jasonwucj@gmail.com>
5081 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
5082 (nds32*-*-*): Add float and fpu_config into supported_defaults.
5083 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
5084 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
5085 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
5086 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
5087 * config/nds32/constraints.md: New constraints and checking for hard
5088 float configuration.
5089 * config/nds32/iterators.md: New mode iterator and attribute for hard
5090 float configuration.
5091 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
5093 * config/nds32/nds32-fpu.md: New file.
5094 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
5095 deal with hard float code generation.
5096 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
5098 (abi_type, float_reg_number): New enum type.
5099 * config/nds32/nds32-predicates.c: New predicates for hard float.
5100 * config/nds32/nds32-protos.h: Declare functions for hard float.
5101 * config/nds32/nds32.c: Implementation for hard float configuration.
5102 * config/nds32/nds32.h: Definitions for hard float configuration.
5103 * config/nds32/nds32.md: Include hard float machine description and
5104 modify patterns for hard float configuration.
5105 * config/nds32/nds32.opt: New options for hard float configuration.
5106 * config/nds32/predicates.md: New predicates for hard float
5109 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5111 * common/config/nds32/nds32-common.c
5112 (nds32_option_optimization_table): Enable -mreleax-hint by default.
5114 2018-04-05 Jakub Jelinek <jakub@redhat.com>
5117 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
5118 CONSTRUCTOR_ELT (ctor, ...)->value.
5120 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
5123 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
5125 2018-04-05 Tom de Vries <tom@codesourcery.com>
5128 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
5131 2018-04-05 Shiva Chen <shiva0217@gmail.com>
5132 Kito Cheng <kito.cheng@gmail.com>
5134 * config/nds32/constraints.md (U33): Fine-tune checking condition.
5135 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
5136 * config/nds32/nds32.h (nds32_16bit_address_type): Add
5137 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
5139 2018-04-05 Shiva Chen <shiva0217@gmail.com>
5140 Kito Cheng <kito.cheng@gmail.com>
5142 * config/nds32/constraints.md (Ufe): New memory constraint.
5143 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
5144 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
5145 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
5147 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
5148 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
5150 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5152 * config/nds32/nds32.md: Use optimize_size in the condition for
5153 alu-shift instructions.
5155 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5157 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
5159 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5161 * config/nds32/nds32.md (negsi2): Refine pattern.
5163 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
5164 Chung-Ju Wu <jasonwucj@gmail.com>
5166 * config/nds32/iterators.md (shift_rotate): New code iterator.
5167 (shift): New code attribute.
5168 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
5169 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
5170 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
5171 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
5172 bit-wise operations.
5173 (andsi3, *andsi3): Ditto.
5174 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
5175 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
5176 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
5177 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
5178 nds32_ior_operand, nds32_xor_operand): New predicates.
5180 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5182 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
5183 (addsi3, subsi3): ... this.
5185 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5187 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
5189 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5191 * config/nds32/nds32.md: Adjust indention.
5193 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
5195 * config/nds32/nds32.md (feature): New attribute.
5197 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5199 * config/nds32/nds32.md (subtype): New attribute.
5201 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
5204 * config/arm/arm-builtins.c (arm_expand_builtin): Change
5205 expansion to perform a bitwise AND of the argument followed by a
5206 boolean negation of the result.
5208 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
5210 PR rtl-optimization/84878
5211 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
5212 the basic block. Assert the use reference is not artificial and that
5213 it has an associated insn.
5215 2018-04-04 Michael Matz <matz@suse.de>
5217 * builtins.c (compute_objsize): Pass correct operand
5218 to array_at_struct_end_p.
5220 2018-04-04 Richard Biener <rguenther@suse.de>
5223 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
5224 from contexts for DINFO_LEVEL_TERSE and below.
5226 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
5228 * config/nds32/nds32-doubleword.md (move_<mode>): Require
5229 resiter_operand condition.
5230 * config/nds32/nds32.md (*move<mode>): Ditto.
5232 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
5233 Monk Chiang <sh.chiang04@gmail.com>
5235 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
5237 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5239 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
5241 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5242 Kito Cheng <kito.cheng@gmail.com>
5244 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
5245 nds32_cond_code_str, output_cond_branch,
5246 output_cond_branch_compare_zero, nds32_expand_cbranch,
5247 nds32_expand_cstore, nds32_expand_movcc,
5248 nds32_output_cbranchsi4_equality_zero,
5249 nds32_output_cbranchsi4_equality_reg,
5250 nds32_output_cbranchsi4_equality_reg_or_const_int,
5251 nds32_output_cbranchsi4_greater_less_zero: New functions.
5252 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
5253 nds32_expand_cstore, nds32_expand_movcc,
5254 nds32_output_cbranchsi4_equality_zero,
5255 nds32_output_cbranchsi4_equality_reg,
5256 nds32_output_cbranchsi4_equality_reg_or_const_int,
5257 nds32_output_cbranchsi4_greater_less_zero): Declare.
5258 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
5259 nds32_rimm11s_operand): New predicates.
5260 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
5261 * config/nds32/nds32.md: Rewrite all the branch and conditional move
5264 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
5266 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
5267 * config/nds32/nds32.md: Ditto.
5268 * config/nds32/pipelines.md: Ditto.
5270 2018-04-04 Richard Biener <rguenther@suse.de>
5272 PR tree-optimization/85168
5273 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
5274 propagating abnormals.
5276 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5278 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
5280 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5281 Kito Cheng <kito.cheng@gmail.com>
5283 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
5284 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
5285 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
5286 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
5287 * config/nds32/nds32.md (sibcall_internal): New.
5288 (sibcall_register): Remove.
5289 (sibcall_immediate): Remove.
5290 (sibcall_value_internal): New.
5291 (sibcall_value_register): Remove.
5292 (sibcall_value_immediate): Remove.
5293 * config/nds32/predicates.md (nds32_general_register_operand): New.
5294 (nds32_call_address_operand): New.
5296 2018-04-03 Jakub Jelinek <jakub@redhat.com>
5298 PR rtl-optimization/85167
5299 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
5300 bb_defs if *split_p, instead preinitialize it to NULL.
5302 PR tree-optimization/85156
5303 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
5304 evaluating the argument multiple times.
5306 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
5308 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
5310 (_mm_cvtpd_ps): Likewise.
5311 (_mm_cvttpd_epi32): Likewise.
5312 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
5313 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
5314 vector, pixel, and bool following altivec.h include.
5316 2018-04-03 Martin Sebor <msebor@redhat.com>
5318 * doc/extend.texi (Common Function Attributes): Clarify.
5319 (const attribute): Likewise.
5320 (pure attribute): Likewise.
5322 2018-04-03 Jakub Jelinek <jakub@redhat.com>
5325 * config/i386/i386.c (ix86_expand_vector_set): Use
5326 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
5328 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
5330 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
5331 instructions when changing rounding bits to preserve precision bits
5332 in the x87 control word.
5334 2018-04-03 Martin Liska <mliska@suse.cz>
5336 PR tree-optimization/82491
5337 * rtl.h (strip_offset_and_add): Replace += suboffset with
5338 poly_uint64 () + suboffset.
5340 2018-03-29 Martin Liska <mliska@suse.cz>
5341 Martin Jambor <mjambor@suse.cz>
5344 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
5345 param_type is not an integral or pointer type.
5347 2018-04-03 Richard Biener <rguenther@suse.de>
5349 * sese.h (recompute_all_dominators): Remove.
5351 2018-04-02 Martin Sebor <msebor@redhat.com>
5353 * doc/invoke.texi (-Wrestrict): Fix typos.
5355 2018-04-02 Jim Wilson <jimw@sifive.com>
5357 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
5358 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
5359 (<optab>di3, <optab>si3_extend): Likewise.
5360 (<optab>si3_mask, <optab>si3_mask_1): New.
5361 (<optab>di3_mask, <optab>di3_mask_1): New.
5362 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
5363 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
5364 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
5366 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
5368 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
5371 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
5373 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
5374 (nds32_canonicalize_comparison): New function.
5376 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5377 Kito Cheng <kito.cheng@gmail.com>
5378 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5380 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
5381 * config/nds32/constants.md (unspec_volatile_element): Add
5382 UNSPEC_VOLATILE_RELAX_GROUP.
5383 * config/nds32/nds32-relax-opt.c: New file.
5384 * config/nds32/nds32-predicates.c
5385 (nds32_symbol_load_store_p): New function.
5386 * config/nds32/nds32-protos.h
5387 (nds32_symbol_load_store_p): Declare function.
5388 (make_pass_nds32_relax_opt): Declare new rtl pass function.
5389 * config/nds32/nds32.c
5390 (nds32_register_pass): New function to register pass.
5391 (nds32_register_passes): New function to register passes.
5392 * config/nds32/nds32.md (relax_group): New pattern.
5393 * config/nds32/nds32.opt (mrelax-hint): New option.
5394 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
5396 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
5398 * config/nds32/t-nds32: Modify files dependency.
5400 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5402 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
5403 (PROFILE_HOOK): Define its implementation.
5405 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5407 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
5408 type and 32-bit size.
5410 2018-04-01 Jakub Jelinek <jakub@redhat.com>
5413 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
5414 (V_128_256): New mode iterator.
5415 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
5416 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
5417 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
5419 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
5420 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
5422 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
5425 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
5426 NaN inputs correctly.
5428 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
5431 * config/rs6000/vsx.md (??r): New mode attribute.
5432 (*vsx_mov<mode>_64bit): Use it.
5433 (*vsx_mov<mode>_32bit): Likewise.
5435 2018-03-30 Martin Sebor <msebor@redhat.com>
5437 PR tree-optimization/84818
5438 * builtins.c (check_access): Use warning_n.
5440 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5443 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
5445 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
5448 2018-03-30 Julia Koval <julia.koval@intel.com>
5451 * x86-tune.def (movx, partial_reg_dependency): Enable for
5454 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
5457 * lra-constraints.c (process_alt_operands): Move setting
5458 this_alternative_matches below.
5460 2018-03-29 Martin Liska <mliska@suse.cz>
5463 * doc/invoke.texi: Document how LTO works with debug info.
5464 Describe auto-load support of binutils. Mention 'x86-64'
5465 as valid option value of -march option.
5467 2018-03-29 Jakub Jelinek <jakub@redhat.com>
5469 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
5472 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
5473 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
5474 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
5477 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
5480 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
5481 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
5482 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
5483 for RS6000_BTM_POWERPC64.
5484 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
5485 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
5486 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
5491 2018-03-28 Carl Love <cel@us.ibm.com>
5494 2017-09-27 Carl Love <cel@us.ibm.com>
5496 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
5497 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
5498 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
5501 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5503 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
5504 instead of __vector bool.
5505 (_mm_max_pu8): Likewise.
5506 (_mm_min_pi16): Likewise.
5508 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
5511 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
5512 (DIVWEUO): Likewise.
5514 (DIVDEUO): Likewise.
5515 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
5516 DIVWEUO and DIVDEUO.
5517 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
5518 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
5519 (div_extend): Likewise.
5520 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
5522 (__builtin_divweuo): Likewise.
5523 (__builtin_divdeo): Likewise.
5524 (__builtin_divdeuo): Likewise.
5526 2018-03-28 Jakub Jelinek <jakub@redhat.com>
5529 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
5530 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
5532 PR tree-optimization/82004
5533 * gimple-match-head.c (optimize_pow_to_exp): New function.
5534 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
5535 Don't fold to exp if optimize_pow_to_exp is false.
5537 2018-03-28 Martin Liska <mliska@suse.cz>
5540 * calls.c (initialize_argument_information): Fix trailing space.
5541 * common.opt: Fix typo and provide better explanation for
5542 -fsanitize-coverage option.
5543 * config/i386/i386.opt: Fix typo.
5545 2018-03-28 Jakub Jelinek <jakub@redhat.com>
5546 Martin Liska <mliska@suse.cz>
5549 * gimplify.c (asan_poison_variable): Don't do the check for
5550 gimplify_omp_ctxp here.
5551 (gimplify_decl_expr): Do it here.
5552 (gimplify_target_expr): Likewise.
5554 2018-03-28 Martin Liska <mliska@suse.cz>
5557 * config/i386/i386.c (ix86_function_arg_advance): Do not call
5558 chkp_type_bounds_count if MPX is not enabled.
5560 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
5562 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
5564 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
5567 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
5568 function to create the function decl for complex long double
5569 multiply and divide for -mabi=ieeelongdouble.
5570 (init_float128_ieee): Call it.
5572 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
5575 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
5576 -fcf-protection=branch -mibt.
5577 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
5579 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5582 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
5584 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
5587 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
5588 extern array declarations.
5590 2018-03-27 Richard Biener <rguenther@suse.de>
5593 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
5594 explicit single_use checks.
5596 2018-03-27 Richard Biener <rguenther@suse.de>
5598 PR tree-optimization/85082
5599 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
5602 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5604 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
5605 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
5606 Turn on fasynchronous-unwind-tables and funwind-tables.
5608 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
5611 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
5612 (*bmi_blsr_<mode>_ccz): Ditto.
5614 2018-03-26 Tom de Vries <tom@codesourcery.com>
5616 PR tree-optimization/85063
5617 * omp-general.c (offloading_function_p): New function. Factor out
5619 * omp-offload.c (pass_omp_target_link::gate): ... here.
5620 * omp-general.h (offloading_function_p): Declare.
5621 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
5622 with attribute omp declare target for offloading functions.
5624 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
5626 PR tree-optimization/84005
5627 * tree-data-ref.h (get_base_for_alignment): Declare.
5628 * tree-data-ref.c (get_base_for_alignment_1): New function.
5629 (get_base_for_alignment): Likewise.
5630 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
5631 get_base_for_alignment to find a suitable base object, instead
5632 of always using drb->base_address.
5634 2018-03-23 Jakub Jelinek <jakub@redhat.com>
5637 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
5638 known size by default.
5640 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
5643 * lra-constraints.c (process_alt_operands): Don't match BLKmode
5644 and non BLKmode operands.
5646 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5649 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
5650 Clean up attributes.
5652 2018-03-23 Richard Biener <rguenther@suse.de>
5655 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
5656 we are going to emit early debug for LTO.
5658 2018-03-23 Jakub Jelinek <jakub@redhat.com>
5661 * function.c (match_asm_constraints_1): Don't optimize if input
5662 doesn't satisfy general_operand predicate for output's mode.
5665 * alias.c (write_dependence_p): Don't require for x_canonicalized
5666 non-VOIDmode if x has VOIDmode.
5669 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
5670 just don't try to optimize it rather than assert it never happens.
5672 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
5674 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
5675 macro expansions for definition of ST_INTERNAL_<mode> and
5676 LD_INTERNAL_<mode> builtins.
5677 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
5679 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
5681 (altivec_expand_st_builtin): Likewise.
5682 (altivec_expand_builtin): Remove calls to deleted functions.
5683 (rs6000_address_for_altivec): Delete this function.
5684 * config/rs6000/vector.md: Remove expands for
5685 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
5687 2018-03-22 Sudakshina Das <sudi.das@arm.com>
5690 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
5691 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
5692 re-computing once computed.
5693 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
5694 (arm_init_machine_status): Initialize
5695 machine->static_chain_stack_bytes.
5697 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
5700 * doc/extend.texi: Add four new prototypes for vec_ld.
5701 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
5702 definitions for more logical presentation.
5703 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
5704 entries for V1TI variants of __builtin_altivec_ld builtin.
5705 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
5706 handling of V1TI variant of LVX icode pattern.
5707 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
5708 (rs6000_gimple_fold_builtin): Likewise.
5709 (altivec_init_builtins): Add code to define
5710 __builtin_altivec_lvx_v1ti function.
5712 2018-03-22 Jakub Jelinek <jakub@redhat.com>
5715 * function.c (match_asm_constraints_1): Don't do the optimization
5716 if input isn't a REG, SUBREG, MEM or constant.
5718 2018-03-22 Tom de Vries <tom@codesourcery.com>
5720 PR tree-optimization/84956
5721 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
5722 bb_has_abnormal_pred.
5724 2018-03-22 Jakub Jelinek <jakub@redhat.com>
5727 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
5728 DECL_INITIAL (decl) to decl at the end.
5729 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
5732 2018-03-21 Joseph Myers <joseph@codesourcery.com>
5734 * doc/extend.texi (__builtin_tgmath): Document when complex
5735 integer types are treated as _Complex _Float64.
5737 2018-03-21 Tom de Vries <tom@codesourcery.com>
5739 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
5741 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5743 PR tree-optimization/84960
5744 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
5745 if it is ENTRY block, move them into single succ of ENTRY in that case.
5747 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
5749 PR tree-optimization/84811
5750 * poly-int.h (poly_span_traits): Remove the T3 parameter and
5751 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
5752 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
5753 (known_subrange_p): Update accordingly. Cast each value involved
5754 in the size comparison, rather than casting the result of the
5757 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5759 PR tree-optimization/84982
5760 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
5761 by flipping the least significant bit rather than all bits from
5762 bitpos to bitpos + bitsize - 1.
5764 2018-03-21 Nathan Sidwell <nathan@acm.org>
5766 * doc/extend.texi (Deprecated Features): Remove mention of
5767 long-deleted deprecations.
5769 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5772 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
5773 * configure: Regenerate.
5775 2018-03-21 Tom de Vries <tom@codesourcery.com>
5777 PR tree-optimization/83126
5778 * tree-parloops.c (num_phis): New function.
5779 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
5781 2018-03-21 Nathan Sidwell <nathan@acm.org>
5783 * doc/extend.texi (Deprecated Features): Update deprecated flags,
5784 mention anon-struct/union members and trailing attributes.
5786 2018-03-21 Bin Cheng <bin.cheng@arm.com>
5788 PR tree-optimization/84969
5789 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
5790 builtin memset partitions if they set different rhs values.
5792 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5794 PR rtl-optimization/84989
5795 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
5796 VEC_DUPLICATE with scalar result mode.
5798 2018-03-21 Martin Liska <mliska@suse.cz>
5801 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
5802 not intended return statement.
5804 2018-03-21 Martin Liska <mliska@suse.cz>
5807 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
5808 (chkp_find_bound_slots_1): Limit number of iterations.
5810 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
5813 * Minor grammar fixes for x86 options.
5815 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5818 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
5819 holding REG_CFA_RESTORE notes, instead turn them into a USE.
5821 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
5824 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
5825 (altivec_lvx_<mode>_1op): Likewise.
5826 (altivec_stvx_<mode>_2op): Likewise.
5827 (altivec_stvx_<mode>_1op): Likewise.
5828 (altivec_lvx_<VM2:mode>): New define_expand.
5829 (altivec_stvx_<VM2:mode>): Likewise.
5830 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
5831 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
5832 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
5833 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
5834 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
5835 (rs6000_gen_lvx): Likewise.
5836 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
5837 (altivec_expand_stv_builtin): Likewise.
5838 (altivec_expand_builtin): Likewise.
5839 * config/rs6000/vector.md: Likewise.
5841 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5844 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
5847 2018-03-20 Richard Biener <rguenther@suse.de>
5850 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
5851 sign-conversions as zero, fall back to standard scalar_stmt
5854 2018-03-20 Martin Liska <mliska@suse.cz>
5857 * predict.c (rebuild_frequencies): Handle case when we have
5858 PROFILE_ABSENT, but flag_guess_branch_prob is false.
5860 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5863 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
5864 flag_section_anchors.
5865 * varasm.c (use_blocks_for_decl_p): Remove hack for
5866 dw2_force_const_mem.
5869 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
5871 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
5872 be created, use lowpart_subreg of operands[0] rather than operands[0]
5874 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
5875 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
5876 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
5877 and n constraint instead of aarch64_shift_imm_di and Usd.
5878 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
5879 (*aarch64_<optab>_reg_minus<mode>3): ... this.
5881 2018-03-20 Sudakshina Das <sudi.das@arm.com>
5884 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
5885 to favor GPR over NEON registers.
5886 (<shift>di3_neon): Likewise.
5888 2018-03-20 Tom de Vries <tom@codesourcery.com>
5891 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
5892 (nvptx_process_pars): Emit bar.sync asap and alap.
5894 2018-03-20 Tom de Vries <tom@codesourcery.com>
5897 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
5898 seen_label if seen_label is already set.
5900 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5903 * config/i386/i386.c (fold_builtin_cpu): For features above 31
5904 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
5905 Use 1U instead of 1. Formatting fixes.
5908 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
5909 instead of TREE_TYPE (s1) for the return value.
5911 2018-03-19 Jakub Jelinek <jakub@redhat.com>
5913 PR tree-optimization/84946
5914 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
5915 bitsize + bitsize in poly_uint64 rather than poly_int64.
5918 * dwarf2asm.c: Include fold-const.c.
5919 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
5920 of decl rather than decl itself.
5922 PR rtl-optimization/84643
5923 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
5925 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
5928 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
5929 calling assemble_variable.
5931 2018-03-19 Sudakshina Das <sudi.das@arm.com>
5934 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
5935 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
5937 2018-03-19 Jim Wilson <jimw@sifive.com>
5940 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
5941 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
5942 (riscv_first_stack_step): Likewise.
5943 (riscv_option_override): Use STACK_BOUNDARY instead of
5945 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
5947 (BIGGEST_ALIGNMENT): Set to 128.
5948 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
5949 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
5952 2018-03-19 Richard Biener <rguenther@suse.de>
5954 PR tree-optimization/84933
5955 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
5956 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
5958 2018-03-19 Richard Biener <rguenther@suse.de>
5960 PR tree-optimization/84859
5961 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
5962 (cond_if_else_store_replacement): Perform sinking operation on
5963 single-store BBs regardless of MAX_STORES_TO_SINK setting.
5964 Generalize what a BB with a single eligible store is.
5966 2018-03-19 Richard Biener <rguenther@suse.de>
5968 PR tree-optimization/84929
5969 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
5970 chrec_is_positive against non-chrec arg.
5972 2018-03-19 Tamar Christina <tamar.christina@arm.com>
5975 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
5977 2018-03-18 Martin Liska <mliska@suse.cz>
5979 PR rtl-optimization/84635
5980 * regrename.c (build_def_use): Use matches_mode only when
5983 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
5985 PR tree-optimization/84913
5986 * tree-vect-loop.c (vectorizable_reduction): Don't try to
5987 vectorize chains of COND_EXPRs.
5989 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
5991 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
5993 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
5995 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
5997 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
5999 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
6001 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
6002 Kito Cheng <kito.cheng@gmail.com>
6004 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
6005 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
6006 (nds32_adjust_reg_alloc_order): New function.
6007 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
6009 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
6011 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
6012 nds32_print_operand, nds32_print_operand_address): Use
6013 HOST_WIDE_INT_PRINT_DEC instead.
6015 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
6017 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
6019 2018-03-17 Jakub Jelinek <jakub@redhat.com>
6022 * config/i386/i386.c (initial_ix86_tune_features,
6023 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
6025 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
6026 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
6027 rather than 1u << ix86_tune. Formatting fix.
6028 (ix86_option_override_internal): Change ix86_arch_mask from
6029 unsigned int to unsigned HOST_WIDE_INT, initialize to
6030 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
6031 (ix86_function_specific_restore): Likewise.
6033 2018-03-16 Jakub Jelinek <jakub@redhat.com>
6036 * postreload.c (reload_combine_recognize_pattern): Perform
6037 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
6038 truncate_int_for_mode the result for the destination's mode.
6041 * hsa-gen.c (mem_type_for_type): Fix comment typo.
6042 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
6044 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
6047 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
6050 * lra-assigns.c (lra_split_hard_reg_for): Don't use
6051 regno_allocno_class_array and sorted_pseudos.
6052 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
6053 insns where regno is used.
6055 2018-03-16 Martin Liska <mliska@suse.cz>
6058 * multiple_target.c (create_dispatcher_calls): Redirect
6059 reference in the symbol table.
6061 2018-03-16 Martin Liska <mliska@suse.cz>
6064 * multiple_target.c (create_dispatcher_calls): Redirect also
6067 2018-03-16 Jakub Jelinek <jakub@redhat.com>
6071 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
6072 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
6073 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
6075 2018-03-16 Julia Koval <julia.koval@intel.com>
6077 * doc/invoke.texi (Skylake Server): Add CLWB.
6078 Cannonlake): Remove CLWB.
6080 2018-03-16 Jakub Jelinek <jakub@redhat.com>
6082 PR tree-optimization/84841
6083 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
6085 (FLOAT_ONE_CONST_TYPE): Define.
6086 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
6087 (sort_by_operand_rank): Put entries with higher constant_type last
6088 rather than first to match comments.
6090 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
6092 * config/nios2/nios2.md (movsi_internal): Fix thinko in
6095 2018-03-15 Jakub Jelinek <jakub@redhat.com>
6098 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
6099 check and use address of target always.
6101 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
6104 * config/i386/i386.c (indirect_thunk_needed): Update comments.
6105 (indirect_thunk_bnd_needed): Likewise.
6106 (indirect_thunks_used): Likewise.
6107 (indirect_thunks_bnd_used): Likewise.
6108 (indirect_return_needed): New.
6109 (indirect_return_bnd_needed): Likewise.
6110 (output_indirect_thunk_function): Add a bool argument for
6112 (output_indirect_thunk_function): Don't generate alias for
6113 function return thunk.
6114 (ix86_code_end): Call output_indirect_thunk_function to generate
6115 function return thunks.
6116 (ix86_output_function_return): Set indirect_return_bnd_needed
6117 and indirect_return_needed instead of indirect_thunk_bnd_needed
6118 and indirect_thunk_needed.
6120 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
6122 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
6123 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
6124 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
6126 2018-03-15 David Malcolm <dmalcolm@redhat.com>
6127 Paul Hua <paul.hua.gm@gmail.com>
6130 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
6132 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
6134 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
6135 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
6138 2018-03-15 Tamar Christina <tamar.christina@arm.com>
6141 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
6142 instead of GET_MODE_SIZE when comparing Units.
6144 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
6147 * varasm.c (hash_section): Return an unchangeble hash value
6148 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
6149 Return !aarch64_can_use_per_function_literal_pools_p ().
6151 2018-03-15 Jakub Jelinek <jakub@redhat.com>
6154 * optabs.c (emit_conditional_move): Pass address of cmode's copy
6155 rather than address of cmode as last argument to prepare_cmp_insn.
6157 2018-03-15 Julia Koval <julia.koval@intel.com>
6159 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
6160 F_AVX512VNNI, F_AVX512BITALG): New.
6162 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
6165 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
6166 insn for floating-point loads and stores.
6168 2018-03-14 Carl Love <cel@us.ibm.com>
6170 * config/rs6000/rs6000-c.c: Add macro definitions for
6171 ALTIVEC_BUILTIN_VEC_PERMXOR.
6172 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
6173 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
6174 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
6175 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
6177 * config/doc/extend.texi: Add prototypes for vec_permxor.
6179 2018-03-14 David Malcolm <dmalcolm@redhat.com>
6182 * diagnostic-show-locus.c (class layout_point): Convert m_line
6183 from int to linenum_type.
6184 (line_span::comparator): Use linenum "compare" function when
6185 comparing line numbers.
6186 (test_line_span): New function.
6187 (layout_range::contains_point): Convert param "row" from int to
6189 (layout_range::intersects_line_p): Likewise.
6190 (layout::will_show_line_p): Likewise.
6191 (layout::print_source_line): Likewise.
6192 (layout::should_print_annotation_line_p): Likewise.
6193 (layout::print_annotation_line): Likewise.
6194 (layout::print_leading_fixits): Likewise.
6195 (layout::annotation_line_showed_range_p): Likewise.
6196 (struct line_corrections): Likewise for field m_row.
6197 (line_corrections::line_corrections): Likewise for param "row".
6198 (layout::print_trailing_fixits): Likewise.
6199 (layout::get_state_at_point): Likewise.
6200 (layout::get_x_bound_for_row): Likewise.
6201 (layout::print_line): Likewise.
6202 (diagnostic_show_locus): Likewise for locals "last_line" and
6204 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
6205 * input.c (selftest::test_linenum_comparisons): New function.
6206 (selftest::input_c_tests): Call it.
6207 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
6208 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
6209 * selftest.h (ASSERT_GT): New macro.
6210 (ASSERT_GT_AT): New macro.
6211 (ASSERT_LT): New macro.
6212 (ASSERT_LT_AT): New macro.
6214 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
6216 PR rtl-optimization/84780
6217 * combine.c (distribute_links): Don't make a link based on pc_rtx.
6219 2018-03-14 Martin Liska <mliska@suse.cz>
6221 * tree.c (record_node_allocation_statistics): Use
6222 get_stats_node_kind.
6223 (get_stats_node_kind): New function extracted from
6224 record_node_allocation_statistics.
6225 (free_node): Use get_stats_node_kind.
6227 2018-03-14 Richard Biener <rguenther@suse.de>
6229 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
6230 that the value-set of ANTIC_IN doesn't grow.
6233 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
6235 (BB_VISITED_WITH_VISITED_SUCCS): New define.
6236 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
6238 2018-03-14 Julia Koval <julia.koval@intel.com>
6240 * config.gcc (icelake-client, icelake-server): New.
6242 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
6243 (initial_ix86_arch_features): Ditto.
6244 (PTA_SKYLAKE): Add SGX.
6245 (PTA_ICELAKE): Remove.
6246 (PTA_ICELAKE_CLIENT): New.
6247 (PTA_ICELAKE_SERVER): New.
6248 (ix86_option_override_internal): Split up icelake on icelake client and
6250 (get_builtin_code_for_version): Ditto.
6251 (fold_builtin_cpu): Ditto.
6252 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
6253 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
6254 * config/i386/i386.h (processor_type): Ditto.
6255 * doc/invoke.texi: Ditto.
6257 2018-03-14 Jakub Jelinek <jakub@redhat.com>
6260 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
6261 INTEGER_CST offset, add it together with bitpos / 8 and
6262 sign extend based on POINTER_SIZE.
6266 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
6269 * config/i386/constraints.md (Yc): New register constraint.
6270 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
6271 Use Yc constraint for alternative 2 of operand 0. Remove
6272 preferred_for_speed attribute.
6274 2018-03-14 Richard Biener <rguenther@suse.de>
6276 PR tree-optimization/84830
6277 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
6278 with the old one to avoid oscillations.
6280 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
6283 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
6285 (assign_by_spills): Return a flag of reload assignment failure.
6286 Do not process the reload assignment failures. Do not spill other
6287 reload pseudos if they has the same reg class. Update n if
6289 (lra_assign): Add a return arg. Set up from the result of
6290 assign_by_spills call.
6291 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6292 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6293 usage_insns if it is not NULL.
6294 (spill_hard_reg_in_range): New function.
6295 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6296 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6297 function prototypes.
6298 (lra_assign): Change prototype.
6299 * lra.c (lra): Add code to deal with fails by splitting hard reg
6302 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
6304 * config/riscv/riscv.opt (mrelax): New option.
6305 * config/riscv/riscv.c (riscv_file_start): Emit ".option
6306 "norelax" when riscv_mrelax is disabled.
6307 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
6309 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6312 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
6313 reassociation for int modes.
6315 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6317 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
6318 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
6320 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
6321 * config/aarch64/aarch64-sve.md
6322 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
6323 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
6324 (*extend<mode><Vwide>2): Rename to...
6325 (aarch64_sve_extend<mode><Vwide>2): ...this.
6326 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
6327 renaming the old pattern to...
6328 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
6330 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
6331 define_expand, renaming the old pattern to...
6332 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
6333 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
6334 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
6335 account when deciding which SVE instruction the optab should use.
6336 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
6338 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6340 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
6341 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
6342 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
6343 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
6344 (tlsdesc_small_<mode>): Turn a define_expand and use
6345 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
6346 (tlsdesc_small_advsimd_<mode>): ...this.
6347 (tlsdesc_small_sve_<mode>): New pattern.
6349 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6351 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
6352 (UNSPEC_UMUL_HIGHPART): New constants.
6353 (MUL_HIGHPART): New int iteraor.
6354 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
6355 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
6357 (*<su>mul<mode>3_highpart): New define_insn.
6359 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
6362 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
6365 2018-03-13 Martin Liska <mliska@suse.cz>
6368 * (sem_item_optimizer::sem_item_optimizer): Initialize new
6370 (sem_item_optimizer::~sem_item_optimizer): Release it.
6371 (sem_item_optimizer::merge_classes): Register variable aliases.
6372 (sem_item_optimizer::fixup_pt_set): New function.
6373 (sem_item_optimizer::fixup_points_to_sets): Likewise.
6374 * ipa-icf.h: Declare new variables and functions.
6376 2018-03-13 Jakub Jelinek <jakub@redhat.com>
6379 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
6380 integer_pow2p@2 and test integer_pow2p in condition.
6381 (A < 0 ? C : 0): Similarly for @1.
6384 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
6385 characters starting at p contain '\0' character, don't look beyond
6389 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
6390 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
6393 * reg-stack.c (change_stack): Change update_end var from int to
6394 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
6395 also call set_block_for_insn on the newly added insns and rescan.
6398 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
6399 on the last operand.
6402 * tree.c (stabilize_reference_1): Return save_expr (e) for
6403 STATEMENT_LIST even if it doesn't have side-effects.
6405 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
6407 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
6409 2018-03-12 Renlin Li <renlin.li@arm.com>
6411 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
6412 aarch64_output_scalar_simd_mov_immediate.
6414 2018-03-12 Martin Sebor <msebor@redhat.com>
6416 PR tree-optimization/83456
6417 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
6418 for perfectly overlapping calls to memcpy.
6419 (gimple_fold_builtin_memory_chk): Same.
6420 (gimple_fold_builtin_strcpy): Handle no-warning.
6421 (gimple_fold_builtin_stxcpy_chk): Same.
6422 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
6424 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
6426 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
6427 parameter. Use it for SFmode.
6428 (rs6000_function_arg_advance_1): Adjust.
6429 (rs6000_function_arg): Adjust.
6430 (rs6000_gimplify_va_arg): Pass false for that new parameter.
6432 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
6434 PR rtl-optimization/84169
6435 PR rtl-optimization/84780
6436 * combine.c (can_combine_p): Check for a 2-insn combination whether
6437 the destination register is used between the two insns, too.
6439 2018-03-12 Richard Biener <rguenther@suse.de>
6441 PR tree-optimization/84803
6442 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
6443 for refs DR analysis didn't process.
6445 2018-03-12 Richard Biener <rguenther@suse.de>
6447 PR tree-optimization/84777
6448 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
6449 force-vectorize loops ignore whether we are optimizing for size.
6451 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
6453 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
6454 (TARGET_MD_ASM_ADJUST): Define.
6456 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
6457 Kito Cheng <kito.cheng@gmail.com>
6458 Chung-Ju Wu <jasonwucj@gmail.com>
6460 * config/nds32/nds32.c (nds32_compute_stack_frame,
6461 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
6462 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
6463 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
6464 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
6465 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
6466 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
6467 * config/nds32/nds32.md (prologue, epilogue): Use macro
6468 NDS32_V3PUSH_AVAILABLE_P to do checking.
6470 2018-03-11 Jakub Jelinek <jakub@redhat.com>
6473 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
6474 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
6475 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
6476 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
6477 addition of most attributes on !orig_type_die or the attribute not
6478 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
6480 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6481 Chung-Ju Wu <jasonwucj@gmail.com>
6483 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
6485 * config/nds32/nds32.opt (mvh): New option.
6487 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6488 Chung-Ju Wu <jasonwucj@gmail.com>
6490 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
6492 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
6493 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
6496 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6497 Chung-Ju Wu <jasonwucj@gmail.com>
6499 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
6501 * config/nds32/nds32-multiple.md (strlensi): New pattern.
6502 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
6504 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
6505 Kito Cheng <kito.cheng@gmail.com>
6506 Chung-Ju Wu <jasonwucj@gmail.com>
6508 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
6509 UNSPEC_FFMISM and UNSPEC_FLMISM.
6510 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
6511 for ffb, ffmism and flmism.
6512 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
6513 (unspec_ffmism): Ditto.
6514 (unspec_flmism): Ditto.
6515 (nds32_expand_builtin_impl): Check if string extension is available.
6516 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
6517 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
6519 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
6522 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
6525 * lra-assigns.c (assign_by_spills): Return a flag of reload
6526 assignment failure. Do not process the reload assignment
6527 failures. Do not spill other reload pseudos if they has the same
6529 (lra_assign): Add a return arg. Set up from the result of
6530 assign_by_spills call.
6531 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6532 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6533 usage_insns if it is not NULL.
6534 (spill_hard_reg_in_range): New function.
6535 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6536 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6537 function prototypes.
6538 (lra_assign): Change prototype.
6539 * lra.c (lra): Add code to deal with fails by splitting hard reg
6542 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
6545 * config/i386/i386.opt: Replace Enforcment with Enforcement.
6547 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
6550 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
6551 (dw_val_node): Add val_symbolic_view.
6552 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
6553 (symview_upper_bound): New.
6554 (new_line_info_table): Initialize symviews_since_reset.
6555 (dwarf2out_source_line): Count symviews_since_reset and set
6556 symview_upper_bound.
6557 (dw_val_equal_p): Handle symview.
6558 (add_AT_symview): New.
6559 (print_dw_val): Handle symview.
