1 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
2 Kito Cheng <kito.cheng@gmail.com>
4 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
5 (nds32le-*-*, nds32be-*-*): Integrate checking process.
6 (nds32*-*-*): Add glibc and uclibc conditions.
7 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
8 (TARGET_EXCEPT_UNWIND_INFO): Define.
9 * config/nds32/elf.h: New file.
10 * config/nds32/linux.h: New file.
11 * config/nds32/nds32-elf.opt: New file.
12 * config/nds32/nds32-linux.opt: New file.
13 * config/nds32/nds32-fp-as-gp.c
14 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
15 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
17 (nds32_asm_file_end): Ditto.
18 (nds32_print_operand): Ditto.
19 (nds32_insert_attributes): Ditto.
20 (nds32_init_libfuncs): New function.
21 (TARGET_HAVE_TLS): Define.
22 (TARGET_INIT_LIBFUNCS): Define.
23 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
25 (TARGET_ELF): Apply different mcmodel setting.
26 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
27 been migrated into elf.h and linux.h files.
28 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
29 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
30 (mcmodel): The content has been migrated into nds32-elf.opt and
31 nds32-linux.opt files.
32 * config/nds32/t-elf: New file.
33 * config/nds32/t-linux: New file.
35 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
36 Shiva Chen <shiva0217@gmail.com>
38 * config/nds32/constants.md (unspec_volatile_element): Add
39 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
40 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
42 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
43 (make_pass_nds32_fp_as_gp): Declare.
44 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
46 (nds32_asm_function_end_prologue): Remove unused asm output.
47 (nds32_asm_function_begin_epilogue): Remove unused asm output.
48 (nds32_asm_file_start): Output necessary fp_as_gp information.
49 (nds32_option_override): Adjust register usage.
50 (nds32_expand_prologue): Consider fp_as_gp situation.
51 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
52 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
55 (simple_return): Ditto.
56 (omit_fp_begin): Output special directive for fp_as_gp.
57 (omit_fp_end): Output special directive for fp_as_gp.
58 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
59 mforbid-fp-as-gp): New options.
61 2018-06-01 Mark Wielaard <mark@klomp.org>
63 * dwarf2out.c (dwarf2out_finish): Remove generation of
66 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
68 * gimple-ssa-store-merging.c: Include gimple-fold.h.
69 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
70 (struct merged_store_group): Add bit_insertion field.
71 (dump_char_array): Use standard hexadecimal format.
72 (merged_store_group::merged_store_group): Set bit_insertion to false.
73 (merged_store_group::apply_stores): Use optimal buffer size. Deal
74 with BIT_INSERT_EXPR stores. Move up code updating the mask and
75 also print the mask in the dump file.
76 (pass_store_merging::gate): Minor tweak.
77 (imm_store_chain_info::coalesce_immediate): Fix wrong association
78 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
79 stores with INTEGER_CST stores.
80 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
81 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
82 and use it throughout. Generate bit insertion sequences if need be.
83 (pass_store_merging::process_store): Remove redundant condition.
84 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
86 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
88 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
89 the 128-bit floating point types. Fix function comment.
91 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
93 * config/aarch64/aarch64-simd.md
94 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
96 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
99 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
101 PR tree-optimization/85989
102 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
104 (backprop::intersect_uses): Check it when deciding whether this
105 is a backedge reference.
106 (backprop::process_block): Add each phi to m_visited_phis
107 after visiting it, then clear it at the end.
109 2018-06-01 Richard Biener <rguenther@suse.de>
111 * tree-vectorizer.h (vect_dr_stmt): New function.
112 (vect_get_load_cost): Adjust.
113 (vect_get_store_cost): Likewise.
114 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
115 Use vect_dr_stmt instead of DR_SMTT.
116 (vect_record_base_alignments): Likewise.
117 (vect_calculate_target_alignment): Likewise.
118 (vect_compute_data_ref_alignment): Likewise and make static.
119 (vect_update_misalignment_for_peel): Likewise.
120 (vect_verify_datarefs_alignment): Likewise.
121 (vector_alignment_reachable_p): Likewise.
122 (vect_get_data_access_cost): Likewise. Pass down
123 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
124 (vect_get_peeling_costs_all_drs): Likewise.
125 (vect_peeling_hash_get_lowest_cost): Likewise.
126 (vect_enhance_data_refs_alignment): Likewise.
127 (vect_find_same_alignment_drs): Likewise.
128 (vect_analyze_data_refs_alignment): Likewise.
129 (vect_analyze_group_access_1): Likewise.
130 (vect_analyze_group_access): Likewise.
131 (vect_analyze_data_ref_access): Likewise.
132 (vect_analyze_data_ref_accesses): Likewise.
133 (vect_vfa_segment_size): Likewise.
134 (vect_small_gap_p): Likewise.
135 (vectorizable_with_step_bound_p): Likewise.
136 (vect_prune_runtime_alias_test_list): Likewise.
137 (vect_analyze_data_refs): Likewise.
138 (vect_supportable_dr_alignment): Likewise.
139 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
140 (vect_gen_prolog_loop_niters): Likewise.
141 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
142 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
144 (vect_recog_mask_conversion_pattern): Likewise.
145 (vect_try_gather_scatter_pattern): Likewise.
146 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
147 to vect_get_store_cost.
148 (vect_get_store_cost): Get stmt_info instead of DR.
149 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
150 (vect_get_load_cost): Get stmt_info instead of DR.
152 2018-06-01 Richard Biener <rguenther@suse.de>
155 * gimple-fold.c (var_decl_component_p): Also allow offsetted
156 vars wrapped in MEM_REFs.
158 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
160 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
161 Fix subreg tests so that we only return a choice between
162 GENERAL_REGS and FP_REGS if the original classes included both.
164 2018-06-01 Richard Biener <rguenther@suse.de>
167 * tree-ssa-structalias.c (get_function_part_constraint):
168 Handle NULL fi->decl.
169 (find_func_aliases_for_call): Properly handle indirect
171 (find_func_clobbers): Likewise.
172 (ipa_pta_execute): Likewise.
173 (create_variable_info_for): For functions that are ifunc_resolver
174 resolve to a varinfo that contains the result of the resolver
176 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
179 2018-05-31 Michael Collison <michael.collison@arm.com>
181 * config/aarch64/aarch64.md:
182 (*fix_to_zero_extenddfdi2): New pattern.
183 * gcc.target/aarch64/fix_extend1.c: New testcase.
185 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
189 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
190 and BUILT_IN_STRNCMP_EQ.
191 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
193 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
194 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
195 (gimple_fold_builtin): Likewise.
196 * tree-ssa-strlen.c (compute_string_length): New function.
197 (determine_min_obsize): New function.
198 (handle_builtin_string_cmp): New function to handle calls to
199 string compare functions.
200 (strlen_optimize_stmt): Add handling to builtin string compare
202 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
203 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
204 * tree.c (build_common_builtin_nodes): Add new defines of
205 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
207 2018-05-31 Jakub Jelinek <jakub@redhat.com>
210 * bb-reorder.c (pass_partition_blocks::gate): Return false for
211 functions with naked attribute.
213 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
215 * config/i386/sse.md (avx_vec_concat<mode>):
216 Substitute concat_tg_mode mode attribute with xtg_mode.
217 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
218 (concat_tg_mode): Remove mode attribute.
220 2018-05-31 Martin Sebor <msebor@redhat.com>
223 * calls.c (alloc_max_size): Correct a logic error/typo.
224 Treat excessive arguments as infinite. Warn for invalid arguments.
225 * doc/invoke.texi (-Walloc-size-larger-than): Update.
227 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
230 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
231 and movx for Haswell.
233 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
234 Cesar Philippidis <cesar@codesourcery.com>
237 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
238 when emitting error on private/firstprivate reductions.
239 * omp-low.c (lower_omp_target): Avoid reference-type processing
240 on pointers for firstprivate clause.
242 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
244 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
247 * config/aarch64/aarch64-simd.md
248 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
249 (aarch64_ld1_x3_<mode>): Likewise
250 (aarch64_st1x2<VALLDIF:mode>): Likewise
251 (aarch64_st1_x2_<mode>): Likewise
252 (aarch64_st1x3<VALLDIF:mode>): Likewise
253 (aarch64_st1_x3_<mode>): Likewise
254 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
255 (vld1_s8_x3): Likewise.
256 (vld1_u16_x3): Likewise.
257 (vld1_s16_x3): Likewise.
258 (vld1_u32_x3): Likewise.
259 (vld1_s32_x3): Likewise.
260 (vld1_u64_x3): Likewise.
261 (vld1_s64_x3): Likewise.
262 (vld1_f16_x3): Likewise.
263 (vld1_f32_x3): Likewise.
264 (vld1_f64_x3): Likewise.
265 (vld1_p8_x3): Likewise.
266 (vld1_p16_x3): Likewise.
267 (vld1_p64_x3): Likewise.
268 (vld1q_u8_x3): Likewise.
269 (vld1q_s8_x3): Likewise.
270 (vld1q_u16_x3): Likewise.
271 (vld1q_s16_x3): Likewise.
272 (vld1q_u32_x3): Likewise.
273 (vld1q_s32_x3): Likewise.
274 (vld1q_u64_x3): Likewise.
275 (vld1q_s64_x3): Likewise.
276 (vld1q_f16_x3): Likewise.
277 (vld1q_f32_x3): Likewise.
278 (vld1q_f64_x3): Likewise.
279 (vld1q_p8_x3): Likewise.
280 (vld1q_p16_x3): Likewise.
281 (vld1q_p64_x3): Likewise.
282 (vst1_s64_x2): Likewise.
283 (vst1_u64_x2): Likewise.
284 (vst1_f64_x2): Likewise.
285 (vst1_s8_x2): Likewise.
286 (vst1_p8_x2): Likewise.
287 (vst1_s16_x2): Likewise.
288 (vst1_p16_x2): Likewise.
289 (vst1_s32_x2): Likewise.
290 (vst1_u8_x2): Likewise.
291 (vst1_u16_x2): Likewise.
292 (vst1_u32_x2): Likewise.
293 (vst1_f16_x2): Likewise.
294 (vst1_f32_x2): Likewise.
295 (vst1_p64_x2): Likewise.
296 (vst1q_s8_x2): Likewise.
297 (vst1q_p8_x2): Likewise.
298 (vst1q_s16_x2): Likewise.
299 (vst1q_p16_x2): Likewise.
300 (vst1q_s32_x2): Likewise.
301 (vst1q_s64_x2): Likewise.
302 (vst1q_u8_x2): Likewise.
303 (vst1q_u16_x2): Likewise.
304 (vst1q_u32_x2): Likewise.
305 (vst1q_u64_x2): Likewise.
306 (vst1q_f16_x2): Likewise.
307 (vst1q_f32_x2): Likewise.
308 (vst1q_f64_x2): Likewise.
309 (vst1q_p64_x2): Likewise.
310 (vst1_s64_x3): Likewise.
311 (vst1_u64_x3): Likewise.
312 (vst1_f64_x3): Likewise.
313 (vst1_s8_x3): Likewise.
314 (vst1_p8_x3): Likewise.
315 (vst1_s16_x3): Likewise.
316 (vst1_p16_x3): Likewise.
317 (vst1_s32_x3): Likewise.
318 (vst1_u8_x3): Likewise.
319 (vst1_u16_x3): Likewise.
320 (vst1_u32_x3): Likewise.
321 (vst1_f16_x3): Likewise.
322 (vst1_f32_x3): Likewise.
323 (vst1_p64_x3): Likewise.
324 (vst1q_s8_x3): Likewise.
325 (vst1q_p8_x3): Likewise.
326 (vst1q_s16_x3): Likewise.
327 (vst1q_p16_x3): Likewise.
328 (vst1q_s32_x3): Likewise.
329 (vst1q_s64_x3): Likewise.
330 (vst1q_u8_x3): Likewise.
331 (vst1q_u16_x3): Likewise.
332 (vst1q_u32_x3): Likewise.
333 (vst1q_u64_x3): Likewise.
334 (vst1q_f16_x3): Likewise.
335 (vst1q_f32_x3): Likewise.
336 (vst1q_f64_x3): Likewise.
337 (vst1q_p64_x3): Likewise.
339 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
341 * config/msp430/msp430.c (msp430_output_labelref): Prepend
342 user_label_prefix to name.
344 * tree-core.h: Update comment about the format of NAME string
345 passed to handler in attribute_spec.