6560 (attr_checksum, attr_checksum_ordered): Likewise.
6561 (same_dw_val_p, size_of_die): Likewise.
6562 (value_format, output_die): Likewise.
6563 (add_high_low_attributes): Use add_AT_symview for entry_view.
6564 (dwarf2out_finish): Reset symview_upper_bound, clear
6567 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
6570 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
6571 Add strict argument and use it.
6572 (rs6000_split_multireg_move): Update for new strict argument.
6573 (mem_operand_gpr): Disallow all non-offsettable addresses.
6574 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
6576 2018-03-09 Jakub Jelinek <jakub@redhat.com>
6579 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
6580 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
6581 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
6584 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
6585 decl, use remap_type if we want to use the type.
6587 2018-03-09 Martin Sebor <msebor@redhat.com>
6589 PR tree-optimization/84526
6590 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
6592 (builtin_access::generic_overlap): Be prepared to handle non-array
6595 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
6597 PR rtl-optimization/84682
6598 * lra-constraints.c (process_address_1): Check is_address flag
6599 for address constraints.
6600 (process_alt_operands): Likewise.
6601 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
6602 preprocess_constraints.
6603 * recog.h (preprocess_constraints): Add oploc parameter.
6605 * recog.c (preprocess_constraints): Test address_operand for
6606 CT_ADDRESS constraints.
6608 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
6611 * lra-assigns.c (assign_by_spills): Return a flag of reload
6612 assignment failure. Do not process the reload assignment
6613 failures. Do not spill other reload pseudos if they has the same
6615 (lra_assign): Add a return arg. Set up from the result of
6616 assign_by_spills call.
6617 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6618 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6619 usage_insns if it is not NULL.
6620 (spill_hard_reg_in_range): New function.
6621 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6622 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6623 function prototypes.
6624 (lra_assign): Change prototype.
6625 * lra.c (lra): Add code to deal with fails by splitting hard reg
6628 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6631 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
6632 Accept complain bool parameter. Only emit errors if it is true.
6633 (arm_parse_cpu_option_name): Likewise.
6634 (arm_target_thumb_only): Adjust callers of the above.
6635 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
6636 prototype to take a default true bool parameter.
6637 (arm_parse_arch_option_name): Likewise.
6639 2018-03-09 David Malcolm <dmalcolm@redhat.com>
6640 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
6644 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
6645 * configure: Regenerate.
6646 * configure.ac ("linker --version-script option"): New.
6647 ("linker soname option"): New.
6649 2018-03-09 Richard Biener <rguenther@suse.de>
6651 PR tree-optimization/84775
6652 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
6653 immediate uses of predicate stmts and mark them modified.
6656 PR tree-optimization/84178
6657 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
6659 (version_loop_for_if_conversion): Delay update_ssa call.
6660 (tree_if_conversion): Delay update_ssa until after predicate
6663 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
6666 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
6667 when the function accesses prior frames.
6669 2018-03-08 Jakub Jelinek <jakub@redhat.com>
6672 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
6673 gen_llsym, otherwise call maybe_gen_llsym.
6676 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
6677 has ',' character inside of it.
6679 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6682 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
6683 as clobbering CC_REGNUM.
6685 2018-03-08 Richard Biener <rguenther@suse.de>
6688 * tree-scalar-evolution.c: Include tree-into-ssa.h.
6689 (follow_copies_to_constant): Do not follow SSA names registered
6692 2018-03-08 Richard Biener <rguenther@suse.de>
6694 PR tree-optimization/84178
6695 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
6697 (version_loop_for_if_conversion): Delay update_ssa call.
6698 (tree_if_conversion): Delay update_ssa until after predicate
6701 2018-03-08 David Malcolm <dmalcolm@redhat.com>
6703 PR tree-optimization/84178
6704 * tree-if-conv.c (release_bb_predicate): Remove the
6705 the assertion that the stmts have NULL use_ops.
6706 Discard the statements, asserting that they haven't
6707 yet been added to a BB.
6709 2018-03-08 Richard Biener <rguenther@suse.de>
6711 PR tree-optimization/84746
6712 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
6713 (phi_translate): Pass in destination ANTIC_OUT set.
6714 (phi_translate_1): Likewise. For a simplified result lookup
6715 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
6716 (phi_translate_set): Adjust.
6717 (do_pre_regular_insertion): Likewise.
6718 (do_pre_partial_partial_insertion): Likewise.
6720 2018-03-08 Martin Liska <mliska@suse.cz>
6722 PR gcov-profile/84735
6723 * doc/gcov.texi: Document usage of profile files.
6724 * gcov-io.h: Document changes in the format.
6726 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
6730 * dwarf2out.c (struct dw_line_info_table): Update comments for
6732 (FORCE_RESET_NEXT_VIEW): New.
6733 (FORCE_RESETTING_VIEW_P): New.
6734 (RESETTING_VIEW_P): Check for -1 too.
6735 (ZERO_VIEW_P): Likewise.
6736 (new_line_info_table): Force-reset next view.
6737 (dwarf2out_begin_function): Likewise.
6738 (dwarf2out_source_line): Simplify zero_view_p initialization.
6739 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
6740 view directly. Omit view when omitting .loc at line 0.
6742 2018-03-08 Jakub Jelinek <jakub@redhat.com>
6744 PR tree-optimization/84740
6745 * tree-switch-conversion.c (process_switch): Call build_constructors
6746 only if info.phi_count is non-zero.
6748 PR tree-optimization/84739
6749 * tree-tailcall.c (find_tail_calls): Check call arguments against
6750 DECL_ARGUMENTS (current_function_decl) rather than
6751 DECL_ARGUMENTS (func) when checking for tail recursion.
6753 2018-03-07 Jakub Jelinek <jakub@redhat.com>
6755 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
6756 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
6757 Volker Reichelt's entry and add entries for people that perform
6758 GCC fuzzy testing and report numerous bugs.
6760 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
6763 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
6764 readonly data in sdata, if that is disabled.
6765 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
6766 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
6767 -mreadonly-in-sdata option.
6769 2018-03-07 Martin Sebor <msebor@redhat.com>
6771 PR tree-optimization/84468
6772 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
6773 basic block when looking for nul assignment.
6775 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
6778 * except.h (output_function_exception_table): Adjust prototype.
6779 * except.c (output_function_exception_table): Remove FNNAME parameter
6780 and add SECTION parameter. Ouput one part of the table at a time.
6781 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
6782 the first part of the exception table and emit unwind directives.
6783 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
6784 (i386_pe_seh_cold_init): Likewise.
6785 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
6786 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
6787 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
6788 (ix86_output_call_insn): Emit a nop in one more case for SEH.
6789 * config/i386/winnt.c: Include except.h.
6790 (struct seh_frame_state): Add reg_offset, after_prologue and
6791 in_cold_section fields.
6792 (i386_pe_seh_end_prologue): Set seh->after_prologue.
6793 (i386_pe_seh_cold_init): New function.
6794 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
6795 to seh->in_cold_section.
6796 (seh_emit_push): Record the offset of the push.
6797 (seh_emit_save): Record the offet of the save.
6798 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
6799 Test seh->after_prologue to disregard the epilogue.
6800 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
6801 (i386_pe_end_cold_function): New function.
6803 2018-03-07 Jakub Jelinek <jakub@redhat.com>
6806 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
6807 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
6810 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
6812 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
6813 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
6816 * multiple_target.c: Include tree-inline.h and intl.h.
6817 (expand_target_clones): Diagnose and fail if node->definition and
6818 !tree_versionable_function_p (node->decl).
6820 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
6822 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
6824 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
6825 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
6826 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
6828 2018-03-06 Jakub Jelinek <jakub@redhat.com>
6831 * combine.c (try_combine): Use reg_or_subregno instead of handling
6832 just paradoxical SUBREGs and REGs.
6834 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
6836 * config/arc/arc.c (arc_finalize_pic): Remove function.
6837 (arc_must_save_register): We use single base PIC register, remove
6838 checks to save/restore the PIC register.
6839 (arc_expand_prologue): Likewise.
6840 * config/arc/arc-protos.h (arc_set_default_type_attributes):
6842 (arc_verify_short): Likewise.
6843 (arc_attr_type): Likewise.
6844 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
6845 (walk_stores): Likewise.
6846 (arc_address_cost): Make it static.
6847 (arc_verify_short): Likewise.
6848 (branch_dest): Likewise.
6849 (arc_attr_type): Likewise.
6850 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
6851 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
6852 (arc_final_prescan_insn): Remove inserting the nops due to
6853 hardware hazards. It is done in reorg step.
6854 (insn_length_variant_t): Remove.
6855 (insn_length_parameters_t): Likewise.
6856 (arc_insn_length_parameters): Likewise.
6857 (arc_get_insn_variants): Likewise.
6858 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
6860 2018-03-06 Jakub Jelinek <jakub@redhat.com>
6863 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
6866 PR tree-optimization/84687
6867 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
6869 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
6871 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6873 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
6874 Rename to ppc_speculation_barrier.
6875 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
6876 __builtin_ppc_speculation_barrier.
6878 2018-03-05 Jakub Jelinek <jakub@redhat.com>
6881 * combine.c (combine_simplify_rtx): Don't try to simplify if
6882 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
6885 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
6887 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
6888 to 32 bytes when compiling for POWER9.
6890 2018-03-05 Jakub Jelinek <jakub@redhat.com>
6893 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
6894 regparm >= 3 with no arg reg available also for calls with
6895 flag_force_indirect_call. Pass decl to ix86_function_regparm.
6898 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
6900 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
6902 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
6905 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
6907 2018-03-05 Richard Biener <rguenther@suse.de>
6909 PR tree-optimization/84486
6910 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
6911 When inserting a __builtin_assume_aligned call set the LHS
6912 SSA name alignment info accordingly.
6914 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
6916 PR tree-optimization/84114
6917 * config/aarch64/aarch64.c (aarch64_reassociation_width)
6918 Avoid reassociation of FLOAT_MODE addition.
6920 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
6922 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
6923 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
6924 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
6925 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
6926 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
6927 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
6928 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
6930 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
6931 __builtin_ia32_wbinvd): New builtins.
6932 (SPECIAL_ARGS2): New.
6933 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
6934 (SPECIAL_ARGS2): New.
6935 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
6936 (ix86_valid_target_attribute_inner_p): Ditto.
6937 (ix86_init_mmx_sse_builtins): Add special_args2.
6938 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
6939 TARGET_WBNOINVD_P): New.
6940 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
6941 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
6942 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
6943 * config/i386/immintrin.h (_wbinvd): New intrinsic.
6944 * config/i386/pconfigintrin.h: New file.
6945 * config/i386/wbnoinvdintrin.h: Ditto.
6946 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
6947 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
6949 2018-03-05 Richard Biener <rguenther@suse.de>
6951 PR tree-optimization/84670
6952 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
6954 (BB_VISITED_WITH_VISITED_SUCCS): New define.
6955 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
6956 (compute_antic_aux): Only assert the number of values in ANTIC_IN
6957 doesn't grow if all successors (recursively) were visited at least
6960 2018-03-05 Richard Biener <rguenther@suse.de>
6962 PR tree-optimization/84650
6963 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
6964 if executed in the loop pipeline.
6966 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
6968 * doc/configfiles.texi (Configuration Files): Move info about
6969 conditionalizing $target-protos.h to...
6970 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
6971 differs from $target-protos.h.
6973 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
6974 Chung-Ju Wu <jasonwucj@gmail.com>
6976 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
6977 * config/nds32/nds32-multiple.md (setmemsi): Define.
6978 * config/nds32/nds32-memory-manipulation.c
6979 (nds32_gen_dup_4_byte_to_word_value): New.
6980 (emit_setmem_word_loop): New.
6981 (emit_setmem_byte_loop): New.
6982 (nds32_expand_setmem_loop): New.
6983 (nds32_expand_setmem_loop_v3m): New.
6984 (nds32_expand_setmem_unroll): New.
6985 (nds32_expand_setmem): New.
6987 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
6988 Chung-Ju Wu <jasonwucj@gmail.com>
6990 * config/nds32/nds32-memory-manipulation.c
6991 (nds32_emit_load_store): New.
6992 (nds32_emit_post_inc_load_store): New.
6993 (nds32_emit_mem_move): New.
6994 (nds32_emit_mem_move_block): New.
6995 (nds32_expand_movmemsi_loop_unknown_size): New.
6996 (nds32_expand_movmemsi_loop_known_size): New.
6997 (nds32_expand_movmemsi_loop): New.
6998 (nds32_expand_movmemsi_unroll): New.
6999 (nds32_expand_movmemqi): Rename ...
7000 (nds32_expand_movmemsi): ... to this.
7001 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
7002 (movmemsi): ... to this.
7003 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
7004 (nds32_expand_movmemsi): ... to this.
7006 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
7007 Monk Chiang <sh.chiang04@gmail.com>
7008 Chung-Ju Wu <jasonwucj@gmail.com>
7010 * config/nds32/nds32-protos.h
7011 (nds32_expand_load_multiple): New arguments.
7012 (nds32_expand_store_multiple): Ditto.
7013 (nds32_valid_multiple_load_store): Rename ...
7014 (nds32_valid_multiple_load_store_p): ... to this.
7015 * config/nds32/nds32-memory-manipulation.c
7016 (nds32_expand_load_multiple): Refine implementation.
7017 (nds32_expand_store_multiple): Ditto.
7018 * config/nds32/nds32-multiple.md
7019 (load_multiple): Update nds32_expand_load_multiple interface.
7020 (store_multiple): Update nds32_expand_store_multiple interface.
7021 * config/nds32/nds32-predicates.c
7022 (nds32_valid_multiple_load_store): Rename ...
7023 (nds32_valid_multiple_load_store_p): ... to this and refine
7025 * config/nds32/predicates.md
7026 (nds32_load_multiple_and_update_address_operation): New predicate.
7027 (nds32_store_multiple_and_update_address_operation): New predicate.
7029 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
7030 Chung-Ju Wu <jasonwucj@gmail.com>
7032 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
7033 (combo): New attribute.
7034 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
7036 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
7038 * config/nds32/nds32.opt: Change -mcmodel= default value.
7040 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
7041 Monk Chiang <sh.chiang04@gmail.com>
7042 Chung-Ju Wu <jasonwucj@gmail.com>
7044 * config/nds32/constants.md (unspec_element): New enum.
7045 * config/nds32/constraints.md (Umw): New constraint.
7046 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
7047 * config/nds32/nds32-intrinsic.md: Likewise.
7048 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
7049 (nds32_valid_smw_lwm_base_p): New.
7050 (nds32_output_smw_single_word): New.
7051 (nds32_output_lmw_single_word): New.
7052 (nds32_expand_unaligned_load): New.
7053 (nds32_expand_unaligned_store): New.
7054 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
7055 (nds32_output_smw_single_word): Declare.
7056 (nds32_output_lmw_single_word): Declare.
7057 (nds32_expand_unaligned_load): Declare.
7058 (nds32_expand_unaligned_store): Declare.
7059 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
7060 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
7061 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
7062 NDS32_BUILTIN_UASTORE_DW.
7063 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
7066 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
7067 Kito Cheng <kito.cheng@gmail.com>
7068 Chung-Ju Wu <jasonwucj@gmail.com>
7070 * config/nds32/nds32-intrinsic.c
7071 (nds32_expand_builtin_null_ftype_reg): Delete.
7072 (nds32_expand_builtin_reg_ftype_imm): Ditto.
7073 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
7074 (nds32_read_argument): New.
7075 (nds32_legitimize_target): Ditto.
7076 (nds32_legitimize_argument): Ditto.
7077 (nds32_check_constant_argument): Ditto.
7078 (nds32_expand_unop_builtin): Ditto.
7079 (nds32_expand_unopimm_builtin): Ditto.
7080 (nds32_expand_binop_builtin): Ditto.
7081 (nds32_builtin_decl_impl): Ditto.
7082 (builtin_description): Ditto.
7083 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
7084 (nds32_init_builtins_impl): Ditto.
7085 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
7086 (nds32_builtin_decl): New.
7087 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
7088 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
7090 2018-03-02 Jeff Law <law@redhat.com>
7092 * reorg.c (stop_search_p): Handle DEBUG_INSN.
7093 (redundant_insn, fill_simple_delay_slots): Likewise.
7094 (fill_slots_from_thread): Likewise.
7095 * resource.c (mark_referenced_resources): Likewise.
7096 (mark_set_resources, find_dead_or_set_registers): Likewise.
7098 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7100 * substring-locations.h (format_warning_va): Formatting fix for
7102 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
7104 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
7105 * substring-locations.c: Include intl.h.
7106 (format_warning_va): Turned into small wrapper around
7107 format_warning_n_va, renamed to ...
7108 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
7109 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
7111 (format_warning_at_substring_n): New function.
7112 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
7113 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
7114 format_warning_at_substring with just a shorter name instead of
7115 const function pointer.
7116 (fmtwarn_n): New function.
7117 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
7118 appropriate, get rid of all the fmtstr temporaries, move conditionals
7119 with G_() wrapped string literals directly into fmtwarn arguments,
7120 cast dir.len to (int), formatting fixes.
7122 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
7124 * doc/invoke.texi: Remove "Cilk Plus" references.
7126 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7127 Richard Biener <rguenther@suse.de>
7130 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
7131 for error or warning attributes if CALL_FROM_THUNK_P is set.
7134 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7137 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
7138 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
7141 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
7142 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
7144 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
7145 instead of -1U in last predictors element's probability member.
7147 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
7150 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
7151 arguments if they are comparable.
7153 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
7155 PR tree-optimization/84634
7156 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
7157 masks and masked_loop_p with a single loop_masks, making sure it's
7158 null for bb vectorization.
7160 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
7162 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
7163 (vect_analyze_data_ref_access): Use loop->safe_len rather than
7164 loop->force_vectorize to check whether there is no alias.
7166 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7169 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
7171 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
7173 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
7174 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
7175 instead of a loop around prev_real_insn.
7176 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
7180 * config/i386/i386.c (ix86_print_operand): Use conditional
7181 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
7184 2018-03-02 Richard Biener <rguenther@suse.de>
7186 PR tree-optimization/84427
7187 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
7188 (bitmap_set_subtract_values): Rewrite to handle multiple
7191 (prune_clobbered_mems): Likewise.
7192 (phi_translate): Take edge instead of pred/phiblock.
7193 (phi_translate_1): Likewise.
7194 (phi_translate_set): Likewise. Insert all translated
7195 exprs for a value into the set, keeping possibly multiple
7196 expressions per value.
7197 (compute_antic_aux): Adjust for phi_translate changes.
7198 When intersecting union the expressions and prune those
7199 not in the final value set, keeping possibly multiple
7200 expressions per value. Do not use value-insertion
7201 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
7202 all expressions. Add verification that the value-sets
7203 only shrink during iteration.
7204 (compute_partial_antic_aux): Adjust for the phi_translate changes.
7205 (do_pre_regular_insertion): Likewise.
7206 (do_pre_partial_partial_insertion): Likewise.
7208 2018-03-02 Richard Biener <rguenther@suse.de>
7211 * config/darwin.c (saved_debug_info_level): New static global.
7212 (darwin_asm_lto_start): Disable debug info generation for LTO out.
7213 (darwin_asm_lto_end): Restore debug info generation settings.
7215 2018-03-01 Martin Liska <mliska@suse.cz>
7218 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
7221 2018-03-01 Richard Biener <rguenther@suse.de>
7224 * dwarf2out.c (gen_variable_die): Properly handle late VLA
7225 type annotation with LTO when debug was disabled at compile-time.
7227 2018-03-01 Matthew Fortune <mfortune@gmail.com>
7229 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
7231 (mips_final_postscan_insn): Likewise.
7233 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
7235 PR rtl-optimization/84528
7236 * alias.c (init_alias_target): Add commentary.
7237 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
7238 a unique base value if the frame pointer is not eliminated
7239 to the stack pointer.
7241 2018-03-01 Tom de Vries <tom@codesourcery.com>
7243 PR rtl-optimization/83327
7244 * lra-int.h (hard_regs_spilled_into): Declare.
7245 * lra.c (hard_regs_spilled_into): Define.
7246 (init_reg_info): Init hard_regs_spilled_into.
7247 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
7248 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
7249 (process_bb_lives): Handle hard_regs_spilled_into.
7250 (lra_create_live_ranges_1): Before doing liveness propagation, clear
7251 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
7253 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
7255 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
7256 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
7257 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
7258 * config/rs6000/aix72.h: New file.
7260 2018-02-28 Jakub Jelinek <jakub@redhat.com>
7262 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
7263 instead of warning_at with conditional singular and plural messages
7267 * stor-layout.c (update_alignment_for_field): For
7268 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
7269 && !DECL_PACKED (field), do the alignment update, just use
7270 only desired_align instead of MAX (type_align, desired_align)
7272 (place_field): Don't do known_align < desired_align handling
7273 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
7274 is non-NULL, instead do it after rli->prev_field handling and
7275 only if not within a bitfield word. For DECL_PACKED (field)
7276 use type_align of BITS_PER_UNIT.
7278 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
7280 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
7281 superfluous parentheses and trailing spaces.
7283 2018-02-28 Richard Biener <rguenther@suse.de>
7285 PR tree-optimization/84584
7286 * graphite-scop-detection.c (scop_detection::add_scop): Discard
7287 SCoPs with fake exit edge.
7289 2018-02-28 Martin Liska <mliska@suse.cz>
7292 * timevar.c (timer::print): Fix format to properly print 100%
7295 2018-02-28 Richard Biener <rguenther@suse.de>
7298 * genmatch.c (capture_info::walk_match): Do not mark
7299 captured expressions without operands as expr_p given
7300 they act more like predicates and should be subject to
7301 "lost tail" side-effect preserving.
7303 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
7305 PR rtl-optimization/81611
7306 * auto-inc-dec.c (attempt_change): Move dead note from
7307 mem_insn if it's the next use of regno
7308 (find_address): Take address use of reg holding
7309 non-incremented value. Add parm to limit search to the named
7311 (merge_in_block): Attempt to use a mem insn that is the next
7312 use of the original regno.
7314 2018-02-27 Martin Sebor <msebor@redhat.com>
7317 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
7318 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
7320 2018-02-27 Martin Sebor <msebor@redhat.com>
7322 PR translation/84207
7323 * diagnostic-core.h (warning_n, error_n, inform_n): Change
7324 n argument to unsigned HOST_WIDE_INT.
7325 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
7326 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
7327 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
7328 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
7330 2018-02-27 Richard Biener <rguenther@suse.de>
7332 PR tree-optimization/84512
7333 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
7334 Do not use the estimate returned from record_stmt_cost for
7335 the scalar iteration cost but sum properly using add_stmt_cost.
7337 2018-02-27 Richard Biener <rguenther@suse.de>
7339 PR tree-optimization/84466
7340 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
7341 Adjust last change to less strictly validate use operands.
7343 2018-02-27 Martin Liska <mliska@suse.cz>
7345 PR gcov-profile/84548
7346 * gcov.c (process_file): Allow partial overlap and consider it
7347 also as group functions.
7348 (output_lines): Properly calculate range of lines for a group.
7350 2018-02-27 Martin Liska <mliska@suse.cz>
7352 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
7353 'ggc' suffixes. Change first column width.
7354 (timer::print): Fix formatting of the column.
7356 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
7358 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
7359 preserve inline entry blocks for the sake of debug inline
7360 entry point markers alone.
7361 (remove_unused_locals): Suggest in comments a better place to
7362 force the preservation of inline entry blocks that are
7363 otherwise unused, but do not preserve them.
7365 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7367 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
7369 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7372 * config/i386/constraints.md (Bs): Replace
7373 ix86_indirect_branch_register with
7374 TARGET_INDIRECT_BRANCH_REGISTER.
7376 * config/i386/i386.md (indirect_jump): Likewise.
7377 (tablejump): Likewise.
7378 (*sibcall_memory): Likewise.
7379 (*sibcall_value_memory): Likewise.
7380 Peepholes of indirect call and jump via memory: Likewise.
7381 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
7382 (*sibcall_value_GOT_32): Likewise.
7383 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7384 (GOT_memory_operand): Likewise.
7385 (call_insn_operand): Likewise.
7386 (sibcall_insn_operand): Likewise.
7387 (GOT32_symbol_operand): Likewise.
7388 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
7390 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
7392 PR rtl-optimization/83496
7393 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
7394 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
7395 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
7396 redundant insn, if any.
7397 (relax_delay_slots): Likewise.
7398 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
7400 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
7402 PR tree-optimization/83965
7403 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
7404 that grouped statements are part of a reduction chain. Return
7405 true if the statement is not marked as a reduction itself but
7407 (vect_recog_dot_prod_pattern): Don't check whether the statement
7408 is part of a group here.
7409 (vect_recog_sad_pattern): Likewise.
7410 (vect_recog_widen_sum_pattern): Likewise.
7412 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
7415 * final.c (rest_of_clean_state): Also look for calls inside sequences.
7417 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7420 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
7422 (ix86_output_indirect_function_return): New prototype.
7423 (ix86_split_simple_return_pop_internal): Likewise.
7424 * config/i386/i386.c (indirect_return_via_cx): New.
7425 (indirect_return_via_cx_bnd): Likewise.
7426 (indirect_thunk_name): Handle return va CX_REG.
7427 (output_indirect_thunk_function): Create alias for
7428 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
7429 (ix86_output_indirect_jmp): Remove the bool argument.
7430 (ix86_output_indirect_function_return): New function.
7431 (ix86_split_simple_return_pop_internal): Likewise.
7432 * config/i386/i386.md (*indirect_jump): Don't pass false
7433 to ix86_output_indirect_jmp.
7434 (*tablejump_1): Likewise.
7435 (simple_return_pop_internal): Change it to define_insn_and_split.
7436 Call ix86_split_simple_return_pop_internal to split it for
7438 (simple_return_indirect_internal): Call
7439 ix86_output_indirect_function_return instead of
7440 ix86_output_indirect_jmp.
7442 2018-02-26 Jakub Jelinek <jakub@redhat.com>
7445 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
7446 memset and value initialization afterwards.
7448 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
7450 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
7452 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
7455 * common/config/aarch64/aarch64-common.c
7456 (aarch_option_optimization_table[]): Switch
7457 off fomit-frame-pointer
7459 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
7460 Chung-Ju Wu <jasonwucj@gmail.com>
7462 * config/nds32/nds32-multiple.md (load_multiple): Disallow
7464 (store_multiple): Ditto.
7466 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
7468 * config.gcc: Add --with-cpu support for nds32 target.
7469 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
7470 * config/nds32/nds32.opt: Add -mcpu= option.
7472 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
7474 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
7475 isel=yes): Warn for these deprecated options.
7477 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
7479 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
7480 ISA_2_5_MASKS_EMBEDDED.
7482 2018-02-23 Jakub Jelinek <jakub@redhat.com>
7484 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
7485 p->max as pointers rather than using iterative_hash_expr.
7487 2018-02-23 Carl Love <cel@us.ibm.com>
7489 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
7490 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
7492 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
7493 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
7494 P8V_BUILTIN_VEC_VUNSIGNED2.
7496 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
7499 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
7500 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
7502 * lra-constraints.c (curr_insn_transform): Set up
7503 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
7505 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
7506 * lra-eliminations.c (spill_pseudos): Ditto.
7507 (process_insn_for_elimination): Ditto.
7508 * lra-lives.c (reg_early_clobber_p): Use the new macros.
7509 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
7510 LRA_NON_CLOBBERED_ALT.
7512 2018-02-22 Martin Sebor <msebor@redhat.com>
7514 PR tree-optimization/84480
7515 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
7516 to maybe_diag_stxncpy_trunc. Call it.
7517 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
7518 from gimple_fold_builtin_strcpy. Print inlining stack.
7519 (handle_builtin_stxncpy): Print inlining stack.
7520 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
7522 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
7525 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
7526 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
7527 and -fcheck-pointer-bounds are used together.
7528 (indirect_thunk_prefix): New enum.
7529 (indirect_thunk_need_prefix): New function.
7530 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
7531 "_nt" instead of "_bnd" for NOTRACK prefix.
7532 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
7533 (output_indirect_thunk_function): Likewise.
7535 (ix86_code_end): Update output_indirect_thunk_function calls.
7536 (ix86_output_indirect_branch_via_reg): Replace
7537 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
7538 (ix86_output_indirect_branch_via_push): Likewise.
7539 (ix86_output_function_return): Likewise.
7540 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
7541 incompatible with -fcf-protection=branch and
7542 -fcheck-pointer-bounds.
7544 2018-02-22 Steve Ellcey <sellcey@cavium.com>
7547 * config/aarch64/aarch64.c (aarch64_print_address_internal):
7548 Change gcc_assert call to output_operand_lossage.
7550 2018-02-22 Steve Ellcey <sellcey@cavium.com>
7552 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
7554 2018-02-22 DJ Delorie <dj@redhat.com>
7555 Sebastian Perta <sebastian.perta@renesas.com>
7556 Oleg Endo <olegendo@gcc.gnu.org>
7558 * config/rx/rx.c (rx_rtx_costs): New function.
7559 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
7561 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
7563 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
7565 2018-02-22 Martin Liska <mliska@suse.cz>
7568 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
7569 Add "native" as a possible value.
7571 2018-02-22 Martin Liska <mliska@suse.cz>
7574 * config/i386/i386.c (ix86_option_override_internal):
7575 Add "native" as a possible value for -march and -mtune.
7577 2018-02-22 Jakub Jelinek <jakub@redhat.com>
7580 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
7581 to all type variants.
7583 PR tree-optimization/84503
7584 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
7585 width as info->bitpos + info->bitsize - start.
7586 (merged_store_group::merge_overlapping): Simplify width computation.
7587 (check_no_overlap): New function.
7588 (imm_store_chain_info::try_coalesce_bswap): Compute expected
7589 start + width and last_order of the group, fail if check_no_overlap
7591 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
7592 to group if check_no_overlap fails.
7594 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
7596 * config/rs6000/altivec.md: Delete contraint arguments to
7597 define_expand, define_split, and define_peephole2, and in
7598 define_insn_and_split if always unused.
7599 * config/rs6000/darwin.md: Ditto.
7600 * config/rs6000/dfp.md: Ditto.
7601 * config/rs6000/rs6000.md: Ditto.
7602 * config/rs6000/sync.md: Ditto.
7603 * config/rs6000/vector.md: Ditto.
7604 * config/rs6000/vsx.md: Ditto.
7606 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
7608 * config/rs6000/altivec.md: Write output control strings as braced
7609 blocks instead of double-quoted strings.
7610 * config/rs6000/darwin.md: Ditto.
7611 * config/rs6000/rs6000.md: Ditto.
7612 * config/rs6000/vector.md: Ditto.
7613 * config/rs6000/vsx.md: Ditto.
7615 2018-02-21 Jason Merrill <jason@redhat.com>
7617 PR c++/84314 - ICE with templates and fastcall attribute.
7618 * attribs.c (build_type_attribute_qual_variant): Remove assert.
7620 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
7622 * ipa-cp.c (determine_versionability): Fix comment typos.
7624 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
7627 * ipa-cp.c (determine_versionability): Do not version functions caling
7630 2018-02-21 Martin Liska <mliska@suse.cz>
7633 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
7634 Add "native" as a possible value.
7635 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
7636 the macro when native cpu detection is available.
7638 2018-02-21 Martin Liska <mliska@suse.cz>
7641 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
7642 Add "native" as a possible value.
7643 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
7644 when native cpu detection is available.
7646 2018-02-21 Jakub Jelinek <jakub@redhat.com>
7647 Martin Sebor <msebor@redhat.com>
7649 PR tree-optimization/84478
7650 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
7652 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
7653 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
7654 support which is conservatively correct, for 2 only stay conservative
7655 for maxlen. Formatting and comment capitalization fixes. Add STRICT
7656 argument to the 2 argument get_range_strlen, adjust 6 arg
7657 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
7659 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
7660 (gimple_fold_builtin_strlen): Pass true as last argument to
7663 2018-02-20 Martin Sebor <msebor@redhat.com>
7666 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
7667 (builtin_memref::set_base_and_offset): Same. Handle inner references.
7668 (builtin_memref::builtin_memref): Factor out parts into
7669 set_base_and_offset and call it.
7671 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
7674 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
7675 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
7676 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
7677 search at the associated MODE_INT.
7679 2018-02-20 Jeff Law <law@redhat.com>
7682 PR tree-optimization/81592
7684 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
7685 for range data rather than using global data.
7686 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
7687 range data rather than using global data.
7688 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
7689 pass it to children as needed.
7690 (struct directive::fmtresult): Similarly.
7691 (struct directive::set_width): Similarly.
7692 (struct directive::set_precision): Similarly.
7693 (format_integer, format_directive, parse_directive): Similarly.
7694 (format_none): Accept unnamed vr_values parameter.
7695 (format_percent, format_floating, format_character): Similarly.
7696 (format_string, format_plain): Similarly.
7697 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
7698 the EVRP range analyzer for range data rather than using global data.
7699 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
7700 gimple-ssa-evrp-analyze.h
7701 (class sprintf_dom_walker): Add after_dom_children member function.
7702 Add evrp_range_analyzer member.
7703 (sprintf_dom_walker::before_dom_children): Call into the EVRP
7704 range analyzer as needed.
7705 (sprintf_dom_walker::after_dom_children): New member function.
7706 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
7708 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
7709 (evrp_range_analyzer::pop_to_marker): Likewise.
7711 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
7713 PR tree-optimization/84419
7714 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
7715 with the required type if its current type is compatible but
7718 2018-02-20 Jakub Jelinek <jakub@redhat.com>
7721 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
7722 after vectorization.
7724 2018-02-20 Martin Liska <mliska@suse.cz>
7727 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
7728 possible values if we don't have a hint.
7730 2018-02-20 Martin Liska <mliska@suse.cz>
7734 * final.c (shorten_branches): Build align_tab array with one
7736 * opts.c (finish_options): Add alignment option limit check.
7737 (MAX_CODE_ALIGN): Likewise.
7738 (MAX_CODE_ALIGN_VALUE): Likewise.
7739 * doc/invoke.texi: Document maximum allowed option value for
7740 all -falign-* options.
7742 2018-02-19 Jakub Jelinek <jakub@redhat.com>
7745 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
7746 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
7747 * var-tracking.c (emit_note_insn_var_location): Remove all references
7748 to NOTE_INSN_CALL_ARG_LOCATION.
7749 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
7750 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
7751 Use copy_rtx_if_shared.
7752 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
7753 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
7754 (dwarf2out_var_location): Remove handling of
7755 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
7757 * final.c (final_scan_insn): Remove all references to
7758 NOTE_INSN_CALL_ARG_LOCATION.
7759 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
7760 before dumping final insns.
7761 * except.c (emit_note_eh_region_end): Remove all references to
7762 NOTE_INSN_CALL_ARG_LOCATION.
7763 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
7764 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
7765 * config/arc/arc.c (hwloop_optimize): Likewise.
7766 * config/arm/arm.c (create_fix_barrier): Likewise.
7767 * config/s390/s390.c (s390_chunkify_start): Likewise.
7768 * config/sh/sh.c (find_barrier): Likewise.
7769 * config/i386/i386.c (rest_of_insert_endbranch,
7770 ix86_seh_fixup_eh_fallthru): Likewise.
7771 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
7772 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
7773 * config/frv/frv.c (frv_function_prologue): Likewise.
7774 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
7776 (note_outside_basic_block_p): Remove all references to
7777 NOTE_INSN_CALL_ARG_LOCATION.
7778 * gengtype.c (adjust_field_rtx_def): Likewise.
7779 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
7781 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
7782 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
7785 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
7788 PR tree-optimization/84452
7789 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
7790 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
7793 2018-02-19 Martin Liska <mliska@suse.cz>
7796 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
7798 2018-02-19 Martin Liska <mliska@suse.cz>
7799 Richard Sandiford <richard.sandiford@linaro.org>
7801 PR tree-optimization/82491
7802 * gimple-fold.c (get_base_constructor): Make earlier bail out
7805 2018-02-19 Carl Love <cel@us.ibm.com>
7807 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
7808 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
7810 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
7811 P8V_BUILTIN_VEC_NEG.
7813 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
7815 * config/rl78/rl78.md (movdf): New define expand.
7817 2018-02-19 Martin Liska <mliska@suse.cz>
7820 * doc/invoke.texi: Fix typo.
7821 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
7823 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
7825 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
7826 handle rs6000_single_float and rs6000_double_float specially for
7829 2018-02-16 Jeff Law <law@redhat.com>
7831 * config/rx/rx.c (add_pop_cfi_notes): New function.;
7834 2018-02-16 Jakub Jelinek <jakub@redhat.com>
7837 * ipa-inline.c (inline_small_functions): Fix a typo.
7839 2018-02-16 Nathan Sidwell <nathan@acm.org>
7841 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
7843 2018-02-16 Carl Love <cel@us.ibm.com>
7845 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
7846 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
7847 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
7848 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
7849 expansion to P8V_BUILTIN_VEC_FLOAT2.
7851 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
7853 PR rtl-optimization/70023
7854 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
7855 src_regno into account.
7857 2018-02-16 Carl Love <cel@us.ibm.com>
7859 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
7860 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
7861 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
7862 * config/rs6000/rs6000.c: Remove case statements for
7863 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
7864 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
7865 and P9V_BUILTIN_VEC_VINSERT4B.
7866 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
7867 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
7868 * config/rs6000/vsx.md:
7869 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
7872 2018-02-16 Carl Love <cel@us.ibm.com>
7874 * config/rs6000/altivec.h: Add builtin names vec_extract4b
7876 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
7878 * config/rs6000/rs6000-c.c: Add the definitions for
7879 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
7880 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
7881 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
7882 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
7883 definition for insert4b and define insn *insert3b_internal.
7884 * doc/extend.texi: Add documentation for vec_extract4b.
7886 2018-02-16 Nathan Sidwell <nathan@acm.org>
7888 * doc/extend.texi (Backwards Compatibility): Mention friend
7889 injection. Note for-scope is deprecated.
7890 * doc/invoke.texi (-ffriend-injection): Deprecate.
7892 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
7894 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
7895 that moved to I2, also allow destinations that are a paradoxical
7896 subreg (instead of a normal reg).
7898 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
7901 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
7904 2018-02-16 Richard Biener <rguenther@suse.de>
7906 PR tree-optimization/84037
7907 PR tree-optimization/84016
7909 * config/i386/i386.c (ix86_builtin_vectorization_cost):
7910 Adjust vec_construct for the fact we need additional higher latency
7911 128bit inserts for AVX256 and AVX512 vector builds.
7912 (ix86_add_stmt_cost): Scale vector construction cost for
7915 2018-02-16 Richard Biener <rguenther@suse.de>
7917 PR tree-optimization/84417
7918 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
7919 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
7920 (non_rewritable_lvalue_p): Likewise, use poly-ints.
7922 2018-02-16 Martin Liska <mliska@suse.cz>
7925 * internal-fn.def (ASAN_CHECK): Set proper flags.
7926 (ASAN_MARK): Likewise.
7928 2018-02-16 Julia Koval <julia.koval@intel.com>
7930 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
7931 from PTA_CANNONLAKE.
7933 2018-02-16 Jakub Jelinek <jakub@redhat.com>
7936 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
7937 Use ++iter rather than iter++ for std::list iterators.
7938 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
7939 defer deleting them until all nodes in the forest are processed. Do
7940 free even leaf nodes. Change to_process into auto_vec.
7943 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
7944 * vec.h (vec_default_construct): Use memset instead of placement new
7945 if BROKEN_VALUE_INITIALIZATION is defined.
7946 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
7947 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
7950 PR rtl-optimization/83723
7951 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
7952 * lra.c (lra_substitute_pseudo): Likewise. If true, use
7953 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
7955 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
7957 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
7959 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
7961 PR rtl-optimization/81443
7962 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
7963 from inner REGs to paradoxical SUBREGs.
7965 2018-02-16 Richard Biener <rguenther@suse.de>
7967 PR tree-optimization/84399
7968 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
7969 For operands we can analyze at their definition make sure we can
7970 analyze them at each use as well.
7972 2018-02-16 Richard Biener <rguenther@suse.de>
7974 PR tree-optimization/84190
7975 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
7976 volatile accesses if the decl isn't volatile.
7978 2018-02-15 Jason Merrill <jason@redhat.com>
7980 PR c++/84314 - ICE with templates and fastcall attribute.
7981 * attribs.c (build_type_attribute_qual_variant): Don't clobber
7982 TYPE_CANONICAL on an existing type.
7984 2018-02-15 Jakub Jelinek <jakub@redhat.com>
7986 PR tree-optimization/84383
7987 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
7988 dstoff nor call operand_equal_p if dstbase is NULL.
7990 PR tree-optimization/84334
7991 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
7992 also a CONSTANT_CLASS_P, punt.
7994 2018-02-14 Jim Wilson <jimw@sifive.com>
7996 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
7997 first SMALL_OPERAND check. New local min_second_step. Move assert
7998 to where locals are set. Add TARGET_RVC support.
7999 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
8001 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
8003 * doc/invoke.texi: Correct -Wformat-overflow code sample.
8005 2018-02-14 Martin Sebor <msebor@redhat.com>
8007 PR tree-optimization/83698
8008 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
8009 arrays constrain the offset range to their bounds.
8010 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
8011 (builtin_access::overlap): Avoid setting the size of overlap if it's
8013 (maybe_diag_overlap): Also consider arrays when deciding what values
8014 of offsets to include in diagnostics.
8016 2018-02-14 Martin Sebor <msebor@redhat.com>
8019 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
8020 that correspond to the kind of a declaration.
8022 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
8025 * config/pa/pa.md: Load address of PIC label using the linkage table
8026 if the label is nonlocal.
8028 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
8030 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
8031 warning message if user requests -maltivec=be.
8032 * doc/invoke.texi: Document deprecation of -maltivec=be.
8034 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
8037 * config/rs6000/rs6000-c.c: Update definitions for
8038 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
8039 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
8041 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
8044 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
8045 add _get_ssp intrinsics. Remove argument from
8046 __builtin_ia32_rdssp[d|q].
8047 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
8048 * config/i386/i386-builtin.def: Remove argument from
8049 __builtin_ia32_rdssp[d|q].
8050 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
8051 ix86_expand_special_args_builtin for _rdssp[d|q].
8052 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
8053 Clear register before usage.
8054 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
8055 Add documentation for new _get_ssp and _inc_ssp intrinsics.
8057 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
8059 PR tree-optimization/84357
8060 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
8061 operand 1 of an ARRAY_REF too.
8063 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
8066 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
8067 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
8069 (set_of_reg): New struct.
8070 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
8071 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
8072 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
8074 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
8075 Split into bitclr, bitset, bitinvert patterns if appropriate.
8076 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
8077 use rx_fuse_in_memory_bitop.
8078 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
8079 to named insn, correct maximum insn length.
8081 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
8084 * machmode.def: Define a complex mode for PARTIAL_INT.
8085 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
8087 * doc/rtl.texi: Document CSPImode.
8088 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
8090 (msp430_hard_regno_nregs_with_padding): Likewise.
8092 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
8095 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
8097 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
8099 PR rtl-optimization/84169
8100 * combine.c (try_combine): New variable split_i2i3. Set it to true if
8101 we generated a parallel as new i3 and we split that to new i2 and i3
8102 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
8103 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
8104 those to i2, not i1. Partially rewrite this scan code.
8106 2018-02-13 Jakub Jelinek <jakub@redhat.com>
8109 * stor-layout.c (place_field): For variable length fields, adjust
8110 offset_align afterwards not just based on the field's alignment,
8111 but also on the size.
8114 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
8115 of exps and logs in the use_exp2 case.
8117 2018-02-13 Jeff Law <law@redhat.com>
8119 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
8122 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
8125 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
8129 * common.opt (gas-loc-support, gas-locview-support): New.
8130 (ginline-points, ginternal-reset-location-views): New.
8131 * doc/invoke.texi: Document them. Use @itemx where intended.
8132 (gvariable-location-views): Adjust.
8133 * target.def (reset_location_view): New.
8134 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
8135 (TARGET_RESET_LOCATION_VIEW): New.
8136 * doc/tm.texi: Rebuilt.
8137 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
8138 (dwarf2out_default_as_locview_support): New.
8139 (output_asm_line_debug_info): Use option variables.
8140 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
8141 (output_loc_list): Likewise.
8142 (add_high_low_attributes): Check option variables.
8143 Don't output entry view attribute in strict mode.
8144 (gen_inlined_subroutine_die): Check option variables.
8145 (dwarf2out_inline_entry): Likewise.
8146 (init_sections_and_labels): Likewise.
8147 (dwarf2out_early_finish): Likewise.
8148 (maybe_reset_location_view): New, from...
8149 (dwarf2out_var_location): ... here. Call it.
8150 * debug.h (dwarf2out_default_as_loc_support): Declare.
8151 (dwarf2out_default_as_locview_support): Declare.
8152 * hooks.c (hook_int_rtx_insn_0): New.
8153 * hooks.h (hook_int_rtx_insn_0): Declare.
8154 * toplev.c (process_options): Take -gas-loc-support and
8155 -gas-locview-support from dwarf2out. Enable
8156 -gvariable-location-views by default only with locview
8157 assembler support. Enable -ginternal-reset-location-views by
8158 default only if the target defines the corresponding hook.
8159 Enable -ginline-points by default if location views are
8160 enabled; force it disabled if statement frontiers are
8162 * tree-inline.c (expand_call_inline): Check option variables.
8163 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
8165 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
8167 PR tree-optimization/84321
8168 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
8169 handling. Also check whether the anti-range contains any values
8170 that satisfy the mask; switch to a VR_RANGE if not.
8172 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
8175 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
8177 2018-02-13 Martin Jambor <mjambor@suse.cz>
8180 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
8181 of call statements, also set location of a load to a temporary.
8183 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
8185 * config/rl78/rl78.c (add_vector_labels): New function.
8186 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
8187 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
8188 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
8189 which checks that no arguments are passed.
8190 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
8191 * doc/extend.texi: Documentation for the new attribute.
8193 2018-02-13 Andreas Schwab <schwab@suse.de>
8195 * config/riscv/linux.h (CPP_SPEC): Define.
8197 2018-02-13 Jakub Jelinek <jakub@redhat.com>
8200 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
8201 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
8202 OPTION_MASK_ISA_AES as first argument to def_builtin_const
8203 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
8204 instead of OPTION_MASK_ISA_PCLMUL as first argument to
8205 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
8206 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
8207 temporarily for AES and PCLMUL builtins.
8209 PR tree-optimization/84339
8210 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
8211 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
8215 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
8216 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
8217 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
8219 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
8221 * omp-simd-clone.h: New file.
8222 * omp-simd-clone.c: Include omp-simd-clone.h.
8223 (expand_simd_clones): No longer static.
8224 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
8225 cgraph.h and omp-simd-clone.h.
8226 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
8227 (vect_recog_widen_shift_pattern): Formatting fix.
8228 (vect_pattern_recog_1): Don't check optab for calls.
8231 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
8232 operands[2] into a REG before using gen_lowpart on it.
8234 2018-02-12 Jeff Law <law@redhat.com>
8237 * config/sh/sh.c (find_barrier): Consider a sibling call
8240 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
8241 successfully back substituting a reg.
8243 2018-02-12 Richard Biener <rguenther@suse.de>
8245 PR tree-optimization/84037
8246 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
8247 parameter, move visited init to caller.
8248 (vect_slp_analyze_operations): Separate cost from validity
8249 check, initialize visited once for all instances.
8250 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
8252 * tree-vect-stmts.c (vect_model_simple_cost): Make early
8254 (vect_model_promotion_demotion_cost): Likewise.
8255 (vectorizable_bswap): Guard cost modeling with !slp_node
8256 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
8258 (vectorizable_call): Likewise.
8259 (vectorizable_conversion): Likewise.
8260 (vectorizable_assignment): Likewise.
8261 (vectorizable_shift): Likewise.
8262 (vectorizable_operation): Likewise.
8263 (vectorizable_store): Likewise.
8264 (vectorizable_load): Likewise.
8265 (vectorizable_condition): Likewise.
8266 (vectorizable_comparison): Likewise.
8268 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
8271 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
8272 (ASAN_MARK): Fix fnspec to account for return value, change pointer
8273 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
8275 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
8278 * params.def (inline-min-speedup): Increase from 8 to 15.
8279 (max-inline-insns-auto): Decrease from 40 to 30.
8280 * ipa-split.c (consider_split): Add some buffer for function to
8281 be considered inlining candidate.
8282 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
8285 2018-02-12 Richard Biener <rguenther@suse.de>
8287 PR tree-optimization/84037
8288 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
8289 matched stmts if we cannot swap the non-matched ones.
8291 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
8293 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
8294 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
8295 _mm_maskz_scalef_round_ss): New intrinsics.
8296 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
8297 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
8298 __builtin_ia32_scalefss_round): Remove.
8299 (__builtin_ia32_scalefsd_mask_round,
8300 __builtin_ia32_scalefss_mask_round): New intrinsics.
8301 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
8302 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
8303 ((match_operand:VF_128 2 "<round_nimm_predicate>"
8304 "<round_constraint>")): Changed to ...
8305 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
8306 "<round_scalar_constraint>")): ... this.
8307 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
8308 %0, %1, %2<round_op3>}"): Changed to ...
8309 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
8310 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
8311 %2<round_scalar_mask_op3>}"): ... this.
8312 * config/i386/subst.md (round_scalar_nimm_predicate): New.
8314 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
8316 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
8317 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
8318 (_mm_maskz_sqrt_round_ss): New intrinsics.
8319 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
8320 (__builtin_ia32_sqrtsd_mask_round)
8321 (__builtin_ia32_sqrtss_mask_round): New builtins.
8322 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
8323 (__builtin_ia32_sqrtss_round): Remove.
8324 (__builtin_ia32_sqrtsd_mask_round)
8325 (__builtin_ia32_sqrtss_mask_round): New builtins.
8326 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
8327 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
8328 ((match_operand:VF_128 1 "vector_operand"
8329 "xBm,<round_constraint>")): Changed to ...
8330 ((match_operand:VF_128 1 "vector_operand"
8331 "xBm,<round_scalar_constraint>")): ... this.
8332 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
8333 %0, %2, %<iptr>1<round_op3>}): Changed to ...
8334 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
8335 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
8336 %<iptr>1<round_scalar_mask_op3>}): ... this.
8337 ((set_attr "prefix" "<round_prefix>")): Changed to ...
8338 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
8340 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
8343 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
8344 Cast vec_cmpeq result to correct type.
8345 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
8346 Cast vec_cmpgt result to correct type.
8348 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
8350 * final.c (final_scan_insn_1): Renamed from...
8351 (final_scan_insn): ... this. New wrapper, to recover
8352 seen from the outermost call in recursive ones.
8353 * config/sparc/sparc.c (output_return): Drop seen from call.
8354 (output_sibcall): Likewise.
8355 * config/visium/visium.c (output_branch): Likewise.
8357 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
8359 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
8362 2018-02-10 Alan Modra <amodra@gmail.com>
8365 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
8366 Specify LR as an input.
8368 2018-02-10 Jakub Jelinek <jakub@redhat.com>
8371 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
8372 remove_member_access_dummy_vars): New functions.
8373 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
8374 lower_omp_1, execute_lower_omp): Use them.
8376 PR rtl-optimization/84308
8377 * shrink-wrap.c (spread_components): Release todo vector.
8379 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
8381 PR rtl-optimization/57193
8382 * ira-color.c (struct allocno_color_data): Add member
8383 conflict_allocno_hard_prefs.
8384 (update_conflict_allocno_hard_prefs): New.
8385 (bucket_allocno_compare_func): Add a preference based on
8386 conflict_allocno_hard_prefs.
8387 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
8388 (color_allocnos): Remove a dead code. Initiate
8389 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
8391 2018-02-09 Jakub Jelinek <jakub@redhat.com>
8394 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
8395 constraint from =wa to wa. Avoid a subreg on the output operand,
8396 instead use a pseudo and subreg it in a move.
8397 (p9_xxbrd_<mode>): Changed to ...
8398 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
8399 (p9_xxbrd_v2df): New expander.
8400 (p9_xxbrw_<mode>): Changed to ...
8401 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
8402 (p9_xxbrw_v4sf): New expander.
8404 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
8406 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
8408 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
8411 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
8412 multiply in 32-bit mode.
8413 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
8414 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
8417 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
8419 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
8420 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
8421 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
8422 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
8424 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
8427 * dwarf2out.c (is_trivial_indirect_ref): New function.
8428 (dwarf2out_late_global_decl): Do not generate a location
8429 attribute for variables that have a non-trivial DECL_VALUE_EXPR
8430 and that are not defined in the current unit.
8432 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
8434 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
8435 instead of a libcall for UNORDERED.
8437 2018-02-09 Tamar Christina <tamar.christina@arm.com>
8440 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
8441 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
8443 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8446 * config/s390/s390.c (s390_set_current_function): Invoke
8447 s390_indirect_branch_settings also if fndecl didn't change.
8449 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
8451 * config/rs6000/rs6000.md (blockage): Set length to zero.
8453 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
8455 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
8457 2018-02-09 Jakub Jelinek <jakub@redhat.com>
8460 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
8461 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
8465 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
8466 PARALLEL incoming that failed vt_get_decl_and_offset check.
8469 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
8470 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
8472 (get_variable_section): For decls in named .bss* sections pass true as
8473 second argument to bss_initializer_p.
8475 2018-02-09 Marek Polacek <polacek@redhat.com>
8476 Jakub Jelinek <jakub@redhat.com>
8479 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
8480 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
8481 Sync some changes from cxx_fold_indirect_ref.
8483 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
8485 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
8487 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
8488 (BLOCK_INLINE_ENTRY_LABEL): New.
8489 (dwarf2out_var_location): Disregard inline entry markers.
8490 (inline_entry_data): New struct.
8491 (inline_entry_data_hasher): New hashtable type.
8492 (inline_entry_data_hasher::hash): New.
8493 (inline_entry_data_hasher::equal): New.
8494 (inline_entry_data_table): New variable.
8495 (add_high_low_attributes): Add DW_AT_entry_pc and
8496 DW_AT_GNU_entry_view attributes if a pending entry is found
8497 in inline_entry_data_table. Add old entry_pc attribute only
8498 if debug nonbinding markers are disabled.
8499 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
8500 markers are enabled.
8501 (block_within_block_p, dwarf2out_inline_entry): New.
8502 (dwarf2out_finish): Check that no entries remained in
8503 inline_entry_data_table.
8504 * final.c (reemit_insn_block_notes): Handle inline entry notes.
8505 (final_scan_insn, notice_source_line): Likewise.
8506 (rest_of_clean_state): Skip inline entry markers.
8507 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
8509 * gimple.c (gimple_build_debug_inline_entry): New.
8510 * gimple.h (enum gimple_debug_subcode): Add
8511 GIMPLE_DEBUG_INLINE_ENTRY.
8512 (gimple_build_debug_inline_entry): Declare.
8513 (gimple_debug_inline_entry_p): New.
8514 (gimple_debug_nonbind_marker_p): Adjust.
8515 * insn-notes.def (INLINE_ENTRY): New.
8516 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
8517 inline entry marker notes.
8518 (print_insn): Likewise.
8519 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
8520 (INSN_DEBUG_MARKER_KIND): Likewise.
8521 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
8522 * tree-inline.c (expand_call_inline): Build and insert
8523 debug_inline_entry stmt.
8524 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
8525 inline entry blocks early, if nonbind markers are enabled.
8526 (dump_scope_block): Dump fragment info.
8527 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
8528 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
8529 (gimple_build_debug_inline_entry): New.
8530 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
8531 Enable/disable inline entry points too.
8532 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
8533 (DEBUG_INSN): Describe inline entry markers.
8535 * common.opt (gvariable-location-views): New.
8536 (gvariable-location-views=incompat5): New.
8537 * config.in: Rebuilt.
8538 * configure: Rebuilt.
8539 * configure.ac: Test assembler for view support.
8540 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
8541 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
8542 * dwarf2out.c (var_loc_view): New typedef.
8543 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
8544 (dwarf2out_locviews_in_attribute): New.
8545 (dwarf2out_locviews_in_loclist): New.
8546 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
8547 (enum dw_line_info_opcode): Add LI_adv_address.
8548 (struct dw_line_info_table): Add view.
8549 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
8550 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
8551 (zero_view_p): New variable.
8552 (ZERO_VIEW_P): New macro.
8553 (output_asm_line_debug_info): New.
8554 (struct var_loc_node): Add view.
8555 (add_AT_view_list, AT_loc_list): New.
8556 (add_var_loc_to_decl): Add view param. Test it against last.
8557 (new_loc_list): Add view params. Record them.
8558 (AT_loc_list_ptr): Handle loc and view lists.
8559 (view_list_to_loc_list_val_node): New.
8560 (print_dw_val): Handle dw_val_class_view_list.
8561 (size_of_die): Likewise.
8562 (value_format): Likewise.
8563 (loc_list_has_views): New.
8564 (gen_llsym): Set vl_symbol too.
8565 (maybe_gen_llsym, skip_loc_list_entry): New.
8566 (dwarf2out_maybe_output_loclist_view_pair): New.
8567 (output_loc_list): Output view list or entries too.
8568 (output_view_list_offset): New.
8569 (output_die): Handle dw_val_class_view_list.
8570 (output_dwarf_version): New.
8571 (output_compilation_unit_header): Use it.
8572 (output_skeleton_debug_sections): Likewise.
8573 (output_rnglists, output_line_info): Likewise.
8574 (output_pubnames, output_aranges): Update version comments.
8575 (output_one_line_info_table): Output view numbers in asm comments.
8576 (dw_loc_list): Determine current endview, pass it to new_loc_list.
8577 Call maybe_gen_llsym.
8578 (loc_list_from_tree_1): Adjust.
8579 (add_AT_location_description): Create view list attribute if
8580 needed, check it's absent otherwise.
8581 (convert_cfa_to_fb_loc_list): Adjust.
8582 (maybe_emit_file): Call output_asm_line_debug_info for test.
8583 (dwarf2out_var_location): Reset views as needed. Precompute
8584 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
8585 attribute. Set view.
8586 (new_line_info_table): Reset next view.
8587 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
8588 (dwarf2out_source_line): Likewise. Output view resets and labels to
8589 the assembler, or select appropriate line info opcodes.
8590 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
8591 (optimize_string_length): Catch it. Adjust.
8592 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
8593 dw_val_class_view_list, and remove it if no longer needed.
8594 (hash_loc_list): Hash view numbers.
8595 (loc_list_hasher::equal): Compare them.
8596 (optimize_location_lists): Check whether a view list symbol is
8597 needed, and whether the locview attribute is present, and
8598 whether they match. Remove the locview attribute if no longer
8600 (index_location_lists): Call skip_loc_list_entry for test.
8601 (dwarf2out_finish): Call output_asm_line_debug_info for test.
8602 Use output_dwarf_version.
8603 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
8604 (struct dw_val_node): Add val_view_list.
8605 * final.c (SEEN_NEXT_VIEW): New.
8606 (set_next_view_needed): New.
8607 (clear_next_view_needed): New.
8608 (maybe_output_next_view): New.
8609 (final_start_function): Rename to...
8610 (final_start_function_1): ... this. Take pointer to FIRST,
8611 add SEEN parameter. Emit param bindings in the initial view.
8612 (final_start_function): Reintroduce SEEN-less interface.
8613 (final): Rename to...
8614 (final_1): ... this. Take SEEN parameter. Output final pending
8615 next view at the end.
8616 (final): Reintroduce seen-less interface.
8617 (final_scan_insn): Output pending next view before switching
8618 sections or ending a block. Mark the next view as needed when
8619 outputting variable locations. Notify debug backend of section
8620 changes, and of location view changes.
8621 (rest_of_handle_final): Adjust.
8622 * toplev.c (process_options): Autodetect value for debug variable
8623 location views option. Warn on incompat5 without -gdwarf-5.
8624 * doc/invoke.texi (gvariable-location-views): New.
8625 (gvariable-location-views=incompat5): New.
8626 (gno-variable-location-views): New.
8628 2018-02-08 David Malcolm <dmalcolm@redhat.com>
8630 PR tree-optimization/84136
8631 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
8632 that the result of find_edge is non-NULL.
8634 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
8637 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
8638 storing integer register in SImode. Fix cost of 256 and 512
8639 byte aligned SSE register store.
8641 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
8643 * config/i386/i386.c (ix86_multiplication_cost): Fix
8644 multiplication cost for TARGET_AVX512DQ.
8646 2018-02-08 Marek Polacek <polacek@redhat.com>
8648 PR tree-optimization/84238
8649 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
8652 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8654 PR tree-optimization/84265
8655 * tree-vect-stmts.c (vectorizable_store): Don't treat
8656 VMAT_CONTIGUOUS accesses as grouped.
8657 (vectorizable_load): Likewise.
8659 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8661 PR tree-optimization/81635
8662 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
8663 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
8664 (test_round_for_mask): New functions.
8665 (wide_int_cc_tests): Call test_round_for_mask.
8666 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
8667 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
8668 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
8669 range returned by get_range_info.
8671 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
8674 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
8675 * symtab.c: Include builtins.h
8676 (symtab_node::output_to_lto_symbol_table_p): Move here
8677 from lto-streamer-out.c:output_symbol_p.
8678 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
8679 (output_symbol_p): Move all logic to symtab.c
8680 (produce_symtab): Update.
8682 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8684 * config/s390/s390-opts.h (enum indirect_branch): Define.
8685 * config/s390/s390-protos.h (s390_return_addr_from_memory)
8686 (s390_indirect_branch_via_thunk)
8687 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
8688 (enum s390_indirect_branch_type): Define.
8689 * config/s390/s390.c (struct s390_frame_layout, struct
8690 machine_function): Remove.
8691 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
8692 (indirect_branch_table_label_no, indirect_branch_table_name):
8694 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
8695 (enum s390_indirect_branch_option): Define.
8696 (s390_return_addr_from_memory): New function.
8697 (s390_handle_string_attribute): New function.
8698 (s390_attribute_table): Add new attribute handler.
8699 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
8700 (s390_indirect_branch_via_thunk): New function.
8701 (s390_indirect_branch_via_inline_thunk): New function.
8702 (s390_function_ok_for_sibcall): When jumping via thunk disallow
8703 sibling call optimization for non z10 compiles.
8704 (s390_emit_call): Force indirect branch target to be a single
8705 register. Add r1 clobber for non-z10 compiles.
8706 (s390_emit_epilogue): Emit return jump via return_use expander.
8707 (s390_reorg): Handle JUMP_INSNs as execute targets.
8708 (s390_option_override_internal): Perform validity checks for the
8709 new command line options.
8710 (s390_indirect_branch_attrvalue): New function.
8711 (s390_indirect_branch_settings): New function.
8712 (s390_set_current_function): Invoke s390_indirect_branch_settings.
8713 (s390_output_indirect_thunk_function): New function.
8714 (s390_code_end): Implement target hook.
8715 (s390_case_values_threshold): Implement target hook.
8716 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
8718 * config/s390/s390.h (struct s390_frame_layout)
8719 (struct machine_function): Move here from s390.c.
8720 (TARGET_INDIRECT_BRANCH_NOBP_RET)
8721 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
8722 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
8723 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
8724 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
8725 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
8726 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
8727 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
8728 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
8729 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
8730 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
8731 (mnemonic attribute): Add values which aren't recognized
8733 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
8734 pattern for branch conversion. Fix mnemonic attribute.
8735 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
8736 indirect branch via thunk if requested.
8737 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
8738 ("*indirect_jump"): Disable for branch conversion using out of
8740 ("indirect_jump_via_thunk<mode>_z10")
8741 ("indirect_jump_via_thunk<mode>")
8742 ("indirect_jump_via_inlinethunk<mode>_z10")
8743 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
8744 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
8745 ("casesi_jump_via_inlinethunk<mode>_z10")
8746 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
8747 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
8748 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
8749 ("*indirect2_jump"): Disable for branch conversion.
8750 ("casesi_jump"): Turn into expander and expand patterns for branch
8752 ("return_use"): New expander.
8753 ("*return"): Emit return via thunk and rename it to ...
8754 ("*return<mode>"): ... this one.
8755 * config/s390/s390.opt: Add new options and and enum for the
8758 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8760 * lra-constraints.c (match_reload): Unconditionally use
8761 gen_lowpart_SUBREG, rather than selecting between that
8762 and equivalent gen_rtx_SUBREG code.
8764 2018-02-08 Richard Biener <rguenther@suse.de>
8766 PR tree-optimization/84233
8767 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
8768 changed flag instead of boguously re-using phi_inserted.
8770 2018-02-08 Martin Jambor <mjambor@suse.cz>
8772 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
8773 static local variables.
8775 2018-02-08 Richard Biener <rguenther@suse.de>
8777 PR tree-optimization/84278
8778 * tree-vect-stmts.c (vectorizable_store): When looking for
8779 smaller vector types to perform grouped strided loads/stores
8780 make sure the mode is supported by the target.
8781 (vectorizable_load): Likewise.
8783 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
8785 * config/aarch64/aarch64.c (aarch64_components_for_bb):
8786 Increase LDP/STP opportunities by adding adjacent callee-saves.
8788 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
8790 PR rtl-optimization/84068
8791 PR rtl-optimization/83459
8792 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
8794 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
8796 PR tree-optimization/84224
8797 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
8798 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
8801 2018-02-07 Iain Sandoe <iain@codesourcery.com>
8804 * config/rs6000/altivec.md (*restore_world): Remove LR use.
8805 * config/rs6000/predicates.md (restore_world_operation): Adjust op
8806 count, remove one USE.
8808 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
8810 * doc/install.texi (Configuration): Document the
8811 --with-long-double-format={ibm,ieee} PowerPC configuration
8815 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
8816 Convert from define_expand to be define_insn_and_split. Rework
8817 float/double/_Float128 conversions to QI/HI/SImode to work with
8818 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
8819 conversions to QI/HImode types did a store and then a load to
8820 truncate the value. For conversions to VSX registers, don't split
8821 the insn, instead emit the code directly. Use the code iterator
8822 any_fix to combine signed and unsigned conversions.
8823 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
8824 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
8825 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
8826 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
8827 (fix_<mode>di2_hw): Likewise.
8828 (fixuns_<mode>di2_hw): Likewise.
8829 (fix_<mode>si2_hw): Likewise.
8830 (fixuns_<mode>si2_hw): Likewise.
8831 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
8832 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
8833 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
8834 fix<uns>_trunc<SFDF:mode>si2_p8.
8835 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
8837 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
8838 (fix<uns>_<mode>_mem): Likewise.
8839 (fctiw<u>z_<mode>_mem): Likewise.
8840 (fix<uns>_<mode>_mem): Likewise.
8841 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
8842 the register allocator from doing a direct move to the GPRs to do
8843 a store, and instead use the ISA 3.0 store byte/half-word from
8844 vector register instruction. For IEEE 128-bit floating point,
8845 also optimize stores of 32-bit ints.
8846 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
8848 2018-02-07 Alan Hayward <alan.hayward@arm.com>
8850 * genextract.c (push_pathstr_operand): New function to support
8852 (walk_rtx): Call push_pathstr_operand.
8853 (print_path): Support [a-zA-Z].
8855 2018-02-07 Richard Biener <rguenther@suse.de>
8857 PR tree-optimization/84037
8858 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
8859 (cse_and_gimplify_to_preheader): Declare.
8860 (vect_get_place_in_interleaving_chain): Likewise.
8861 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
8863 (_loop_vec_info::~_loop_vec_info): Delete it.
8864 (cse_and_gimplify_to_preheader): New function.
8865 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
8866 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
8867 (vectorizable_load): Likewise. For grouped stores always base
8868 the IV on the first element.
8869 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
8870 condition before gimplifying.
8872 2018-02-07 Jakub Jelinek <jakub@redhat.com>
8874 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
8875 *DIV_EXPR and *MOD_EXPR.
8877 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
8880 * config/i386/i386.c (ix86_option_override_internal): Mask out
8881 the CF_SET bit when checking -fcf-protection.
8883 2018-02-07 Tom de Vries <tom@codesourcery.com>
8886 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
8889 2018-02-07 Richard Biener <rguenther@suse.de>
8891 PR tree-optimization/84204
8892 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
8895 PR tree-optimization/84205
8896 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
8897 special-case isl_ast_op_zdiv_r.
8899 PR tree-optimization/84223
8900 * graphite-scop-detection.c (gather_bbs::before_dom_children):
8901 Only add conditions from within the region.
8902 (gather_bbs::after_dom_children): Adjust.
8904 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
8907 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
8908 * config/avr/avr.md: Only post-reload split REG-REG moves if
8909 either register is GENERAL_REG_P.
8911 2018-02-07 Jakub Jelinek <jakub@redhat.com>
8913 PR tree-optimization/84235
8914 * tree-ssa-scopedtables.c
8915 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
8916 if the subtraction is performed in floating point type where NaNs are
8917 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
8918 build 1. Formatting fix.
8920 2018-02-06 Jakub Jelinek <jakub@redhat.com>
8923 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
8924 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
8925 and skip it regardless of bb boundaries. Use CALL_P macro,
8926 don't test INSN_P (insn) together with CALL_P or JUMP_P check
8927 unnecessarily, formatting fix.
8929 2018-02-06 Michael Collison <michael.collison@arm.com>
8931 * config/arm/thumb2.md:
8932 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
8933 (*thumb_mov_notscc): Ditto.
8935 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
8938 * config/rs6000/rs6000.md (su code attribute): Use "u" for
8939 unsigned_fix, not "s".
8941 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8943 * configure.ac (gcc_fn_eh_frame_ro): New function.
8944 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
8945 correct .eh_frame permissions.
8946 * configure: Regenerate.
8948 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
8950 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
8953 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8955 * config/rs6000/rs6000.c (rs6000_option_override_internal):
8956 Display warning message for -mno-speculate-indirect-jumps.
8958 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
8960 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
8962 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
8964 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
8966 PR tree-optimization/84225
8967 * tree-eh.c (find_trapping_overflow): Only call
8968 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
8970 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
8973 * config/i386/i386.c: Reimplement the check of possible options
8974 -mibt/-mshstk conbination. Change error messages.
8975 * doc/invoke.texi: Fix a typo: remove extra '='.
8977 2018-02-06 Marek Polacek <polacek@redhat.com>
8979 PR tree-optimization/84228
8980 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
8982 2018-02-06 Tamar Christina <tamar.christina@arm.com>
8985 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
8986 emitted arch directives.
8987 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
8988 __ARM_FEATURE_COPROC before changing architectures.
8990 2018-02-06 Richard Biener <rguenther@suse.de>
8992 * config/i386/i386.c (print_reg): Fix typo.
8993 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
8995 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
8997 * configure: Regenerate.
8999 2018-02-05 Martin Sebor <msebor@redhat.com>
9001 PR tree-optimization/83369
9002 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
9005 2018-02-05 Martin Liska <mliska@suse.cz>
9007 * doc/invoke.texi: Cherry-pick upstream r323995.
9009 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
9011 * ira.c (ira_init_register_move_cost): Adjust comment.
9013 2018-02-05 Martin Liska <mliska@suse.cz>
9015 PR gcov-profile/84137
9016 * doc/gcov.texi: Fix typo in documentation.
9018 2018-02-05 Martin Liska <mliska@suse.cz>
9020 PR gcov-profile/83879
9021 * doc/gcov.texi: Document necessity of --dynamic-list-data when
9022 using dlopen functionality.
9024 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
9026 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
9027 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
9028 _mm_maskz_range_ss, _mm_mask_range_round_ss,
9029 _mm_maskz_range_round_ss): New intrinsics.
9030 (__builtin_ia32_rangesd128_round)
9031 (__builtin_ia32_rangess128_round): Remove.
9032 (__builtin_ia32_rangesd128_mask_round,
9033 __builtin_ia32_rangess128_mask_round): New builtins.
9034 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
9035 __builtin_ia32_rangess128_round): Remove.
9036 (__builtin_ia32_rangesd128_mask_round,
9037 __builtin_ia32_rangess128_mask_round): New builtins.
9038 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
9039 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
9040 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
9041 "<round_saeonly_constraint>")): Changed to ...
9042 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
9043 "<round_saeonly_scalar_constraint>")): ... this.
9044 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
9045 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
9046 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
9047 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
9048 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
9050 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
9052 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
9054 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
9055 Remove all values except native, 8540 and 8548.
9057 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
9059 * config/i386/i386.c (ix86_output_function_return): Pass
9060 INVALID_REGNUM, instead of -1, as invalid register number to
9061 indirect_thunk_name and output_indirect_thunk.
9063 2018-02-02 Julia Koval <julia.koval@intel.com>
9065 * config.gcc: Add -march=icelake.
9066 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
9067 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
9068 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
9069 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
9070 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
9071 (processor_target_table): Add icelake.
9072 (ix86_option_override_internal): Handle new PTAs.
9073 (get_builtin_code_for_version): Handle icelake.
9074 (M_INTEL_COREI7_ICELAKE): New.
9075 (fold_builtin_cpu): Handle icelake.
9076 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
9077 * doc/invoke.texi: Add -march=icelake.
9079 2018-02-02 Julia Koval <julia.koval@intel.com>
9081 * config/i386/i386.c (ix86_option_override_internal): Change flags type
9082 to wide_int_bitmask.
9083 * wide-int-bitmask.h: New.
9085 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
9088 * config/i386/i386.md: Replace Pmode with word_mode in
9089 builtin_setjmp_setup and builtin_longjmp to support x32.
9091 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
9095 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
9097 (rs6000_supported_cpu_names): New static variable.
9098 (linux_cpu_translation_table): Likewise.
9099 (elf_platform) <cpu>: Define new static variable and use it.
9100 Translate kernel AT_PLATFORM name to canonical name if needed.
9101 Error if platform name is unknown.
9103 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
9106 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
9108 2018-02-01 Jeff Law <law@redhat.com>
9111 * config/i386/i386.c (release_scratch_register_on_entry): Add new
9112 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
9113 the scratch if RELEASE_VIA_POP is false.
9114 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
9115 If we have to save a temporary register, decrement SIZE appropriately.
9116 Pass new arguments to release_scratch_register_on_entry.
9117 (ix86_adjust_stack_and_probe): Likewise.