347 * config/msp430/msp430.md: Remove erroneous subreg expression from
348 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
349 zero_extend{q,h}isi2.
351 2018-05-30 Borislav Petkov <bp@suse.de>
353 * doc/extend.texi: Document some architecture specific
354 constraints and sort entries.
356 2018-05-30 Martin Sebor <msebor@redhat.com>
359 * builtins.c (expand_builtin_stpcpy_1): New function.
360 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
361 only if the former succeeds.
363 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
365 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
368 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
370 * doc/invoke.texi (-flinker-output): Document
372 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
374 * passes.c (ipa_write_summaries): Only modify statements if body
376 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
377 incrementally linking.
378 (ipa_passes): Likewise.
379 * lto-cgraph.c (lto_output_node): When incrementally linking do not
380 pass down resolution info.
381 * common.opt (flag_incremental_link): Update info.
382 * gcc.c (plugin specs): Turn flinker-output=* to
383 -plugin-opt=-linker-output-known
384 * toplev.c (compile_file): Also cut compilation when doing incremental
386 * flag-types. (enum lto_partition_model): Add
387 LTO_LINKER_OUTPUT_NOLTOREL.
388 (invoke.texi): Add -flinker-output docs.
389 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
390 link same way as WPA; do not stream in dead initializers.
392 * dwarf2out.c (dwarf2out_die_ref_for_decl,
393 darf2out_register_external_decl): Support incremental link.
395 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
397 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
399 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
401 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
402 it down to simple_object_copy_lto_debug_sections.
403 (run_gcc): Determine incremental LTO link time and configure
404 lto1 into non-wpa mode, disable renaming of debug sections.
406 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
408 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
409 descriptions of various incorrectly documented functions.
411 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
414 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
415 address check not strict.
417 2018-05-30 Richard Biener <rguenther@suse.de>
419 PR tree-optimization/85964
420 * tracer.c (better_p): Drop initialized count check, we only
421 call the function with initialized counts now.
422 (find_best_successor): Do find a best edge if one
423 has uninitialized count.
424 (find_best_predecessor): Likewise. Do BB frequency check only
425 if count is initialized.
427 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
429 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
430 (aarch64_ldrstr_offset_compare): New.
431 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
432 load/store orderings.
433 (aarch64_gen_adjusted_ldpstp): Likewise.
435 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
437 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
438 Check for subset of GENERAL_REGS and FP_REGS.
439 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
442 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
444 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
445 and wi::to_poly_offset. Add the current offset and then check
446 whether the sum fits, rather than using an unchecked addition of
447 a checked term. Check for a shwi rather than a uhwi.
448 * expr.c (get_bit_range): Use tree_to_poly_uint64.
449 (store_constructor): Use poly_int_tree_p.
450 (expand_expr_real_1): Likewise.
451 * function.c (assign_temp): Likewise.
452 * fold-const.c (const_binop): Use poly_int_tree_p and
454 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
456 * ipa-icf-gimple.c (func_checker::compare_operand): Use
457 to_poly_offset for MEM offsets.
458 * ipa-icf.c (sem_variable::equals): Likewise.
459 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
460 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
461 wi::to_poly_offset for BIT_FIELD_REF offsets.
462 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
464 * var-tracking.c (emit_note_insn_var_location): Use
467 2018-05-29 Jim Wilson <jimw@sifive.com>
469 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
471 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
474 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
475 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
477 (sse4_1_round<mode>2): Use nonimmediate_operand
478 for operand 1 predicate.
480 2018-05-29 Martin Sebor <msebor@redhat.com>
481 Richard Biener <rguenther@suse.de>
484 * calls.c (get_size_range): Call determine_value_range instead
486 * tree-vrp.h (determine_value_range): Declared new function.
487 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
489 2018-05-29 Richard Biener <rguenther@suse.de>
491 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
492 sure to use non-pattern stmts for get_earlier_stmt arguments.
493 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
494 called on pattern stmts.
495 (get_later_stmt): Likewise.
497 2018-05-29 Martin Liska <mliska@suse.cz>
499 PR gcov-profile/85759
500 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
503 2018-05-29 Jakub Jelinek <jakub@redhat.com>
505 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
507 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
511 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
512 VEC_PACK_FLOAT_EXPR): New tree codes.
513 * tree-pretty-print.c (op_code_prio): Handle
514 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
515 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
516 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
517 * tree-inline.c (estimate_operator_cost): Likewise.
518 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
519 * fold-const.c (const_binop): Likewise.
520 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
521 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
522 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
523 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
524 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
525 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
526 * expr.c (expand_expr_real_2): Likewise.
527 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
528 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
529 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
531 * optabs.c (expand_widen_pattern_expr): For
532 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
533 sign from result type rather than operand's type.
534 (expand_binop_directly): For vec_packu_float_optab and
535 vec_packs_float_optab allow result type to be different from operand's
537 * optabs-tree.c (optab_for_tree_code): Handle
538 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
539 VEC_PACK_FLOAT_EXPR. Formatting fixes.
540 * tree-vect-generic.c (expand_vector_operations_1): Handle
541 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
543 * tree-vect-stmts.c (supportable_widening_operation): Handle
545 (supportable_narrowing_operation): Handle FLOAT_EXPR.
546 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
547 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
548 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
549 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
551 (vec_pack<floatprefix>_float_<mode>): New expander.
552 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
554 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
555 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
556 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
557 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
558 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
560 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
561 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
562 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
563 VEC_PACK_FLOAT_EXPR): Document.
565 2018-05-29 Richard Biener <rguenther@suse.de>
567 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
569 (stmt_vec_info_vec): Make pointer.
570 (init_stmt_vec_info_vec): Remove.
571 (free_stmt_vec_info_vec): Likewise.
572 (set_stmt_vec_info_vec): New function.
573 (free_stmt_vec_infos): Likewise.
574 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
575 (set_vinfo_for_stmt): Likewise.
576 (get_earlier_stmt): Likewise.
577 (get_later_stmt): Likewise.
578 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
579 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
580 (vec_info::~vec_info): Free stmt_vec_infos.
581 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
582 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
583 (pass_slp_vectorize::execute): Likewise.
584 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
585 (free_stmt_vec_info_vec): Likewise.
586 (set_stmt_vec_info_vec): New function.
587 (free_stmt_vec_infos): Likewise.
588 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
589 the global stmt_vec_info_vec.
590 * tree-parloops.c (gather_scalar_reductions): Use
591 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
594 2018-05-29 Richard Biener <rguenther@suse.de>
596 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
598 2018-05-29 Martin Liska <mliska@suse.cz>
599 David Malcolm <dmalcolm@redhat.com>
601 * vec.c (test_reverse): New.
602 (vec_c_tests): Add new test.
603 * vec.h (vl_ptr>::reverse): New function.
605 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
607 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
609 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
612 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
614 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
616 2018-05-28 Richard Biener <rguenther@suse.de>
618 PR tree-optimization/85933
619 * tree-vect-data-refs.c (vect_record_base_alignments): Only
620 look at stmts marked as vectorizable.
622 2018-05-28 Richard Biener <rguenther@suse.de>
624 PR tree-optimization/85934
625 * tree-vect-generic.c (expand_vector_operations_1): Hoist
626 vector boolean check before scalar optimization.
628 2018-05-28 Jakub Jelinek <jakub@redhat.com>
630 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
633 2018-05-28 Mark Wielaard <mark@klomp.org>
635 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
636 if it is an expression containing a minus sign.
638 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
640 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
642 2018-05-27 Paul Koning <ni1d@arrl.net>
644 * config/pdp11/pdp11.md (truncsihi2): Remove.
646 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
647 Chung-Ju Wu <jasonwucj@gmail.com>
649 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
651 (unaligned_store_dw): Ditto.
652 * config/nds32/nds32-memory-manipulation.c
653 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
654 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
655 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
656 (emit_setmem_word_loop): Rename to ...
657 (emit_setmem_doubleword_loop): ... this.
658 (nds32_gen_dup_4_byte_to_word_value): New function.
659 (nds32_gen_dup_8_byte_to_double_word_value): New function.
660 (nds32_expand_setmem_loop): Refine implementation.
661 (nds32_expand_setmem_loop_v3m): Ditto.
662 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
665 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
667 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
669 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
671 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
672 (nds32_init_machine_status): Initialize machine->attr_naked_p and
673 machine->attr_no_prologue_p.
674 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
675 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
676 (nds32_expand_epilogue): Consider attr_naked_p.
677 (nds32_expand_epilogue_v3pop): Likewise.
678 (nds32_can_use_return_insn): Likewise.
679 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
680 attr_no_prologue_p fields.
681 * config/nds32/nds32.opt (mret-in-naked-func): New option.
683 2018-05-27 Jakub Jelinek <jakub@redhat.com>
686 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
689 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
691 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
693 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
695 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
697 (*<floatsuffix>floatv2div2sf2): Rename to ...
698 (*float<floatunssuffix>v2div2sf2): ... this.
699 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
700 (float<floatunssuffix>v2div2sf2_mask): ... this.
701 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
702 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
703 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
705 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
707 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
709 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
711 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
713 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
715 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
716 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
717 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
718 gen_ufix_truncv8dfv8si2.
719 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
720 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
721 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
722 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
723 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
724 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
725 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
726 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
728 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
732 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
734 2018-05-25 Jim Wilson <jimw@sifive.com>
736 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
737 * config/riscv/riscv.c (struct machine_function): Add
738 interrupt_handler_p and attribute_checked_p fields.
739 (riscv_attribute_table): Add interrupt.
740 (riscv_interrupt_type_p): New.
741 (riscv_save_reg_p): Save extra regs for interrupt handler.
742 (riscv_use_save_libcall): Return false for interrupt handler.
743 (riscv_first_stack_step): Add forward declaration.
744 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
745 for interrupt handler with large frame. Use it for saved reg list.
746 (riscv_expand_prologue): Move flag_stack_usage_info support to
747 eliminate duplication.
748 (riscv_expand_epilogue): Generate mret for interrupt handler.
749 (riscv_epilogue_uses): New.
750 (riscv_can_use_return_insn): Return false for interrupt handler.
751 (riscv_function_ok_for_sibcall): Likewise.
752 (riscv_set_current_function): Add interrupt handler support.
753 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
754 * config/riscv/riscv.md (UNSPECV_MRET): New.
756 (riscv_frflags, riscv_fsflags): Use tab after opcode.
758 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
760 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
762 PR tree-optimization/85712
763 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
764 this candidate has already been replaced in-situ by a copy.
766 2018-05-25 Jason Merrill <jason@redhat.com>
768 PR c++/80485 - inline function non-zero address.
769 * symtab.c (nonzero_address): Check DECL_COMDAT.
771 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
774 * config/alpha/alpha.md (ashlsi3): New insn pattern.
775 (*ashlsi_se): Rename from *ashldi_se. Define as sign
776 extension of SImode operation. Use const123_operand predicate.
778 (*saddl_se_1): Ditto.
780 (*ssubl_se_1): Ditto.
781 * config/alpha/predicates.md (const123_operand): New predicate.
782 * config/alpha/constraints.md (P): Use IN_RANGE.
784 2018-05-25 Richard Biener <rguenther@suse.de>
786 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
788 (ref_maybe_used_by_stmt_p): Likewise.
789 (stmt_may_clobber_ref_p): Likewise.
790 (stmt_may_clobber_ref_p_1): Likewise.
791 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
793 (ref_maybe_used_by_stmt_p): Likewise.
794 (stmt_may_clobber_ref_p): Likewise.
795 (stmt_may_clobber_ref_p_1): Likewise.
796 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
797 the alias oracle to disambiguate DRs with stmts DR analysis
799 (vect_analyze_data_refs): Do not give up on not analyzable
800 DRs for BB vectorization. Remove code truncating the dataref
803 2018-05-25 Jakub Jelinek <jakub@redhat.com>
806 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
807 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
808 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
810 2018-05-25 Richard Biener <rguenther@suse.de>
812 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
813 function, combining stmt data ref gathering and fatal analysis
815 (vect_analyze_data_refs): Remove now redudnant code and simplify.
816 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
817 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
818 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
819 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
821 2018-05-25 Bin Cheng <bin.cheng@arm.com>
823 PR tree-optimization/85720
824 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
825 SCC if all partitions are builtins.
826 (version_loop_by_alias_check): New parameter. Generate cancelable
827 runtime alias check if all partitions are builtins.
828 (distribute_loop): Update call to above function.
830 2018-05-25 Bin Cheng <bin.cheng@arm.com>
832 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
833 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
834 (parm_default_def_partition_arg): Ditto.