9118 (ix86_emit_probe_stack_range): Pass new arguments to
9119 release_scratch_register_on_entry.
9121 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
9123 PR rtl-optimization/84157
9124 * combine.c (change_zero_ext): Use REG_P predicate in
9125 front of HARD_REGISTER_P predicate.
9127 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
9129 * config/avr/avr.c (avr_option_override): Move disabling of
9130 -fdelete-null-pointer-checks to...
9131 * common/config/avr/avr-common.c (avr_option_optimization_table):
9134 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9136 PR tree-optimization/81635
9137 * tree-data-ref.c (split_constant_offset_1): For types that
9138 wrap on overflow, try to use range info to prove that wrapping
9141 2018-02-01 Renlin Li <renlin.li@arm.com>
9144 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
9146 (aarch64_register_move_cost): Likewise.
9147 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
9149 (REG_CLASS_NAMES): Likewise.
9150 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
9151 TAILCALL_ADDR_REGS. Remove IP registers.
9152 * config/aarch64/aarch64.md (Ucs): Update register constraint.
9154 2018-02-01 Richard Biener <rguenther@suse.de>
9156 * domwalk.h (dom_walker::dom_walker): Add additional constructor
9157 for specifying RPO order and allow NULL for that.
9158 * domwalk.c (dom_walker::dom_walker): Likewise.
9159 (dom_walker::walk): Handle NULL RPO order.
9160 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
9162 (rewrite_update_dom_walker): Likewise.
9163 (mark_def_dom_walker): Likewise.
9165 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9167 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
9168 (aarch64_maybe_expand_sve_subreg_move): Declare.
9169 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
9170 * config/aarch64/predicates.md (aarch64_any_register_operand): New
9172 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
9173 that are semantically a reverse operation.
9174 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
9175 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
9176 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
9178 (aarch64_can_change_mode_class): For big-endian, forbid changes
9179 between two SVE modes if they have different element sizes.
9181 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9183 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
9184 the TImode handling for big-endian targets.
9186 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9188 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
9189 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
9191 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
9192 Remove BSWAP handing for big-endian targets and use the form of
9193 LD1RQ appropariate for the mode.
9195 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9197 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
9198 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
9201 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9204 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
9205 check for operands that need to go through aarch64_sve_reload_be.
9207 2018-02-01 Jakub Jelinek <jakub@redhat.com>
9209 PR tree-optimization/81661
9210 PR tree-optimization/84117
9211 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
9212 * tree-eh.c: Include gimplify.h.
9213 (find_trapping_overflow, replace_trapping_overflow,
9214 rewrite_to_non_trapping_overflow): New functions.
9215 * tree-vect-loop.c: Include tree-eh.h.
9216 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
9217 * tree-data-ref.c: Include tree-eh.h.
9218 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
9220 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
9222 PR rtl-optimization/84123
9223 * combine.c (change_zero_ext): Check if hard register satisfies
9224 can_change_dest_mode before calling gen_lowpart_SUBREG.
9226 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
9229 * ira.c (ira_init_register_move_cost): Remove assert.
9231 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
9233 PR rtl-optimization/84071
9234 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
9235 * doc/tm.texi: Regenerate.
9237 2018-01-31 Richard Biener <rguenther@suse.de>
9239 PR tree-optimization/84132
9240 * tree-data-ref.c (analyze_miv_subscript): Properly
9241 check whether evolution_function_is_affine_multivariate_p
9242 before calling gcd_of_steps_may_divide_p.
9244 2018-01-31 Julia Koval <julia.koval@intel.com>
9247 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
9248 * config/i386/i386.md (rdpid_rex64) New.
9249 (rdpid): Make 32bit only.
9251 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
9254 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
9255 an IDENTIFIER_NODE for FUNCTION_TYPE's.
9257 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
9260 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
9262 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
9264 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
9266 PR rtl-optimization/84071
9267 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
9268 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
9270 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
9272 * config/arc/arc.c (arc_handle_aux_attribute): New function.
9273 (arc_attribute_table): Add 'aux' attribute.
9274 (arc_in_small_data_p): Consider aux like variables.
9275 (arc_is_aux_reg_p): New function.
9276 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
9277 (arc_get_aux_arg): New function.
9278 (prepare_move_operands): Handle aux-register access.
9279 (arc_handle_aux_attribute): New function.
9280 * doc/extend.texi (ARC Variable attributes): Add subsection.
9282 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
9284 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
9285 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
9286 (arc_attribute_table): Add 'uncached' attribute.
9287 (arc_print_operand): Print '.di' flag for uncached memory
9289 (arc_in_small_data_p): Do not consider for small data the uncached
9291 (arc_is_uncached_mem_p): New function.
9292 * config/arc/predicates.md (compact_store_memory_operand): Check
9293 for uncached memory accesses.
9294 (nonvol_nonimm_operand): Likewise.
9295 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
9297 2018-01-31 Jakub Jelinek <jakub@redhat.com>
9300 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
9301 falign-loops=): Add Optimization flag.
9303 2018-01-30 Jeff Law <law@redhat.com>
9306 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
9307 INT_REGISTERS_SAVED. Check it prior to calling
9308 get_scratch_register_on_entry.
9309 (ix86_adjust_stack_and_probe): Similarly.
9310 (ix86_emit_probe_stack_range): Similarly.
9311 (ix86_expand_prologue): Corresponding changes.
9313 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9316 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
9317 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
9319 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
9322 * lra-constraints.c (curr_insn_transform): Process AND in the
9325 2018-01-30 Jakub Jelinek <jakub@redhat.com>
9327 PR rtl-optimization/83986
9328 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
9329 dependence against last_pending_memory_flush in addition to
9332 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
9334 PR tree-optimization/81611
9335 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
9338 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9341 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
9344 2018-01-30 Richard Biener <rguenther@suse.de>
9345 Jakub Jelinek <jakub@redhat.com>
9347 PR tree-optimization/84111
9348 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
9349 inner loops added during recursion, as they don't have up-to-date
9352 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
9355 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
9356 (can_inline_edge_by_limits_p): ... here.
9357 (can_early_inline_edge_p, check_callers,
9358 update_caller_keys, update_callee_keys, recursive_inlining,
9359 add_new_edges_to_heap, speculation_useful_p,
9360 inline_small_functions,
9361 inline_small_functions, flatten_function,
9362 inline_to_all_callers_1): Update.
9364 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
9366 * profile-count.c (profile_count::combine_with_ipa_count): Handle
9369 2018-01-30 Richard Biener <rguenther@suse.de>
9371 PR tree-optimization/83008
9372 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
9373 invariant and constant vector uses in stmts when they need
9376 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9379 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
9380 * configure: Regenerate.
9382 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9384 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
9386 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
9387 Use gen_rtx_REG rather than gen_lowpart.
9389 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9391 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
9392 rather than 0 when creating partial subregs.
9394 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9396 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
9399 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
9402 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
9403 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
9404 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
9405 flags. This restores the settings used before the 2017-07-24.
9406 Turning off pre increment/decrement/modify allows IVOPTS to
9407 optimize DF/SF loops where the index is an int.
9409 2018-01-29 Richard Biener <rguenther@suse.de>
9410 Kelvin Nilsen <kelvin@gcc.gnu.org>
9413 * tree-vect-stmts.c (vectorizable_call): Don't call
9414 targetm.vectorize_builtin_md_vectorized_function if callee is
9417 2018-01-22 Carl Love <cel@us.ibm.com>
9419 * doc/extend.tex: Fix typo in second arg in
9420 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
9422 2018-01-29 Richard Biener <rguenther@suse.de>
9424 PR tree-optimization/84086
9425 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
9426 (flush_ssaname_freelist): When SSA names were released reset
9427 the SCEV hash table.
9429 2018-01-29 Richard Biener <rguenther@suse.de>
9431 PR tree-optimization/84057
9432 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
9433 removed paths when removing edges.
9435 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
9437 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
9438 -mfunction-return=@var{choice}.
9440 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
9443 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
9445 (layout::print_source_line): Likewise.
9446 (test_get_line_width_without_trailing_whitespace): Add test cases.
9448 2018-01-27 Jakub Jelinek <jakub@redhat.com>
9451 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
9454 2018-01-26 Jim Wilson <jimw@sifive.com>
9456 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
9458 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
9461 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9463 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
9464 and CMP + SUB-immediate -> SUBS.
9466 2018-01-26 Martin Sebor <msebor@redhat.com>
9468 PR tree-optimization/83896
9469 * tree-ssa-strlen.c (get_string_len): Rename...
9470 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
9471 Avoid assuming length is constant.
9472 (handle_char_store): Use HOST_WIDE_INT for string length.
9474 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
9477 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
9478 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
9480 2018-01-26 Richard Biener <rguenther@suse.de>
9482 PR rtl-optimization/84003
9483 * dse.c (record_store): Only record redundant stores when
9484 the earlier store aliases at least all accesses the later one does.
9486 2018-01-26 Jakub Jelinek <jakub@redhat.com>
9488 PR rtl-optimization/83985
9489 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
9490 REG_CFA_RESTORE insns.
9491 (delete_unmarked_insns): Don't ignore separate shrink wrapping
9492 REG_CFA_RESTORE insns here.
9495 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
9496 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
9498 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9500 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
9501 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
9502 (arc_init): Likewise.
9503 (arc_override_options): Likewise.
9504 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
9506 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
9508 * config/arc/arc.h (TARGET_DBNZ): Define.
9509 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
9510 properly set the tune attribute.
9511 (dbnz): Use TARGET_DBNZ guard.
9512 * config/arc/arc.opt (mtune): Add core3 option.
9514 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9516 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
9517 recognize new pic like addresses.
9518 (arc_delegitimize_address): Clean up.
9520 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9522 * config/arc/arc-arches.def: Option mrf16 valid for all
9524 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
9525 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
9526 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
9527 * config/arc/arc-tables.opt: Regenerate.
9528 * config/arc/arc.c (arc_conditional_register_usage): Handle
9529 reduced register file case.
9530 (arc_file_start): Set must have build attributes.
9531 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
9533 * config/arc/arc.opt (mrf16): Add new option.
9534 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
9535 * config/arc/genmultilib.awk: Handle new mrf16 option.
9536 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
9537 * config/arc/t-multilib: Regenerate.
9538 * doc/invoke.texi (ARC Options): Document mrf16 option.
9540 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9542 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
9543 * config/arc/arc.c (arc_handle_secure_attribute): New function.
9544 (arc_attribute_table): Add 'secure_call' attribute.
9545 (arc_print_operand): Print secure call operand.
9546 (arc_function_ok_for_sibcall): Don't optimize tail calls when
9548 (arc_is_secure_call_p): New function. * config/arc/arc.md
9549 (call_i): Add support for sjli instruction.
9550 (call_value_i): Likewise.
9551 * config/arc/constraints.md (Csc): New constraint.
9553 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9554 John Eric Martin <John.Martin@emmicro-us.com>
9556 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
9557 * config/arc/arc.c (_arc_jli_section): New struct.
9558 (arc_jli_section): New type.
9559 (rc_jli_sections): New static variable.
9560 (arc_handle_jli_attribute): New function.
9561 (arc_attribute_table): Add jli_always and jli_fixed attribute.
9562 (arc_file_end): New function.
9563 (TARGET_ASM_FILE_END): Define.
9564 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
9565 (arc_add_jli_section): New function.
9566 (jli_call_scan): Likewise.
9567 (arc_reorg): Call jli_call_scan.
9568 (arc_output_addsi): Remove 'S' from printing asm operand.
9569 (arc_is_jli_call_p): New function.
9570 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
9572 (movhi_insn): Likewise.
9573 (movsi_insn): Likewise.
9574 (movsi_set_cc_insn): Likewise.
9575 (loadqi_update): Likewise.
9576 (load_zeroextendqisi_update): Likewise.
9577 (load_signextendqisi_update): Likewise.
9578 (loadhi_update): Likewise.
9579 (load_zeroextendhisi_update): Likewise.
9580 (load_signextendhisi_update): Likewise.
9581 (loadsi_update): Likewise.
9582 (loadsf_update): Likewise.
9583 (movsicc_insn): Likewise.
9584 (bset_insn): Likewise.
9585 (bxor_insn): Likewise.
9586 (bclr_insn): Likewise.
9587 (bmsk_insn): Likewise.
9588 (bicsi3_insn): Likewise.
9589 (cmpsi_cc_c_insn): Likewise.
9590 (movsi_ne): Likewise.
9591 (movsi_cond_exec): Likewise.
9592 (clrsbsi2): Likewise.
9603 (call_i): Remove 'S' asm letter, add jli instruction.
9604 (call_value_i): Likewise.
9605 * config/arc/arc.op (mjli-always): New option.
9606 * config/arc/constraints.md (Cji): New constraint.
9607 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
9609 (subsf3_fpx): Likewise.
9610 (mulsf3_fpx): Likewise.
9611 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
9613 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
9615 * doc/invoke.texi (ARC): Document mjli-always option.
9617 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
9619 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
9620 avoid addition with 0 and use incw and decw where possible.
9622 2018-01-26 Richard Biener <rguenther@suse.de>
9624 PR tree-optimization/81082
9625 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
9626 association if it requires casting to unsigned.
9627 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
9628 from fold_plusminus_mult_expr to catch important cases late when
9629 range info is available.
9631 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9633 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
9634 * configure.ac (hidden_linkonce): New test.
9635 * configure: Regenerate.
9636 * config.in: Regenerate.
9638 2018-01-26 Julia Koval <julia.koval@intel.com>
9640 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
9641 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
9642 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
9643 _mm_mask_bitshuffle_epi64_mask): Fix type.
9644 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
9645 USI_FTYPE_V4DI_V4DI_USI): Remove.
9646 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
9647 __builtin_ia32_vpshufbitqmb256_mask,
9648 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
9649 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
9650 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
9652 2018-01-26 Alan Modra <amodra@gmail.com>
9655 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
9656 UNSPEC_VBPERMQ. Sort other unspecs.
9658 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
9660 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
9662 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
9665 * predict.c (drop_profile): Do not push/pop cfun; update also
9667 (handle_missing_profiles): Fix logic looking for zero profiles.
9669 2018-01-25 Jakub Jelinek <jakub@redhat.com>
9672 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
9673 on functions with #pragma omp declare simd or functions with simd
9675 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
9676 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
9677 Remove trailing \n from warning_at calls.
9679 2018-01-25 Tom de Vries <tom@codesourcery.com>
9682 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
9683 for neutered workers.
9685 2018-01-24 Joseph Myers <joseph@codesourcery.com>
9688 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
9689 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
9691 2018-01-24 Jeff Law <law@redhat.com>
9694 * i386.c (get_probe_interval): Move to earlier point.
9695 (ix86_compute_frame_layout): If -fstack-clash-protection and
9696 the frame is larger than the probe interval, then use pushes
9697 to save registers rather than reg->mem moves.
9698 (ix86_expand_prologue): Remove conditional for int_registers_saved
9701 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
9704 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
9705 min/max for never referenced object.
9707 2018-01-24 Jakub Jelinek <jakub@redhat.com>
9710 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
9712 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
9713 attributes from DECL_ATTRIBUTES (decl) without affecting
9714 DECL_ATTRIBUTES (current_function_decl).
9715 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
9716 functions with non-NULL DECL_ABSTRACT_ORIGIN.
9718 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
9720 PR tree-optimization/83979
9721 * fold-const.c (fold_comparison): Use constant_boolean_node
9722 instead of boolean_{true,false}_node.
9724 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
9726 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
9729 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9731 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9732 Simplify the clause that sets the length attribute.
9733 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9734 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
9735 clause that sets the length attribute.
9736 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9738 2018-01-24 Tom de Vries <tom@codesourcery.com>
9741 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
9742 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
9743 Add strict parameter.
9744 (prevent_branch_around_nothing): Insert dummy insn between branch to
9745 label and label with no ptx insn inbetween.
9746 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
9748 2018-01-24 Tom de Vries <tom@codesourcery.com>
9751 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
9752 for neutered threads in warp.
9753 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
9755 2018-01-24 Richard Biener <rguenther@suse.de>
9757 PR tree-optimization/83176
9758 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
9761 2018-01-24 Richard Biener <rguenther@suse.de>
9763 PR tree-optimization/82819
9764 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
9765 code generating pluses that are no-ops in the target precision.
9767 2018-01-24 Richard Biener <rguenther@suse.de>
9770 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
9772 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9774 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
9775 to merge probabilities.
9776 * predict.c (probably_never_executed): Also mark as cold functions
9777 with global 0 profile and guessed local profile.
9778 * profile-count.c (profile_probability::combine_with_count): New
9780 * profile-count.h (profile_probability::operator*,
9781 profile_probability::operator*=, profile_probability::operator/,
9782 profile_probability::operator/=): Reduce precision to adjusted
9783 and set value to guessed on contradictory divisions.
9784 (profile_probability::combine_with_freq): Remove.
9785 (profile_probability::combine_wiht_count): Declare.
9786 (profile_count::force_nonzero):: Set to adjusted.
9787 (profile_count::probability_in):: Set quality to adjusted.
9788 * tree-ssa-tail-merge.c (replace_block_by): Use
9791 2018-01-23 Andrew Waterman <andrew@sifive.com>
9792 Jim Wilson <jimw@sifive.com>
9794 * config/riscv/riscv.c (riscv_stack_boundary): New.
9795 (riscv_option_override): Set riscv_stack_boundary. Handle
9796 riscv_preferred_stack_boundary_arg.
9797 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
9798 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
9799 (STACK_BOUNDARY): Set to riscv_stack_boundary.
9800 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
9801 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
9802 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
9804 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
9807 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
9808 of struct ix86_frame.
9809 (ix86_expand_epilogue): Likewise. Add a local variable for
9810 the reg_save_offset field in struct ix86_frame.
9812 2018-01-23 Bin Cheng <bin.cheng@arm.com>
9814 PR tree-optimization/82604
9815 * tree-loop-distribution.c (enum partition_kind): New enum item
9816 PKIND_PARTIAL_MEMSET.
9817 (partition_builtin_p): Support above new enum item.
9818 (generate_code_for_partition): Ditto.
9819 (compute_access_range): Differentiate cases that equality can be
9820 proven at all loops, the innermost loops or no loops.
9821 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
9822 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
9823 (finalize_partitions, distribute_loop): Don't fuse partition of
9824 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
9825 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
9828 2018-01-23 Martin Liska <mliska@suse.cz>
9830 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
9831 order to ignore the predictor.
9832 (PRED_POLYMORPHIC_CALL): Likewise.
9833 (PRED_RECURSIVE_CALL): Likewise.
9835 2018-01-23 Martin Liska <mliska@suse.cz>
9837 * tree-profile.c (tree_profiling): Print function header to
9838 aware reader which function we are working on.
9839 * value-prof.c (gimple_find_values_to_profile): Do not print
9840 not interesting value histograms.
9842 2018-01-23 Martin Liska <mliska@suse.cz>
9844 * profile-count.h (enum profile_quality): Add
9845 profile_uninitialized as the first value. Do not number values
9846 as they are zero based.
9847 (profile_count::verify): Update sanity check.
9848 (profile_probability::verify): Likewise.
9850 2018-01-23 Nathan Sidwell <nathan@acm.org>
9852 * doc/invoke.texi (ffor-scope): Deprecate.
9854 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9856 PR tree-optimization/83510
9857 * domwalk.c (set_all_edges_as_executable): New function.
9858 (dom_walker::dom_walker): Convert bool param
9859 "skip_unreachable_blocks" to enum reachability. Move setup of
9860 edge flags to set_all_edges_as_executable and only do it when
9861 reachability is REACHABLE_BLOCKS.
9862 * domwalk.h (enum dom_walker::reachability): New enum.
9863 (dom_walker::dom_walker): Convert bool param
9864 "skip_unreachable_blocks" to enum reachability.
9865 (set_all_edges_as_executable): New decl.
9866 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
9867 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
9869 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
9870 but converting true to REACHABLE_BLOCKS.
9871 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
9873 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
9874 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
9875 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
9877 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
9878 if check_all_array_refs will be called.
9880 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9882 * tree.c (selftest::test_location_wrappers): Add more test
9885 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9887 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
9888 (selftest::test_bit_in_range): Likewise.
9890 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
9893 * doc/sourcebuild.texi (vect_float): Say that the selector
9894 only describes the situation when -funsafe-math-optimizations is on.
9895 (vect_float_strict): Document.
9897 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
9899 PR tree-optimization/83965
9900 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
9901 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
9902 instead of checking only for a reduction.
9903 (vect_recog_widen_sum_pattern): Likewise.
9905 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9907 * predict.c (probably_never_executed): Only use precise profile info.
9908 (compute_function_frequency): Skip after inlining hack since we now
9909 have quality checking.
9911 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9913 * profile-count.h (profile_probability::very_unlikely,
9914 profile_probability::unlikely, profile_probability::even): Set
9915 precision to guessed.
9917 2018-01-23 Richard Biener <rguenther@suse.de>
9919 PR tree-optimization/83963
9920 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
9921 Properly terminate dominator walk when crossing the exit edge not
9922 when visiting its source block.
9924 2018-01-23 Jakub Jelinek <jakub@redhat.com>
9927 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
9928 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
9930 2018-01-22 Jakub Jelinek <jakub@redhat.com>
9932 PR tree-optimization/83957
9933 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
9934 semicolon after for body surrounded by braces.
9936 PR tree-optimization/83081
9937 * profile-count.h (profile_probability::split): New method.
9938 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
9939 Use profile_probability::split.
9940 (do_compare_rtx_and_jump): Fix adjustment of probabilities
9941 when splitting a single conditional jump into 2.
9943 2018-01-22 David Malcolm <dmalcolm@redhat.com>
9945 PR tree-optimization/69452
9946 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
9949 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9951 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
9952 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
9953 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
9955 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9957 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
9958 * config/rl78/rl78.md (movdi): New define_expand.
9959 * config/rl78/rl78.c (rl78_split_movdi): New function.
9961 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
9964 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
9966 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
9967 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
9968 128-bit to produce an UNSPEC move to get the double word with the
9969 signbit and then a shift directly to do signbit.
9970 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
9971 implementation with a new version that just does either a direct
9972 move or a regular move. Move memory interface to separate insns.
9973 Move insns so they are next to the expander.
9974 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
9975 with signbit move. Split big and little endian case.
9976 (signbit<mode>2_dm_mem_le): Likewise.
9977 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
9978 (signbit<mode>2_dm2): Likewise.
9980 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9982 * config/rl78/rl78.md (anddi3): New define_expand.
9984 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9986 * config/rl78/rl78.md (umindi3): New define_expand.
9988 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9990 * config/rl78/rl78.md (smindi3): New define_expand.
9992 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9994 * config/rl78/rl78.md (smaxdi3): New define_expand.
9996 2018-01-22 Carl Love <cel@us.ibm.com>
9998 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
9999 LVX_V1TI): Add macro expansion.
10000 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
10001 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
10002 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
10003 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
10004 Change check to determine if the instruction is a byte reversing
10005 entry. Fix typo in comment.
10006 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
10007 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
10008 Add def_builtin calls for new builtins.
10009 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
10010 Add define_insn expansion.
10012 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10014 * config/rl78/rl78.md (umaxdi3): New define_expand.
10016 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10018 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
10019 for non-QImode registers.
10021 2018-01-22 Richard Biener <rguenther@suse.de>
10023 PR tree-optimization/83963
10024 * graphite-scop-detection.c (scop_detection::get_sese): Delay
10025 including the loop exit block.
10026 (scop_detection::merge_sese): Likewise.
10027 (scop_detection::add_scop): Do it here instead.
10029 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10031 * doc/sourcebuild.texi (arm_softfloat): Document.
10033 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
10036 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
10037 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
10038 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
10040 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10041 David Edelsohn <dje.gcc@gmail.com>
10044 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
10045 Change "crset eq" to "crset 2".
10046 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
10047 (*call_indirect_aix<mode>_nospec): Likewise.
10048 (*call_value_indirect_aix<mode>_nospec): Likewise.
10049 (*call_indirect_elfv2<mode>_nospec): Likewise.
10050 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
10051 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
10052 change assembly output from . to $.
10053 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
10054 (indirect_jump<mode>_nospec): Change assembly output from . to $.
10055 (*tablejump<mode>_internal1_nospec): Likewise.
10057 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
10060 * config/sh/sh_optimize_sett_clrt.cc:
10061 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
10063 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
10065 PR tree-optimization/83940
10066 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
10067 offset_dt to vect_constant_def rather than vect_unknown_def_type.
10068 (vect_check_load_store_mask): Add a mask_dt_out parameter and
10069 use it to pass back the definition type.
10070 (vect_check_store_rhs): Likewise rhs_dt_out.
10071 (vect_build_gather_load_calls): Add a mask_dt argument and use
10072 it instead of a call to vect_is_simple_use.
10073 (vectorizable_store): Update calls to vect_check_load_store_mask
10074 and vect_check_store_rhs. Use the dt returned by the latter instead
10075 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
10076 instead of calls to vect_is_simple_use. Pass the scalar rather
10077 than the vector operand to vect_is_simple_use when handling
10078 second and subsequent copies of an rhs value.
10079 (vectorizable_load): Update calls to vect_check_load_store_mask
10080 and vect_build_gather_load_calls. Use the cached mask_dt and
10081 gs_info.offset_dt instead of calls to vect_is_simple_use.
10083 2018-01-20 Jakub Jelinek <jakub@redhat.com>
10085 PR middle-end/83945
10086 * tree-emutls.c: Include gimplify.h.
10087 (lower_emutls_2): New function.
10088 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
10089 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
10090 it before further processing.
10093 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
10094 UINTVAL (trueop1) instead of INTVAL (op1).
10096 2018-01-19 Jakub Jelinek <jakub@redhat.com>
10100 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
10101 INCOMING_FRAME_SP_OFFSET if not defined.
10102 (scan_trace): Add ENTRY argument. If true and
10103 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
10104 emit a note to adjust the CFA offset.
10105 (create_cfi_notes): Adjust scan_trace callers.
10106 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
10107 INCOMING_FRAME_SP_OFFSET in the CIE.
10108 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
10109 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
10111 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
10112 * doc/tm.texi: Regenerated.
10114 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
10116 PR rtl-optimization/83147
10117 * lra-constraints.c (remove_inheritance_pseudos): Use
10118 lra_substitute_pseudo_within_insn.
10120 2018-01-19 Tom de Vries <tom@codesourcery.com>
10121 Cesar Philippidis <cesar@codesourcery.com>
10124 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
10126 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
10129 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
10130 spaces for function labels.
10132 2018-01-19 Martin Liska <mliska@suse.cz>
10134 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
10135 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
10136 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
10137 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
10138 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
10139 (PRED_CONST_RETURN): Change from 69 to 65.
10140 (PRED_NULL_RETURN): Change from 91 to 71.
10141 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
10142 (PRED_LOOP_GUARD): Change from 66 to 73.
10144 2018-01-19 Martin Liska <mliska@suse.cz>
10146 * predict.c (predict_insn_def): Add new assert.
10147 (struct branch_predictor): Change type to signed integer.
10148 (test_prediction_value_range): Amend test to cover
10149 PROB_UNINITIALIZED.
10150 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
10151 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
10152 (PRED_LOOP_ITERATIONS_MAX): Likewise.
10153 (PRED_LOOP_IV_COMPARE): Likewise.
10154 * predict.h (PROB_UNINITIALIZED): Define new constant.
10156 2018-01-19 Martin Liska <mliska@suse.cz>
10158 * predict.c (dump_prediction): Add new format for
10159 analyze_brprob.py script which is enabled with -details
10161 * profile-count.h (precise_p): New function.
10163 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
10165 PR tree-optimization/83922
10166 * tree-vect-loop.c (vect_verify_full_masking): Return false if
10167 there are no statements that need masking.
10168 (vect_active_double_reduction_p): New function.
10169 (vect_analyze_loop_operations): Use it when handling phis that
10170 are not in the loop header.
10172 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
10174 PR tree-optimization/83914
10175 * tree-vect-loop.c (vectorizable_induction): Don't convert
10176 init_expr or apply the peeling adjustment for inductions
10177 that are nested within the vectorized loop.
10179 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10181 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
10184 2018-01-18 Jakub Jelinek <jakub@redhat.com>
10188 * function.h (gimplify_parameters): Add gimple_seq * argument.
10189 * function.c: Include gimple.h and options.h.
10190 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
10191 for the added local temporaries if needed.
10192 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
10193 if there are any parameter cleanups, wrap whole body into a
10194 try/finally with the cleanups.
10196 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
10199 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
10200 Use GET_MODE_CLASS for scalar floating point.
10202 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
10206 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
10207 Fix call of call_cgraph_insertion_hooks.
10209 2018-01-18 Martin Sebor <msebor@redhat.com>
10211 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
10213 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
10216 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
10219 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
10222 * common.opt: (-ffile-prefix-map): New option.
10223 * opts.c (common_handle_option): Defer it.
10224 * opts-global.c (handle_common_deferred_options): Handle it.
10225 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
10226 * file-prefix-map.h: New file.
10227 (remap_debug_filename, add_debug_prefix_map): ...here.
10228 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
10229 * final.c (debug_prefix_map, add_debug_prefix_map
10230 remap_debug_filename): Move to...
10231 * file-prefix-map.c: New file.
10232 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
10233 generalize, get rid of alloca(), use strrchr() instead of strchr().
10234 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
10235 Implement in terms of add_prefix_map().
10236 (remap_macro_filename, remap_debug_filename): Implement in term of
10238 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
10239 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
10240 * dbxout.c: Include file-prefix-map.h.
10241 * varasm.c: Likewise.
10242 * vmsdbgout.c: Likewise.
10243 * xcoffout.c: Likewise.
10244 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
10245 * doc/cppopts.texi (-fmacro-prefix-map): Document.
10246 * doc/invoke.texi (-ffile-prefix-map): Document.
10247 (-fdebug-prefix-map): Update description.
10249 2018-01-18 Martin Liska <mliska@suse.cz>
10251 * config/i386/i386.c (indirect_thunk_name): Document that also
10253 (output_indirect_thunk): Document why both instructions
10254 (pause and lfence) are generated.
10256 2018-01-18 Richard Biener <rguenther@suse.de>
10258 PR tree-optimization/83887
10259 * graphite-scop-detection.c
10260 (scop_detection::get_nearest_dom_with_single_entry): Remove.
10261 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
10262 (scop_detection::merge_sese): Re-implement with a flood-fill
10263 algorithm that properly finds a SESE region if it exists.
10265 2018-01-18 Jakub Jelinek <jakub@redhat.com>
10268 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
10269 pointer_diff optimizations use view_convert instead of convert.
10271 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10273 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
10274 Generate different code for -mno-speculate-indirect-jumps.
10275 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
10276 (*call_indirect_aix<mode>): Disable for
10277 -mno-speculate-indirect-jumps.
10278 (*call_indirect_aix<mode>_nospec): New define_insn.
10279 (*call_value_indirect_aix<mode>): Disable for
10280 -mno-speculate-indirect-jumps.
10281 (*call_value_indirect_aix<mode>_nospec): New define_insn.
10282 (*sibcall_nonlocal_sysv<mode>): Generate different code for
10283 -mno-speculate-indirect-jumps.
10284 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
10286 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
10288 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
10289 long double type, set the flags for noting the default long double
10290 type, even if we don't pass or return a long double type.
10292 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
10295 * ipa-inline.c (flatten_function): Do not overwrite final inlining
10298 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
10300 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
10301 support for merge[hl].
10302 (fold_mergehl_helper): New helper function.
10303 (tree-vector-builder.h): New #include for tree_vector_builder usage.
10304 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
10305 (altivec_vmrglw_direct): Add xxmrglw insn.
10307 2018-01-17 Andrew Waterman <andrew@sifive.com>
10309 * config/riscv/riscv.c (riscv_conditional_register_usage): If
10310 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
10312 2018-01-17 David Malcolm <dmalcolm@redhat.com>
10315 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
10316 call the lto_location_cache before reading the
10317 DECL_SOURCE_LOCATION of the types.
10319 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
10320 Richard Sandiford <richard.sandiford@linaro.org>
10322 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
10323 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
10324 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
10325 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
10326 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
10328 * config/aarch64/constraints.md (aarch64_movti_operand):
10330 * config/aarch64/predicates.md (Uti): Add new constraint.
10332 2018-01-17 Carl Love <cel@us.ibm.com>
10334 * config/rs6000/vsx.md (define_expand xl_len_r,
10335 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
10336 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
10338 (define_expand, define_insn): Move the shift left from the
10339 define_insn to the define_expand for lxvl and stxvl instructions.
10340 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
10341 and XL_LEN_R definitions to PURE.
10343 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
10345 * config/i386/i386.c (indirect_thunk_name): Declare regno
10346 as unsigned int. Compare regno with INVALID_REGNUM.
10347 (output_indirect_thunk): Ditto.
10348 (output_indirect_thunk_function): Ditto.
10349 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
10350 in the call to output_indirect_thunk_function.
10352 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
10354 PR middle-end/83884
10355 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
10356 rather than the size of inner_type to determine the stack slot size
10357 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
10359 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
10362 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
10365 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
10367 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
10368 endian Linux systems to optionally enable multilibs for selecting
10369 the long double type if the user configured an explicit type.
10370 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
10371 have no long double multilibs if not defined.
10372 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
10373 warn if the user used -mabi={ieee,ibm}longdouble and we built
10374 multilibs for long double.
10375 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
10376 appropriate multilib option.
10377 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
10379 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
10380 for building long double multilibs.
10381 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
10383 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
10385 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
10388 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
10390 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
10393 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
10396 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
10399 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
10401 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
10402 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
10404 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
10406 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
10407 different rtl trees depending on TARGET_64BIT.
10408 (rs6000_gen_lvx): Likewise.
10410 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
10412 * config/visium/visium.md (nop): Tweak comment.
10413 (hazard_nop): Likewise.
10415 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10417 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
10418 -mspeculate-indirect-jumps.
10419 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
10420 for -mno-speculate-indirect-jumps.
10421 (*call_indirect_elfv2<mode>_nospec): New define_insn.
10422 (*call_value_indirect_elfv2<mode>): Disable for
10423 -mno-speculate-indirect-jumps.
10424 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
10425 (indirect_jump): Emit different RTL for
10426 -mno-speculate-indirect-jumps.
10427 (*indirect_jump<mode>): Disable for
10428 -mno-speculate-indirect-jumps.
10429 (*indirect_jump<mode>_nospec): New define_insn.
10430 (tablejump): Emit different RTL for
10431 -mno-speculate-indirect-jumps.
10432 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
10433 (tablejumpsi_nospec): New define_expand.
10434 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
10435 (tablejumpdi_nospec): New define_expand.
10436 (*tablejump<mode>_internal1): Disable for
10437 -mno-speculate-indirect-jumps.
10438 (*tablejump<mode>_internal1_nospec): New define_insn.
10439 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
10442 2018-01-16 Artyom Skrobov tyomitch@gmail.com
10444 * caller-save.c (insert_save): Drop unnecessary parameter. All
10447 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10448 Richard Biener <rguenth@suse.de>
10451 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
10452 return early, inline manually is_gimple_sizepos. Make sure if we
10453 call gimplify_expr we don't end up with a gimple constant.
10454 * tree.c (variably_modified_type_p): Don't return true for
10455 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
10456 * gimplify.h (is_gimple_sizepos): Remove.
10458 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10460 PR tree-optimization/83857
10461 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
10462 vectorizable_live_operation for pure SLP statements.
10463 (vectorizable_live_operation): Handle PHIs.
10465 2018-01-16 Richard Biener <rguenther@suse.de>
10467 PR tree-optimization/83867
10468 * tree-vect-stmts.c (vect_transform_stmt): Precompute
10469 nested_in_vect_loop_p since the scalar stmt may get invalidated.
10471 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10474 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
10475 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
10476 If off is not INTEGER_CST, issue a may not be aligned warning
10477 rather than isn't aligned. Use isn%'t rather than isn't.
10478 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
10480 <case MULT_EXPR>: Improve the case when bottom and one of the
10481 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
10482 operand, in that case check if the other operand is multiple of
10483 bottom divided by the INTEGER_CST operand.
10485 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10488 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
10489 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
10490 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
10491 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
10492 * config/pa/pa.c (pa_function_arg_advance): Likewise.
10493 (pa_function_arg, pa_arg_partial_bytes): Likewise.
10494 (pa_function_arg_size): New function.
10496 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10498 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
10499 in a separate statement.
10501 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10503 PR tree-optimization/83847
10504 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
10505 group gathers and scatters.
10507 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10509 PR rtl-optimization/86620
10510 * params.def (max-sched-ready-insns): Bump minimum value to 1.
10512 PR rtl-optimization/83213
10513 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
10514 to last if both are JUMP_INSNs.
10516 PR tree-optimization/83843
10517 * gimple-ssa-store-merging.c
10518 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
10519 store_immediate_info for bswap/nop orig_stores.
10521 2018-01-15 Andrew Waterman <andrew@sifive.com>
10523 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
10525 <UDIV>: Increase cost if !TARGET_DIV.
10527 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
10529 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
10530 (define_attr "cr_logical_3op"): New.