835 (set_parm_default_def_partition): Ditto.
836 (get_parm_default_def_partitions): Ditto and make it static.
837 (get_undefined_value_partitions): Ditto and make it static.
838 (remove_ssa_form): Refactor call to init_var_map here.
839 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
840 computation for loop region.
841 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
842 (register_default_def): Delete.
843 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
844 (parm_default_def_partition_arg): Ditto.
845 (set_parm_default_def_partition): Ditto.
846 (get_parm_default_def_partitions): Ditto and make it static.
847 (get_undefined_value_partitions): Ditto and make it static.
848 (coalesce_with_default, coalesce_with_default): Update comment.
849 (create_coalesce_list_for_region): New func factored out from
850 create_outofssa_var_map.
851 (populate_coalesce_list_for_outofssa): New func factored out from
852 create_outofssa_var_map and coalesce_ssa_name.
853 (create_outofssa_var_map): Delete.
854 (coalesce_ssa_name): Refactor to support live range computation.
855 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
856 (get_parm_default_def_partitions): Delete.
857 (get_undefined_value_partitions): Ditto.
858 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
859 computation for loop region.
860 (new_tree_live_info, loe_visit_block): Ditto.
861 (live_worklist, set_var_live_on_entry): Ditto.
862 (calculate_live_on_exit, verify_live_on_entry): Ditto.
863 * tree-ssa-live.h (struct _var_map): New fields.
864 (init_var_map): Change decl.
865 (region_contains_p): New.
867 2018-05-25 Bin Cheng <bin.cheng@arm.com>
869 * tree-ssa-live.h (live_merge_and_clear): Delete.
871 2018-05-25 Richard Biener <rguenther@suse.de>
874 * tree-dump.c (dequeue_and_dump): Remove access to removed
875 operand 2 of a SWITCH_EXPR.
877 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
879 * doc/sourcebuild.texi (vect_double_cond_arith): Include
880 multiplication and division.
881 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
882 (cond_udiv@var{m}, cond_umod@var{m}): Document.
883 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
884 (cond_udiv_optab, cond_umod_optab): New optabs.
885 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
886 (IFN_COND_RDIV): New internal functions.
887 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
888 TRUNC_MOD_EXPR and RDIV_EXPR.
889 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
890 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
892 (SVE_INT_BINARY): Include mult.
893 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
894 (optab, sve_int_op): Handle mult.
895 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
897 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
898 for SVE_INT_BINARY_SD.
900 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
902 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
903 (optab, sve_int_op): Handle div and udiv.
904 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
905 for SVE_INT_BINARY_SD.
906 (*<optab><mode>3): New insn for the same.
908 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
910 * tree-vect-patterns.c: Include predict.h.
911 (vect_recog_divmod_pattern): Restrict check for division support
912 to when optimizing for size.
914 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
916 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
917 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
918 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
919 (gimple_match_op::set_op): Likewise.
920 (gimple_resimplify4): Declare.
921 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
922 (expr::gen_transform): Likewise.
923 (decision_tree::gen): Generate a simplification routine for 4 operands.
924 * gimple-match-head.c (gimple_simplify): Add an overload for
925 4 operands. In the top-level function, handle up to 4 call
926 arguments and call gimple_resimplify4.
927 (gimple_resimplify4): New function.
928 (build_call_internal): Pass a fourth operand.
929 (maybe_push_to_seq): Likewise.
930 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
931 Fold VEC_COND_EXPRs of an operation and a default value into
932 an IFN_COND_* function if possible.
933 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
935 (SVE_COND_FP_BINARY): Include them.
936 (optab, sve_fp_op): Handle them.
937 (SVE_INT_BINARY_REV): New code iterator.
938 (SVE_COND_FP_BINARY_REV): New int iterator.
939 (commutative): New int attribute.
940 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
942 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
944 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
945 (*cond_<optab><mode>): New patterns for reversed operands.
947 2018-05-25 Richard Biener <rguenther@suse.de>
949 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
950 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
951 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
952 (STMT_VINFO_GROUPED_ACCESS): Adjust.
953 * tree-vect-data-refs.c (everywhere): Adjust users.
954 * tree-vect-loop.c (everywhere): Likewise.
955 * tree-vect-slp.c (everywhere): Likewise.
956 * tree-vect-stmts.c (everywhere): Likewise.
957 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
959 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
961 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
963 (gcc_cv_as_section_exclude): ... this.
964 Try Solaris as #exclude syntax.
965 * configure: Regenerate.
966 * config.in: Regenerate.
967 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
969 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
970 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
972 * varasm.c (default_elf_asm_named_section): Don't check if
973 HAVE_GAS_SECTION_EXCLUDE is defined.
975 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
977 * doc/md.texi: Update the documentation of the cond_* optabs
978 to mention the new final operand. Fix GET_MODE_NUNITS call.
979 Describe the scalar case too.
980 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
981 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
983 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
984 (get_conditional_internal_fn): Update comment.
985 * tree-vect-loop.c (vectorizable_reduction): Pass the original
986 accumulator value as a final argument to conditional functions.
987 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
988 a define_expand and add an "else" operand. Assert for now that
989 the else operand is equal to operand 2. Use SVE_INT_BINARY and
990 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
991 (*cond_<optab><mode>): New patterns.
992 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
993 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
994 (UNSPEC_COND_EOR): Delete.
995 (optab): Remove associated mappings.
996 (SVE_INT_BINARY): New code iterator.
997 (sve_int_op): Remove int attribute and add "minus" to the code
999 (SVE_COND_INT_OP): Delete.
1000 (SVE_COND_FP_OP): Rename to...
1001 (SVE_COND_FP_BINARY): ...this.
1003 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1005 * optabs.c (can_reuse_operands_p): New function.
1006 (maybe_legitimize_operands): Try to reuse the results for
1009 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
1011 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
1012 Add {q} suffix to insn mnemonic.
1014 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1016 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
1017 (msp430_warn_func_return): New.
1019 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
1021 * fold-const.c (tree_nonzero_bits): New function.
1022 * fold-const.h (tree_nonzero_bits): Likewise.
1023 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
1024 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
1026 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
1030 * varasm.c (assemble_alias): Check ifunc_resolver only on
1033 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
1036 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
1037 when memory input operand is handled.
1039 2018-05-24 Luis Machado <luis.machado@linaro.org>
1041 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
1043 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
1045 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
1047 * match.pd: Delay FMA folds until after vectorization.
1049 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
1052 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
1053 address check not strict.
1055 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
1057 * gimple-match.h (gimple_match_op): New class.
1058 (mprts_hook): Replace parameters with a gimple_match_op *.
1059 (maybe_build_generic_op): Likewise.
1060 (gimple_simplified_result_is_gimple_val): Replace parameters with
1061 a const gimple_match_op *.
1062 (gimple_simplify): Replace code_helper * and tree * parameters with
1063 a gimple_match_op * parameter.
1064 (gimple_resimplify1): Replace code_helper *, tree and tree *
1065 parameters with a gimple_match_op * parameter.
1066 (gimple_resimplify2): Likewise.
1067 (gimple_resimplify3): Likewise.
1068 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
1069 parameters with a gimple_match_op * parameter.
1070 * gimple-match-head.c (gimple_simplify): Change prototypes of
1071 auto-generated functions to take a gimple_match_op * instead of
1072 separate code_helper * and tree * parameters. Make the same
1073 change in the top-level overload and update calls to the
1074 gimple_resimplify routines. Update calls to the auto-generated
1075 functions and to maybe_push_res_to_seq in the publicly-facing
1076 operation-specific gimple_simplify overloads.
1077 (gimple_match_op::MAX_NUM_OPS): Define.
1078 (gimple_resimplify1): Replace rcode and ops with a single res_op
1079 parameter. Update call to gimple_simplify.
1080 (gimple_resimplify2): Likewise.
1081 (gimple_resimplify3): Likewise.
1082 (mprts_hook): Replace parameters with a gimple_match_op *.
1083 (maybe_build_generic_op): Likewise.
1084 (build_call_internal): Replace type, nargs and ops with
1085 a gimple_match_op *.
1086 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
1087 with a single gimple_match_op *. Update calls to mprts_hook,
1088 build_call_internal and gimple_simplified_result_is_gimple_val.
1089 Factor out code that is common to the tree_code and combined_fn cases.
1090 * genmatch.c (expr::gen_transform): Replace tem_code and
1091 tem_ops with a gimple_match_op called tem_op. Update calls
1092 to the gimple_resimplify functions and maybe_push_res_to_seq.
1093 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
1094 res_ops. Update call to the gimple_resimplify functions.
1095 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
1096 (decision_tree::gen): Make the functions take a gimple_match_op *
1097 called res_op instead of separate res_code and res_ops parameters.
1098 Update call accordingly.
1099 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
1100 and ops with a single res_op parameter. Update calls to
1101 maybe_build_generic_op and maybe_push_res_to_seq.
1102 (fold_stmt_1): Update calls to gimple_simplify and
1103 replace_stmt_with_simplification.
1104 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
1105 and gimple_simplified_result_is_gimple_val.
1106 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
1108 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
1109 with a gimple_match_op *.
1110 (vn_nary_build_or_lookup): Likewise. Update call to
1111 vn_nary_build_or_lookup_1.
1112 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
1113 gimple_match_op *. Update calls to the gimple_resimplify routines
1114 and to gimple_simplified_result_is_gimple_val.
1115 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
1116 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
1117 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
1118 (visit_nary_op): Likewise.
1119 (visit_reference_op_load): Likewise.
1121 2018-05-23 Luis Machado <luis.machado@linaro.org>
1123 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
1124 modifier for printing the step amount.
1126 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
1129 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
1132 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
1134 * doc/sourcebuild.texi (Endianness): New subsubsection.
1136 2018-05-23 Luis Machado <luis.machado@linaro.org>
1138 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1139 <prefetch_dynamic_strides>: New const bool field.
1140 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1141 prefetch_dynamic_strides.
1142 (exynosm1_prefetch_tune): Likewise.
1143 (thunderxt88_prefetch_tune): Likewise.
1144 (thunderx_prefetch_tune): Likewise.
1145 (thunderx2t99_prefetch_tune): Likewise.
1146 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
1148 (aarch64_override_options_internal): Update to set
1149 PARAM_PREFETCH_DYNAMIC_STRIDES.
1150 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1151 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1152 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1153 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1154 prefetch-dynamic-strides setting.
1156 2018-05-23 Luis Machado <luis.machado@linaro.org>
1158 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1159 <minimum_stride>: New const int field.
1160 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1161 minimum_stride field defaulting to -1.
1162 (exynosm1_prefetch_tune): Likewise.
1163 (thunderxt88_prefetch_tune): Likewise.
1164 (thunderx_prefetch_tune): Likewise.
1165 (thunderx2t99_prefetch_tune): Likewise.
1166 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
1167 <default_opt_level>: Set to 3.
1168 (aarch64_override_options_internal): Update to set
1169 PARAM_PREFETCH_MINIMUM_STRIDE.
1170 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1171 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1172 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1173 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1174 stride is constant and is below the minimum stride threshold.
1176 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1178 * config/arm/arm-cpus.in (mode26): Delete.
1179 (armv4): Delete mode26 reference.
1180 * config/arm/arm.c (arm_configure_build_target): Delete use of
1183 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
1185 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
1187 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
1188 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
1190 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
1191 (floatunsdidf2): Ditto.
1193 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
1195 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
1196 (fixuns_trunc<mode>si2_avx512f): Ditto.
1197 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
1198 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
1199 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
1201 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
1203 PR rtl-optimization/79985
1204 * df-scan.c (df_insn_refs_collect): Remove special case for
1205 global registers and asm statements.
1207 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
1209 * extend.texi (Global Register Variables): Rewrite the bullet list.
1210 Note that the register is available for allocation. Note that access
1211 via inline asm must use constraints. Add note about async-signal
1212 handlers. Remove paragraph about automagic register selection.
1214 2018-05-23 Richard Biener <rguenther@suse.de>
1216 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
1217 of fixed offset from memset VN.
1219 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
1221 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
1223 (alloc_cand_and_find_basis): Initialize first_interp field.
1224 (slsr_process_mul): Modify first_interp field.
1225 (slsr_process_add): Likewise.
1226 (slsr_process_cast): Modify first_interp field for each new
1228 (slsr_process_copy): Likewise.
1229 (dump_candidate): Dump first_interp field.