10531 (cceq_ior_compare): Adjust.
10532 (cceq_ior_compare_complement): Adjust.
10533 (*cceq_rev_compare): Adjust.
10534 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
10535 (is_cracked_insn): Adjust.
10536 (insn_must_be_first_in_group): Adjust.
10537 * config/rs6000/40x.md: Adjust.
10538 * config/rs6000/440.md: Adjust.
10539 * config/rs6000/476.md: Adjust.
10540 * config/rs6000/601.md: Adjust.
10541 * config/rs6000/603.md: Adjust.
10542 * config/rs6000/6xx.md: Adjust.
10543 * config/rs6000/7450.md: Adjust.
10544 * config/rs6000/7xx.md: Adjust.
10545 * config/rs6000/8540.md: Adjust.
10546 * config/rs6000/cell.md: Adjust.
10547 * config/rs6000/e300c2c3.md: Adjust.
10548 * config/rs6000/e500mc.md: Adjust.
10549 * config/rs6000/e500mc64.md: Adjust.
10550 * config/rs6000/e5500.md: Adjust.
10551 * config/rs6000/e6500.md: Adjust.
10552 * config/rs6000/mpc.md: Adjust.
10553 * config/rs6000/power4.md: Adjust.
10554 * config/rs6000/power5.md: Adjust.
10555 * config/rs6000/power6.md: Adjust.
10556 * config/rs6000/power7.md: Adjust.
10557 * config/rs6000/power8.md: Adjust.
10558 * config/rs6000/power9.md: Adjust.
10559 * config/rs6000/rs64.md: Adjust.
10560 * config/rs6000/titan.md: Adjust.
10562 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10564 * config/i386/predicates.md (indirect_branch_operand): Rewrite
10565 ix86_indirect_branch_register logic.
10567 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10569 * config/i386/constraints.md (Bs): Update
10570 ix86_indirect_branch_register check. Don't check
10571 ix86_indirect_branch_register with GOT_memory_operand.
10573 * config/i386/predicates.md (GOT_memory_operand): Don't check
10574 ix86_indirect_branch_register here.
10575 (GOT32_symbol_operand): Likewise.
10577 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10579 * config/i386/predicates.md (constant_call_address_operand):
10580 Rewrite ix86_indirect_branch_register logic.
10581 (sibcall_insn_operand): Likewise.
10583 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10585 * config/i386/constraints.md (Bs): Replace
10586 ix86_indirect_branch_thunk_register with
10587 ix86_indirect_branch_register.
10589 * config/i386/i386.md (indirect_jump): Likewise.
10590 (tablejump): Likewise.
10591 (*sibcall_memory): Likewise.
10592 (*sibcall_value_memory): Likewise.
10593 Peepholes of indirect call and jump via memory: Likewise.
10594 * config/i386/i386.opt: Likewise.
10595 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10596 (GOT_memory_operand): Likewise.
10597 (call_insn_operand): Likewise.
10598 (sibcall_insn_operand): Likewise.
10599 (GOT32_symbol_operand): Likewise.
10601 2018-01-15 Jakub Jelinek <jakub@redhat.com>
10603 PR middle-end/83837
10604 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
10605 type rather than type addr's type points to.
10606 (expand_omp_atomic_mutex): Likewise.
10607 (expand_omp_atomic): Likewise.
10609 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10612 * config/i386/i386.c (output_indirect_thunk_function): Use
10613 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
10614 for __x86_return_thunk.
10616 2018-01-15 Richard Biener <rguenther@suse.de>
10618 PR middle-end/83850
10619 * expmed.c (extract_bit_field_1): Fix typo.
10621 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10624 * config/arm/iterators.md (VF): New mode iterator.
10625 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
10626 Remove integer-related logic from pattern.
10627 (neon_vabd<mode>_3): Likewise.
10629 2018-01-15 Jakub Jelinek <jakub@redhat.com>
10631 PR middle-end/82694
10632 * common.opt (fstrict-overflow): No longer an alias.
10633 (fwrapv-pointer): New option.
10634 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
10635 also for pointer types based on flag_wrapv_pointer.
10636 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
10637 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
10638 opts->x_flag_wrapv got set.
10639 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
10640 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
10641 POINTER_TYPE_OVERFLOW_UNDEFINED.
10642 * match.pd: Likewise in address comparison pattern.
10643 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
10645 2018-01-15 Richard Biener <rguenther@suse.de>
10648 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
10649 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
10650 Reset type names to their identifier if their TYPE_DECL doesn't
10651 have linkage (and thus is used for ODR and devirt).
10652 (save_debug_info_for_decl): Remove.
10653 (save_debug_info_for_type): Likewise.
10654 (add_tree_to_fld_list): Adjust.
10655 * tree-pretty-print.c (dump_generic_node): Make dumping of
10656 type names more robust.
10658 2018-01-15 Richard Biener <rguenther@suse.de>
10660 * BASE-VER: Bump to 8.0.1.
10662 2018-01-14 Martin Sebor <msebor@redhat.com>
10665 * builtins.c (check_access): Avoid warning when the no-warning bit
10668 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
10670 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
10671 * ira-color (allocno_hard_regs_compare): Likewise.
10673 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
10676 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
10677 Use .pushsection/.popsection.
10679 2018-01-14 Martin Sebor <msebor@redhat.com>
10682 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
10684 2018-01-14 Jakub Jelinek <jakub@redhat.com>
10686 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
10687 entry from extra_headers.
10688 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
10689 extra_headers, make the list bitwise identical to the i?86-*-* one.
10691 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10693 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
10694 -mcmodel=large with -mindirect-branch=thunk,
10695 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
10696 -mfunction-return=thunk-extern.
10697 * doc/invoke.texi: Document -mcmodel=large is incompatible with
10698 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
10699 -mfunction-return=thunk and -mfunction-return=thunk-extern.
10701 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10703 * config/i386/i386.c (print_reg): Print the name of the full
10704 integer register without '%'.
10705 (ix86_print_operand): Handle 'V'.
10706 * doc/extend.texi: Document 'V' modifier.
10708 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10710 * config/i386/constraints.md (Bs): Disallow memory operand for
10711 -mindirect-branch-register.
10713 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10714 (GOT_memory_operand): Likewise.
10715 (call_insn_operand): Likewise.
10716 (sibcall_insn_operand): Likewise.
10717 (GOT32_symbol_operand): Likewise.
10718 * config/i386/i386.md (indirect_jump): Call convert_memory_address
10719 for -mindirect-branch-register.
10720 (tablejump): Likewise.
10721 (*sibcall_memory): Likewise.
10722 (*sibcall_value_memory): Likewise.
10723 Disallow peepholes of indirect call and jump via memory for
10724 -mindirect-branch-register.
10725 (*call_pop): Replace m with Bw.
10726 (*call_value_pop): Likewise.
10727 (*sibcall_pop_memory): Replace m with Bs.
10728 * config/i386/i386.opt (mindirect-branch-register): New option.
10729 * doc/invoke.texi: Document -mindirect-branch-register option.
10731 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10733 * config/i386/i386-protos.h (ix86_output_function_return): New.
10734 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
10735 set function_return_type.
10736 (indirect_thunk_name): Add ret_p to indicate thunk for function
10738 (output_indirect_thunk_function): Pass false to
10739 indirect_thunk_name.
10740 (ix86_output_indirect_branch_via_reg): Likewise.
10741 (ix86_output_indirect_branch_via_push): Likewise.
10742 (output_indirect_thunk_function): Create alias for function
10743 return thunk if regno < 0.
10744 (ix86_output_function_return): New function.
10745 (ix86_handle_fndecl_attribute): Handle function_return.
10746 (ix86_attribute_table): Add function_return.
10747 * config/i386/i386.h (machine_function): Add
10748 function_return_type.
10749 * config/i386/i386.md (simple_return_internal): Use
10750 ix86_output_function_return.
10751 (simple_return_internal_long): Likewise.
10752 * config/i386/i386.opt (mfunction-return=): New option.
10753 (indirect_branch): Mention -mfunction-return=.
10754 * doc/extend.texi: Document function_return function attribute.
10755 * doc/invoke.texi: Document -mfunction-return= option.
10757 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10759 * config/i386/i386-opts.h (indirect_branch): New.
10760 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
10761 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
10762 with local indirect jump when converting indirect call and jump.
10763 (ix86_set_indirect_branch_type): New.
10764 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
10765 (indirectlabelno): New.
10766 (indirect_thunk_needed): Likewise.
10767 (indirect_thunk_bnd_needed): Likewise.
10768 (indirect_thunks_used): Likewise.
10769 (indirect_thunks_bnd_used): Likewise.
10770 (INDIRECT_LABEL): Likewise.
10771 (indirect_thunk_name): Likewise.
10772 (output_indirect_thunk): Likewise.
10773 (output_indirect_thunk_function): Likewise.
10774 (ix86_output_indirect_branch_via_reg): Likewise.
10775 (ix86_output_indirect_branch_via_push): Likewise.
10776 (ix86_output_indirect_branch): Likewise.
10777 (ix86_output_indirect_jmp): Likewise.
10778 (ix86_code_end): Call output_indirect_thunk_function if needed.
10779 (ix86_output_call_insn): Call ix86_output_indirect_branch if
10781 (ix86_handle_fndecl_attribute): Handle indirect_branch.
10782 (ix86_attribute_table): Add indirect_branch.
10783 * config/i386/i386.h (machine_function): Add indirect_branch_type
10784 and has_local_indirect_jump.
10785 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
10787 (tablejump): Likewise.
10788 (*indirect_jump): Use ix86_output_indirect_jmp.
10789 (*tablejump_1): Likewise.
10790 (simple_return_indirect_internal): Likewise.
10791 * config/i386/i386.opt (mindirect-branch=): New option.
10792 (indirect_branch): New.
10795 (thunk-inline): Likewise.
10796 (thunk-extern): Likewise.
10797 * doc/extend.texi: Document indirect_branch function attribute.
10798 * doc/invoke.texi: Document -mindirect-branch= option.
10800 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
10803 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
10805 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
10807 * ipa-inline.c (want_inline_small_function_p): Return false if
10808 inlining has already failed with CIF_FINAL_ERROR.
10809 (update_caller_keys): Call want_inline_small_function_p before
10811 (update_callee_keys): Likewise.
10813 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
10815 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
10817 (rs6000_quadword_masked_address_p): Likewise.
10818 (quad_aligned_load_p): Likewise.
10819 (quad_aligned_store_p): Likewise.
10820 (const_load_sequence_p): Add comment to describe the outer-most loop.
10821 (mimic_memory_attributes_and_flags): New function.
10822 (rs6000_gen_stvx): Likewise.
10823 (replace_swapped_aligned_store): Likewise.
10824 (rs6000_gen_lvx): Likewise.
10825 (replace_swapped_aligned_load): Likewise.
10826 (replace_swapped_load_constant): Capitalize argument name in
10827 comment describing this function.
10828 (rs6000_analyze_swaps): Add a third pass to search for vector loads
10829 and stores that access quad-word aligned addresses and replace
10830 with stvx or lvx instructions when appropriate.
10831 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
10832 New function prototype.
10833 (rs6000_quadword_masked_address_p): Likewise.
10834 (rs6000_gen_lvx): Likewise.
10835 (rs6000_gen_stvx): Likewise.
10836 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
10837 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
10838 when memory address is aligned.
10839 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
10840 this split to select lvx instruction when memory address is aligned.
10841 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
10842 instruction when memory address is aligned.
10843 (*vsx_le_perm_load_v16qi): Likewise.
10844 (four unnamed splitters): Modify to select the stvx instruction
10845 when memory is aligned.
10847 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
10849 * predict.c (determine_unlikely_bbs): Handle correctly BBs
10850 which appears in the queue multiple times.
10852 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10853 Alan Hayward <alan.hayward@arm.com>
10854 David Sherwood <david.sherwood@arm.com>
10856 * tree-vectorizer.h (vec_lower_bound): New structure.
10857 (_loop_vec_info): Add check_nonzero and lower_bounds.
10858 (LOOP_VINFO_CHECK_NONZERO): New macro.
10859 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
10860 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
10861 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
10862 fields. Make seg_len the distance travelled, not including the
10864 (dr_direction_indicator): Declare.
10865 (dr_zero_step_indicator): Likewise.
10866 (dr_known_forward_stride_p): Likewise.
10867 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
10869 (runtime_alias_check_p): Allow runtime alias checks with
10871 (operator ==): Compare access_size and align.
10872 (prune_runtime_alias_test_list): Rework for new distinction between
10873 the access_size and seg_len.
10874 (create_intersect_range_checks_index): Likewise. Cope with polynomial
10876 (get_segment_min_max): New function.
10877 (create_intersect_range_checks): Use it.
10878 (dr_step_indicator): New function.
10879 (dr_direction_indicator): Likewise.
10880 (dr_zero_step_indicator): Likewise.
10881 (dr_known_forward_stride_p): Likewise.
10882 * tree-loop-distribution.c (data_ref_segment_size): Return
10883 DR_STEP * (niters - 1).
10884 (compute_alias_check_pairs): Update call to the dr_with_seg_len
10886 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
10887 (vect_preserves_scalar_order_p): New function, split out from...
10888 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
10889 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
10890 (vect_vfa_access_size): New function.
10891 (vect_vfa_align): Likewise.
10892 (vect_compile_time_alias): Take access_size_a and access_b arguments.
10893 (dump_lower_bound): New function.
10894 (vect_check_lower_bound): Likewise.
10895 (vect_small_gap_p): Likewise.
10896 (vectorizable_with_step_bound_p): Likewise.
10897 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
10898 depencies if the vectorization factor is 1. Convert the checks
10899 for nonzero steps into checks on the bounds of DR_STEP. Try using
10900 a bunds check for variable steps if the minimum required step is
10901 relatively small. Update calls to the dr_with_seg_len
10902 constructor and to vect_compile_time_alias.
10903 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
10905 (vect_loop_versioning): Call it.
10906 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
10908 (vect_estimate_min_profitable_iters): Account for any bounds checks.
10910 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10911 Alan Hayward <alan.hayward@arm.com>
10912 David Sherwood <david.sherwood@arm.com>
10914 * doc/sourcebuild.texi (vect_scatter_store): Document.
10915 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
10917 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
10919 * genopinit.c (main): Add supports_vec_scatter_store and
10920 supports_vec_scatter_store_cached to target_optabs.
10921 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
10922 IFN_MASK_SCATTER_STORE.
10923 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
10925 * internal-fn.h (internal_store_fn_p): Declare.
10926 (internal_fn_stored_value_index): Likewise.
10927 * internal-fn.c (scatter_store_direct): New macro.
10928 (expand_scatter_store_optab_fn): New function.
10929 (direct_scatter_store_optab_supported_p): New macro.
10930 (internal_store_fn_p): New function.
10931 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
10932 IFN_MASK_SCATTER_STORE.
10933 (internal_fn_mask_index): Likewise.
10934 (internal_fn_stored_value_index): New function.
10935 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
10936 for scatter stores.
10937 * optabs-query.h (supports_vec_scatter_store_p): Declare.
10938 * optabs-query.c (supports_vec_scatter_store_p): New function.
10939 * tree-vectorizer.h (vect_get_store_rhs): Declare.
10940 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
10941 true for scatter stores.
10942 (vect_gather_scatter_fn_p): Handle scatter stores too.
10943 (vect_check_gather_scatter): Consider using scatter stores if
10944 supports_vec_scatter_store_p.
10945 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
10946 scatter stores too.
10947 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
10948 internal_fn_stored_value_index.
10949 (check_load_store_masking): Handle scatter stores too.
10950 (vect_get_store_rhs): Make public.
10951 (vectorizable_call): Use internal_store_fn_p.
10952 (vectorizable_store): Handle scatter store internal functions.
10953 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
10954 when deciding whether the end of the group has been reached.
10955 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
10956 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
10957 (mask_scatter_store<mode>): New insns.
10959 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10960 Alan Hayward <alan.hayward@arm.com>
10961 David Sherwood <david.sherwood@arm.com>
10963 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
10964 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
10965 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
10967 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
10968 Use vect_truncate_gather_scatter_offset if we can't treat the
10969 operation as a normal gather load or scatter store.
10970 (get_group_load_store_type): Take the gather_scatter_info
10971 as argument. Try using a gather load or scatter store for
10972 single-element groups.
10973 (get_load_store_type): Update calls to get_group_load_store_type
10974 and vect_use_strided_gather_scatters_p.
10976 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10977 Alan Hayward <alan.hayward@arm.com>
10978 David Sherwood <david.sherwood@arm.com>
10980 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
10981 optional tree argument.
10982 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
10984 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
10985 but continue to use the current value as a fallback.
10986 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
10987 to compare the updates.
10988 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
10989 (get_load_store_type): Use it when handling a strided access.
10990 (vect_get_strided_load_store_ops): New function.
10991 (vect_get_data_ptr_increment): Likewise.
10992 (vectorizable_load): Handle strided gather loads. Always pass
10993 a step to vect_create_data_ref_ptr and bump_vector_ptr.
10995 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10996 Alan Hayward <alan.hayward@arm.com>
10997 David Sherwood <david.sherwood@arm.com>
10999 * doc/md.texi (gather_load@var{m}): Document.
11000 (mask_gather_load@var{m}): Likewise.
11001 * genopinit.c (main): Add supports_vec_gather_load and
11002 supports_vec_gather_load_cached to target_optabs.
11003 * optabs-tree.c (init_tree_optimization_optabs): Use
11004 ggc_cleared_alloc to allocate target_optabs.
11005 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
11006 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
11008 * internal-fn.h (internal_load_fn_p): Declare.
11009 (internal_gather_scatter_fn_p): Likewise.
11010 (internal_fn_mask_index): Likewise.
11011 (internal_gather_scatter_fn_supported_p): Likewise.
11012 * internal-fn.c (gather_load_direct): New macro.
11013 (expand_gather_load_optab_fn): New function.
11014 (direct_gather_load_optab_supported_p): New macro.
11015 (direct_internal_fn_optab): New function.
11016 (internal_load_fn_p): Likewise.
11017 (internal_gather_scatter_fn_p): Likewise.
11018 (internal_fn_mask_index): Likewise.
11019 (internal_gather_scatter_fn_supported_p): Likewise.
11020 * optabs-query.c (supports_at_least_one_mode_p): New function.
11021 (supports_vec_gather_load_p): Likewise.
11022 * optabs-query.h (supports_vec_gather_load_p): Declare.
11023 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
11024 and memory_type field.
11025 (NUM_PATTERNS): Bump to 15.
11026 * tree-vect-data-refs.c: Include internal-fn.h.
11027 (vect_gather_scatter_fn_p): New function.
11028 (vect_describe_gather_scatter_call): Likewise.
11029 (vect_check_gather_scatter): Try using internal functions for
11030 gather loads. Recognize existing calls to a gather load function.
11031 (vect_analyze_data_refs): Consider using gather loads if
11032 supports_vec_gather_load_p.
11033 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
11034 (vect_get_gather_scatter_offset_type): Likewise.
11035 (vect_convert_mask_for_vectype): Likewise.
11036 (vect_add_conversion_to_patterm): Likewise.
11037 (vect_try_gather_scatter_pattern): Likewise.
11038 (vect_recog_gather_scatter_pattern): New pattern recognizer.
11039 (vect_vect_recog_func_ptrs): Add it.
11040 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
11041 internal_fn_mask_index and internal_gather_scatter_fn_p.
11042 (check_load_store_masking): Take the gather_scatter_info as an
11043 argument and handle gather loads.
11044 (vect_get_gather_scatter_ops): New function.
11045 (vectorizable_call): Check internal_load_fn_p.
11046 (vectorizable_load): Likewise. Handle gather load internal
11048 (vectorizable_store): Update call to check_load_store_masking.
11049 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
11050 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
11051 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
11052 (aarch64_gather_scale_operand_d): New predicates.
11053 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
11054 (mask_gather_load<mode>): New insns.
11056 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11057 Alan Hayward <alan.hayward@arm.com>
11058 David Sherwood <david.sherwood@arm.com>
11060 * optabs.def (fold_left_plus_optab): New optab.
11061 * doc/md.texi (fold_left_plus_@var{m}): Document.
11062 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
11063 * internal-fn.c (fold_left_direct): Define.
11064 (expand_fold_left_optab_fn): Likewise.
11065 (direct_fold_left_optab_supported_p): Likewise.
11066 * fold-const-call.c (fold_const_fold_left): New function.
11067 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
11068 * tree-parloops.c (valid_reduction_p): New function.
11069 (gather_scalar_reductions): Use it.
11070 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
11071 (vect_finish_replace_stmt): Declare.
11072 * tree-vect-loop.c (fold_left_reduction_fn): New function.
11073 (needs_fold_left_reduction_p): New function, split out from...
11074 (vect_is_simple_reduction): ...here. Accept reductions that
11075 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
11076 (vect_force_simple_reduction): Also store the reduction type in
11077 the assignment's STMT_VINFO_REDUC_TYPE.
11078 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
11079 (merge_with_identity): New function.
11080 (vect_expand_fold_left): Likewise.
11081 (vectorize_fold_left_reduction): Likewise.
11082 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
11083 scalar phi in place for it. Check for target support and reject
11084 cases that would reassociate the operation. Defer the transform
11085 phase to vectorize_fold_left_reduction.
11086 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
11087 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
11088 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
11090 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11092 * tree-if-conv.c (predicate_mem_writes): Remove redundant
11093 call to ifc_temp_var.
11095 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11096 Alan Hayward <alan.hayward@arm.com>
11097 David Sherwood <david.sherwood@arm.com>
11099 * target.def (legitimize_address_displacement): Take the original
11100 offset as a poly_int.
11101 * targhooks.h (default_legitimize_address_displacement): Update
11103 * targhooks.c (default_legitimize_address_displacement): Likewise.
11104 * doc/tm.texi: Regenerate.
11105 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
11106 as an argument, moving assert of ad->disp == ad->disp_term to...
11107 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
11108 Try calling targetm.legitimize_address_displacement before expanding
11109 the address rather than afterwards, and adjust for the new interface.
11110 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
11111 Match the new hook interface. Handle SVE addresses.
11112 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
11113 new hook interface.
11115 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11117 * Makefile.in (OBJS): Add early-remat.o.
11118 * target.def (select_early_remat_modes): New hook.
11119 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
11120 * doc/tm.texi: Regenerate.
11121 * targhooks.h (default_select_early_remat_modes): Declare.
11122 * targhooks.c (default_select_early_remat_modes): New function.
11123 * timevar.def (TV_EARLY_REMAT): New timevar.
11124 * passes.def (pass_early_remat): New pass.
11125 * tree-pass.h (make_pass_early_remat): Declare.
11126 * early-remat.c: New file.
11127 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
11129 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
11131 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11132 Alan Hayward <alan.hayward@arm.com>
11133 David Sherwood <david.sherwood@arm.com>
11135 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
11136 vfm1 with a bound_epilog parameter.
11137 (vect_do_peeling): Update calls accordingly, and move the prologue
11138 call earlier in the function. Treat the base bound_epilog as 0 for
11139 fully-masked loops and retain vf - 1 for other loops. Add 1 to
11140 this base when peeling for gaps.
11141 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
11142 with fully-masked loops.
11143 (vect_estimate_min_profitable_iters): Handle the single peeled
11144 iteration in that case.
11146 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11147 Alan Hayward <alan.hayward@arm.com>
11148 David Sherwood <david.sherwood@arm.com>
11150 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
11151 single-element interleaving even if the size is not a power of 2.
11152 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
11153 accesses for single-element interleaving if the group size is
11156 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11157 Alan Hayward <alan.hayward@arm.com>
11158 David Sherwood <david.sherwood@arm.com>
11160 * doc/md.texi (fold_extract_last_@var{m}): Document.
11161 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
11162 * optabs.def (fold_extract_last_optab): New optab.
11163 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
11164 * internal-fn.c (fold_extract_direct): New macro.
11165 (expand_fold_extract_optab_fn): Likewise.
11166 (direct_fold_extract_optab_supported_p): Likewise.
11167 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
11168 * tree-vect-loop.c (vect_model_reduction_cost): Handle
11169 EXTRACT_LAST_REDUCTION.
11170 (get_initial_def_for_reduction): Do not create an initial vector
11171 for EXTRACT_LAST_REDUCTION reductions.
11172 (vectorizable_reduction): Leave the scalar phi in place for
11173 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
11174 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
11175 epilogue code for EXTRACT_LAST_REDUCTION and defer the
11176 transform phase to vectorizable_condition.
11177 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
11179 (vect_finish_stmt_generation): ...here.
11180 (vect_finish_replace_stmt): New function.
11181 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
11182 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
11184 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
11186 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11187 Alan Hayward <alan.hayward@arm.com>
11188 David Sherwood <david.sherwood@arm.com>
11190 * doc/md.texi (extract_last_@var{m}): Document.
11191 * optabs.def (extract_last_optab): New optab.
11192 * internal-fn.def (EXTRACT_LAST): New internal function.
11193 * internal-fn.c (cond_unary_direct): New macro.
11194 (expand_cond_unary_optab_fn): Likewise.
11195 (direct_cond_unary_optab_supported_p): Likewise.
11196 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
11197 loops using EXTRACT_LAST.
11198 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
11199 (extract_last_<mode>): ...this optab.
11200 (vec_extract<mode><Vel>): Update accordingly.
11202 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11203 Alan Hayward <alan.hayward@arm.com>
11204 David Sherwood <david.sherwood@arm.com>
11206 * target.def (empty_mask_is_expensive): New hook.
11207 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
11208 * doc/tm.texi: Regenerate.
11209 * targhooks.h (default_empty_mask_is_expensive): Declare.
11210 * targhooks.c (default_empty_mask_is_expensive): New function.
11211 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
11212 if the target says that empty masks are expensive.
11213 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
11215 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
11217 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11218 Alan Hayward <alan.hayward@arm.com>
11219 David Sherwood <david.sherwood@arm.com>
11221 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
11222 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
11223 (vect_use_loop_mask_for_alignment_p): New function.
11224 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
11225 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
11226 niters_skip argument. Make sure that the first niters_skip elements
11227 of the first iteration are inactive.
11228 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
11229 Update call to vect_set_loop_masks_directly.
11230 (get_misalign_in_elems): New function, split out from...
11231 (vect_gen_prolog_loop_niters): ...here.
11232 (vect_update_init_of_dr): Take a code argument that specifies whether
11233 the adjustment should be added or subtracted.
11234 (vect_update_init_of_drs): Likewise.
11235 (vect_prepare_for_masked_peels): New function.
11236 (vect_do_peeling): Skip prologue peeling if we're using a mask
11237 instead. Update call to vect_update_inits_of_drs.
11238 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11240 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
11241 alignment. Do not include the number of peeled iterations in
11242 the minimum threshold in that case.
11243 (vectorizable_induction): Adjust the start value down by
11244 LOOP_VINFO_MASK_SKIP_NITERS iterations.
11245 (vect_transform_loop): Call vect_prepare_for_masked_peels.
11246 Take the number of skipped iterations into account when calculating
11248 * tree-vect-stmts.c (vect_gen_while_not): New function.
11250 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11251 Alan Hayward <alan.hayward@arm.com>
11252 David Sherwood <david.sherwood@arm.com>
11254 * doc/sourcebuild.texi (vect_fully_masked): Document.
11255 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
11256 default value to 0.
11257 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
11259 (vect_analyze_loop_2): ...here. Don't check the vectorization
11260 factor against the number of loop iterations if the loop is
11263 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11264 Alan Hayward <alan.hayward@arm.com>
11265 David Sherwood <david.sherwood@arm.com>
11267 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
11268 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
11269 (dump_groups): Update accordingly.
11270 (iv_use::mem_type): New member variable.
11271 (address_p): New function.
11272 (record_use): Add a mem_type argument and initialize the new
11274 (record_group_use): Add a mem_type argument. Use address_p.
11275 Remove obsolete null checks of base_object. Update call to record_use.
11276 (find_interesting_uses_op): Update call to record_group_use.
11277 (find_interesting_uses_cond): Likewise.
11278 (find_interesting_uses_address): Likewise.
11279 (get_mem_type_for_internal_fn): New function.
11280 (find_address_like_use): Likewise.
11281 (find_interesting_uses_stmt): Try find_address_like_use before
11282 calling find_interesting_uses_op.
11283 (addr_offset_valid_p): Use the iv mem_type field as the type
11284 of the addressed memory.
11285 (add_autoinc_candidates): Likewise.
11286 (get_address_cost): Likewise.
11287 (split_small_address_groups_p): Use address_p.
11288 (split_address_groups): Likewise.
11289 (add_iv_candidate_for_use): Likewise.
11290 (autoinc_possible_for_pair): Likewise.
11291 (rewrite_groups): Likewise.
11292 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
11293 (determine_group_iv_cost): Update after split of USE_ADDRESS.
11294 (get_alias_ptr_type_for_ptr_address): New function.
11295 (rewrite_use_address): Rewrite address uses in calls that were
11296 identified by find_address_like_use.
11298 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11299 Alan Hayward <alan.hayward@arm.com>
11300 David Sherwood <david.sherwood@arm.com>
11302 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
11304 * gimple-expr.h (is_gimple_addressable: Likewise.
11305 * gimple-expr.c (is_gimple_address): Likewise.
11306 * internal-fn.c (expand_call_mem_ref): New function.
11307 (expand_mask_load_optab_fn): Use it.
11308 (expand_mask_store_optab_fn): Likewise.
11310 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11311 Alan Hayward <alan.hayward@arm.com>
11312 David Sherwood <david.sherwood@arm.com>
11314 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
11315 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
11316 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
11317 (cond_umax@var{mode}): Document.
11318 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
11319 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
11320 (cond_umin_optab, cond_umax_optab): New optabs.
11321 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
11322 (COND_IOR, COND_XOR): New internal functions.
11323 * internal-fn.h (get_conditional_internal_fn): Declare.
11324 * internal-fn.c (cond_binary_direct): New macro.
11325 (expand_cond_binary_optab_fn): Likewise.
11326 (direct_cond_binary_optab_supported_p): Likewise.
11327 (get_conditional_internal_fn): New function.
11328 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
11329 Cope with reduction statements that are vectorized as calls rather
11331 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
11332 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
11333 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
11334 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
11335 (UNSPEC_COND_EOR): New unspecs.
11336 (optab): Add mappings for them.
11337 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
11338 (sve_int_op, sve_fp_op): New int attributes.
11340 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11341 Alan Hayward <alan.hayward@arm.com>
11342 David Sherwood <david.sherwood@arm.com>
11344 * optabs.def (while_ult_optab): New optab.
11345 * doc/md.texi (while_ult@var{m}@var{n}): Document.
11346 * internal-fn.def (WHILE_ULT): New internal function.
11347 * internal-fn.h (direct_internal_fn_supported_p): New override
11348 that takes two types as argument.
11349 * internal-fn.c (while_direct): New macro.
11350 (expand_while_optab_fn): New function.
11351 (convert_optab_supported_p): Likewise.
11352 (direct_while_optab_supported_p): New macro.
11353 * wide-int.h (wi::udiv_ceil): New function.
11354 * tree-vectorizer.h (rgroup_masks): New structure.
11355 (vec_loop_masks): New typedef.
11356 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
11357 and fully_masked_p.
11358 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
11359 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
11360 (vect_max_vf): New function.
11361 (slpeel_make_loop_iterate_ntimes): Delete.
11362 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
11363 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
11364 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
11365 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
11366 internal-fn.h, stor-layout.h and optabs-query.h.
11367 (vect_set_loop_mask): New function.
11368 (add_preheader_seq): Likewise.
11369 (add_header_seq): Likewise.
11370 (interleave_supported_p): Likewise.
11371 (vect_maybe_permute_loop_masks): Likewise.
11372 (vect_set_loop_masks_directly): Likewise.
11373 (vect_set_loop_condition_masked): Likewise.
11374 (vect_set_loop_condition_unmasked): New function, split out from
11375 slpeel_make_loop_iterate_ntimes.
11376 (slpeel_make_loop_iterate_ntimes): Rename to..
11377 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
11378 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
11379 (vect_do_peeling): Update call accordingly.
11380 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
11382 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11383 mask_compare_type, can_fully_mask_p and fully_masked_p.
11384 (release_vec_loop_masks): New function.
11385 (_loop_vec_info): Use it to free the loop masks.
11386 (can_produce_all_loop_masks_p): New function.
11387 (vect_get_max_nscalars_per_iter): Likewise.
11388 (vect_verify_full_masking): Likewise.
11389 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
11390 retries, and free the mask rgroups before retrying. Check loop-wide
11391 reasons for disallowing fully-masked loops. Make the final decision
11392 about whether use a fully-masked loop or not.
11393 (vect_estimate_min_profitable_iters): Do not assume that peeling
11394 for the number of iterations will be needed for fully-masked loops.
11395 (vectorizable_reduction): Disable fully-masked loops.
11396 (vectorizable_live_operation): Likewise.
11397 (vect_halve_mask_nunits): New function.
11398 (vect_double_mask_nunits): Likewise.
11399 (vect_record_loop_mask): Likewise.
11400 (vect_get_loop_mask): Likewise.
11401 (vect_transform_loop): Handle the case in which the final loop
11402 iteration might handle a partial vector. Call vect_set_loop_condition
11403 instead of slpeel_make_loop_iterate_ntimes.
11404 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
11405 (check_load_store_masking): New function.
11406 (prepare_load_store_mask): Likewise.
11407 (vectorizable_store): Handle fully-masked loops.
11408 (vectorizable_load): Likewise.
11409 (supportable_widening_operation): Use vect_halve_mask_nunits for
11411 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
11412 (vect_gen_while): New function.
11413 * config/aarch64/aarch64.md (umax<mode>3): New expander.
11414 (aarch64_uqdec<mode>): New insn.
11416 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11417 Alan Hayward <alan.hayward@arm.com>
11418 David Sherwood <david.sherwood@arm.com>
11420 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
11421 (reduc_xor_scal_optab): New optabs.
11422 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
11423 (reduc_xor_scal_@var{m}): Document.
11424 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
11425 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
11426 internal functions.
11427 * fold-const-call.c (fold_const_call): Handle them.
11428 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
11429 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
11430 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
11431 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
11432 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
11433 (UNSPEC_XORV): New unspecs.
11434 (optab): Add entries for them.
11435 (BITWISEV): New int iterator.
11436 (bit_reduc_op): New int attributes.
11438 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11439 Alan Hayward <alan.hayward@arm.com>
11440 David Sherwood <david.sherwood@arm.com>
11442 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
11443 * internal-fn.def (VEC_SHL_INSERT): New internal function.
11444 * optabs.def (vec_shl_insert_optab): New optab.
11445 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
11446 (duplicate_and_interleave): Likewise.
11447 * tree-vect-loop.c: Include internal-fn.h.
11448 (neutral_op_for_slp_reduction): New function, split out from
11449 get_initial_defs_for_reduction.
11450 (get_initial_def_for_reduction): Handle option 2 for variable-length
11451 vectors by loading the neutral value into a vector and then shifting
11452 the initial value into element 0.
11453 (get_initial_defs_for_reduction): Replace the code argument with
11454 the neutral value calculated by neutral_op_for_slp_reduction.
11455 Use gimple_build_vector for constant-length vectors.
11456 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
11457 but the first group_size elements have a neutral value.
11458 Use duplicate_and_interleave otherwise.
11459 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
11460 Update call to get_initial_defs_for_reduction. Handle SLP
11461 reductions for variable-length vectors by creating one vector
11462 result for each scalar result, with the elements associated
11463 with other scalar results stubbed out with the neutral value.
11464 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
11465 Require IFN_VEC_SHL_INSERT for double reductions on
11466 variable-length vectors, or SLP reductions that have
11467 a neutral value. Require can_duplicate_and_interleave_p
11468 support for variable-length unchained SLP reductions if there
11469 is no neutral value, such as for MIN/MAX reductions. Also require
11470 the number of vector elements to be a multiple of the number of
11471 SLP statements when doing variable-length unchained SLP reductions.
11472 Update call to vect_create_epilog_for_reduction.
11473 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
11474 and remove initial values.
11475 (duplicate_and_interleave): Make public.
11476 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
11477 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
11479 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11480 Alan Hayward <alan.hayward@arm.com>
11481 David Sherwood <david.sherwood@arm.com>
11483 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
11484 (can_duplicate_and_interleave_p): New function.
11485 (vect_get_and_check_slp_defs): Take the vector of statements
11486 rather than just the current one. Remove excess parentheses.
11487 Restriction rejectinon of vect_constant_def and vect_external_def
11488 for variable-length vectors to boolean types, or types for which
11489 can_duplicate_and_interleave_p is false.
11490 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
11491 (duplicate_and_interleave): New function.
11492 (vect_get_constant_vectors): Use gimple_build_vector for
11493 constant-length vectors and suitable variable-length constant
11494 vectors. Use duplicate_and_interleave for other variable-length
11495 vectors. Don't defer the update when inserting new statements.
11497 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11498 Alan Hayward <alan.hayward@arm.com>
11499 David Sherwood <david.sherwood@arm.com>
11501 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
11502 min_profitable_iters doesn't go negative.