1230 (replace_mult_candidate): Process all interpretations, not just
1232 (replace_rhs_if_not_dup): Likewise.
1233 (replace_one_candidate): Likewise.
1235 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
1237 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
1239 (aarch64_needs_frame_chain): New function.
1240 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
1242 2018-05-23 Sudakshina Das <sudi.das@arm.com>
1245 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
1246 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
1247 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
1248 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
1249 as true for strict-align.
1250 (aarch64_can_inline_p): Perform checks even when callee has no
1251 attributes to check for strict alignment.
1252 * doc/extend.texi (AArch64 Function Attributes): Document
1254 * doc/invoke.texi: (AArch64 Options): Likewise.
1256 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
1258 PR tree-optimization/85853
1259 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
1260 the handling of the root of the node to...
1261 (vect_slp_analyze_node_operations_1): ...this new function,
1262 and run the whole thing with the child nodes' def types
1263 set according to their SLP node's def type.
1265 2018-05-23 Richard Biener <rguenther@suse.de>
1268 * tree-data-ref.c (create_runtime_alias_checks): Defer
1269 and ignore overflow warnings.
1271 2018-05-23 Yury Gribov <tetra2005@gmail.com>
1273 PR tree-optimization/85822
1274 * tree-vrp.c (is_masked_range_test): Fix handling of negative
1277 2018-05-23 Richard Biener <rguenther@suse.de>
1279 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
1280 memset constants via native_interpret_expr.
1282 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
1285 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
1287 (cgraph_node::create_alias): Likewise.
1288 (cgraph_node::get_availability): Check ifunc_resolver instead
1289 of looking up ifunc attribute.
1290 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
1291 * varasm.c (do_assemble_alias): Likewise.
1292 (assemble_alias): Likewise.
1293 (default_binds_local_p_3): Likewise.
1294 * cgraph.h (cgraph_node): Add ifunc_resolver.
1295 (cgraph_node::only_called_directly_or_aliased_p): Return false
1297 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
1299 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
1300 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
1301 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
1302 instead of looking up ifunc attribute.
1304 2018-05-22 Luis Machado <luis.machado@linaro.org>
1306 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
1308 2018-05-22 Martin Sebor <msebor@redhat.com>
1311 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
1312 only when expasion succeeds.
1313 (expand_builtin_strcmp): Same.
1314 (expand_builtin_strncmp): Same.
1316 2018-05-22 Martin Sebor <msebor@redhat.com>
1318 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
1320 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
1321 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1323 * config/aarch64/aarch64-ldpstp.md: Replace uses of
1324 aarch64_mem_pair_operand with memory_operand and delete operand swapping
1326 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
1327 Add check for legitimate_address.
1328 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
1329 (aarch64_swap_ldrstr_operands): New.
1330 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
1333 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
1334 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1336 * config/aarch64/aarch64.md: New patterns to generate stp
1338 (store_pair_sw, store_pair_dw): New patterns to generate stp for
1339 single words and double words.
1340 (load_pair_sw, load_pair_dw): Likewise.
1341 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
1343 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
1345 * config/aarch64/aarch64-ldpstp.md: Modify peephole
1346 for different mode ldpstp and add peephole for merged zero stores.
1348 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
1350 (aarch64_gen_store_pair): Rename calls to match new patterns.
1351 (aarch64_gen_load_pair): Rename calls to match new patterns.
1352 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
1353 (load_pair<DREG:mode><DREG2:mode>): ... This.
1354 (store_pair<mode>): Rename to...
1355 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
1356 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
1358 (V_INT_EQUIV): Handle SImode.
1359 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
1362 2018-05-22 Martin Sebor <msebor@redhat.com>
1365 * calls.c (maybe_warn_nonstring_arg): Use string length to set
1366 or ajust the presumed bound on an operation to avoid unnecessary
1369 2018-05-22 Martin Sebor <msebor@redhat.com>
1371 PR tree-optimization/85826
1372 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
1373 assuming that a DECL necesarily has a constant size.
1375 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
1378 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
1380 2018-05-22 Richard Biener <rguenther@suse.de>
1382 PR tree-optimization/85834
1383 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
1384 non-constant and non-zero memset arguments.
1386 2018-05-22 Martin Liska <mliska@suse.cz>
1389 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
1391 2018-05-22 Richard Biener <rguenther@suse.de>
1393 PR tree-optimization/85863
1394 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
1395 comparisons when vectype is specified.
1396 (vectorizable_condition): Do not specify vectype for
1397 vect_is_simple_cond when SLP vectorizing.
1399 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
1402 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
1403 define __ibm128 as long double.
1404 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
1405 as a distinct type when IEEE 128-bit support is enabled.
1406 (init_float128_ieee): Fix up conversions between IFmode and IEEE
1407 128-bit types to use the correct functions.
1408 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
1409 convert between 128-bit floating point types that have different
1410 modes but the same representation, instead of using gen_lowpart to
1412 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
1414 (IFKF_reg): New attributes to give the register constraints for
1416 (extend<mode>tf2_internal): New insns to mark an explicit
1417 conversion between 128-bit floating point types that have a
1418 different mode but share the same representation.
1420 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
1422 PR tree-optimization/85814
1423 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
1424 a null return from get_strinfo when unsharing the next
1425 strinfo in the chain.
1427 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
1430 * varasm.c (weak_finish): Clean up weak_decls.
1432 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1434 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
1435 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
1436 UNSPEC_UADALP values.
1437 * config/aarch64/iterators.md (ABAL): New int iterator.
1440 (sur): Add mappings for the above.
1441 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
1443 (aarch64_<sur>abal<mode>_4): Likewise.
1444 (aarch64_<sur>adalp<mode>_3): Likewise.
1445 (<sur>sadv16qi): New define_expand.
1447 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
1449 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
1450 (*movdf_internal): Ditto.
1451 (*rcpsf2_sse): Ditto.
1452 (*rsqrtsf2_sse): Ditto.
1453 (*sqrt<mode>2_sse): Ditto.
1455 2018-05-21 Tamar Christina <tamar.christina@arm.com>
1457 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
1459 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
1460 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
1461 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
1462 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
1464 * config/aarch64/arm_neon.h: Likewise.
1465 * config/aarch64/iterators.md (VQ_I): New.
1467 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
1469 * config.gcc: Add arc/t-multilib-linux to tmake_file for
1471 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
1474 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
1476 * config/nds32/constraints.md (S): New constraint.
1477 * config/nds32/nds32.md (call_internal): Use constraint S.
1478 (call_value_internal): Likewise.
1479 (sibcall_internal): Likewise.
1480 (sibcall_value_internal): Likewise.
1482 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
1483 Chung-Ju Wu <jasonwucj@gmail.com>
1485 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
1488 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
1489 Chung-Ju Wu <jasonwucj@gmail.com>
1491 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
1492 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
1493 (nds32_rtx_costs_impl): Simplify.
1494 (nds32_address_cost_impl): Simplify.
1495 (nds32_init_rtx_costs): New function.
1496 (nds32_rtx_costs_speed_prefer): Likewise.
1497 (nds32_rtx_costs_size_prefer): Likewise.
1498 (nds32_address_cost_speed_prefer): Likewise.
1499 (nds32_address_cost_speed_fwprop): Likewise.
1500 (nds32_address_cost_size_prefer): Likewise.
1501 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
1502 * config/nds32/nds32.c (nds32_option_override): Use
1503 nds32_init_rtx_costs function.
1505 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
1507 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
1508 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
1509 (TARGET_PIPELINE_N8): Likewise.
1510 (TARGET_PIPELINE_N10): Likewise.
1511 (TARGET_PIPELINE_N13): Likewise.
1512 (TARGET_PIPELINE_GRAYWOLF): Likewise.
1514 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
1516 * config/nds32/nds32-fpu.md: Update copyright year.
1518 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1520 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
1522 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1524 * config/nds32/nds32.c
1525 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
1526 * config/nds32/nds32.opt (minline-asm-r15): New option.
1528 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1530 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
1532 * config/nds32/nds32.md (abssi2): New pattern.
1534 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
1536 * config/i386/i386.md (rex64namesuffix): New mode attribute.
1537 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
1538 Merge insn pattern from sse_cvtsi2ss<round_name> and
1539 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
1540 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
1541 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
1542 using SWI48 mode iterator.
1543 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
1544 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
1545 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
1546 pattern from sse_cvttss2si<round_saeonly_name>
1547 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
1548 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
1549 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
1550 using SWI48 mode iterator.
1551 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
1552 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
1553 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
1554 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
1555 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
1556 using SWI48 mode iterator.
1557 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
1558 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
1559 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
1560 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
1561 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
1562 SWI48 mode iterator.
1563 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
1564 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
1565 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
1566 pattern from sse_cvttsd2si<round_saeonly_name>
1567 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
1569 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1571 * config/nds32/nds32-md-auxiliary.c
1572 (nds32_valid_smw_lwm_base_p): Refine.
1573 (nds32_output_smw_single_word): Refine.
1574 (nds32_output_smw_double_word): New.
1575 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
1577 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1579 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
1580 (nds32_output_stack_pop): Refine.
1581 (nds32_expand_unaligned_load): Refine.
1582 (nds32_expand_unaligned_store): Refine.
1584 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1585 Chung-Ju Wu <jasonwucj@gmail.com>
1587 * config/nds32/constants.md: Add TP_REGNUM constant.
1588 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
1589 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
1591 * config/nds32/nds32-doubleword.md: Consider flag_pic.
1592 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
1593 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
1594 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
1595 and PIC code generation.
1596 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
1598 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
1600 * config/nds32/nds32.md: Support TLS and PIC.
1601 * config/nds32/nds32.c: Support TLS and PIC.
1602 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
1603 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
1606 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1608 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
1609 mode with E_ prefix.
1611 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1612 Chung-Ju Wu <jasonwucj@gmail.com>
1614 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
1615 * config/nds32/nds32-md-auxiliary.c
1616 (symbolic_reference_mentioned_p): New.
1617 (nds32_legitimize_ict_address): New.
1618 (nds32_expand_ict_move): New.
1619 (nds32_indirect_call_referenced_p): New.
1620 (nds32_symbol_binds_local_p): Delete.
1621 (nds32_long_call_p): Modify.
1622 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
1623 * config/nds32/nds32-protos.h
1624 (symbolic_reference_mentioned_p): Declare.
1625 (nds32_legitimize_ict_address): Declare.
1626 (nds32_expand_ict_move): Declare.
1627 (nds32_indirect_call_referenced_p): Declare.
1628 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
1629 (nds32_relax_group): Use nds32_ict_const_p as condition.
1630 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
1631 (nds32_asm_file_start): Output ict_model directive in asm code.
1632 (nds32_legitimate_address_p): Consider indirect call.
1633 (nds32_print_operand): Consider indirect call.
1634 (nds32_print_operand_address): Consider indirect call.
1635 (nds32_insert_attributes): Handle "indirect_call" attribute.
1636 (TARGET_LEGITIMATE_ADDRESS_P): Define.
1637 (TARGET_LEGITIMATE_CONSTANT_P): Define.
1638 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
1639 (TARGET_DELEGITIMIZE_ADDRESS): Define.
1640 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
1641 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
1642 (TARGET_ICT_MODEL_SMALL): Define.
1643 (TARGET_ICT_MODEL_LARGE): Define.
1644 * config/nds32/nds32.md (movsi): Consider ict model.
1645 (call, call_value): Consider ict model.
1646 (sibcall, sibcall_value): Consider ict model.
1647 * config/nds32/nds32.opt (mict-model): New option.
1648 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
1651 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
1652 Monk Chiang <sh.chiang04@gmail.com>
1653 Jim Wilson <jimw@sifive.com>
1655 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
1656 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
1657 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
1658 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
1659 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
1660 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
1661 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
1662 compute save_libcall_adjustment properly.
1663 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
1664 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
1665 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
1666 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
1667 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
1668 (ABI_SPEC): Handle mabi=ilp32e.
1669 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
1670 (RVE): Add RVE mask.
1671 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
1672 <-march>: Add rv32e as an example.
1674 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
1677 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
1678 (intra_create_variable_infos): Handle C++ constructors.
1680 2018-05-18 Martin Liska <mliska@suse.cz>
1682 * passes.def: Remove a redundant pass.
1684 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
1687 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
1689 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1691 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
1693 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
1695 (armv2, armv2a, armv3, armv3m): Delete architectures.
1696 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
1697 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
1698 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
1700 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
1701 (*mulsidi3adddi): Likewise.
1702 (mulsidi3): Likewise.
1703 (*mulsidi3_nov6): Likewise.
1704 (umulsidi3): Likewise.
1705 (umulsidi3_nov6): Likewise.
1706 (umaddsidi4): Likewise.
1707 (*umulsidi3adddi): Likewise.
1708 (smulsi3_highpart): Likewise.
1709 (*smulsi3_highpart_nov6): Likewise.
1710 (umulsi3_highpart): Likewise.
1711 (*umulsi3_highpart_nov6): Likewise.
1712 * config/arm/arm.h (arm_arch3m): Delete.
1713 * config/arm/arm.c (arm_arch3m): Delete.
1714 (arm_option_override_internal): Update armv3-related comment.
1715 (arm_configure_build_target): Delete use of isa_bit_mode32.
1716 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
1717 (arm_rtx_costs_internal): Delete check of arm_arch3m.
1718 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
1720 (mulusa3): Likewise.
1721 * config/arm/arm-protos.h (arm_arch3m): Delete.
1722 * config/arm/arm-tables.opt: Regenerate.
1723 * config/arm/arm-tune.md: Likewise.
1724 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
1725 deleted architectures.
1727 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1729 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
1730 (armv5t, armv5te): New features.
1731 (ARMv5, ARMv5e): Delete fgroups.
1732 (ARMv5t, ARMv5te): Adjust for above changes.
1734 (armv5, armv5e): Delete arches.
1735 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
1737 (*call_reg_arm): Likewise.
1738 (*call_value_reg_armv5): Likewise.
1739 (*call_value_reg_arm): Likewise.
1740 (*call_symbol): Likewise.
1741 (*call_value_symbol): Likewise.
1742 (*sibcall_insn): Likewise.
1743 (*sibcall_value_insn): Likewise.
1745 (prefetch): Likewise.
1746 (define_split and define_peephole2 dependent on arm_arch5):
1748 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
1750 (TARGET_ARM_QBIT): Likewise.
1751 (TARGET_DSP_MULTIPLY): Likewise.
1752 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
1753 (arm_arch5, arm_arch5e): Delete.
1754 (arm_arch5t, arm_arch5te): Declare.
1755 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
1756 (arm_arch5t): Declare.
1757 (arm_option_reconfigure_globals): Update for the above.
1758 (arm_options_perform_arch_sanity_checks): Update comment, replace
1759 use of arm_arch5 with arm_arch5t.
1760 (use_return_insn): Likewise.
1761 (arm_emit_call_insn): Likewise.
1762 (output_return_instruction): Likewise.
1763 (arm_final_prescan_insn): Likewise.
1764 (arm_coproc_builtin_available): Likewise.
1765 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
1766 arm_arch5e with arm_arch5t and arm_arch5te.
1767 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
1768 (arm_arch5t, arm_arch5te): Declare.
1769 * config/arm/arm-tables.opt: Regenerate.
1770 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
1771 * config/arm/t-multilib: Likewise.
1772 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
1773 instead of arm_arch5.
1774 (*call_reg_thumb1): Likewise.
1775 (*call_value_reg_thumb1_v5): Likewise.
1776 (*call_value_reg_thumb1): Likewise.
1777 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
1779 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
1781 2018-05-18 Martin Liska <mliska@suse.cz>
1783 PR gcov-profile/84846
1784 * doc/gcov.texi: Document -t option of gcov tool.
1786 2018-05-18 Martin Liska <mliska@suse.cz>
1788 PR gcov-profile/84846
1789 * gcov.c (print_usage): Add new -t option.
1790 (process_args): Handle the option.
1791 (generate_results): Use stdout as output when requested by
1794 2018-05-18 Martin Liska <mliska@suse.cz>
1796 PR gcov-profile/84846
1797 * coverage.c (coverage_init): Write PWD to .gcno file.
1798 * doc/gcov.texi: Document how working directory is printed.
1799 * gcov-dump.c (dump_gcov_file): Print PWD.
1800 * gcov.c (output_intermediate_file): Likewise.
1801 (read_graph_file): Read PWD string.
1802 (output_lines): Print PWD.
1804 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1807 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
1808 for retval and return false if all args to phi are zero.
1810 2018-05-18 Richard Biener <rguenther@suse.de>
1812 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
1814 (evrp_dom_walker::before_dom_children): Call it.
1816 2018-05-18 Richard Biener <rguenther@suse.de>
1818 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
1819 results when processing array refs with variable index.
1821 2018-05-18 Toon Moene <toon@moene.org>
1823 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
1824 directly after that of -floop-interchange. Indicate that both
1825 options are enabled by default when specifying -O3.
1827 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1829 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
1830 iterator. Delete separate integer-mode vec_set<mode> expander.
1831 (aarch64_simd_vec_setv2di): Delete.
1832 (vec_setv2di): Delete.
1833 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
1834 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
1835 the "w, r" alternative.
1837 2018-05-18 Martin Liska <mliska@suse.cz>
1839 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
1840 * tree-pass.h (make_pass_lower_switch_O0): New function.
1841 * tree-switch-conversion.c (node_has_low_bound): Remove.
1842 (node_has_high_bound): Likewise.
1843 (node_is_bounded): Likewise.
1844 (class pass_lower_switch): Make it a template type and create
1846 (pass_lower_switch::execute): Add template argument.
1847 (make_pass_lower_switch): New function.
1848 (make_pass_lower_switch_O0): New function.
1849 (do_jump_if_equal): Remove.
1850 (emit_case_nodes): Simplify to just handle all 3 cases and leave
1851 all the hard work to tree optimization passes.
1853 2018-05-18 Martin Liska <mliska@suse.cz>
1855 * dbgcnt.c (limit_low): Renamed from limit.
1856 (limit_high): New variable.
1857 (dbg_cnt_is_enabled): Check for upper limit.
1858 (dbg_cnt): Adjust dumping.
1859 (dbg_cnt_set_limit_by_index): Add new argument for high
1861 (dbg_cnt_set_limit_by_name): Likewise.
1862 (dbg_cnt_process_single_pair): Parse new format.
1863 (dbg_cnt_process_opt): Use strtok.
1864 (dbg_cnt_list_all_counters): Remove 'value' and add
1866 * doc/invoke.texi: Document changes.
1868 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
1870 * doc/sourcebuild.texi (scalar_all_fma): Document.
1871 * tree.def (FMA_EXPR): Delete.
1872 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
1873 * internal-fn.c (ternary_direct): New macro.
1874 (expand_ternary_optab_fn): Likewise.
1875 (direct_ternary_optab_supported_p): Likewise.
1876 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
1877 * builtins.c (fold_builtin_fma): Delete.
1878 (fold_builtin_3): Don't call it.
1879 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
1880 * expr.c (expand_expr_real_2): Likewise.
1881 * fold-const.c (operand_equal_p): Likewise.
1882 (fold_ternary_loc): Likewise.
1883 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
1884 * gimple.c (DEFTREECODE): Likewise.
1885 * gimplify.c (gimplify_expr): Likewise.
1886 * optabs-tree.c (optab_for_tree_code): Likewise.
1887 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
1888 * tree-eh.c (operation_could_trap_p): Likewise.
1889 (stmt_could_throw_1_p): Likewise.
1890 * tree-inline.c (estimate_operator_cost): Likewise.
1891 * tree-pretty-print.c (dump_generic_node): Likewise.
1892 (op_code_prio): Likewise.
1893 * tree-ssa-loop-im.c (stmt_cost): Likewise.
1894 * tree-ssa-operands.c (get_expr_operands): Likewise.
1895 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
1896 * fold-const-call.h (fold_fma): Delete.
1897 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
1898 CFN_FNMA and CFN_FNMS.
1900 * genmatch.c (combined_fn): New enum.
1901 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
1902 (commutative_op): New function.
1903 (commutate): Use it. Handle more than 2 operands.
1904 (dt_operand::gen_gimple_expr): Use commutative_op.
1905 (parser::parse_expr): Allow :c to be used with non-binary
1906 operators if the commutative operand is known.
1907 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
1908 CFN_FMS, CFN_FNMA and CFN_FNMS.
1909 (backprop::process_assign_use): Remove FMA_EXPR handling.
1910 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
1911 (gen_hsa_fma): New function.
1912 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
1913 IFN_FNMA and IFN_FNMS.
1914 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
1915 * gimple-fold.h (follow_all_ssa_edges): Declare.
1916 * gimple-fold.c (follow_all_ssa_edges): New function.
1917 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
1918 gimple_build interface and use follow_all_ssa_edges to fold the result.
1919 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
1920 instead of checking for optabs directly.
1921 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
1922 rather than FMA_EXPRs.
1923 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
1924 call to IFN_FMA instead of an FMA_EXPR.
1926 2018-05-17 Jim Wilson <jimw@sifive.com>
1928 * expr.c (do_tablejump): When converting index to Pmode, if we have a
1929 sign extended promoted subreg, and the range does not have the sign bit
1930 set, then do a sign extend.
1932 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
1933 test, check for sign extended subreg and/or constant operands, and
1934 do a sign extend in that case.
1936 2018-05-17 Steve Ellcey <sellcey@cavium.com>
1938 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
1939 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
1941 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
1942 Change logics_shift_reg to logics_shift_imm.
1943 (thunderx2t99_fp_loadpair_basic): Delete.
1944 (thunderx2t99_fp_storepair_basic): Delete.
1945 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
1946 (thunderx2t99_asimd_polynomial): Delete.
1947 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
1948 and neon_fp_mul_d_scalar_q.
1949 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
1950 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
1951 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
1952 (thunderx2t99_asimd_lut): Add missing tbl types.
1953 (thunderx2t99_asimd_ext): Delete.
1954 (thunderx2t99_asimd_load1_1_mult): Delete.
1955 (thunderx2t99_asimd_load1_2_mult): Delete.
1956 (thunderx2t99_asimd_load1_ldp): New.
1957 (thunderx2t99_asimd_load1): New.
1958 (thunderx2t99_asimd_load2): Add missing *load2* types.
1959 (thunderx2t99_asimd_load3): New.
1960 (thunderx2t99_asimd_load4): New.
1961 (thunderx2t99_asimd_store1_1_mult): Delete.
1962 (thunderx2t99_asimd_store1_2_mult): Delete.
1963 (thunderx2t99_asimd_store2_mult): Delete.
1964 (thunderx2t99_asimd_store2_onelane): Delete.
1965 (thunderx2t99_asimd_store_stp): New.
1966 (thunderx2t99_asimd_store1): New.
1967 (thunderx2t99_asimd_store2): New.
1968 (thunderx2t99_asimd_store3): New.
1969 (thunderx2t99_asimd_store4): New.
1971 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
1973 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
1974 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
1976 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
1977 Segher Boessenkool <segher@kernel.crashing.org>
1980 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
1983 2018-05-17 Richard Biener <rguenther@suse.de>
1985 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
1986 for pruning loop and prune defs feeding only already visited PHIs.
1988 2018-05-17 Richard Biener <rguenther@suse.de>
1990 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
1992 2018-05-17 Bin Cheng <bin.cheng@arm.com>
1993 Richard Biener <rguenther@suse.de>
1995 PR tree-optimization/85793
1996 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
1997 for VMAT_ELEMENTWISE.
1999 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
2001 * internal-fn.h (lookup_internal_fn): Declare
2002 * internal-fn.c (lookup_internal_fn): New function.
2003 * gimple.c (gimple_build_call_from_tree): Handle calls to
2005 * gimple-pretty-print.c (dump_gimple_call): Print "." before
2006 internal function names.
2007 * tree-pretty-print.c (dump_generic_node): Likewise.
2008 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
2010 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
2012 * gimple-fold.h (gimple_build): Make the function forms take
2013 combined_fn rather than built_in_function.
2014 (gimple_simplify): Likewise.
2015 * gimple-match-head.c (gimple_simplify): Likewise.
2016 * gimple-fold.c (gimple_build): Likewise.
2017 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
2018 rather than gimple_build_call_internal.
2019 (get_initial_defs_for_reduction): Likewise.
2020 (vect_create_epilog_for_reduction): Likewise.
2021 (vectorizable_live_operation): Likewise.
2023 2018-05-17 Martin Liska <mliska@suse.cz>
2025 * gimple-ssa-sprintf.c (format_directive): Do not use
2026 space in between 'G_' and '('.
2028 2018-05-17 Jakub Jelinek <jakub@redhat.com>
2031 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
2032 even if the mask is not all ones.