11504 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11505 Alan Hayward <alan.hayward@arm.com>
11506 David Sherwood <david.sherwood@arm.com>
11508 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
11509 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
11510 * optabs.def (vec_mask_load_lanes_optab): New optab.
11511 (vec_mask_store_lanes_optab): Likewise.
11512 * internal-fn.def (MASK_LOAD_LANES): New internal function.
11513 (MASK_STORE_LANES): Likewise.
11514 * internal-fn.c (mask_load_lanes_direct): New macro.
11515 (mask_store_lanes_direct): Likewise.
11516 (expand_mask_load_optab_fn): Handle masked operations.
11517 (expand_mask_load_lanes_optab_fn): New macro.
11518 (expand_mask_store_optab_fn): Handle masked operations.
11519 (expand_mask_store_lanes_optab_fn): New macro.
11520 (direct_mask_load_lanes_optab_supported_p): Likewise.
11521 (direct_mask_store_lanes_optab_supported_p): Likewise.
11522 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
11524 (vect_load_lanes_supported): Likewise.
11525 * tree-vect-data-refs.c (strip_conversion): New function.
11526 (can_group_stmts_p): Likewise.
11527 (vect_analyze_data_ref_accesses): Use it instead of checking
11528 for a pair of assignments.
11529 (vect_store_lanes_supported): Take a masked_p parameter.
11530 (vect_load_lanes_supported): Likewise.
11531 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
11532 vect_store_lanes_supported and vect_load_lanes_supported.
11533 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
11534 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
11535 parameter. Don't allow gaps for masked accesses.
11536 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
11537 and vect_load_lanes_supported.
11538 (get_load_store_type): Take a masked_p parameter and update
11539 call to get_group_load_store_type.
11540 (vectorizable_store): Update call to get_load_store_type.
11541 Handle IFN_MASK_STORE_LANES.
11542 (vectorizable_load): Update call to get_load_store_type.
11543 Handle IFN_MASK_LOAD_LANES.
11545 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11546 Alan Hayward <alan.hayward@arm.com>
11547 David Sherwood <david.sherwood@arm.com>
11549 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
11551 * config/aarch64/aarch64-protos.h
11552 (aarch64_sve_struct_memory_operand_p): Declare.
11553 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
11554 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
11555 (VPRED, vpred): Handle SVE structure modes.
11556 * config/aarch64/constraints.md (Utx): New constraint.
11557 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
11558 (aarch64_sve_struct_nonimmediate_operand): New predicates.
11559 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
11560 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
11561 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
11562 structure modes. Split into pieces after RA.
11563 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
11564 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
11566 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
11567 SVE structure modes.
11568 (aarch64_classify_address): Likewise.
11569 (sizetochar): Move earlier in file.
11570 (aarch64_print_operand): Handle SVE register lists.
11571 (aarch64_array_mode): New function.
11572 (aarch64_sve_struct_memory_operand_p): Likewise.
11573 (TARGET_ARRAY_MODE): Redefine.
11575 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11576 Alan Hayward <alan.hayward@arm.com>
11577 David Sherwood <david.sherwood@arm.com>
11579 * target.def (array_mode): New target hook.
11580 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
11581 * doc/tm.texi: Regenerate.
11582 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
11583 * hooks.c (hook_optmode_mode_uhwi_none): New function.
11584 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
11585 targetm.array_mode.
11586 * stor-layout.c (mode_for_array): Likewise. Support polynomial
11589 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11590 Alan Hayward <alan.hayward@arm.com>
11591 David Sherwood <david.sherwood@arm.com>
11593 * fold-const.c (fold_binary_loc): Check the argument types
11594 rather than the result type when testing for a vector operation.
11596 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11598 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
11599 * doc/tm.texi: Regenerate.
11601 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11602 Alan Hayward <alan.hayward@arm.com>
11603 David Sherwood <david.sherwood@arm.com>
11605 * doc/invoke.texi (-msve-vector-bits=): Document new option.
11606 (sve): Document new AArch64 extension.
11607 * doc/md.texi (w): Extend the description of the AArch64
11608 constraint to include SVE vectors.
11609 (Upl, Upa): Document new AArch64 predicate constraints.
11610 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
11612 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
11613 (msve-vector-bits=): New option.
11614 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
11615 SVE when these are disabled.
11616 (sve): New extension.
11617 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
11618 modes. Adjust their number of units based on aarch64_sve_vg.
11619 (MAX_BITSIZE_MODE_ANY_MODE): Define.
11620 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
11621 aarch64_addr_query_type.
11622 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
11623 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
11624 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
11625 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
11626 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
11627 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
11628 (aarch64_simd_imm_zero_p): Delete.
11629 (aarch64_check_zero_based_sve_index_immediate): Declare.
11630 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
11631 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
11632 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
11633 (aarch64_sve_float_mul_immediate_p): Likewise.
11634 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
11635 rather than an rtx.
11636 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
11637 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
11638 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
11639 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
11640 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
11641 (aarch64_regmode_natural_size): Likewise.
11642 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
11643 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
11645 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
11646 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
11647 for VG and the SVE predicate registers.
11648 (V_ALIASES): Add a "z"-prefixed alias.
11649 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
11650 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
11651 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
11652 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
11653 (REG_CLASS_NAMES): Add entries for them.
11654 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
11655 and the predicate registers.
11656 (aarch64_sve_vg): Declare.
11657 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
11658 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
11659 (REGMODE_NATURAL_SIZE): Define.
11660 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
11662 * config/aarch64/aarch64.c: Include cfgrtl.h.
11663 (simd_immediate_info): Add a constructor for series vectors,
11664 and an associated step field.
11665 (aarch64_sve_vg): New variable.
11666 (aarch64_dbx_register_number): Handle VG and the predicate registers.
11667 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
11668 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
11669 (VEC_ANY_DATA, VEC_STRUCT): New constants.
11670 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
11671 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
11672 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
11673 (aarch64_get_mask_mode): New functions.
11674 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
11675 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
11676 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
11677 predicate modes and predicate registers. Explicitly restrict
11678 GPRs to modes of 16 bytes or smaller. Only allow FP registers
11679 to store a vector mode if it is recognized by
11680 aarch64_classify_vector_mode.
11681 (aarch64_regmode_natural_size): New function.
11682 (aarch64_hard_regno_caller_save_mode): Return the original mode
11684 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
11685 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
11686 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
11687 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
11689 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
11690 does not overlap dest if the function is frame-related. Handle
11692 (aarch64_split_add_offset): New function.
11693 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
11694 them aarch64_add_offset.
11695 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
11696 and update call to aarch64_sub_sp.
11697 (aarch64_add_cfa_expression): New function.
11698 (aarch64_expand_prologue): Pass extra temporary registers to the
11699 functions above. Handle the case in which we need to emit new
11700 DW_CFA_expressions for registers that were originally saved
11701 relative to the stack pointer, but now have to be expressed
11702 relative to the frame pointer.
11703 (aarch64_output_mi_thunk): Pass extra temporary registers to the
11705 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
11706 IP0 and IP1 values for SVE frames.
11707 (aarch64_expand_vec_series): New function.
11708 (aarch64_expand_sve_widened_duplicate): Likewise.
11709 (aarch64_expand_sve_const_vector): Likewise.
11710 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
11711 Handle SVE constants. Use emit_move_insn to move a force_const_mem
11712 into the register, rather than emitting a SET directly.
11713 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
11714 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
11715 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
11716 (offset_9bit_signed_scaled_p): New functions.
11717 (aarch64_replicate_bitmask_imm): New function.
11718 (aarch64_bitmask_imm): Use it.
11719 (aarch64_cannot_force_const_mem): Reject expressions involving
11720 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
11721 (aarch64_classify_index): Handle SVE indices, by requiring
11722 a plain register index with a scale that matches the element size.
11723 (aarch64_classify_address): Handle SVE addresses. Assert that
11724 the mode of the address is VOIDmode or an integer mode.
11725 Update call to aarch64_classify_symbol.
11726 (aarch64_classify_symbolic_expression): Update call to
11727 aarch64_classify_symbol.
11728 (aarch64_const_vec_all_in_range_p): New function.
11729 (aarch64_print_vector_float_operand): Likewise.
11730 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
11731 "vN" for FP registers with SVE modes. Handle (const ...) vectors
11732 and the FP immediates 1.0 and 0.5.
11733 (aarch64_print_address_internal): Handle SVE addresses.
11734 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
11735 (aarch64_regno_regclass): Handle predicate registers.
11736 (aarch64_secondary_reload): Handle big-endian reloads of SVE
11738 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
11739 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
11740 (aarch64_convert_sve_vector_bits): New function.
11741 (aarch64_override_options): Use it to handle -msve-vector-bits=.
11742 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
11743 rather than an rtx.
11744 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
11745 Handle SVE vector and predicate modes. Accept VL-based constants
11746 that need only one temporary register, and VL offsets that require
11747 no temporary registers.
11748 (aarch64_conditional_register_usage): Mark the predicate registers
11749 as fixed if SVE isn't available.
11750 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
11751 Return true for SVE vector and predicate modes.
11752 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
11753 rather than an unsigned int. Handle SVE modes.
11754 (aarch64_preferred_simd_mode): Update call accordingly. Handle
11756 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
11758 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
11759 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
11760 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
11761 (aarch64_sve_float_mul_immediate_p): New functions.
11762 (aarch64_sve_valid_immediate): New function.
11763 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
11764 Explicitly reject structure modes. Check for INDEX constants.
11765 Handle PTRUE and PFALSE constants.
11766 (aarch64_check_zero_based_sve_index_immediate): New function.
11767 (aarch64_simd_imm_zero_p): Delete.
11768 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
11769 vector modes. Accept constants in the range of CNT[BHWD].
11770 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
11771 ask for an Advanced SIMD mode.
11772 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
11773 (aarch64_simd_vector_alignment): Handle SVE predicates.
11774 (aarch64_vectorize_preferred_vector_alignment): New function.
11775 (aarch64_simd_vector_alignment_reachable): Use it instead of
11777 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
11778 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
11780 (MAX_VECT_LEN): Delete.
11781 (expand_vec_perm_d): Add a vec_flags field.
11782 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
11783 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
11784 (aarch64_evpc_ext): Don't apply a big-endian lane correction
11786 (aarch64_evpc_rev): Rename to...
11787 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
11788 (aarch64_evpc_rev_global): New function.
11789 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
11790 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
11792 (aarch64_evpc_sve_tbl): New function.
11793 (aarch64_expand_vec_perm_const_1): Update after rename of
11794 aarch64_evpc_rev. Handle SVE permutes too, trying
11795 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
11796 than aarch64_evpc_tbl.
11797 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
11798 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
11799 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
11800 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
11801 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
11802 (aarch64_expand_sve_vcond): New functions.
11803 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
11804 of aarch64_vector_mode_p.
11805 (aarch64_dwarf_poly_indeterminate_value): New function.
11806 (aarch64_compute_pressure_classes): Likewise.
11807 (aarch64_can_change_mode_class): Likewise.
11808 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
11809 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
11810 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
11811 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
11812 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
11813 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
11814 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
11815 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
11817 (Dn, Dl, Dr): Accept const as well as const_vector.
11818 (Dz): Likewise. Compare against CONST0_RTX.
11819 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
11820 of "vector" where appropriate.
11821 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
11822 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
11823 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
11824 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
11825 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
11826 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
11827 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
11828 (v_int_equiv): Extend to SVE modes.
11829 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
11831 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
11832 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
11833 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
11834 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
11835 (SVE_COND_FP_CMP): New int iterators.
11836 (perm_hilo): Handle the new unpack unspecs.
11837 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
11839 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
11840 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
11841 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
11842 (aarch64_equality_operator, aarch64_constant_vector_operand)
11843 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
11844 (aarch64_sve_nonimmediate_operand): Likewise.
11845 (aarch64_sve_general_operand): Likewise.
11846 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
11847 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
11848 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
11849 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
11850 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
11851 (aarch64_sve_float_arith_immediate): Likewise.
11852 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
11853 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
11854 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
11855 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
11856 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
11857 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
11858 (aarch64_sve_float_arith_operand): Likewise.
11859 (aarch64_sve_float_arith_with_sub_operand): Likewise.
11860 (aarch64_sve_float_mul_operand): Likewise.
11861 (aarch64_sve_vec_perm_operand): Likewise.
11862 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
11863 (aarch64_mov_operand): Accept const_poly_int and const_vector.
11864 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
11865 as well as const_vector.
11866 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
11867 in file. Use CONST0_RTX and CONSTM1_RTX.
11868 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
11869 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
11870 Use aarch64_simd_imm_zero.
11871 * config/aarch64/aarch64-sve.md: New file.
11872 * config/aarch64/aarch64.md: Include it.
11873 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
11874 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
11875 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
11876 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
11877 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
11878 (sve): New attribute.
11879 (enabled): Disable instructions with the sve attribute unless
11881 (movqi, movhi): Pass CONST_POLY_INT operaneds through
11882 aarch64_expand_mov_immediate.
11883 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
11884 CNT[BHSD] immediates.
11885 (movti): Split CONST_POLY_INT moves into two halves.
11886 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
11887 Split additions that need a temporary here if the destination
11888 is the stack pointer.
11889 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
11890 (*add<mode>3_poly_1): New instruction.
11891 (set_clobber_cc): New expander.
11893 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11895 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
11896 parameter and use it instead of GET_MODE_SIZE (innermode). Use
11897 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
11898 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
11899 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
11900 Change innermode from fixed_mode_size to machine_mode.
11901 (simplify_subreg): Update call accordingly. Handle a constant-sized
11902 subreg of a variable-length CONST_VECTOR.
11904 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11905 Alan Hayward <alan.hayward@arm.com>
11906 David Sherwood <david.sherwood@arm.com>
11908 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
11909 (add_offset_to_base): New function, split out from...
11910 (create_mem_ref): ...here. When handling a scale other than 1,
11911 check first whether the address is valid without the offset.
11912 Add it into the base if so, leaving the index and scale as-is.
11914 2018-01-12 Jakub Jelinek <jakub@redhat.com>
11917 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
11918 fold_for_warn before checking if arg2 is INTEGER_CST.
11920 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
11922 * config/rs6000/predicates.md (load_multiple_operation): Delete.
11923 (store_multiple_operation): Delete.
11924 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
11925 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
11926 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
11927 guarded by TARGET_STRING.
11928 (rs6000_output_load_multiple): Delete.
11929 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
11930 OPTION_MASK_STRING / TARGET_STRING handling.
11931 (print_operand) <'N', 'O'>: Add comment that these are unused now.
11932 (const rs6000_opt_masks) <"string">: Change mask to 0.
11933 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
11934 (MASK_STRING): Delete.
11935 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
11937 (load_multiple): Delete.
11944 (store_multiple): Delete.
11951 (movmemsi_8reg): Delete.
11952 (corresponding unnamed define_insn): Delete.
11953 (movmemsi_6reg): Delete.
11954 (corresponding unnamed define_insn): Delete.
11955 (movmemsi_4reg): Delete.
11956 (corresponding unnamed define_insn): Delete.
11957 (movmemsi_2reg): Delete.
11958 (corresponding unnamed define_insn): Delete.
11959 (movmemsi_1reg): Delete.
11960 (corresponding unnamed define_insn): Delete.
11961 * config/rs6000/rs6000.opt (mno-string): New.
11962 (mstring): Replace by deprecation warning stub.
11963 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
11965 2018-01-12 Jakub Jelinek <jakub@redhat.com>
11967 * regrename.c (regrename_do_replace): If replacing the same
11968 reg multiple times, try to reuse last created gen_raw_REG.
11971 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
11972 main to workaround a bug in GDB.
11974 2018-01-12 Tom de Vries <tom@codesourcery.com>
11977 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
11979 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
11981 PR rtl-optimization/80481
11982 * ira-color.c (get_cap_member): New function.
11983 (allocnos_conflict_by_live_ranges_p): Use it.
11984 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
11985 (setup_slot_coalesced_allocno_live_ranges): Ditto.
11987 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
11990 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
11991 (*saddl_se_1): Ditto.
11992 (*ssubsi_1): Ditto.
11993 (*ssubl_se_1): Ditto.
11995 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
11997 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
11998 rather than wi::to_widest for DR_INITs.
11999 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
12000 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
12001 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
12003 (vect_analyze_group_access_1): Note that here.
12005 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
12007 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
12008 polynomial type sizes.
12010 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
12012 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
12013 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
12014 (gimple_add_tmp_var): Likewise.
12016 2018-01-12 Martin Liska <mliska@suse.cz>
12018 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
12019 (gimple_alloc_sizes): Likewise.
12020 (dump_gimple_statistics): Use PRIu64 in printf format.
12021 * gimple.h: Change uint64_t to int.
12023 2018-01-12 Martin Liska <mliska@suse.cz>
12025 * tree-core.h: Use uint64_t instead of int.
12026 * tree.c (tree_node_counts): Likewise.
12027 (tree_node_sizes): Likewise.
12028 (dump_tree_statistics): Use PRIu64 in printf format.
12030 2018-01-12 Martin Liska <mliska@suse.cz>
12032 * Makefile.in: As qsort_chk is implemented in vec.c, add
12033 vec.o to linkage of gencfn-macros.
12034 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
12035 passing the info to record_node_allocation_statistics.
12036 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
12038 * ggc-common.c (struct ggc_usage): Add operator== and use
12039 it in operator< and compare function.
12040 * mem-stats.h (struct mem_usage): Likewise.
12041 * vec.c (struct vec_usage): Remove operator< and compare
12042 function. Can be simply inherited.
12044 2018-01-12 Martin Jambor <mjambor@suse.cz>
12047 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
12048 * tree-ssa-math-opts.c: Include domwalk.h.
12049 (convert_mult_to_fma_1): New function.
12050 (fma_transformation_info): New type.
12051 (fma_deferring_state): Likewise.
12052 (cancel_fma_deferring): New function.
12053 (result_of_phi): Likewise.
12054 (last_fma_candidate_feeds_initial_phi): Likewise.
12055 (convert_mult_to_fma): Added deferring logic, split actual
12056 transformation to convert_mult_to_fma_1.
12057 (math_opts_dom_walker): New type.
12058 (math_opts_dom_walker::after_dom_children): New method, body moved
12059 here from pass_optimize_widening_mul::execute, added deferring logic
12061 (pass_optimize_widening_mul::execute): Moved most of code to
12062 math_opts_dom_walker::after_dom_children.
12063 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
12064 * config/i386/i386.c (ix86_option_override_internal): Added
12065 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
12067 2018-01-12 Richard Biener <rguenther@suse.de>
12070 * dwarf2out.c (gen_variable_die): Do not reset old_die for
12071 inline instance vars.
12073 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
12076 * config/rx/rx.c (rx_is_restricted_memory_address):
12077 Handle SUBREG case.
12079 2018-01-12 Richard Biener <rguenther@suse.de>
12081 PR tree-optimization/80846
12082 * target.def (split_reduction): New target hook.
12083 * targhooks.c (default_split_reduction): New function.
12084 * targhooks.h (default_split_reduction): Declare.
12085 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
12086 target requests first reduce vectors by combining low and high
12088 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
12089 (get_vectype_for_scalar_type_and_size): Export.
12090 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
12091 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
12092 * doc/tm.texi: Regenerate.
12093 * config/i386/i386.c (ix86_split_reduction): Implement
12094 TARGET_VECTORIZE_SPLIT_REDUCTION.
12096 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
12099 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
12100 in PIC mode except for TARGET_VXWORKS_RTP.
12101 * config/sparc/sparc.c: Include cfgrtl.h.
12102 (TARGET_INIT_PIC_REG): Define.
12103 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
12104 (sparc_pic_register_p): New predicate.
12105 (sparc_legitimate_address_p): Use it.
12106 (sparc_legitimize_pic_address): Likewise.
12107 (sparc_delegitimize_address): Likewise.
12108 (sparc_mode_dependent_address_p): Likewise.
12109 (gen_load_pcrel_sym): Remove 4th parameter.
12110 (load_got_register): Adjust call to above. Remove obsolete stuff.
12111 (sparc_expand_prologue): Do not call load_got_register here.
12112 (sparc_flat_expand_prologue): Likewise.
12113 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
12114 (sparc_use_pseudo_pic_reg): New function.
12115 (sparc_init_pic_reg): Likewise.
12116 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
12117 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
12119 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
12121 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
12122 Add item for branch_cost.
12124 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
12126 PR rtl-optimization/83565
12127 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
12128 not extend the result to a larger mode for rotate operations.
12129 (num_sign_bit_copies1): Likewise.
12131 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12134 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
12136 Use values-Xc.o for -pedantic.
12137 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
12139 2018-01-12 Martin Liska <mliska@suse.cz>
12142 * ipa-devirt.c (final_warning_record::grow_type_warnings):
12144 (possible_polymorphic_call_targets): Use it.
12145 (ipa_devirt): Likewise.
12147 2018-01-12 Martin Liska <mliska@suse.cz>
12149 * profile-count.h (enum profile_quality): Use 0 as invalid
12150 enum value of profile_quality.
12152 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
12154 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
12155 -mext-string options.
12157 2018-01-12 Richard Biener <rguenther@suse.de>
12159 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
12160 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
12161 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
12163 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
12165 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
12167 * configure.ac (--with-long-double-format): Add support for the
12168 configuration option to change the default long double format on
12170 * config.gcc (powerpc*-linux*-*): Likewise.
12171 * configure: Regenerate.
12172 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
12173 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
12174 used without modification.
12176 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12178 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
12179 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
12180 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
12181 MISC_BUILTIN_SPEC_BARRIER.
12182 (rs6000_init_builtins): Likewise.
12183 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
12185 (speculation_barrier): New define_insn.
12186 * doc/extend.texi: Document __builtin_speculation_barrier.
12188 2018-01-11 Jakub Jelinek <jakub@redhat.com>
12191 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
12192 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
12193 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
12195 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
12196 integral modes instead of "ss" and "sd".
12197 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
12198 vectors with 32-bit and 64-bit elements.
12199 (vecdupssescalarmodesuffix): New mode attribute.
12200 (vec_dup<mode>): Use it.
12202 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
12205 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
12206 frame if argument is passed on stack.
12208 2018-01-11 Jakub Jelinek <jakub@redhat.com>
12211 * ree.c (combine_reaching_defs): Optimize also
12212 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
12213 reg2=any_extend(exp); reg1=reg2;, formatting fix.
12215 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
12217 PR middle-end/83189
12218 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
12220 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
12222 PR middle-end/83718
12223 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
12224 after they are computed.
12226 2018-01-11 Bin Cheng <bin.cheng@arm.com>
12228 PR tree-optimization/83695
12229 * gimple-loop-linterchange.cc
12230 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
12231 reset cached scev information after interchange.
12232 (pass_linterchange::execute): Remove call to scev_reset_htab.
12234 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12236 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
12237 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
12238 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
12239 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
12240 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
12241 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
12242 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
12243 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
12244 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
12245 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
12246 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
12247 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
12248 (V_lane_reg): Likewise.
12249 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
12251 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
12252 (vfmal_lane_low<mode>_intrinsic,
12253 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
12254 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
12255 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
12256 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
12257 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
12258 vfmsl_lane_high<mode>_intrinsic): New define_insns.
12260 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12262 * config/arm/arm-cpus.in (fp16fml): New feature.
12263 (ALL_SIMD): Add fp16fml.
12264 (armv8.2-a): Add fp16fml as an option.
12265 (armv8.3-a): Likewise.
12266 (armv8.4-a): Add fp16fml as part of fp16.
12267 * config/arm/arm.h (TARGET_FP16FML): Define.
12268 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
12270 * config/arm/arm-modes.def (V2HF): Define.
12271 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
12272 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
12273 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
12274 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
12275 vfmsl_low, vfmsl_high): New set of builtins.
12276 * config/arm/iterators.md (PLUSMINUS): New code iterator.
12277 (vfml_op): New code attribute.
12278 (VFMLHALVES): New int iterator.
12279 (VFML, VFMLSEL): New mode attributes.
12280 (V_reg): Define mapping for V2HF.
12281 (V_hi, V_lo): New mode attributes.
12282 (VF_constraint): Likewise.
12283 (vfml_half, vfml_half_selector): New int attributes.
12284 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
12286 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
12287 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
12289 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
12290 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
12291 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
12292 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
12294 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
12295 Document new effective target and option set.
12297 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12299 * config/arm/arm-cpus.in (armv8_4): New feature.
12300 (ARMv8_4a): New fgroup.
12301 (armv8.4-a): New arch.
12302 * config/arm/arm-tables.opt: Regenerate.
12303 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
12304 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
12305 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
12306 Add matching rules for -march=armv8.4-a and extensions.
12307 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
12309 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
12312 * config/rx/rx.md (BW): New mode attribute.
12313 (sync_lock_test_and_setsi): Add mode suffix to insn output.
12315 2018-01-11 Richard Biener <rguenther@suse.de>
12317 PR tree-optimization/83435
12318 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
12319 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
12320 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
12322 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12323 Alan Hayward <alan.hayward@arm.com>
12324 David Sherwood <david.sherwood@arm.com>
12326 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
12328 (aarch64_classify_address): Initialize it. Track polynomial offsets.
12329 (aarch64_print_address_internal): Use it to check for a zero offset.
12331 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12332 Alan Hayward <alan.hayward@arm.com>
12333 David Sherwood <david.sherwood@arm.com>
12335 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
12336 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
12337 Return a poly_int64 rather than a HOST_WIDE_INT.
12338 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
12339 rather than a HOST_WIDE_INT.
12340 * config/aarch64/aarch64.h (aarch64_frame): Protect with
12341 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
12342 hard_fp_offset, frame_size, initial_adjust, callee_offset and
12343 final_offset from HOST_WIDE_INT to poly_int64.
12344 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
12345 to_constant when getting the number of units in an Advanced SIMD
12347 (aarch64_builtin_vectorized_function): Check for a constant number
12349 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
12351 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
12352 attribute instead of GET_MODE_NUNITS.
12353 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
12354 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
12355 GET_MODE_SIZE for fixed-size registers.
12356 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
12357 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
12358 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
12359 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
12360 (aarch64_print_operand, aarch64_print_address_internal)
12361 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
12362 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
12363 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
12364 Handle polynomial GET_MODE_SIZE.
12365 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
12366 wider than SImode without modification.
12367 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
12368 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
12369 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
12370 passing and returning SVE modes.
12371 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
12372 rather than GEN_INT.
12373 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
12374 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
12375 (aarch64_allocate_and_probe_stack_space): Likewise.
12376 (aarch64_layout_frame): Cope with polynomial offsets.
12377 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
12378 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
12379 polynomial offsets.
12380 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
12381 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
12382 poly_int64 rather than a HOST_WIDE_INT.
12383 (aarch64_get_separate_components, aarch64_process_components)
12384 (aarch64_expand_prologue, aarch64_expand_epilogue)
12385 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
12386 (aarch64_anchor_offset): New function, split out from...
12387 (aarch64_legitimize_address): ...here.
12388 (aarch64_builtin_vectorization_cost): Handle polynomial
12389 TYPE_VECTOR_SUBPARTS.
12390 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
12392 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
12393 number of elements from the PARALLEL rather than the mode.
12394 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
12395 rather than GET_MODE_BITSIZE.
12396 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
12397 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
12398 (aarch64_expand_vec_perm_const_1): Handle polynomial
12399 d->perm.length () and d->perm elements.
12400 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
12401 Apply to_constant to d->perm elements.
12402 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
12403 polynomial CONST_VECTOR_NUNITS.
12404 (aarch64_move_pointer): Take amount as a poly_int64 rather
12406 (aarch64_progress_pointer): Avoid temporary variable.
12407 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
12408 the mode attribute instead of GET_MODE.
12410 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12411 Alan Hayward <alan.hayward@arm.com>
12412 David Sherwood <david.sherwood@arm.com>
12414 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
12415 x exists before using it.
12416 (aarch64_add_constant_internal): Rename to...
12417 (aarch64_add_offset_1): ...this. Replace regnum with separate
12418 src and dest rtxes. Handle the case in which they're different,
12419 including when the offset is zero. Replace scratchreg with an rtx.
12420 Use 2 additions if there is no spare register into which we can
12421 move a 16-bit constant.
12422 (aarch64_add_constant): Delete.
12423 (aarch64_add_offset): Replace reg with separate src and dest
12424 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
12425 Use aarch64_add_offset_1.
12426 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
12427 an rtx rather than an int. Take the delta as a poly_int64
12428 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
12429 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
12430 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
12431 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
12432 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
12433 and aarch64_add_sp.
12434 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
12435 aarch64_add_constant.
12437 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12439 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
12440 Use scalar_float_mode.
12442 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12444 * config/aarch64/aarch64-simd.md
12445 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
12446 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
12447 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
12448 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
12449 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
12450 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
12451 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
12452 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
12453 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
12454 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
12456 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12459 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
12460 targ_options->x_arm_arch_string is non NULL.
12462 2018-01-11 Tamar Christina <tamar.christina@arm.com>
12464 * config/aarch64/aarch64.h
12465 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
12467 2018-01-11 Sudakshina Das <sudi.das@arm.com>
12470 * expmed.c (emit_store_flag_force): Swap if const op0
12471 and change VOIDmode to mode of op0.
12473 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12475 PR rtl-optimization/83761
12476 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
12477 than bytes to mode_for_size.
12479 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12481 PR middle-end/83189
12482 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
12483 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
12486 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12488 PR middle-end/83575
12489 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
12490 when in layout mode.
12491 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
12492 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
12495 2018-01-10 Michael Collison <michael.collison@arm.com>
12497 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
12498 * config/aarch64/aarch64-option-extension.def: Add
12499 AARCH64_OPT_EXTENSION of 'fp16fml'.
12500 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12501 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
12502 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
12503 * config/aarch64/constraints.md (Ui7): New constraint.
12504 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
12505 (VFMLA_SEL_W): Ditto.
12508 (VFMLA16_LOW): New int iterator.
12509 (VFMLA16_HIGH): Ditto.
12510 (UNSPEC_FMLAL): New unspec.
12511 (UNSPEC_FMLSL): Ditto.
12512 (UNSPEC_FMLAL2): Ditto.
12513 (UNSPEC_FMLSL2): Ditto.
12514 (f16mac): New code attribute.
12515 * config/aarch64/aarch64-simd-builtins.def
12516 (aarch64_fmlal_lowv2sf): Ditto.
12517 (aarch64_fmlsl_lowv2sf): Ditto.
12518 (aarch64_fmlalq_lowv4sf): Ditto.
12519 (aarch64_fmlslq_lowv4sf): Ditto.
12520 (aarch64_fmlal_highv2sf): Ditto.
12521 (aarch64_fmlsl_highv2sf): Ditto.
12522 (aarch64_fmlalq_highv4sf): Ditto.
12523 (aarch64_fmlslq_highv4sf): Ditto.
12524 (aarch64_fmlal_lane_lowv2sf): Ditto.
12525 (aarch64_fmlsl_lane_lowv2sf): Ditto.
12526 (aarch64_fmlal_laneq_lowv2sf): Ditto.
12527 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
12528 (aarch64_fmlalq_lane_lowv4sf): Ditto.
12529 (aarch64_fmlsl_lane_lowv4sf): Ditto.
12530 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
12531 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
12532 (aarch64_fmlal_lane_highv2sf): Ditto.
12533 (aarch64_fmlsl_lane_highv2sf): Ditto.
12534 (aarch64_fmlal_laneq_highv2sf): Ditto.
12535 (aarch64_fmlsl_laneq_highv2sf): Ditto.
12536 (aarch64_fmlalq_lane_highv4sf): Ditto.
12537 (aarch64_fmlsl_lane_highv4sf): Ditto.
12538 (aarch64_fmlalq_laneq_highv4sf): Ditto.
12539 (aarch64_fmlsl_laneq_highv4sf): Ditto.
12540 * config/aarch64/aarch64-simd.md:
12541 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
12542 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
12543 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
12544 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
12545 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
12546 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
12547 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
12548 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
12549 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
12550 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
12551 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
12552 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
12553 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
12554 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
12555 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
12556 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
12557 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
12558 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
12559 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
12560 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
12561 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
12562 (vfmlsl_low_u32): Ditto.
12563 (vfmlalq_low_u32): Ditto.
12564 (vfmlslq_low_u32): Ditto.
12565 (vfmlal_high_u32): Ditto.
12566 (vfmlsl_high_u32): Ditto.
12567 (vfmlalq_high_u32): Ditto.
12568 (vfmlslq_high_u32): Ditto.
12569 (vfmlal_lane_low_u32): Ditto.
12570 (vfmlsl_lane_low_u32): Ditto.
12571 (vfmlal_laneq_low_u32): Ditto.
12572 (vfmlsl_laneq_low_u32): Ditto.
12573 (vfmlalq_lane_low_u32): Ditto.
12574 (vfmlslq_lane_low_u32): Ditto.
12575 (vfmlalq_laneq_low_u32): Ditto.
12576 (vfmlslq_laneq_low_u32): Ditto.
12577 (vfmlal_lane_high_u32): Ditto.
12578 (vfmlsl_lane_high_u32): Ditto.
12579 (vfmlal_laneq_high_u32): Ditto.
12580 (vfmlsl_laneq_high_u32): Ditto.
12581 (vfmlalq_lane_high_u32): Ditto.
12582 (vfmlslq_lane_high_u32): Ditto.
12583 (vfmlalq_laneq_high_u32): Ditto.
12584 (vfmlslq_laneq_high_u32): Ditto.
12585 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
12586 (AARCH64_FL_FOR_ARCH8_4): New.
12587 (AARCH64_ISA_F16FML): New ISA flag.
12588 (TARGET_F16FML): New feature flag for fp16fml.
12589 (doc/invoke.texi): Document new fp16fml option.
12591 2018-01-10 Michael Collison <michael.collison@arm.com>
12593 * config/aarch64/aarch64-builtins.c:
12594 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
12595 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12596 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
12597 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
12598 (AARCH64_ISA_SHA3): New ISA flag.
12599 (TARGET_SHA3): New feature flag for sha3.
12600 * config/aarch64/iterators.md (sha512_op): New int attribute.
12601 (CRYPTO_SHA512): New int iterator.
12602 (UNSPEC_SHA512H): New unspec.
12603 (UNSPEC_SHA512H2): Ditto.
12604 (UNSPEC_SHA512SU0): Ditto.
12605 (UNSPEC_SHA512SU1): Ditto.
12606 * config/aarch64/aarch64-simd-builtins.def
12607 (aarch64_crypto_sha512hqv2di): New builtin.
12608 (aarch64_crypto_sha512h2qv2di): Ditto.
12609 (aarch64_crypto_sha512su0qv2di): Ditto.
12610 (aarch64_crypto_sha512su1qv2di): Ditto.
12611 (aarch64_eor3qv8hi): Ditto.
12612 (aarch64_rax1qv2di): Ditto.
12613 (aarch64_xarqv2di): Ditto.
12614 (aarch64_bcaxqv8hi): Ditto.
12615 * config/aarch64/aarch64-simd.md:
12616 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
12617 (aarch64_crypto_sha512su0qv2di): Ditto.
12618 (aarch64_crypto_sha512su1qv2di): Ditto.
12619 (aarch64_eor3qv8hi): Ditto.
12620 (aarch64_rax1qv2di): Ditto.
12621 (aarch64_xarqv2di): Ditto.
12622 (aarch64_bcaxqv8hi): Ditto.
12623 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
12624 (vsha512h2q_u64): Ditto.
12625 (vsha512su0q_u64): Ditto.
12626 (vsha512su1q_u64): Ditto.
12627 (veor3q_u16): Ditto.
12628 (vrax1q_u64): Ditto.
12629 (vxarq_u64): Ditto.
12630 (vbcaxq_u16): Ditto.
12631 * config/arm/types.md (crypto_sha512): New type attribute.
12632 (crypto_sha3): Ditto.
12633 (doc/invoke.texi): Document new sha3 option.
12635 2018-01-10 Michael Collison <michael.collison@arm.com>
12637 * config/aarch64/aarch64-builtins.c:
12638 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
12639 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12640 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
12641 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
12642 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
12643 (AARCH64_ISA_SM4): New ISA flag.
12644 (TARGET_SM4): New feature flag for sm4.
12645 * config/aarch64/aarch64-simd-builtins.def
12646 (aarch64_sm3ss1qv4si): Ditto.
12647 (aarch64_sm3tt1aq4si): Ditto.
12648 (aarch64_sm3tt1bq4si): Ditto.
12649 (aarch64_sm3tt2aq4si): Ditto.
12650 (aarch64_sm3tt2bq4si): Ditto.
12651 (aarch64_sm3partw1qv4si): Ditto.
12652 (aarch64_sm3partw2qv4si): Ditto.
12653 (aarch64_sm4eqv4si): Ditto.
12654 (aarch64_sm4ekeyqv4si): Ditto.
12655 * config/aarch64/aarch64-simd.md:
12656 (aarch64_sm3ss1qv4si): Ditto.
12657 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
12658 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
12659 (aarch64_sm4eqv4si): Ditto.
12660 (aarch64_sm4ekeyqv4si): Ditto.
12661 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
12662 (sm3part_op): Ditto.
12663 (CRYPTO_SM3TT): Ditto.
12664 (CRYPTO_SM3PART): Ditto.
12665 (UNSPEC_SM3SS1): New unspec.