2035 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
2037 (ix86_gimple_fold_builtin): Likewise.
2040 * config/i386/i386.c: Include tree-vector-builder.h.
2041 (ix86_vector_shift_count): New function.
2042 (ix86_fold_builtin): Fold shift builtins by scalar count.
2043 (ix86_gimple_fold_builtin): Likewise.
2045 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
2046 _mm512_setzero): New intrinsics.
2048 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
2049 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2051 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
2052 code generation for cases where splatting a value is not useful.
2053 * simplify-rtx.c (simplify_ternary_operation): Simplify
2054 vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
2055 mode to a vec_concat.
2057 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
2059 * config.gcc: Support "goldmont-plus".
2060 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
2062 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2063 PROCESSOR_GOLDMONT_PLUS.
2064 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
2065 (processor_target_table): Add "goldmont-plus".
2066 (PTA_GOLDMONT_PLUS): Define.
2067 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
2068 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
2069 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
2070 (fold_builtin_cpu): Add "goldmont-plus".
2071 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
2072 (ix86_option_override_internal): Add "goldmont-plus".
2073 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
2074 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
2075 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
2076 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
2078 2018-05-17 Richard Biener <rguenther@suse.de>
2080 PR tree-optimization/85757
2081 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
2082 remove defs that only feed that PHI from further processing.
2084 2018-05-16 Jim Wilson <jimw@sifive.com>
2086 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
2088 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
2090 2018-05-16 Mark Wielaard <mark@klomp.org>
2092 * dwarf2out.c (count_index_strings): New function.
2093 (output_indirect_strings): Call count_index_strings and generate
2094 header for dwarf_version >= 5.
2096 2018-05-16 Mark Wielaard <mark@klomp.org>
2098 * dwarf2out.c (dwarf_FORM): New function.
2099 (set_indirect_string): Use dwarf_FORM.
2100 (reset_indirect_string): Likewise.
2101 (size_of_die): Likewise.
2102 (value_format): Likewise.
2103 (output_die): Likewise.
2104 (add_skeleton_AT_string): Likewise.
2105 (output_macinfo_op): Likewise.
2106 (index_string): Likewise.
2107 (output_index_string_offset): Likewise.
2108 (output_index_string): Likewise.
2109 (count_index_strings): Likewise.
2111 2018-05-16 Carl Love <cel@us.ibm.com>
2113 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
2114 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
2116 2018-05-16 Martin Jambor <mjambor@suse.cz>
2118 * ipa-prop.c (ipa_free_all_edge_args): Remove.
2119 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
2121 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
2123 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
2124 (fnma<mode>4): Likewise.
2125 (fms<mode>4): Likewise.
2126 (fnms<mode>4): Likewise.
2127 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
2128 (aarch64_fnma<mode>4): Likewise.
2129 (aarch64_fms<mode>4): Likewise.
2130 (aarch64_fnms<mode>4): Likewise.
2131 (aarch64_fnmadd<mode>4): Likewise.
2133 2018-05-16 Jason Merrill <jason@redhat.com>
2135 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
2137 2018-05-16 Richard Biener <rguenther@suse.de>
2139 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
2140 (dump_stmt_cost): Declare.
2141 (add_stmt_cost): Dump cost we add.
2142 (add_stmt_costs): New function.
2143 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
2145 (vect_analyze_stmt): Adjust prototype.
2146 (vectorizable_condition): Likewise.
2147 (vectorizable_live_operation): Likewise.
2148 (vectorizable_reduction): Likewise.
2149 (vectorizable_induction): Likewise.
2150 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
2151 cost vector to pass to vectorizable_ and record afterwards.
2152 (vect_model_reduction_cost): Take cost vector argument and adjust.
2153 (vect_model_induction_cost): Likewise.
2154 (vectorizable_reduction): Likewise.
2155 (vectorizable_induction): Likewise.
2156 (vectorizable_live_operation): Likewise.
2157 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
2158 SLP_TREE_NUMBER_OF_VEC_STMTS.
2159 (vect_analyze_slp_cost_1): Remove.
2160 (vect_analyze_slp_cost): Likewise.
2161 (vect_slp_analyze_node_operations): Take visited args and
2162 a target cost vector. Avoid processing already visited stmt sets.
2163 (vect_slp_analyze_operations): Use a local cost vector to gather
2164 costs and register those of non-discarded instances.
2165 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
2166 (vect_schedule_slp_instance): Remove copying of
2167 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
2169 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
2170 adding cost. Record cost entry location.
2171 (vect_prologue_cost_for_slp_op): Function to compute cost of
2172 a constant or invariant generated for SLP vect in the prologue,
2173 split out from vect_analyze_slp_cost_1.
2174 (vect_model_simple_cost): Make static. Adjust for SLP costing.
2175 (vect_model_promotion_demotion_cost): Likewise.
2176 (vect_model_store_cost): Likewise, make static.
2177 (vect_model_load_cost): Likewise.
2178 (vectorizable_bswap): Add cost vector arg and adjust.
2179 (vectorizable_call): Likewise.
2180 (vectorizable_simd_clone_call): Likewise.
2181 (vectorizable_conversion): Likewise.
2182 (vectorizable_assignment): Likewise.
2183 (vectorizable_shift): Likewise.
2184 (vectorizable_operation): Likewise.
2185 (vectorizable_store): Likewise.
2186 (vectorizable_load): Likewise.
2187 (vectorizable_condition): Likewise.
2188 (vectorizable_comparison): Likewise.
2189 (can_vectorize_live_stmts): Likewise.
2190 (vect_analyze_stmt): Likewise.
2191 (vect_transform_stmt): Adjust calls to vectorizable_*.
2192 * tree-vectorizer.c: Include gimple-pretty-print.h.
2193 (dump_stmt_cost): New function.
2195 2018-05-16 Richard Biener <rguenther@suse.de>
2197 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
2198 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
2199 * tree-ssa-dse.c: Include tree-ssa-loop.h.
2200 (check_name): New callback.
2201 (dse_classify_store): Track cycles via a visited bitmap of PHI
2202 defs and simplify handling of in-loop and across loop dead stores
2203 and properly fail for loop-variant refs. Handle byte-tracking with
2204 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
2207 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
2209 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
2210 (vect_get_mask_type_for_stmt): Likewise.
2211 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
2213 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
2214 to determine the statement's vector type and the vector type that
2215 should be used for calculating nunits. Deal with cases in which
2216 the type has to be deferred.
2217 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
2218 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
2219 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
2220 (vect_determine_vf_for_stmt): New functions, split out from...
2221 (vect_determine_vectorization_factor): ...here.
2222 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
2223 (vect_get_mask_type_for_stmt): New functions, split out from
2224 vect_determine_vectorization_factor.
2226 2018-05-16 Richard Biener <rguenther@suse.de>
2228 * tree-cfg.c (verify_gimple_assign_ternary): Properly
2229 verify the [VEC_]COND_EXPR embedded comparison.
2231 2018-05-15 Martin Sebor <msebor@redhat.com>
2233 PR tree-optimization/85753
2234 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
2235 RECORD_TYPE in addition to ARRAY_TYPE.
2237 2018-05-15 Martin Sebor <msebor@redhat.com>
2240 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
2242 2018-05-15 Richard Biener <rguenther@suse.de>
2244 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
2245 add by_clobber_p one. Change algorithm to collect all defs
2246 representing uses we need to walk and try reducing them to
2247 a single one before failing.
2248 (dse_dom_walker::dse_optimize_stmt): Adjust.
2250 2018-05-13 Mark Wielaard <mark@klomp.org>
2252 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
2253 (size_of_loc_descr): Likewise.
2254 (output_loc_operands): Likewise.
2255 (output_loc_operands_raw): Likewise.
2256 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
2257 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
2258 (hash_loc_operands): Likewise.
2259 (compare_loc_operands): Likewise.
2261 2018-05-14 Mark Wielaard <mark@klomp.org>
2263 * dwarf2out.c (count_index_addrs): New function.
2264 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
2266 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2268 PR tree-optimization/83648
2269 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
2270 return value as malloc candidate.
2272 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2275 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
2276 as true in call to suggest_attribute.
2278 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
2280 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
2281 -mreadonly-in-sdata.
2283 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2285 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
2287 (aarch64_crypto_aesd_fused): Likewise.
2289 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
2291 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
2292 (movsi_aarch64): Likewise.
2293 (load_pairsi): Likewise.
2294 (load_pairdi): Likewise.
2295 (store_pairsi): Likewise.
2296 (store_pairdi): Likewise.
2297 (load_pairsf): Likewise.
2298 (load_pairdf): Likewise.
2299 (store_pairsf): Likewise.
2300 (store_pairdf): Likewise.
2301 (zero_extend): Likewise.
2302 (trunc): Swap alternatives.
2303 (fcvt_target): Add '?' to prefer w over r.
2305 2018-05-14 Jakub Jelinek <jakub@redhat.com>
2308 * config/i386/i386.md: Disallow non-commutative arithmetics in
2309 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
2310 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
2311 in the peephole2 before it.
2313 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
2315 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
2316 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
2317 (ix86_handle_option): Handle -mcldemote.
2318 * config.gcc: New header.
2319 * config/i386/cldemoteintrin.h: New file.
2320 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
2321 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
2323 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2324 OPTION_MASK_ISA_CLDEMOTE.
2325 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
2326 (ix86_valid_target_attribute_inner_p): Ditto.
2327 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
2328 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
2329 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
2330 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
2331 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
2333 * config/i386/i386.opt: Add -mcldemote.
2334 * config/i386/x86intrin.h: New header.
2335 * doc/invoke.texi: Add -mcldemote.
2337 2018-05-14 Richard Biener <rguenther@suse.de>
2339 * doc/match-and-simplify.texi: Adjust :s documentation.
2341 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
2343 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
2344 intended memcpy size.
2345 (REORDER_45): Likewise.
2347 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
2349 * sort.cc: New file.
2350 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
2351 * vec.c (qsort_chk): Use gcc_qsort.
2352 * Makefile.in (OBJS-libcommon): Add sort.o.
2353 (build/sort.o): New target. Use it...
2354 (BUILD_RTL): ... here, and...
2355 (build/gencfn-macros): ... here, and...
2356 (build/genmatch): ... here.
2358 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
2359 Chung-Ju Wu <jasonwucj@gmail.com>
2361 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
2362 * config/nds32/nds32-graywolf.md: New file.
2363 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
2364 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
2366 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
2367 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
2368 * config/nds32/nds32.md (pipeline_model): Add graywolf.
2369 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
2370 * config/nds32/pipelines.md: Include n15 settings.
2372 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
2373 Chung-Ju Wu <jasonwucj@gmail.com>
2375 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
2376 * config/nds32/nds32-n13.md: New file.
2377 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
2378 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
2380 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
2381 * config/nds32/nds32.md (pipeline_model): Add n13.
2382 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
2383 * config/nds32/pipelines.md: Include n13 settings.
2385 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
2386 Chung-Ju Wu <jasonwucj@gmail.com>
2388 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
2389 * config/nds32/nds32-n10.md: New file.
2390 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
2391 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
2393 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
2394 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
2395 * config/nds32/nds32.md (pipeline_model): Add n10.
2396 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
2397 * config/nds32/pipelines.md: Include n10 settings.
2399 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
2400 Kito Cheng <kito.cheng@gmail.com>
2401 Chung-Ju Wu <jasonwucj@gmail.com>
2403 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
2404 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2405 Add enum values for DSP extension instructions.
2406 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
2408 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
2409 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
2411 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
2412 * config/nds32/nds32-dspext.md: New file for DSP implementation.
2413 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
2414 * config/nds32/nds32-intrinsic.md: Likewise.
2415 * config/nds32/nds32_intrinsic.h: Likewise.
2416 * config/nds32/nds32-md-auxiliary.c: Likewise.
2417 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
2418 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
2419 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
2420 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
2421 * config/nds32/nds32-protos.h: New declarations for DSP extension.
2422 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
2423 TYPE_DMAC in switch statement.
2424 * config/nds32/nds32.c: New checking and implementation for DSP
2425 extension instructions.
2426 * config/nds32/nds32.h: Likewise.
2427 * config/nds32/nds32.md: Likewise.
2428 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
2429 * config/nds32/predicates.md: Implement new predicates for DSP
2432 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
2434 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
2435 Reformat alternatives and attributes so it is easier to identify
2436 which constraints/attributes go with which instruction.
2437 (mov<mode>_hardfloat32, FMOVE64): Likewise.
2438 (mov<mode>_softfloat32, FMOVE64): Likewise.
2439 (mov<mode>_hardfloat64, FMOVE64): Likewise.
2440 (mov<mode>_softfloat64, FMOVE64): Likewise.
2442 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
2444 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2446 (Basic PowerPC Built-in Functions): The new name of the
2447 subsection previously known as "PowerPC Built-in Functions".
2448 (Basic PowerPC Built-in Functions Available on all Configurations):
2450 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
2451 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
2452 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
2453 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
2455 2018-05-11 Martin Jambor <mjambor@suse.cz>
2458 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
2461 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
2464 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
2466 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
2468 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
2469 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
2470 (ix86_handle_option): Handle -mwaitpkg.
2471 * config.gcc: New header.
2472 * config/i386/cpuid.h (bit_WAITPKG): New bit.
2473 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
2474 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
2476 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2477 OPTION_MASK_ISA_WAITPKG.
2478 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
2479 (ix86_option_override_internal): Add PTA_WAITPKG.
2480 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
2481 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
2482 IX86_BUILTIN_TPAUSE.
2483 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
2484 __builtin_ia32_umwait and __builtin_ia32_tpause.
2485 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
2486 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
2487 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
2488 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
2489 UNSPECV_TPAUSE): New.
2490 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
2491 * config/i386/i386.opt: Add -mwaitpkg.
2492 * config/i386/waitpkgintrin.h: New file.
2493 * config/i386/x86intrin.h: New header.
2494 * doc/invoke.texi: Add -mwaitpkg.
2496 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
2499 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
2501 (cortex-m0): Use armv6s-m isa.
2502 (cortex-m0plus): Likewise.
2503 (cortex-m1): Likewise.
2504 (cortex-m0.small-multiply): Likewise.
2505 (cortex-m0plus.small-multiply): Likewise.
2506 (cortex-m1.small-multiply): Likewise.
2508 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
2509 Jakub Jelinek <jakub@redhat.com>
2511 PR tree-optimization/85692
2512 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
2513 source permute as well.
2515 2018-05-11 Martin Liska <mliska@suse.cz>
2518 * doc/extend.texi: Document LLVM style format for no_sanitize
2521 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
2523 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
2524 mode_supports_vsx_dform_quad to mode_supports_dq_form.
2525 (mode_supports_vsx_dform_quad): Likewise.
2526 (mode_supports_vmx_dform): Move these functions to be next to the
2527 other mode_supports functions.
2528 (mode_supports_dq_form): Likewise.
2529 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
2530 mode_supports_dq_form.
2531 (reg_offset_addressing_ok_p): Likewise.
2532 (offsettable_ok_by_alignment): Likewise.
2533 (rs6000_legitimate_offset_address_p): Likewise.
2534 (legitimate_lo_sum_address_p): Likewise.
2535 (rs6000_legitimize_address): Likewise.
2536 (rs6000_legitimize_reload_address): Likewise.
2537 (rs6000_secondary_reload_inner): Likewise.
2538 (rs6000_preferred_reload_class): Likewise.
2539 (rs6000_output_move_128bit): Likewise.
2541 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
2543 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
2544 Generate SImode target register for null target.
2545 <case IX86_BUILTIN_XGETBV>: Ditto.
2546 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
2547 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
2549 2018-05-10 Carl Love <cel@us.ibm.com>
2551 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
2552 dcbtt and dcbtstt if operands[2] is 0.
2554 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
2557 * config/i386/sse.md (usadv64qi): New expander.
2559 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
2561 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
2562 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
2563 -maltivec=be support.
2564 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
2565 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
2566 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
2567 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
2568 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
2569 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
2570 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
2571 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
2572 altivec_vsumsws): Adjust.
2573 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
2574 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
2575 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
2577 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
2578 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
2579 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
2580 (altivec_lve<VI_char>x): Delete expand.
2581 (*altivec_lve<VI_char>x_internal): Rename to...
2582 (altivec_lve<VI_char>x): ... this.
2583 (altivec_lvxl_<mode>): Delete expand.
2584 (*altivec_lvxl_<mode>_internal): Rename to ...
2585 (altivec_lvxl_<mode>): ... this.
2586 (altivec_stvxl_<mode>): Delete expand.
2587 (*altivec_stvxl_<mode>_internal): Rename to ...
2588 (altivec_stvxl_<mode>): ... this.
2589 (altivec_stve<VI_char>x): Delete expand.
2590 (*altivec_stve<VI_char>x_internal): Rename to ...
2591 (altivec_stve<VI_char>x): ... this.
2592 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
2593 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
2594 reduc_plus_scal_<mode>): Adjust.
2595 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
2597 (rs6000_cpu_cpp_builtins): Adjust.
2598 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
2599 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
2600 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
2601 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
2602 -maltivec=be support.
2603 (rs6000_split_vec_extract_var): Adjust.
2604 (rs6000_split_v4si_init): Adjust.
2605 (swap_selector_for_mode): Delete.
2606 (altivec_expand_lvx_be, altivec_expand_stvx_be,
2607 altivec_expand_stvex_be): Delete.
2608 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
2609 -maltivec=be support.
2610 (rs6000_gimple_fold_builtin): Ditto.
2611 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
2613 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
2614 (TARGET_DIRECT_MOVE_64BIT): Adjust.
2615 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
2616 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
2617 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
2618 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
2619 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
2620 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
2621 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
2622 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
2623 anonymous split): Adjust.
2624 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
2625 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
2627 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
2629 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
2630 when --with-gxx-include-dir is also specified.
2631 * configure: Regenerate.
2633 2018-05-09 Jim Wilson <jimw@sifive.com>
2636 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
2637 * config/riscv/t-withmultilib: New.
2638 * config/riscv/withmultilib.h: New.
2639 * doc/install.texi: Document RISC-V --with-multilib-list support.
2641 2018-05-09 Richard Biener <rguenther@suse.de>
2643 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
2645 (vect_bb_vectorization_profitable_p): Adjust. Compute
2646 actual scalar cost using the cost vector and the add_stmt_cost
2649 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2651 PR rtl-optimization/85645
2652 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
2653 in the REG_CFA_REGISTER note for LR, don't leave it empty.
2655 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2657 PR rtl-optimization/85645
2658 * shrink-wrap.c (spread_components): Return a boolean saying if
2659 anything was changed.
2660 (try_shrink_wrapping_separate): Iterate spread_components until
2661 nothing changes anymore.
2663 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2665 PR rtl-optimization/85645
2666 * regrename.c (build_def_use): Also kill the chains that include the
2667 destination of a REG_CFA_REGISTER note.
2669 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2671 PR rtl-optimization/85645
2672 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
2673 insn that has a REG_CFA_REGISTER note.
2675 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
2677 * cfgexpand.c (expand_clobber): New function.
2678 (expand_gimple_stmt_1): Use it.
2679 * tree-vect-stmts.c (vect_clobber_variable): New function,
2681 (vectorizable_simd_clone_call): ...here.
2682 (vectorizable_store): Emit a clobber either side of an
2683 IFN_STORE_LANES sequence.
2684 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
2686 2018-05-09 Tom de Vries <tom@codesourcery.com>
2689 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
2690 (define_insn "trap_if_false"): Add exit after trap.
2692 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
2694 PR rtl-optimization/85638
2695 * bb-reorder.c: Include common/common-target.h.
2696 (create_forwarder_block): New function extracted from...
2697 (fix_up_crossing_landing_pad): ...here. Rename into...
2698 (dw2_fix_up_crossing_landing_pad): ...this.
2699 (sjlj_fix_up_crossing_landing_pad): New function.
2700 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
2701 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
2702 from both partitions and exit the loop after one iteration.
2704 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
2707 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2709 (Basic PowerPC Built-in Functions): The new name of the
2710 subsection previously known as "PowerPC Built-in Functions".
2711 (Basic PowerPC Built-in Functions Available on all Configurations):
2713 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
2715 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
2717 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
2719 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
2722 2018-05-08 Jim Wilson <jimw@sifive.com>
2724 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
2725 (LD_EMUL_SUFFIX): New.
2726 (LINK_SPEC): Use it.
2728 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
2730 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2732 (Basic PowerPC Built-in Functions): The new name of the
2733 subsection previously known as "PowerPC Built-in Functions".
2734 (Basic PowerPC Built-in Functions Available on all Configurations):
2736 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
2738 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
2740 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
2742 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
2745 2018-05-08 Jakub Jelinek <jakub@redhat.com>
2748 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
2749 after cmpelim optimization.
2751 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
2753 * config.gcc: Support "goldmont".
2754 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
2755 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2757 * config/i386/i386.c (m_GOLDMONT): Define.
2758 (processor_target_table): Add "goldmont".
2759 (PTA_GOLDMONT): Define.
2760 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
2761 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
2762 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
2763 (fold_builtin_cpu): Add "goldmont".
2764 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
2765 (ix86_option_override_internal): Add "goldmont".
2766 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
2767 (processor_type): Add PROCESSOR_GOLDMONT.
2768 * config/i386/i386.md: Add CPU "glm".
2769 * config/i386/glm.md: New file.
2770 * config/i386/x86-tune.def: Add m_GOLDMONT.
2771 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
2773 2018-05-08 Jakub Jelinek <jakub@redhat.com>
2776 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
2778 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
2779 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
2780 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
2783 * config/i386/i386.c (ix86_fold_builtin): Handle
2784 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
2787 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
2788 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
2790 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
2793 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
2794 (check_arch): Likewise.
2795 (check_fpu): Return the result rather than printing it.
2796 (end arch): Fix operator precedence.
2797 (end cpu): Likewise.
2798 (END): Print the result from check_fpu.
2800 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
2801 Alan Hayward <alan.hayward@arm.com>
2802 David Sherwood <david.sherwood@arm.com>
2804 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
2805 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
2806 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
2807 (*fcmuo<mode>_and): New patterns.
2809 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
2811 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
2812 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
2813 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
2814 (cmp_op, sve_imm_con): New code attributes.
2815 (SVE_COND_INT_CMP, imm_con): Delete.
2816 (cmp_op): Remove above unspecs from int attribute.
2817 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
2819 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
2820 comparison-specific unspecs.
2821 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
2822 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
2823 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
2824 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
2825 (*vec_fcm<cmp_op><mode>): Rename to...
2826 (*fcm<cmp_op><mode>): ...this and adjust likewise.
2827 (*vec_fcmuo<mode>): Rename to...
2828 (*fcmuo<mode>): ...this and adjust likewise.
2829 (*pred_fcm<cmp_op><mode>): New pattern.
2830 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
2831 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
2833 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
2835 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
2836 (aarch64_emit_sve_predicated_cond): New function.
2837 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
2838 (aarch64_emit_unspec_cond_or): Replace with...
2839 (aarch64_emit_sve_or_conds): ...this new function. Use
2840 aarch64_emit_sve_ptrue_op for the individual comparisons and
2841 aarch64_emit_binop to OR them together.
2842 (aarch64_emit_inverted_unspec_cond): Replace with...
2843 (aarch64_emit_sve_inverted_cond): ...this new function. Use
2844 aarch64_emit_sve_ptrue_op for the comparison and
2845 aarch64_emit_unop to invert the result.
2846 (aarch64_expand_sve_vec_cmp_float): Update after the above
2847 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
2849 2018-05-07 Nathan Sidwell <nathan@acm.org>
2851 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
2852 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
2853 (Backwards Compatibility): Likewise.
2855 2018-05-07 Luis Machado <luis.machado@linaro.org>
2859 2018-05-07 Luis Machado <luis.machado@linaro.org>
2861 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2862 <prefetch_dynamic_strides>: New const bool field.
2863 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2864 prefetch_dynamic_strides.
2865 (exynosm1_prefetch_tune): Likewise.
2866 (thunderxt88_prefetch_tune): Likewise.
2867 (thunderx_prefetch_tune): Likewise.
2868 (thunderx2t99_prefetch_tune): Likewise.
2869 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
2870 (aarch64_override_options_internal): Update to set
2871 PARAM_PREFETCH_DYNAMIC_STRIDES.
2872 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
2873 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
2874 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
2875 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
2876 prefetch-dynamic-strides setting.
2878 2018-05-07 Luis Machado <luis.machado@linaro.org>
2880 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2881 <minimum_stride>: New const int field.
2882 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2883 minimum_stride field.
2884 (exynosm1_prefetch_tune): Likewise.
2885 (thunderxt88_prefetch_tune): Likewise.
2886 (thunderx_prefetch_tune): Likewise.