12666 (UNSPEC_SM3TT1A): Ditto.
12667 (UNSPEC_SM3TT1B): Ditto.
12668 (UNSPEC_SM3TT2A): Ditto.
12669 (UNSPEC_SM3TT2B): Ditto.
12670 (UNSPEC_SM3PARTW1): Ditto.
12671 (UNSPEC_SM3PARTW2): Ditto.
12672 (UNSPEC_SM4E): Ditto.
12673 (UNSPEC_SM4EKEY): Ditto.
12674 * config/aarch64/constraints.md (Ui2): New constraint.
12675 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
12676 * config/arm/types.md (crypto_sm3): New type attribute.
12677 (crypto_sm4): Ditto.
12678 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
12679 (vsm3tt1aq_u32): Ditto.
12680 (vsm3tt1bq_u32): Ditto.
12681 (vsm3tt2aq_u32): Ditto.
12682 (vsm3tt2bq_u32): Ditto.
12683 (vsm3partw1q_u32): Ditto.
12684 (vsm3partw2q_u32): Ditto.
12685 (vsm4eq_u32): Ditto.
12686 (vsm4ekeyq_u32): Ditto.
12687 (doc/invoke.texi): Document new sm4 option.
12689 2018-01-10 Michael Collison <michael.collison@arm.com>
12691 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
12692 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
12693 (AARCH64_FL_FOR_ARCH8_4): New.
12694 (AARCH64_FL_V8_4): New flag.
12695 (doc/invoke.texi): Document new armv8.4-a option.
12697 2018-01-10 Michael Collison <michael.collison@arm.com>
12699 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12700 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
12701 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
12702 * config/aarch64/aarch64-option-extension.def: Add
12703 AARCH64_OPT_EXTENSION of 'sha2'.
12704 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
12705 (crypto): Disable sha2 and aes if crypto disabled.
12706 (crypto): Enable aes and sha2 if enabled.
12707 (simd): Disable sha2 and aes if simd disabled.
12708 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
12710 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
12711 (TARGET_SHA2): New feature flag for sha2.
12712 (TARGET_AES): New feature flag for aes.
12713 * config/aarch64/aarch64-simd.md:
12714 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
12715 conditional on TARGET_AES.
12716 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
12717 (aarch64_crypto_sha1hsi): Make pattern conditional
12719 (aarch64_crypto_sha1hv4si): Ditto.
12720 (aarch64_be_crypto_sha1hv4si): Ditto.
12721 (aarch64_crypto_sha1su1v4si): Ditto.
12722 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
12723 (aarch64_crypto_sha1su0v4si): Ditto.
12724 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
12725 (aarch64_crypto_sha256su0v4si): Ditto.
12726 (aarch64_crypto_sha256su1v4si): Ditto.
12727 (doc/invoke.texi): Document new aes and sha2 options.
12729 2018-01-10 Martin Sebor <msebor@redhat.com>
12731 PR tree-optimization/83781
12732 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
12735 2018-01-11 Martin Sebor <msebor@gmail.com>
12736 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12738 PR tree-optimization/83501
12739 PR tree-optimization/81703
12741 * tree-ssa-strlen.c (get_string_cst): Rename...
12742 (get_string_len): ...to this. Handle global constants.
12743 (handle_char_store): Adjust.
12745 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
12746 Jim Wilson <jimw@sifive.com>
12748 * config/riscv/riscv-protos.h (riscv_output_return): New.
12749 * config/riscv/riscv.c (struct machine_function): New naked_p field.
12750 (riscv_attribute_table, riscv_output_return),
12751 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
12752 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
12753 (riscv_compute_frame_info): Only compute frame->mask if not a naked
12755 (riscv_expand_prologue): Add early return for naked function.
12756 (riscv_expand_epilogue): Likewise.
12757 (riscv_function_ok_for_sibcall): Return false for naked function.
12758 (riscv_set_current_function): New.
12759 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
12760 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
12761 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
12762 * doc/extend.texi (RISC-V Function Attributes): New.
12764 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
12766 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
12767 check for 128-bit long double before checking TCmode.
12768 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
12769 128-bit long doubles before checking TFmode or TCmode.
12770 (FLOAT128_IBM_P): Likewise.
12772 2018-01-10 Martin Sebor <msebor@redhat.com>
12774 PR tree-optimization/83671
12775 * builtins.c (c_strlen): Unconditionally return zero for the empty
12777 Use -Warray-bounds for warnings.
12778 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
12779 for non-constant array indices with COMPONENT_REF, arrays of
12780 arrays, and pointers to arrays.
12781 (gimple_fold_builtin_strlen): Determine and set length range for
12782 non-constant character arrays.
12784 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
12786 PR middle-end/81897
12787 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
12790 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
12792 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
12794 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
12797 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
12798 VECTOR_MEM_ALTIVEC_OR_VSX_P.
12799 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
12800 indexed_or_indirect_operand predicate.
12801 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
12802 (*vsx_le_perm_load_v8hi): Likewise.
12803 (*vsx_le_perm_load_v16qi): Likewise.
12804 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
12805 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
12806 (*vsx_le_perm_store_v8hi): Likewise.
12807 (*vsx_le_perm_store_v16qi): Likewise.
12808 (eight unnamed splitters): Likewise.
12810 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
12812 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
12813 * config/rs6000/emmintrin.h: Likewise.
12814 * config/rs6000/mmintrin.h: Likewise.
12815 * config/rs6000/xmmintrin.h: Likewise.
12817 2018-01-10 David Malcolm <dmalcolm@redhat.com>
12820 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
12822 * tree.c (tree_nop_conversion): Return true for location wrapper
12824 (maybe_wrap_with_location): New function.
12825 (selftest::check_strip_nops): New function.
12826 (selftest::test_location_wrappers): New function.
12827 (selftest::tree_c_tests): Call it.
12828 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
12829 (maybe_wrap_with_location): New decl.
12830 (EXPR_LOCATION_WRAPPER_P): New macro.
12831 (location_wrapper_p): New inline function.
12832 (tree_strip_any_location_wrapper): New inline function.
12834 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
12837 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
12838 stack_realign_offset for the largest alignment of stack slot
12840 (ix86_find_max_used_stack_alignment): New function.
12841 (ix86_finalize_stack_frame_flags): Use it. Set
12842 max_used_stack_alignment if we don't realign stack.
12843 * config/i386/i386.h (machine_function): Add
12844 max_used_stack_alignment.
12846 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
12848 * config/arm/arm.opt (-mbranch-cost): New option.
12849 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
12852 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
12855 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
12856 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
12858 2018-01-10 Richard Biener <rguenther@suse.de>
12861 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
12862 early out so it also covers the case where we have a non-NULL
12865 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
12867 PR tree-optimization/83753
12868 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
12869 for non-strided grouped accesses if the number of elements is 1.
12871 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12874 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
12875 * i386.h (TARGET_USE_GATHER): Define.
12876 * x86-tune.def (X86_TUNE_USE_GATHER): New.
12878 2018-01-10 Martin Liska <mliska@suse.cz>
12881 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
12882 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
12884 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
12885 CLEANUP_NO_PARTITIONING is not set.
12887 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
12889 * doc/rtl.texi: Remove documentation of (const ...) wrappers
12890 for vectors, as a partial revert of r254296.
12891 * rtl.h (const_vec_p): Delete.
12892 (const_vec_duplicate_p): Don't test for vector CONSTs.
12893 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
12894 * expmed.c (make_tree): Likewise.
12897 * common.md (E, F): Use CONSTANT_P instead of checking for
12899 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
12900 checking for CONST_VECTOR.
12902 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
12904 PR middle-end/83575
12905 * predict.c (force_edge_cold): Handle in more sane way edges
12906 with no prediction.
12908 2018-01-09 Carl Love <cel@us.ibm.com>
12910 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
12912 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
12913 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
12914 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
12915 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
12916 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
12917 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
12918 * config/rs6000/rs6000-protos.h: Add extern defition for
12919 rs6000_generate_float2_double_code.
12920 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
12922 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
12923 (float2_v2df): Add define_expand.
12925 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
12928 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
12929 op_mode in the force_to_mode call.
12931 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
12933 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
12934 instead of checking each element individually.
12935 (aarch64_evpc_uzp): Likewise.
12936 (aarch64_evpc_zip): Likewise.
12937 (aarch64_evpc_ext): Likewise.
12938 (aarch64_evpc_rev): Likewise.
12939 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
12940 instead of checking each element individually. Return true without
12942 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
12943 whether all selected elements come from the same input, instead of
12944 checking each element individually. Remove calls to gen_rtx_REG,
12945 start_sequence and end_sequence and instead assert that no rtl is
12948 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
12950 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
12951 order of HIGH and CONST checks.
12953 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
12955 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
12956 if the destination isn't an SSA_NAME.
12958 2018-01-09 Richard Biener <rguenther@suse.de>
12960 PR tree-optimization/83668
12961 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
12963 (canonicalize_loop_form): ... here, renamed from ...
12964 (canonicalize_loop_closed_ssa_form): ... this and amended to
12965 swap successor edges for loop exit blocks to make us use
12966 the RPO order we need for initial schedule generation.
12968 2018-01-09 Joseph Myers <joseph@codesourcery.com>
12970 PR tree-optimization/64811
12971 * match.pd: When optimizing comparisons with Inf, avoid
12972 introducing or losing exceptions from comparisons with NaN.
12974 2018-01-09 Martin Liska <mliska@suse.cz>
12977 * asan.c (shadow_mem_size): Add gcc_assert.
12979 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
12981 Don't save registers in main().
12984 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
12985 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
12986 * config/avr/avr.c (avr_set_current_function): Don't error if
12987 naked, OS_task or OS_main are specified at the same time.
12988 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
12990 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
12992 * common/config/avr/avr-common.c (avr_option_optimization_table):
12993 Switch on -mmain-is-OS_task for optimizing compilations.
12995 2018-01-09 Richard Biener <rguenther@suse.de>
12997 PR tree-optimization/83572
12998 * graphite.c: Include cfganal.h.
12999 (graphite_transform_loops): Connect infinite loops to exit
13000 and remove fake edges at the end.
13002 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
13004 * ipa-inline.c (edge_badness): Revert accidental checkin.
13006 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
13009 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
13010 symbols; not inline clones.
13012 2018-01-09 Jakub Jelinek <jakub@redhat.com>
13015 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
13016 hard registers. Formatting fixes.
13018 PR preprocessor/83722
13019 * gcc.c (try_generate_repro): Pass
13020 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
13021 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
13024 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
13025 Kito Cheng <kito.cheng@gmail.com>
13027 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
13028 (riscv_leaf_function_p): Delete.
13029 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
13031 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13033 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
13035 (do_ifelse): New function.
13036 (do_isel): New function.
13037 (do_sub3): New function.
13038 (do_add3): New function.
13039 (do_load_mask_compare): New function.
13040 (do_overlap_load_compare): New function.
13041 (expand_compare_loop): New function.
13042 (expand_block_compare): Call expand_compare_loop() when appropriate.
13043 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
13044 option description.
13045 (-mblock-compare-inline-loop-limit): New option.
13047 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13050 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
13051 Reverse order of second and third operands in first alternative.
13052 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
13053 of first and second elements in UNSPEC_VPERMR vector.
13054 (altivec_expand_vec_perm_le): Likewise.
13056 2018-01-08 Jeff Law <law@redhat.com>
13058 PR rtl-optimizatin/81308
13059 * tree-switch-conversion.c (cfg_altered): New file scoped static.
13060 (process_switch): If group_case_labels makes a change, then set
13062 (pass_convert_switch::execute): If a switch is converted, then
13063 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
13065 PR rtl-optimization/81308
13066 * recog.c (split_all_insns): Conditionally cleanup the CFG after
13069 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
13071 PR target/83663 - Revert r255946
13072 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
13073 generation for cases where splatting a value is not useful.
13074 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
13075 across a vec_duplicate and a paradoxical subreg forming a vector
13076 mode to a vec_concat.
13078 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13080 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
13081 -march=armv8.3-a variants.
13082 * config/arm/t-multilib: Likewise.
13083 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
13085 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13087 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
13089 (cceq_ior_compare_complement): Give it a name so I can use it, and
13090 change boolean_or_operator predicate to boolean_operator so it can
13091 be used to generate a crand.
13092 (eqne): New code iterator.
13093 (bd/bd_neg): New code_attrs.
13094 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
13095 a single define_insn.
13096 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
13097 decrement (bdnzt/bdnzf/bdzt/bdzf).
13098 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
13099 with the new names of the branch decrement patterns, and added the
13100 names of the branch decrement conditional patterns.
13102 2018-01-08 Richard Biener <rguenther@suse.de>
13104 PR tree-optimization/83563
13105 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
13108 2018-01-08 Richard Biener <rguenther@suse.de>
13110 PR middle-end/83713
13111 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
13113 2018-01-08 Richard Biener <rguenther@suse.de>
13115 PR tree-optimization/83685
13116 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
13117 references to abnormals.
13119 2018-01-08 Richard Biener <rguenther@suse.de>
13122 * dwarf2out.c (output_indirect_strings): Handle empty
13123 skeleton_debug_str_hash.
13124 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
13126 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
13128 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
13129 (emit_store_direct): Likewise.
13130 (arc_trampoline_adjust_address): Likewise.
13131 (arc_asm_trampoline_template): New function.
13132 (arc_initialize_trampoline): Use asm_trampoline_template.
13133 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
13134 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
13135 * config/arc/arc.md (flush_icache): Delete pattern.
13137 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
13139 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
13140 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
13143 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
13146 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
13147 by not USED_FOR_TARGET.
13148 (make_pass_resolve_sw_modes): Likewise.
13150 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
13152 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
13155 2018-01-08 Richard Biener <rguenther@suse.de>
13157 PR middle-end/83580
13158 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
13160 2018-01-08 Richard Biener <rguenther@suse.de>
13162 PR middle-end/83517
13163 * match.pd ((t * 2) / 2) -> t): Add missing :c.
13165 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
13167 PR middle-end/81897
13168 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
13169 basic blocks with a small number of successors.
13170 (convert_control_dep_chain_into_preds): Improve handling of
13172 (dump_predicates): Split apart into...
13173 (dump_pred_chain): ...here...
13174 (dump_pred_info): ...and here.
13175 (can_one_predicate_be_invalidated_p): Add debugging printfs.
13176 (can_chain_union_be_invalidated_p): Improve check for invalidation
13178 (uninit_uses_cannot_happen): Avoid unnecessary if
13179 convert_control_dep_chain_into_preds yielded nothing.
13181 2018-01-06 Martin Sebor <msebor@redhat.com>
13183 PR tree-optimization/83640
13184 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
13185 subtracting negative offset from size.
13186 (builtin_access::overlap): Adjust offset bounds of the access to fall
13187 within the size of the object if possible.
13189 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
13191 PR rtl-optimization/83699
13192 * expmed.c (extract_bit_field_1): Restrict the vector usage of
13193 extract_bit_field_as_subreg to cases in which the extracted
13194 value is also a vector.
13196 * lra-constraints.c (process_alt_operands): Test for the equivalence
13197 substitutions when detecting a possible reload cycle.
13199 2018-01-06 Jakub Jelinek <jakub@redhat.com>
13202 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
13203 by default if flag_selective_schedling{,2}. Formatting fixes.
13205 PR rtl-optimization/83682
13206 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
13207 if it has non-VECTOR_MODE element mode.
13208 (vec_duplicate_p): Likewise.
13210 PR middle-end/83694
13211 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
13212 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
13214 2018-01-05 Jakub Jelinek <jakub@redhat.com>
13217 * config/i386/i386-builtin.def
13218 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
13219 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
13220 Require also OPTION_MASK_ISA_AVX512F in addition to
13221 OPTION_MASK_ISA_GFNI.
13222 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
13223 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
13224 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
13225 to OPTION_MASK_ISA_GFNI.
13226 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
13227 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
13228 OPTION_MASK_ISA_AVX512BW.
13229 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
13230 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
13231 addition to OPTION_MASK_ISA_GFNI.
13232 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
13233 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
13234 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
13235 to OPTION_MASK_ISA_GFNI.
13236 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
13237 a requirement for all ISAs rather than any of them with a few
13239 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
13241 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
13242 bitmasks to be enabled with 3 exceptions, instead of requiring any
13243 enabled ISA with lots of exceptions.
13244 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
13245 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
13246 Change avx512bw in isa attribute to avx512f.
13247 * config/i386/sgxintrin.h: Add license boilerplate.
13248 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
13249 to __AVX512F__ and __AVX512VL to __AVX512VL__.
13250 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
13251 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
13253 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
13254 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
13255 temporarily sse2 rather than sse if not enabled already.
13258 * config/i386/sse.md (VI248_VLBW): Rename to ...
13259 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
13260 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
13261 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
13262 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
13263 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
13264 mode iterator instead of VI248_VLBW.
13266 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
13268 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
13269 (record_modified): Skip clobbers; add debug output.
13270 (param_change_prob): Use sreal frequencies.
13272 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
13274 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
13275 punt for user-aligned variables.
13277 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
13279 * tree-chrec.c (chrec_contains_symbols): Return true for
13282 2018-01-05 Sudakshina Das <sudi.das@arm.com>
13285 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
13286 of (x|y) == x for BICS pattern.
13288 2018-01-05 Jakub Jelinek <jakub@redhat.com>
13290 PR tree-optimization/83605
13291 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
13292 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
13295 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
13297 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
13298 * config/epiphany/rtems.h: New file.
13300 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13301 Uros Bizjak <ubizjak@gmail.com>
13304 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
13305 QIreg_operand instead of register_operand predicate.
13306 * config/i386/i386.c (ix86_rop_should_change_byte_p,
13307 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
13308 comments instead of -fmitigate[-_]rop.
13310 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13313 * cgraphunit.c (symbol_table::compile): Switch to text_section
13314 before calling assembly_start debug hook.
13315 * run-rtl-passes.c (run_rtl_passes): Likewise.
13318 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13320 * tree-vrp.c (extract_range_from_binary_expr_1): Check
13321 range_int_cst_p rather than !symbolic_range_p before calling
13322 extract_range_from_multiplicative_op_1.
13324 2018-01-04 Jeff Law <law@redhat.com>
13326 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
13327 redundant test in assertion.
13329 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13331 * doc/rtl.texi: Document machine_mode wrapper classes.
13333 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13335 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
13336 using tree_to_uhwi.
13338 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13340 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
13341 the VEC_PERM_EXPR fold to fail.
13343 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13346 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
13347 to switched_sections.
13349 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13352 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
13353 test for d.testing.
13355 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
13358 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
13359 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
13361 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13364 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
13365 is BLKmode and bitpos not zero or mode change is needed.
13367 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13370 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
13373 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
13376 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
13377 instead of MULT rtx. Update all corresponding splitters.
13378 (*saddl_se): Ditto.
13379 (*ssub<modesuffix>): Ditto.
13380 (*ssubl_se): Ditto.
13381 (*cmp_sadd_di): Update split patterns.
13382 (*cmp_sadd_si): Ditto.
13383 (*cmp_sadd_sidi): Ditto.
13384 (*cmp_ssub_di): Ditto.
13385 (*cmp_ssub_si): Ditto.
13386 (*cmp_ssub_sidi): Ditto.
13387 * config/alpha/predicates.md (const23_operand): New predicate.
13388 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
13389 Look for ASHIFT, not MULT inner operand.
13390 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
13392 2018-01-04 Martin Liska <mliska@suse.cz>
13394 PR gcov-profile/83669
13395 * gcov.c (output_intermediate_file): Add version to intermediate
13397 * doc/gcov.texi: Document new field 'version' in intermediate
13398 file format. Fix location of '-k' option of gcov command.
13400 2018-01-04 Martin Liska <mliska@suse.cz>
13403 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
13405 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13407 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
13409 2018-01-03 Martin Sebor <msebor@redhat.com>
13411 PR tree-optimization/83655
13412 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
13413 checking calls with invalid arguments.
13415 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13417 * tree-vect-stmts.c (vect_get_store_rhs): New function.
13418 (vectorizable_mask_load_store): Delete.
13419 (vectorizable_call): Return false for masked loads and stores.
13420 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
13421 instead of gimple_assign_rhs1.
13422 (vectorizable_load): Handle IFN_MASK_LOAD.
13423 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
13425 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13427 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
13429 (vectorizable_mask_load_store): ...here.
13430 (vectorizable_load): ...and here.
13432 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13434 * tree-vect-stmts.c (vect_build_all_ones_mask)
13435 (vect_build_zero_merge_argument): New functions, split out from...
13436 (vectorizable_load): ...here.
13438 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13440 * tree-vect-stmts.c (vect_check_store_rhs): New function,
13442 (vectorizable_mask_load_store): ...here.
13443 (vectorizable_store): ...and here.
13445 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13447 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
13449 (vectorizable_mask_load_store): ...here.
13451 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13453 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
13454 (vect_model_store_cost): Take a vec_load_store_type instead of a
13456 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
13457 (vect_model_store_cost): Take a vec_load_store_type instead of a
13459 (vectorizable_mask_load_store): Update accordingly.
13460 (vectorizable_store): Likewise.
13461 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
13463 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13465 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
13466 IFN_MASK_LOAD calls here rather than...
13467 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
13469 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13470 Alan Hayward <alan.hayward@arm.com>
13471 David Sherwood <david.sherwood@arm.com>
13473 * expmed.c (extract_bit_field_1): For vector extracts,
13474 fall back to extract_bit_field_as_subreg if vec_extract
13477 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13478 Alan Hayward <alan.hayward@arm.com>
13479 David Sherwood <david.sherwood@arm.com>
13481 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
13482 they are variable or constant sized.
13483 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
13484 slots for constant-sized data.
13486 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13487 Alan Hayward <alan.hayward@arm.com>
13488 David Sherwood <david.sherwood@arm.com>
13490 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
13491 handling COND_EXPRs with boolean comparisons, try to find a better
13492 basis for the mask type than the boolean itself.
13494 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13496 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
13497 is calculated and how it can be overridden.
13498 * genmodes.c (max_bitsize_mode_any_mode): New variable.
13499 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
13501 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
13504 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13505 Alan Hayward <alan.hayward@arm.com>
13506 David Sherwood <david.sherwood@arm.com>
13508 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
13509 Remove the mode argument.
13510 (aarch64_simd_valid_immediate): Remove the mode and inverse
13512 * config/aarch64/iterators.md (bitsize): New iterator.
13513 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
13514 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
13515 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
13516 aarch64_simd_valid_immediate.
13517 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
13518 (aarch64_reg_or_bic_imm): Likewise.
13519 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
13520 with an insn_type enum and msl with a modifier_type enum.
13521 Replace element_width with a scalar_mode. Change the shift
13522 to unsigned int. Add constructors for scalar_float_mode and
13523 scalar_int_mode elements.
13524 (aarch64_vect_float_const_representable_p): Delete.
13525 (aarch64_can_const_movi_rtx_p)
13526 (aarch64_simd_scalar_immediate_valid_for_move)
13527 (aarch64_simd_make_constant): Update call to
13528 aarch64_simd_valid_immediate.
13529 (aarch64_advsimd_valid_immediate_hs): New function.
13530 (aarch64_advsimd_valid_immediate): Likewise.
13531 (aarch64_simd_valid_immediate): Remove mode and inverse
13532 arguments. Rewrite to use the above. Use const_vec_duplicate_p
13533 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
13534 and aarch64_float_const_representable_p on the result.
13535 (aarch64_output_simd_mov_immediate): Remove mode argument.
13536 Update call to aarch64_simd_valid_immediate and use of
13537 simd_immediate_info.
13538 (aarch64_output_scalar_simd_mov_immediate): Update call
13541 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13542 Alan Hayward <alan.hayward@arm.com>
13543 David Sherwood <david.sherwood@arm.com>
13545 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
13546 (mode_nunits): Likewise CONST_MODE_NUNITS.
13547 * machmode.def (ADJUST_NUNITS): Document.
13548 * genmodes.c (mode_data::need_nunits_adj): New field.
13549 (blank_mode): Update accordingly.
13550 (adj_nunits): New variable.
13551 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
13553 (emit_mode_size_inline): Set need_bytesize_adj for all modes
13554 listed in adj_nunits.
13555 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
13556 listed in adj_nunits. Don't emit case statements for such modes.
13557 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
13558 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
13559 nothing if adj_nunits is nonnull.
13560 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
13561 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
13562 (emit_mode_fbit): Update use of print_maybe_const_decl.
13563 (emit_move_size): Likewise. Treat the array as non-const
13565 (emit_mode_adjustments): Handle adj_nunits.
13567 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13569 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
13570 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
13571 (VECTOR_MODES): Use it.
13572 (make_vector_modes): Take the prefix as an argument.
13574 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13575 Alan Hayward <alan.hayward@arm.com>
13576 David Sherwood <david.sherwood@arm.com>
13578 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
13579 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
13580 for MODE_VECTOR_BOOL.
13581 * machmode.def (VECTOR_BOOL_MODE): Document.
13582 * genmodes.c (VECTOR_BOOL_MODE): New macro.
13583 (make_vector_bool_mode): New function.
13584 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
13586 * lto-streamer-in.c (lto_input_mode_table): Likewise.
13587 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
13589 * stor-layout.c (int_mode_for_mode): Likewise.
13590 * tree.c (build_vector_type_for_mode): Likewise.
13591 * varasm.c (output_constant_pool_2): Likewise.
13592 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
13593 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
13594 for MODE_VECTOR_BOOL.
13595 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
13596 of mode class checks.
13597 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
13598 instead of a list of mode class checks.
13599 (expand_vector_scalar_condition): Likewise.
13600 (type_for_widest_vector_mode): Handle BImode as an inner mode.
13602 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13603 Alan Hayward <alan.hayward@arm.com>
13604 David Sherwood <david.sherwood@arm.com>
13606 * machmode.h (mode_size): Change from unsigned short to
13608 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
13609 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
13610 or if measurement_type is not polynomial.
13611 (fixed_size_mode::includes_p): Check for constant-sized modes.
13612 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
13613 return a poly_uint16 rather than an unsigned short.
13614 (emit_mode_size): Change the type of mode_size from unsigned short
13615 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
13616 (emit_mode_adjustments): Cope with polynomial vector sizes.
13617 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13619 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13621 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
13622 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
13623 * caller-save.c (setup_save_areas): Likewise.
13624 (replace_reg_with_saved_mem): Likewise.
13625 * calls.c (emit_library_call_value_1): Likewise.
13626 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
13627 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
13628 (gen_lowpart_for_combine): Likewise.
13629 * convert.c (convert_to_integer_1): Likewise.
13630 * cse.c (equiv_constant, cse_insn): Likewise.
13631 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
13632 (cselib_subst_to_values): Likewise.
13633 * dce.c (word_dce_process_block): Likewise.
13634 * df-problems.c (df_word_lr_mark_ref): Likewise.
13635 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
13636 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
13637 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
13638 (rtl_for_decl_location): Likewise.
13639 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
13640 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
13641 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
13642 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
13643 (expand_expr_real_1): Likewise.
13644 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
13645 (pad_below): Likewise.
13646 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
13647 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
13648 * ira.c (get_subreg_tracking_sizes): Likewise.
13649 * ira-build.c (ira_create_allocno_objects): Likewise.
13650 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
13651 (ira_sort_regnos_for_alter_reg): Likewise.
13652 * ira-costs.c (record_operand_costs): Likewise.
13653 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
13654 (resolve_simple_move): Likewise.
13655 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
13656 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
13657 (lra_constraints): Likewise.
13658 (CONST_POOL_OK_P): Reject variable-sized modes.
13659 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
13660 (add_pseudo_to_slot, lra_spill): Likewise.
13661 * omp-low.c (omp_clause_aligned_alignment): Likewise.
13662 * optabs-query.c (get_best_extraction_insn): Likewise.
13663 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
13664 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
13665 (expand_mult_highpart, valid_multiword_target_p): Likewise.
13666 * recog.c (offsettable_address_addr_space_p): Likewise.
13667 * regcprop.c (maybe_mode_change): Likewise.
13668 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
13669 * regrename.c (build_def_use): Likewise.
13670 * regstat.c (dump_reg_info): Likewise.
13671 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
13672 (find_reloads, find_reloads_subreg_address): Likewise.
13673 * reload1.c (eliminate_regs_1): Likewise.
13674 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
13675 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
13676 (simplify_binary_operation_1, simplify_subreg): Likewise.
13677 * targhooks.c (default_function_arg_padding): Likewise.
13678 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
13679 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
13680 (verify_gimple_assign_ternary): Likewise.
13681 * tree-inline.c (estimate_move_cost): Likewise.
13682 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13683 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
13684 (get_address_cost_ainc): Likewise.
13685 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
13686 (vect_supportable_dr_alignment): Likewise.
13687 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
13688 (vectorizable_reduction): Likewise.
13689 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
13690 (vectorizable_operation, vectorizable_load): Likewise.
13691 * tree.c (build_same_sized_truth_vector_type): Likewise.
13692 * valtrack.c (cleanup_auto_inc_dec): Likewise.
13693 * var-tracking.c (emit_note_insn_var_location): Likewise.
13694 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
13695 (ADDR_VEC_ALIGN): Likewise.
13697 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13698 Alan Hayward <alan.hayward@arm.com>
13699 David Sherwood <david.sherwood@arm.com>
13701 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
13703 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
13704 or if measurement_type is polynomial.
13705 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
13706 * combine.c (make_extraction): Likewise.
13707 * dse.c (find_shift_sequence): Likewise.
13708 * dwarf2out.c (mem_loc_descriptor): Likewise.
13709 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
13710 (extract_bit_field, extract_low_bits): Likewise.
13711 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
13712 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
13713 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
13714 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
13715 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
13716 * reload.c (find_reloads): Likewise.
13717 * reload1.c (alter_reg): Likewise.
13718 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
13719 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
13720 * tree-if-conv.c (predicate_mem_writes): Likewise.
13721 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
13722 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
13723 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
13724 * valtrack.c (dead_debug_insert_temp): Likewise.
13725 * varasm.c (mergeable_constant_section): Likewise.
13726 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
13728 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13729 Alan Hayward <alan.hayward@arm.com>
13730 David Sherwood <david.sherwood@arm.com>
13732 * expr.c (expand_assignment): Cope with polynomial mode sizes
13733 when assigning to a CONCAT.
13735 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13736 Alan Hayward <alan.hayward@arm.com>
13737 David Sherwood <david.sherwood@arm.com>
13739 * machmode.h (mode_precision): Change from unsigned short to
13741 (mode_to_precision): Return a poly_uint16 rather than an unsigned
13743 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
13744 or if measurement_type is not polynomial.
13745 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
13746 in which the mode is already known to be a scalar_int_mode.
13747 * genmodes.c (emit_mode_precision): Change the type of mode_precision
13748 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
13750 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13751 for GET_MODE_PRECISION.
13752 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13753 for GET_MODE_PRECISION.
13754 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
13756 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
13757 (expand_field_assignment, make_extraction): Likewise.
13758 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
13759 (get_last_value): Likewise.
13760 * convert.c (convert_to_integer_1): Likewise.
13761 * cse.c (cse_insn): Likewise.
13762 * expr.c (expand_expr_real_1): Likewise.
13763 * lra-constraints.c (simplify_operand_subreg): Likewise.
13764 * optabs-query.c (can_atomic_load_p): Likewise.
13765 * optabs.c (expand_atomic_load): Likewise.
13766 (expand_atomic_store): Likewise.
13767 * ree.c (combine_reaching_defs): Likewise.
13768 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
13769 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
13770 * tree.h (type_has_mode_precision_p): Likewise.
13771 * ubsan.c (instrument_si_overflow): Likewise.
13773 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13774 Alan Hayward <alan.hayward@arm.com>
13775 David Sherwood <david.sherwood@arm.com>
13777 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
13778 polynomial numbers of units.
13779 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
13780 (valid_vector_subparts_p): New function.
13781 (build_vector_type): Remove temporary shim and take the number
13782 of units as a poly_uint64 rather than an int.
13783 (build_opaque_vector_type): Take the number of units as a
13784 poly_uint64 rather than an int.
13785 * tree.c (build_vector_from_ctor): Handle polynomial
13786 TYPE_VECTOR_SUBPARTS.
13787 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
13788 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
13789 (build_vector_from_val): If the number of units is variable,
13790 use build_vec_duplicate_cst for constant operands and
13791 VEC_DUPLICATE_EXPR otherwise.
13792 (make_vector_type): Remove temporary is_constant ().
13793 (build_vector_type, build_opaque_vector_type): Take the number of
13794 units as a poly_uint64 rather than an int.
13795 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
13797 * cfgexpand.c (expand_debug_expr): Likewise.
13798 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
13799 (store_constructor, expand_expr_real_1): Likewise.
13800 (const_scalar_mask_from_tree): Likewise.
13801 * fold-const-call.c (fold_const_reduction): Likewise.
13802 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
13803 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
13804 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
13805 (fold_relational_const): Likewise.
13806 (native_interpret_vector): Likewise. Change the size from an
13807 int to an unsigned int.
13808 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
13809 TYPE_VECTOR_SUBPARTS.
13810 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
13811 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
13812 duplicating a non-constant operand into a variable-length vector.
13813 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
13814 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
13815 * ipa-icf.c (sem_variable::equals): Likewise.
13816 * match.pd: Likewise.
13817 * omp-simd-clone.c (simd_clone_subparts): Likewise.
13818 * print-tree.c (print_node): Likewise.
13819 * stor-layout.c (layout_type): Likewise.
13820 * targhooks.c (default_builtin_vectorization_cost): Likewise.
13821 * tree-cfg.c (verify_gimple_comparison): Likewise.
13822 (verify_gimple_assign_binary): Likewise.
13823 (verify_gimple_assign_ternary): Likewise.
13824 (verify_gimple_assign_single): Likewise.
13825 * tree-pretty-print.c (dump_generic_node): Likewise.
13826 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13827 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
13828 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
13829 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
13830 (vect_shift_permute_load_chain): Likewise.
13831 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
13832 (expand_vector_condition, optimize_vector_constructor): Likewise.
13833 (lower_vec_perm, get_compute_type): Likewise.
13834 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
13835 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
13836 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
13837 (vect_recog_mask_conversion_pattern): Likewise.
13838 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
13839 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
13840 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13841 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
13842 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
13843 (vectorizable_shift, vectorizable_operation, vectorizable_store)
13844 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
13845 (supportable_widening_operation): Likewise.
13846 (supportable_narrowing_operation): Likewise.
13847 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
13849 * varasm.c (output_constant): Likewise.
13851 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13852 Alan Hayward <alan.hayward@arm.com>
13853 David Sherwood <david.sherwood@arm.com>
13855 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
13856 so that both the length == 3 and length != 3 cases set up their
13857 own permute vectors. Add comments explaining why we know the
13858 number of elements is constant.
13859 (vect_permute_load_chain): Likewise.
13861 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13862 Alan Hayward <alan.hayward@arm.com>
13863 David Sherwood <david.sherwood@arm.com>
13865 * machmode.h (mode_nunits): Change from unsigned char to
13867 (ONLY_FIXED_SIZE_MODES): New macro.
13868 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
13869 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
13870 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
13872 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
13873 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
13874 or if measurement_type is not polynomial.
13875 * genmodes.c (ZERO_COEFFS): New macro.
13876 (emit_mode_nunits_inline): Make mode_nunits_inline return a
13878 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
13879 Use ZERO_COEFFS when emitting initializers.
13880 * data-streamer.h (bp_pack_poly_value): New function.
13881 (bp_unpack_poly_value): Likewise.
13882 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13883 for GET_MODE_NUNITS.
13884 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13885 for GET_MODE_NUNITS.
13886 * tree.c (make_vector_type): Remove temporary shim and make
13887 the real function take the number of units as a poly_uint64
13888 rather than an int.
13889 (build_vector_type_for_mode): Handle polynomial nunits.
13890 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
13891 * emit-rtl.c (const_vec_series_p_1): Likewise.
13892 (gen_rtx_CONST_VECTOR): Likewise.
13893 * fold-const.c (test_vec_duplicate_folding): Likewise.
13894 * genrecog.c (validate_pattern): Likewise.
13895 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
13896 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
13897 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
13898 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
13899 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
13900 * rtlanal.c (subreg_get_info): Likewise.
13901 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13902 (vect_grouped_load_supported): Likewise.
13903 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
13904 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
13905 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
13906 (simplify_const_unary_operation, simplify_binary_operation_1)
13907 (simplify_const_binary_operation, simplify_ternary_operation)
13908 (test_vector_ops_duplicate, test_vector_ops): Likewise.
13909 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
13910 instead of CONST_VECTOR_NUNITS.
13911 * varasm.c (output_constant_pool_2): Likewise.
13912 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
13913 explicit-encoded elements in the XVEC for variable-length vectors.
13915 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13917 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
13919 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13920 Alan Hayward <alan.hayward@arm.com>
13921 David Sherwood <david.sherwood@arm.com>
13923 * coretypes.h (fixed_size_mode): Declare.
13924 (fixed_size_mode_pod): New typedef.
13925 * builtins.h (target_builtins::x_apply_args_mode)
13926 (target_builtins::x_apply_result_mode): Change type to
13927 fixed_size_mode_pod.
13928 * builtins.c (apply_args_size, apply_result_size, result_vector)
13929 (expand_builtin_apply_args_1, expand_builtin_apply)
13930 (expand_builtin_return): Update accordingly.
13932 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13934 * cse.c (hash_rtx_cb): Hash only the encoded elements.
13935 * cselib.c (cselib_hash_rtx): Likewise.
13936 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
13937 CONST_VECTOR encoding.
13939 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13940 Jeff Law <law@redhat.com>
13943 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
13944 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
13945 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
13946 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
13949 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
13950 explicitly probe *sp in a noreturn function if there were any callee
13951 register saves or frame pointer is needed.
13953 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13956 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
13957 BLKmode for ternary, binary or unary expressions.
13960 * var-tracking.c (delete_vta_debug_insn): New inline function.
13961 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
13962 insns from get_insns () to NULL instead of each bb separately.
13963 Use delete_vta_debug_insn. No longer static.
13964 (vt_debug_insns_local, variable_tracking_main_1): Adjust
13965 delete_vta_debug_insns callers.
13966 * rtl.h (delete_vta_debug_insns): Declare.
13967 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
13968 instead of variable_tracking_main.
13970 2018-01-03 Martin Sebor <msebor@redhat.com>
13972 PR tree-optimization/83603
13973 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
13974 arguments past the endof the argument list in functions declared
13975 without a prototype.
13976 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
13977 Avoid checking when arguments are null.
13979 2018-01-03 Martin Sebor <msebor@redhat.com>
13982 * doc/extend.texi (attribute const): Fix a typo.
13983 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
13984 issuing -Wsuggest-attribute for void functions.
13986 2018-01-03 Martin Sebor <msebor@redhat.com>
13988 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
13989 offset_int::from instead of wide_int::to_shwi.
13990 (maybe_diag_overlap): Remove assertion.
13991 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
13992 * gimple-ssa-sprintf.c (format_directive): Same.
13993 (parse_directive): Same.
13994 (sprintf_dom_walker::compute_format_length): Same.
13995 (try_substitute_return_value): Same.
13997 2018-01-03 Jeff Law <law@redhat.com>
13999 PR middle-end/83654
14000 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
14001 non-constant residual for zero at runtime and avoid probing in
14002 that case. Reorganize code for trailing problem to mirror handling
14005 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14007 PR tree-optimization/83501
14008 * tree-ssa-strlen.c (get_string_cst): New.
14009 (handle_char_store): Call get_string_cst.
14011 2018-01-03 Martin Liska <mliska@suse.cz>
14013 PR tree-optimization/83593
14014 * tree-ssa-strlen.c: Include tree-cfg.h.
14015 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
14016 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
14017 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
14019 (strlen_dom_walker::before_dom_children): Call
14020 gimple_purge_dead_eh_edges. Dump tranformation with details
14022 (strlen_dom_walker::before_dom_children): Update call by adding
14023 new argument cleanup_eh.
14024 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
14026 2018-01-03 Martin Liska <mliska@suse.cz>
14029 * cif-code.def (VARIADIC_THUNK): New enum value.
14030 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
14033 2018-01-03 Jan Beulich <jbeulich@suse.com>
14035 * sse.md (mov<mode>_internal): Tighten condition for when to use
14036 vmovdqu<ssescalarsize> for TI and OI modes.
14038 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14040 Update copyright years.
14042 2018-01-03 Martin Liska <mliska@suse.cz>
14045 * ipa-visibility.c (function_and_variable_visibility): Skip
14046 functions with noipa attribure.
14048 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14050 * gcc.c (process_command): Update copyright notice dates.
14051 * gcov-dump.c (print_version): Ditto.
14052 * gcov.c (print_version): Ditto.
14053 * gcov-tool.c (print_version): Ditto.
14054 * gengtype.c (create_file): Ditto.
14055 * doc/cpp.texi: Bump @copying's copyright year.
14056 * doc/cppinternals.texi: Ditto.
14057 * doc/gcc.texi: Ditto.
14058 * doc/gccint.texi: Ditto.
14059 * doc/gcov.texi: Ditto.
14060 * doc/install.texi: Ditto.
14061 * doc/invoke.texi: Ditto.
14063 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14065 * vector-builder.h (vector_builder::m_full_nelts): Change from
14066 unsigned int to poly_uint64.
14067 (vector_builder::full_nelts): Update prototype accordingly.
14068 (vector_builder::new_vector): Likewise.
14069 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
14070 (vector_builder::operator ==): Likewise.
14071 (vector_builder::finalize): Likewise.
14072 * int-vector-builder.h (int_vector_builder::int_vector_builder):
14073 Take the number of elements as a poly_uint64 rather than an
14075 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
14076 from unsigned int to poly_uint64.
14077 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
14078 (vec_perm_indices::new_vector): Likewise.
14079 (vec_perm_indices::length): Likewise.
14080 (vec_perm_indices::nelts_per_input): Likewise.
14081 (vec_perm_indices::input_nelts): Likewise.
14082 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
14083 number of elements per input as a poly_uint64 rather than an
14084 unsigned int. Use the original encoding for variable-length
14085 vectors, rather than clamping each individual element.
14086 For the second and subsequent elements in each pattern,
14087 clamp the step and base before clamping their sum.
14088 (vec_perm_indices::series_p): Handle polynomial element counts.
14089 (vec_perm_indices::all_in_range_p): Likewise.
14090 (vec_perm_indices_to_tree): Likewise.
14091 (vec_perm_indices_to_rtx): Likewise.
14092 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
14093 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
14094 (tree_vector_builder::new_binary_operation): Handle polynomial
14095 element counts. Return false if we need to know the number
14096 of elements at compile time.
14097 * fold-const.c (fold_vec_perm): Punt if the number of elements
14098 isn't known at compile time.
14100 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14102 * vec-perm-indices.h (vec_perm_builder): Change element type
14103 from HOST_WIDE_INT to poly_int64.
14104 (vec_perm_indices::element_type): Update accordingly.
14105 (vec_perm_indices::clamp): Handle polynomial element_types.
14106 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
14107 (vec_perm_indices::all_in_range_p): Likewise.
14108 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
14110 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
14111 polynomial vec_perm_indices element types.
14112 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
14113 * fold-const.c (fold_vec_perm): Likewise.
14114 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
14115 * tree-vect-generic.c (lower_vec_perm): Likewise.
14116 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
14117 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
14118 element type to HOST_WIDE_INT.
14120 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14121 Alan Hayward <alan.hayward@arm.com>
14122 David Sherwood <david.sherwood@arm.com>
14124 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
14125 rather than an int. Use plus_constant.
14126 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
14127 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
14129 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14130 Alan Hayward <alan.hayward@arm.com>
14131 David Sherwood <david.sherwood@arm.com>
14133 * calls.c (emit_call_1, expand_call): Change struct_value_size from
14134 a HOST_WIDE_INT to a poly_int64.
14136 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14137 Alan Hayward <alan.hayward@arm.com>
14138 David Sherwood <david.sherwood@arm.com>
14140 * calls.c (load_register_parameters): Cope with polynomial
14141 mode sizes. Require a constant size for BLKmode parameters
14142 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
14143 forces a parameter to be padded at the lsb end in order to
14144 fill a complete number of words, require the parameter size
14145 to be ordered wrt UNITS_PER_WORD.
14147 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14148 Alan Hayward <alan.hayward@arm.com>
14149 David Sherwood <david.sherwood@arm.com>
14151 * reload1.c (spill_stack_slot_width): Change element type
14152 from unsigned int to poly_uint64_pod.
14153 (alter_reg): Treat mode sizes as polynomial.
14155 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14156 Alan Hayward <alan.hayward@arm.com>
14157 David Sherwood <david.sherwood@arm.com>
14159 * reload.c (complex_word_subreg_p): New function.
14160 (reload_inner_reg_of_subreg, push_reload): Use it.
14162 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14163 Alan Hayward <alan.hayward@arm.com>
14164 David Sherwood <david.sherwood@arm.com>
14166 * lra-constraints.c (process_alt_operands): Reject matched
14167 operands whose sizes aren't ordered.
14168 (match_reload): Refer to this check here.
14170 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14171 Alan Hayward <alan.hayward@arm.com>
14172 David Sherwood <david.sherwood@arm.com>
14174 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
14175 that the mode size is in the set {1, 2, 4, 8, 16}.
14177 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14178 Alan Hayward <alan.hayward@arm.com>
14179 David Sherwood <david.sherwood@arm.com>
14181 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
14182 Use plus_constant instead of gen_rtx_PLUS.
14184 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14185 Alan Hayward <alan.hayward@arm.com>
14186 David Sherwood <david.sherwood@arm.com>
14188 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
14189 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
14190 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
14191 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
14192 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
14193 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
14194 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
14195 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
14196 * config/i386/i386.c (ix86_push_rounding): ...this new function.
14197 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
14199 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
14200 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
14201 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
14202 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
14203 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
14204 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
14205 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
14206 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
14207 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
14208 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
14210 * expr.c (emit_move_resolve_push): Treat the input and result
14211 of PUSH_ROUNDING as a poly_int64.
14212 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
14213 (emit_push_insn): Likewise.
14214 * lra-eliminations.c (mark_not_eliminable): Likewise.
14215 * recog.c (push_operand): Likewise.
14216 * reload1.c (elimination_effects): Likewise.
14217 * rtlanal.c (nonzero_bits1): Likewise.
14218 * calls.c (store_one_arg): Likewise. Require the padding to be
14219 known at compile time.
14221 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14222 Alan Hayward <alan.hayward@arm.com>
14223 David Sherwood <david.sherwood@arm.com>
14225 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
14226 Use plus_constant instead of gen_rtx_PLUS.
14228 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14229 Alan Hayward <alan.hayward@arm.com>
14230 David Sherwood <david.sherwood@arm.com>
14232 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
14233 rather than an int.
14235 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14236 Alan Hayward <alan.hayward@arm.com>
14237 David Sherwood <david.sherwood@arm.com>
14239 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
14240 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
14241 via stack temporaries. Treat the mode size as polynomial too.
14243 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14244 Alan Hayward <alan.hayward@arm.com>
14245 David Sherwood <david.sherwood@arm.com>
14247 * expr.c (expand_expr_real_2): When handling conversions involving
14248 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
14249 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
14250 as a poly_uint64 too.
14252 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14253 Alan Hayward <alan.hayward@arm.com>
14254 David Sherwood <david.sherwood@arm.com>
14256 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
14258 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14259 Alan Hayward <alan.hayward@arm.com>
14260 David Sherwood <david.sherwood@arm.com>
14262 * combine.c (can_change_dest_mode): Handle polynomial
14263 REGMODE_NATURAL_SIZE.
14264 * expmed.c (store_bit_field_1): Likewise.
14265 * expr.c (store_constructor): Likewise.
14266 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
14267 and polynomial REGMODE_NATURAL_SIZE.
14268 (gen_lowpart_common): Likewise.
14269 * reginfo.c (record_subregs_of_mode): Likewise.
14270 * rtlanal.c (read_modify_subreg_p): Likewise.
14272 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14273 Alan Hayward <alan.hayward@arm.com>
14274 David Sherwood <david.sherwood@arm.com>
14276 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
14277 numbers of elements.
14279 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14280 Alan Hayward <alan.hayward@arm.com>
14281 David Sherwood <david.sherwood@arm.com>
14283 * match.pd: Cope with polynomial numbers of vector elements.
14285 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14286 Alan Hayward <alan.hayward@arm.com>
14287 David Sherwood <david.sherwood@arm.com>
14289 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
14290 in a POINTER_PLUS_EXPR.
14292 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14293 Alan Hayward <alan.hayward@arm.com>
14294 David Sherwood <david.sherwood@arm.com>
14296 * omp-simd-clone.c (simd_clone_subparts): New function.
14297 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
14298 (ipa_simd_modify_function_body): Likewise.
14300 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14301 Alan Hayward <alan.hayward@arm.com>
14302 David Sherwood <david.sherwood@arm.com>
14304 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
14305 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
14306 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
14307 (expand_vector_condition, vector_element): Likewise.
14308 (subparts_gt): New function.
14309 (get_compute_type): Use subparts_gt.
14310 (count_type_subparts): Delete.
14311 (expand_vector_operations_1): Use subparts_gt instead of
14312 count_type_subparts.
14314 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14315 Alan Hayward <alan.hayward@arm.com>
14316 David Sherwood <david.sherwood@arm.com>
14318 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
14319 (vect_compile_time_alias): ...this new function. Do the calculation
14320 on poly_ints rather than trees.
14321 (vect_prune_runtime_alias_test_list): Update call accordingly.
14323 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14324 Alan Hayward <alan.hayward@arm.com>
14325 David Sherwood <david.sherwood@arm.com>
14327 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
14329 (vect_schedule_slp_instance): Likewise.
14331 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14332 Alan Hayward <alan.hayward@arm.com>
14333 David Sherwood <david.sherwood@arm.com>
14335 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
14336 constant and extern definitions for variable-length vectors.
14337 (vect_get_constant_vectors): Note that the number of units
14338 is known to be constant.
14340 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14341 Alan Hayward <alan.hayward@arm.com>
14342 David Sherwood <david.sherwood@arm.com>
14344 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
14345 of units as polynomial. Choose between WIDE and NARROW based
14348 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14349 Alan Hayward <alan.hayward@arm.com>
14350 David Sherwood <david.sherwood@arm.com>
14352 * tree-vect-stmts.c (simd_clone_subparts): New function.
14353 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
14355 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14356 Alan Hayward <alan.hayward@arm.com>
14357 David Sherwood <david.sherwood@arm.com>
14359 * tree-vect-stmts.c (vectorizable_call): Treat the number of
14360 vectors as polynomial. Use build_index_vector for
14361 IFN_GOMP_SIMD_LANE.
14363 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14364 Alan Hayward <alan.hayward@arm.com>
14365 David Sherwood <david.sherwood@arm.com>
14367 * tree-vect-stmts.c (get_load_store_type): Treat the number of
14368 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
14369 for variable-length vectors.
14370 (vectorizable_mask_load_store): Treat the number of units as
14371 polynomial, asserting that it is constant if the condition has
14372 already been enforced.
14373 (vectorizable_store, vectorizable_load): Likewise.
14375 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14376 Alan Hayward <alan.hayward@arm.com>
14377 David Sherwood <david.sherwood@arm.com>
14379 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
14380 of units as polynomial. Punt if we can't tell at compile time
14381 which vector contains the final result.
14383 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14384 Alan Hayward <alan.hayward@arm.com>
14385 David Sherwood <david.sherwood@arm.com>
14387 * tree-vect-loop.c (vectorizable_induction): Treat the number
14388 of units as polynomial. Punt on SLP inductions. Use an integer
14389 VEC_SERIES_EXPR for variable-length integer reductions. Use a
14390 cast of such a series for variable-length floating-point
14393 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14394 Alan Hayward <alan.hayward@arm.com>
14395 David Sherwood <david.sherwood@arm.com>
14397 * tree.h (build_index_vector): Declare.
14398 * tree.c (build_index_vector): New function.
14399 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
14400 of units as polynomial, forcibly converting it to a constant if
14401 vectorizable_reduction has already enforced the condition.
14402 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
14403 to create a {1,2,3,...} vector.
14404 (vectorizable_reduction): Treat the number of units as polynomial.
14405 Choose vectype_in based on the largest scalar element size rather
14406 than the smallest number of units. Enforce the restrictions
14409 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14410 Alan Hayward <alan.hayward@arm.com>
14411 David Sherwood <david.sherwood@arm.com>
14413 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
14414 number of units as polynomial.
14416 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14417 Alan Hayward <alan.hayward@arm.com>
14418 David Sherwood <david.sherwood@arm.com>
14420 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
14421 * target.def (autovectorize_vector_sizes): Return the vector sizes
14422 by pointer, using vector_sizes rather than a bitmask.
14423 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
14424 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
14425 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
14427 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
14428 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
14429 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
14430 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
14431 * omp-general.c (omp_max_vf): Likewise.
14432 * omp-low.c (omp_clause_aligned_alignment): Likewise.
14433 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
14434 * tree-vect-loop.c (vect_analyze_loop): Likewise.
14435 * tree-vect-slp.c (vect_slp_bb): Likewise.
14436 * doc/tm.texi: Regenerate.
14437 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
14439 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
14440 the vector size as a poly_uint64 rather than an unsigned int.
14441 (current_vector_size): Change from an unsigned int to a poly_uint64.
14442 (get_vectype_for_scalar_type): Update accordingly.
14443 * tree.h (build_truth_vector_type): Take the size and number of
14444 units as a poly_uint64 rather than an unsigned int.
14445 (build_vector_type): Add a temporary overload that takes
14446 the number of units as a poly_uint64 rather than an unsigned int.
14447 * tree.c (make_vector_type): Likewise.
14448 (build_truth_vector_type): Take the number of units as a poly_uint64
14449 rather than an unsigned int.
14451 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14452 Alan Hayward <alan.hayward@arm.com>
14453 David Sherwood <david.sherwood@arm.com>
14455 * target.def (get_mask_mode): Take the number of units and length
14456 as poly_uint64s rather than unsigned ints.
14457 * targhooks.h (default_get_mask_mode): Update accordingly.
14458 * targhooks.c (default_get_mask_mode): Likewise.
14459 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
14460 * doc/tm.texi: Regenerate.
14462 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14463 Alan Hayward <alan.hayward@arm.com>
14464 David Sherwood <david.sherwood@arm.com>
14466 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
14467 * omp-general.c (omp_max_vf): Likewise.
14468 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
14469 (expand_omp_simd): Handle polynomial safelen.
14470 * omp-low.c (omplow_simd_context): Add a default constructor.
14471 (omplow_simd_context::max_vf): Change from int to poly_uint64.
14472 (lower_rec_simd_input_clauses): Update accordingly.
14473 (lower_rec_input_clauses): Likewise.
14475 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14476 Alan Hayward <alan.hayward@arm.com>
14477 David Sherwood <david.sherwood@arm.com>
14479 * tree-vectorizer.h (vect_nunits_for_cost): New function.
14480 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
14481 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
14482 (vect_analyze_slp_cost): Likewise.
14483 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
14484 (vect_model_load_cost): Likewise.
14486 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14487 Alan Hayward <alan.hayward@arm.com>
14488 David Sherwood <david.sherwood@arm.com>
14490 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
14491 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
14492 from an unsigned int * to a poly_uint64_pod *.
14493 (calculate_unrolling_factor): New function.
14494 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
14496 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14497 Alan Hayward <alan.hayward@arm.com>
14498 David Sherwood <david.sherwood@arm.com>
14500 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
14501 from an unsigned int to a poly_uint64.
14502 (_loop_vec_info::slp_unrolling_factor): Likewise.
14503 (_loop_vec_info::vectorization_factor): Change from an int
14505 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
14506 (vect_get_num_vectors): New function.
14507 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
14508 (vect_get_num_copies): Use vect_get_num_vectors.
14509 (vect_analyze_data_ref_dependences): Change max_vf from an int *
14510 to an unsigned int *.
14511 (vect_analyze_data_refs): Change min_vf from an int * to a
14513 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
14514 than an unsigned HOST_WIDE_INT.
14515 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
14516 (vect_analyze_data_ref_dependence): Change max_vf from an int *
14517 to an unsigned int *.
14518 (vect_analyze_data_ref_dependences): Likewise.
14519 (vect_compute_data_ref_alignment): Handle polynomial vf.
14520 (vect_enhance_data_refs_alignment): Likewise.
14521 (vect_prune_runtime_alias_test_list): Likewise.
14522 (vect_shift_permute_load_chain): Likewise.
14523 (vect_supportable_dr_alignment): Likewise.
14524 (dependence_distance_ge_vf): Take the vectorization factor as a
14525 poly_uint64 rather than an unsigned HOST_WIDE_INT.
14526 (vect_analyze_data_refs): Change min_vf from an int * to a
14528 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
14529 vfm1 as a poly_uint64 rather than an int. Make the same change
14530 for the returned bound_scalar.
14531 (vect_gen_vector_loop_niters): Handle polynomial vf.
14532 (vect_do_peeling): Likewise. Update call to
14533 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
14534 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
14536 * tree-vect-loop.c (vect_determine_vectorization_factor)
14537 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
14538 (vect_get_known_peeling_cost): Likewise.
14539 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
14540 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
14541 (vect_transform_loop): Likewise. Use the lowest possible VF when
14542 updating the upper bounds of the loop.
14543 (vect_min_worthwhile_factor): Make static. Return an unsigned int
14544 rather than an int.
14545 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
14546 polynomial unroll factors.
14547 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
14548 (vect_make_slp_decision): Likewise.
14549 (vect_supported_load_permutation_p): Likewise, and polynomial
14551 (vect_analyze_slp_cost): Handle polynomial vf.
14552 (vect_slp_analyze_node_operations): Likewise.
14553 (vect_slp_analyze_bb_1): Likewise.
14554 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
14555 than an unsigned HOST_WIDE_INT.
14556 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
14557 (vectorizable_load): Handle polynomial vf.
14558 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
14560 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
14562 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14563 Alan Hayward <alan.hayward@arm.com>
14564 David Sherwood <david.sherwood@arm.com>
14566 * match.pd: Handle bit operations involving three constants
14567 and try to fold one pair.
14569 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14571 * tree-vect-loop-manip.c: Include gimple-fold.h.
14572 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
14573 niters_maybe_zero parameters. Handle other cases besides a step of 1.
14574 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
14575 Add a path that uses a step of VF instead of 1, but disable it
14577 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
14578 and niters_no_overflow parameters. Update calls to
14579 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
14580 Create a new SSA name if the latter choses to use a ste other
14581 than zero, and return it via niters_vector_mult_vf_var.
14582 * tree-vect-loop.c (vect_transform_loop): Update calls to
14583 vect_do_peeling, vect_gen_vector_loop_niters and
14584 slpeel_make_loop_iterate_ntimes.
14585 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
14586 (vect_gen_vector_loop_niters): Update declarations after above changes.
14588 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
14590 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
14591 128-bit round to integer instructions.
14592 (ceil<mode>2): Likewise.
14593 (btrunc<mode>2): Likewise.
14594 (round<mode>2): Likewise.
14596 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14598 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
14599 unaligned VSX load/store on P8/P9.
14600 (expand_block_clear): Allow the use of unaligned VSX
14601 load/store on P8/P9.
14603 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14605 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
14607 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
14608 swap associated with both a load and a store.
14610 2018-01-02 Andrew Waterman <andrew@sifive.com>
14612 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
14613 * config/riscv/riscv.md (clear_cache): Use it.
14615 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
14617 * web.c: Remove out-of-date comment.
14619 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14621 * expr.c (fixup_args_size_notes): Check that any existing
14622 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
14623 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
14624 (emit_single_push_insn): ...here.
14626 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14628 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
14629 (const_vector_encoded_nelts): New function.
14630 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
14631 (const_vector_int_elt, const_vector_elt): Declare.
14632 * emit-rtl.c (const_vector_int_elt_1): New function.
14633 (const_vector_elt): Likewise.
14634 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
14635 of CONST_VECTOR_ELT.
14637 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14639 * expr.c: Include rtx-vector-builder.h.
14640 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
14641 directly on the tree encoding.
14642 (const_vector_from_tree): Likewise.
14643 * optabs.c: Include rtx-vector-builder.h.
14644 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
14645 sequence of "u" values.
14646 * vec-perm-indices.c: Include rtx-vector-builder.h.
14647 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
14648 directly on the vec_perm_indices encoding.
14650 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14652 * doc/rtl.texi (const_vector): Describe new encoding scheme.
14653 * Makefile.in (OBJS): Add rtx-vector-builder.o.
14654 * rtx-vector-builder.h: New file.
14655 * rtx-vector-builder.c: Likewise.
14656 * rtl.h (rtx_def::u2): Add a const_vector field.
14657 (CONST_VECTOR_NPATTERNS): New macro.
14658 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
14659 (CONST_VECTOR_DUPLICATE_P): Likewise.
14660 (CONST_VECTOR_STEPPED_P): Likewise.
14661 (CONST_VECTOR_ENCODED_ELT): Likewise.
14662 (const_vec_duplicate_p): Check for a duplicated vector encoding.
14663 (unwrap_const_vec_duplicate): Likewise.
14664 (const_vec_series_p): Check for a non-duplicated vector encoding.
14665 Say that the function only returns true for integer vectors.
14666 * emit-rtl.c: Include rtx-vector-builder.h.
14667 (gen_const_vec_duplicate_1): Delete.
14668 (gen_const_vector): Call gen_const_vec_duplicate instead of
14669 gen_const_vec_duplicate_1.
14670 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
14671 (gen_const_vec_duplicate): Use rtx_vector_builder.
14672 (gen_const_vec_series): Likewise.
14673 (gen_rtx_CONST_VECTOR): Likewise.
14674 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
14675 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
14676 Build a new vector rather than modifying a CONST_VECTOR in-place.
14677 (handle_special_swappables): Update call accordingly.
14678 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
14679 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
14680 Build a new vector rather than modifying a CONST_VECTOR in-place.
14681 (handle_special_swappables): Update call accordingly.
14683 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14685 * simplify-rtx.c (simplify_const_binary_operation): Use
14686 CONST_VECTOR_ELT instead of XVECEXP.
14688 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14690 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
14691 the selector elements to be different from the data elements
14692 if the selector is a VECTOR_CST.
14693 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
14694 ssizetype for the selector.
14696 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14698 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
14699 before testing each element individually.
14700 * tree-vect-generic.c (lower_vec_perm): Likewise.
14702 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14704 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
14705 * selftest-run-tests.c (selftest::run_tests): Call it.
14706 * vector-builder.h (vector_builder::operator ==): New function.
14707 (vector_builder::operator !=): Likewise.
14708 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
14709 (vec_perm_indices::all_from_input_p): New function.
14710 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
14711 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
14712 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
14713 instead of reading the VECTOR_CST directly. Detect whether both
14714 vector inputs are the same before constructing the vec_perm_indices,
14715 and update the number of inputs argument accordingly. Use the
14716 utility functions added above. Only construct sel2 if we need to.
14718 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14720 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
14721 the broadcast of the low byte.
14722 (expand_mult_highpart): Use an explicit encoding for the permutes.
14723 * optabs-query.c (can_mult_highpart_p): Likewise.
14724 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
14725 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14726 (vectorizable_bswap): Likewise.
14727 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
14728 explicit encoding for the power-of-2 permutes.
14729 (vect_permute_store_chain): Likewise.
14730 (vect_grouped_load_supported): Likewise.
14731 (vect_permute_load_chain): Likewise.
14733 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14735 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
14736 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
14737 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
14738 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
14739 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
14740 (vect_gen_perm_mask_any): Likewise.
14742 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14744 * int-vector-builder.h: New file.
14745 * vec-perm-indices.h: Include int-vector-builder.h.
14746 (vec_perm_indices): Redefine as an int_vector_builder.
14747 (auto_vec_perm_indices): Delete.
14748 (vec_perm_builder): Redefine as a stand-alone class.
14749 (vec_perm_indices::vec_perm_indices): New function.
14750 (vec_perm_indices::clamp): Likewise.
14751 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
14752 (vec_perm_indices::new_vector): New function.
14753 (vec_perm_indices::new_expanded_vector): Update for new
14754 vec_perm_indices class.
14755 (vec_perm_indices::rotate_inputs): New function.
14756 (vec_perm_indices::all_in_range_p): Operate directly on the
14757 encoded form, without computing elided elements.
14758 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
14759 encoding. Update for new vec_perm_indices class.
14760 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
14761 the given vec_perm_builder.
14762 (expand_vec_perm_var): Update vec_perm_builder constructor.
14763 (expand_mult_highpart): Use vec_perm_builder instead of
14764 auto_vec_perm_indices.
14765 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
14766 vec_perm_indices instead of auto_vec_perm_indices. Use a single
14767 or double series encoding as appropriate.
14768 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
14769 vec_perm_indices instead of auto_vec_perm_indices.
14770 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14771 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14772 (vect_permute_store_chain): Likewise.
14773 (vect_grouped_load_supported): Likewise.
14774 (vect_permute_load_chain): Likewise.
14775 (vect_shift_permute_load_chain): Likewise.
14776 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
14777 (vect_transform_slp_perm_load): Likewise.
14778 (vect_schedule_slp_instance): Likewise.
14779 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14780 (vectorizable_mask_load_store): Likewise.
14781 (vectorizable_bswap): Likewise.
14782 (vectorizable_store): Likewise.
14783 (vectorizable_load): Likewise.
14784 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
14785 vec_perm_indices instead of auto_vec_perm_indices. Use
14786 tree_to_vec_perm_builder to read the vector from a tree.
14787 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
14788 vec_perm_builder instead of a vec_perm_indices.
14789 (have_whole_vector_shift): Use vec_perm_builder and
14790 vec_perm_indices instead of auto_vec_perm_indices. Leave the
14791 truncation to calc_vec_perm_mask_for_shift.
14792 (vect_create_epilog_for_reduction): Likewise.
14793 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
14794 from auto_vec_perm_indices to vec_perm_indices.
14795 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
14796 instead of changing individual elements.
14797 (aarch64_vectorize_vec_perm_const): Use new_vector to install
14798 the vector in d.perm.
14799 * config/arm/arm.c (expand_vec_perm_d::perm): Change
14800 from auto_vec_perm_indices to vec_perm_indices.
14801 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
14802 instead of changing individual elements.
14803 (arm_vectorize_vec_perm_const): Use new_vector to install
14804 the vector in d.perm.
14805 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
14806 Update vec_perm_builder constructor.
14807 (rs6000_expand_interleave): Likewise.
14808 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
14809 (rs6000_expand_interleave): Likewise.
14811 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14813 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
14814 to qimode could truncate the indices.
14815 * optabs.c (expand_vec_perm_var): Likewise.
14817 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14819 * Makefile.in (OBJS): Add vec-perm-indices.o.
14820 * vec-perm-indices.h: New file.
14821 * vec-perm-indices.c: Likewise.
14822 * target.h (vec_perm_indices): Replace with a forward class
14824 (auto_vec_perm_indices): Move to vec-perm-indices.h.
14825 * optabs.h: Include vec-perm-indices.h.
14826 (expand_vec_perm): Delete.
14827 (selector_fits_mode_p, expand_vec_perm_var): Declare.
14828 (expand_vec_perm_const): Declare.
14829 * target.def (vec_perm_const_ok): Replace with...
14830 (vec_perm_const): ...this new hook.
14831 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
14832 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
14833 * doc/tm.texi: Regenerate.
14834 * optabs.def (vec_perm_const): Delete.
14835 * doc/md.texi (vec_perm_const): Likewise.
14836 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
14837 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
14838 expand_vec_perm for constant permutation vectors. Assert that
14839 the mode of variable permutation vectors is the integer equivalent
14840 of the mode that is being permuted.
14841 * optabs-query.h (selector_fits_mode_p): Declare.
14842 * optabs-query.c: Include vec-perm-indices.h.
14843 (selector_fits_mode_p): New function.
14844 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
14845 is defined, instead of checking whether the vec_perm_const_optab
14846 exists. Use targetm.vectorize.vec_perm_const instead of
14847 targetm.vectorize.vec_perm_const_ok. Check whether the indices
14848 fit in the vector mode before using a variable permute.
14849 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
14850 vec_perm_indices instead of an rtx.
14851 (expand_vec_perm): Replace with...
14852 (expand_vec_perm_const): ...this new function. Take the selector
14853 as a vec_perm_indices rather than an rtx. Also take the mode of
14854 the selector. Update call to shift_amt_for_vec_perm_mask.
14855 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
14856 Use vec_perm_indices::new_expanded_vector to expand the original
14857 selector into bytes. Check whether the indices fit in the vector
14858 mode before using a variable permute.
14859 (expand_vec_perm_var): Make global.
14860 (expand_mult_highpart): Use expand_vec_perm_const.
14861 * fold-const.c: Includes vec-perm-indices.h.
14862 * tree-ssa-forwprop.c: Likewise.
14863 * tree-vect-data-refs.c: Likewise.
14864 * tree-vect-generic.c: Likewise.
14865 * tree-vect-loop.c: Likewise.
14866 * tree-vect-slp.c: Likewise.
14867 * tree-vect-stmts.c: Likewise.
14868 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
14870 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
14871 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
14872 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
14873 (aarch64_vectorize_vec_perm_const): ...this new function.
14874 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14875 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14876 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
14877 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
14878 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14879 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14880 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
14882 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
14883 check for NEON modes.
14884 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
14885 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
14886 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
14887 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
14889 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
14890 the old VEC_PERM_CONST conditions.
14891 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
14892 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
14893 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
14894 (ia64_vectorize_vec_perm_const_ok): Merge into...
14895 (ia64_vectorize_vec_perm_const): ...this new function.
14896 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
14897 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
14898 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
14899 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
14900 * config/mips/mips.c (mips_expand_vec_perm_const)
14901 (mips_vectorize_vec_perm_const_ok): Merge into...
14902 (mips_vectorize_vec_perm_const): ...this new function.
14903 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
14904 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
14905 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
14906 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
14907 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
14908 (rs6000_expand_vec_perm_const): Delete.
14909 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
14911 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14912 (altivec_expand_vec_perm_const_le): Take each operand individually.
14913 Operate on constant selectors rather than rtxes.
14914 (altivec_expand_vec_perm_const): Likewise. Update call to
14915 altivec_expand_vec_perm_const_le.
14916 (rs6000_expand_vec_perm_const): Delete.
14917 (rs6000_vectorize_vec_perm_const_ok): Delete.
14918 (rs6000_vectorize_vec_perm_const): New function.
14919 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
14920 an element count and rtx array.
14921 (rs6000_expand_extract_even): Update call accordingly.
14922 (rs6000_expand_interleave): Likewise.
14923 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
14924 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
14925 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
14926 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
14927 (rs6000_expand_vec_perm_const): Delete.
14928 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14929 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14930 (altivec_expand_vec_perm_const_le): Take each operand individually.
14931 Operate on constant selectors rather than rtxes.
14932 (altivec_expand_vec_perm_const): Likewise. Update call to
14933 altivec_expand_vec_perm_const_le.
14934 (rs6000_expand_vec_perm_const): Delete.
14935 (rs6000_vectorize_vec_perm_const_ok): Delete.
14936 (rs6000_vectorize_vec_perm_const): New function. Remove stray
14937 reference to the SPE evmerge intructions.
14938 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
14939 an element count and rtx array.
14940 (rs6000_expand_extract_even): Update call accordingly.
14941 (rs6000_expand_interleave): Likewise.
14942 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
14943 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
14945 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14947 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14949 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
14950 vector mode and that that mode matches the mode of the data
14952 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
14953 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
14954 directly using expand_vec_perm_1 when forcing selectors into
14956 (expand_vec_perm_var): New function, split out from expand_vec_perm.
14958 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14960 * optabs-query.h (can_vec_perm_p): Delete.
14961 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
14962 * optabs-query.c (can_vec_perm_p): Split into...
14963 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
14964 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
14965 particular selector is valid.
14966 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14967 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14968 (vect_grouped_load_supported): Likewise.
14969 (vect_shift_permute_load_chain): Likewise.
14970 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
14971 (vect_transform_slp_perm_load): Likewise.
14972 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14973 (vectorizable_bswap): Likewise.
14974 (vect_gen_perm_mask_checked): Likewise.
14975 * fold-const.c (fold_ternary_loc): Likewise. Don't take
14976 implementations of variable permutation vectors into account
14977 when deciding which selector to use.
14978 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
14979 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
14980 with a false third argument.
14981 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
14982 to test whether the constant selector is valid and can_vec_perm_var_p
14983 to test whether a variable selector is valid.
14985 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14987 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
14988 * optabs-query.c (can_vec_perm_p): Likewise.
14989 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
14990 instead of vec_perm_indices.
14991 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
14992 (vect_gen_perm_mask_checked): Likewise,
14993 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
14994 (vect_gen_perm_mask_checked): Likewise,
14996 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14998 * optabs-query.h (qimode_for_vec_perm): Declare.
14999 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
15000 (qimode_for_vec_perm): ...this new function.
15001 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
15003 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15005 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
15006 does not have a conditional at the top.
15008 2018-01-02 Richard Biener <rguenther@suse.de>
15010 * ipa-inline.c (big_speedup_p): Fix expression.
15012 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
15015 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
15018 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
15022 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
15023 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
15024 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
15025 cond_taken_branch_cost 3->4.
15027 2018-01-01 Jakub Jelinek <jakub@redhat.com>
15029 PR tree-optimization/83581
15030 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
15031 TODO_cleanup_cfg if any changes have been made.
15033 PR middle-end/83608
15034 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
15035 convert_modes if target mode has the right side, but different mode
15038 PR middle-end/83609
15039 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
15040 last argument when extracting from CONCAT. If either from_real or
15041 from_imag is NULL, use expansion through memory. If result is not
15042 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
15043 the parts directly to inner mode, if even that fails, use expansion
15046 PR middle-end/83623
15047 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
15048 check for bswap in mode rather than HImode and use that in expand_unop
15051 Copyright (C) 2018 Free Software Foundation, Inc.
15053 Copying and distribution of this file, with or without modification,
15054 are permitted in any medium without royalty provided the copyright
15055 notice and this notice are preserved.