2887 (thunderx2t99_prefetch_tune): Likewise.
2888 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
2889 (aarch64_override_options_internal): Update to set
2890 PARAM_PREFETCH_MINIMUM_STRIDE.
2891 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
2892 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
2893 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
2894 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
2895 stride is constant and is below the minimum stride threshold.
2897 2018-05-07 Luis Machado <luis.machado@linaro.org>
2899 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
2902 2018-05-07 Luis Machado <luis.machado@linaro.org>
2904 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2905 <prefetch_dynamic_strides>: New const bool field.
2906 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2907 prefetch_dynamic_strides.
2908 (exynosm1_prefetch_tune): Likewise.
2909 (thunderxt88_prefetch_tune): Likewise.
2910 (thunderx_prefetch_tune): Likewise.
2911 (thunderx2t99_prefetch_tune): Likewise.
2912 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
2913 (aarch64_override_options_internal): Update to set
2914 PARAM_PREFETCH_DYNAMIC_STRIDES.
2915 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
2916 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
2917 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
2918 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
2919 prefetch-dynamic-strides setting.
2921 2018-05-07 Luis Machado <luis.machado@linaro.org>
2923 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2924 <minimum_stride>: New const int field.
2925 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2926 minimum_stride field.
2927 (exynosm1_prefetch_tune): Likewise.
2928 (thunderxt88_prefetch_tune): Likewise.
2929 (thunderx_prefetch_tune): Likewise.
2930 (thunderx2t99_prefetch_tune): Likewise.
2931 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
2932 (aarch64_override_options_internal): Update to set
2933 PARAM_PREFETCH_MINIMUM_STRIDE.
2934 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
2935 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
2936 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
2937 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
2938 stride is constant and is below the minimum stride threshold.
2940 2018-05-06 Jakub Jelinek <jakub@redhat.com>
2943 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
2944 the type is addressable. Don't force op into register if it has
2947 2018-05-05 Roland McGrath <mcgrathr@google.com>
2950 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
2951 any section for which we don't know a specific type it should have,
2952 regardless of name. Previously this was done only for the exact
2953 names ".init_array", ".fini_array", and ".preinit_array".
2954 (default_elf_asm_named_section): Add comment about
2955 relationship with default_section_type_flags and SECTION_NOTYPE.
2956 (get_section): Don't consider it a type conflict if one side has
2957 SECTION_NOTYPE and the other doesn't, as long as neither has the
2958 SECTION_BSS et al used in the default_section_type_flags logic.
2960 2018-05-05 Tom de Vries <tom@codesourcery.com>
2963 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
2964 (workaround_barsyncs): New function.
2965 (nvptx_reorg): Use workaround_barsyncs.
2966 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
2967 (define_expand "nvptx_membar_cta"): New define_expand.
2968 (define_insn "*nvptx_membar_cta"): New insn.
2970 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
2972 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
2973 To improve optimization opportunities.
2974 * builtin-types.def: The new needed builtin types for the above.
2976 2018-05-04 Richard Biener <rguenther@suse.de>
2978 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
2979 * gimple-ssa-store-merging.c
2980 (imm_store_chain_info::output_merged_store): Remove redundant create,
2981 release split_store vector contents on failure.
2982 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
2983 scalar stmt vector on cache hit.
2985 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
2987 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
2989 * config.gcc (powerpc-xilinx-eabi*): Remove.
2990 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
2992 (fusion_addis_mem_combo_load): Ditto.
2993 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
2995 (rs6000_cpu_cpp_builtins): Ditto.
2996 * config/rs6000/rs6000-linux.c
2997 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
2998 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
2999 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
3001 (rs6000_setup_reg_addr_masks): Ditto.
3002 (rs6000_init_hard_regno_mode_ok): Ditto.
3003 (rs6000_option_override_internal): Ditto.
3004 (legitimate_lo_sum_address_p): Ditto.
3005 (rs6000_legitimize_address): Ditto.
3006 (rs6000_legitimize_reload_address): Ditto.
3007 (rs6000_legitimate_address_p): Ditto.
3008 (abi_v4_pass_in_fpr): Ditto.
3009 (setup_incoming_varargs): Ditto.
3010 (rs6000_gimplify_va_arg): Ditto.
3011 (rs6000_split_multireg_move): Ditto.
3012 (rs6000_savres_strategy): Ditto.
3013 (rs6000_emit_prologue_components): Ditto.
3014 (rs6000_emit_epilogue_components): Ditto.
3015 (rs6000_emit_prologue): Ditto.
3016 (rs6000_emit_epilogue): Ditto.
3017 (rs6000_elf_file_end): Ditto.
3018 (rs6000_function_value): Ditto.
3019 (rs6000_libcall_value): Ditto.
3020 * config/rs6000/rs6000.h: Ditto.
3021 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
3022 (TARGET_MINMAX): ... this. New.
3023 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
3024 * config/rs6000/rs6000.md: Remove Xilinx FP support.
3025 (*movsi_internal1_single): Delete.
3026 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
3027 mfpu=, mxilinx-fpu): Delete.
3028 * config/rs6000/singlefp.h: Delete.
3029 * config/rs6000/sysv4.h: Remove Xilinx FP support.
3030 * config/rs6000/t-rs6000: Ditto.
3031 * config/rs6000/t-xilinx: Delete.
3032 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
3033 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
3034 (VStype_simple): Delete.
3035 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
3036 * config/rs6000/xfpu.h: Delete.
3037 * config/rs6000/xfpu.md: Delete.
3038 * config/rs6000/xilinx.h: Delete.
3039 * config/rs6000/xilinx.opt: Delete.
3040 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
3041 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
3043 2018-05-04 Tom de Vries <tom@codesourcery.com>
3046 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
3049 2018-05-04 Richard Biener <rguenther@suse.de>
3052 * tree-complex.c (update_complex_assignment): We are always in SSA form.
3053 (expand_complex_div_wide): Likewise.
3054 (expand_complex_operations_1): Likewise.
3055 (expand_complex_libcall): Preserve EH info of the original stmt.
3056 (tree_lower_complex): Handle removed blocks.
3057 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
3058 on complex multiplication and division libcall builtins.
3060 2018-05-04 Richard Biener <rguenther@suse.de>
3063 * fold-const.c (negate_expr_p): Restrict negation of operand
3064 zero of a division to when we know that can happen without
3066 (fold_negate_expr_1): Likewise.
3068 2018-05-04 Jakub Jelinek <jakub@redhat.com>
3071 * real.h (real_nextafter): Declare.
3072 * real.c (real_nextafter): New function.
3073 * fold-const-call.c (fold_const_nextafter): New function.
3074 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
3075 CASE_CFN_NEXTTOWARD.
3076 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
3077 even when arg1_mode is different from arg0_mode.
3079 2018-05-03 Nathan Sidwell <nathan@acm.org>
3081 * doc/extend.texi (Deprecated Features): Remove
3083 (Backwards Compatibility): Likewise.
3084 * doc/invoke.texi (C++ Language Options): Likewise.
3085 (C++ Dialect Options): Likewise.
3087 2018-05-03 Jakub Jelinek <jakub@redhat.com>
3090 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
3091 _mm512_mask_mullox_epi64): New intrinsics.
3093 2018-05-03 Tom de Vries <tom@codesourcery.com>
3096 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
3097 dump files): Add offload-tree.
3099 2018-05-03 Richard Biener <rguenther@suse.de>
3101 PR tree-optimization/85615
3102 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
3103 to loops not nested in BBs loop father to avoid creating multi-entry
3106 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3108 PR tree-optimization/70291
3109 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
3110 arguments. Change return type to tree. Emit libcall as a new
3111 statement rather than replacing existing one when inplace_p is true.
3112 (expand_complex_multiplication_components): New function.
3113 (expand_complex_multiplication): Expand floating-point complex
3114 multiplication using the above.
3115 (expand_complex_division): Rename inner_type parameter to type.
3116 Update expand_complex_libcall call-site.
3117 (expand_complex_operations_1): Update expand_complex_multiplication
3118 and expand_complex_division call-sites.
3120 2018-05-02 Jakub Jelinek <jakub@redhat.com>
3123 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
3124 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
3125 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
3126 the highest significant bit of the shift count mask is clear. In
3127 check whether and[sq]i3 is needed verify that all significant bits
3128 of the shift count other than the highest are set.
3130 2018-05-02 Tom de Vries <tom@codesourcery.com>
3133 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
3134 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
3135 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
3136 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
3137 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
3138 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
3139 __builtin_goacc_parlevel_size.
3141 2018-05-02 Richard Biener <rguenther@suse.de>
3143 PR tree-optimization/85597
3144 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
3145 do not use split vect_get_vec_defs call but call vect_get_slp_defs
3148 2018-05-02 Tom de Vries <tom@codesourcery.com>
3151 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
3152 dump files): Add ltrans-tree.
3154 2018-05-02 Tom de Vries <tom@codesourcery.com>
3157 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
3158 dump files): Add wpa-ipa.
3160 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
3162 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
3163 powerpc*-*-linux*paired* target.
3164 * config/rs6000/750cl.h: Delete.
3165 * config/rs6000/paired.h: Delete.
3166 * config/rs6000/paired.md: Delete.
3167 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
3169 * config/rs6000/rs6000-builtin.def: Remove paired float support.
3170 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
3171 comment. Remove paired float support.
3172 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
3173 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
3175 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
3176 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
3178 * config/rs6000/rs6000.c: Remove paired float support.
3179 (paired_expand_vector_init, paired_expand_vector_move,
3180 paired_emit_vector_compare, paired_emit_vector_cond_expr,
3181 (paired_expand_lv_builtin, paired_expand_stv_builtin,
3182 paired_expand_builtin, paired_expand_predicate_builtin,
3183 paired_init_builtins): Delete.
3184 * config/rs6000/rs6000.h: Remove paired float support.
3185 * config/rs6000/rs6000.md: Remove paired float support.
3186 (move_from_CR_ov_bit): Delete.
3187 * config/rs6000/rs6000.opt (mpaired): Delete.
3188 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
3189 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
3191 2018-05-02 Richard Biener <rguenther@suse.de>
3194 * gimplify.c (gimplify_save_expr): When in SSA form allow
3195 SAVE_EXPRs to compute to SSA vars.
3197 2018-05-02 Jakub Jelinek <jakub@redhat.com>
3200 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
3201 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
3202 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
3203 clobber operands[2], instead use a new pseudo. Formatting fixes.
3205 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
3207 PR tree-optimization/85586
3208 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
3209 exit early for statements in the same group if the accesses are
3212 2018-05-02 Tom de Vries <tom@codesourcery.com>
3215 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
3218 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
3220 PR tree-optimization/85143
3221 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
3223 2018-05-01 Tom de Vries <tom@codesourcery.com>
3226 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
3227 not found" error message.
3229 2018-05-01 Tom de Vries <tom@codesourcery.com>
3232 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
3233 * vec.c (test_ordered_remove_if): New function.
3234 (vec_c_tests): Call test_ordered_remove_if.
3235 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
3236 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
3237 * tree-vect-patterns.c (vect_pattern_recog_1): Use
3238 VEC_ORDERED_REMOVE_IF.
3240 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3242 PR tree-optimization/82665
3243 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
3244 pointer subtraction where arguments come from a memchr call.
3246 2018-05-01 Jakub Jelinek <jakub@redhat.com>
3248 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
3249 --push-state --as-needed and --pop-state instead of --as-needed and
3250 --no-as-needed if ld supports it.
3251 * configure: Regenerated.
3254 * doc/install.texi2html: Replace _002d with - and _002a with * in
3255 generated html files using sed.
3257 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3260 * gcc-rich-location.c (blank_line_before_p): New function.
3261 (use_new_line): New function.
3262 (gcc_rich_location::add_fixit_insert_formatted): New function.
3263 * gcc-rich-location.h
3264 (gcc_rich_location::add_fixit_insert_formatted): New function.
3266 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3268 * selftest.c (assert_streq): Rename "expected" and "actual" to
3269 "val1" and "val2". Extend NULL-handling to cover both inputs
3270 symmetrically, while still requiring both to be non-NULL for a pass.
3271 * selftest.h (assert_streq): Rename "expected" and "actual" to
3273 (ASSERT_EQ): Likewise.
3274 (ASSERT_EQ_AT): Likewise.
3275 (ASSERT_KNOWN_EQ): Likewise.
3276 (ASSERT_KNOWN_EQ_AT): Likewise.
3277 (ASSERT_NE): Likewise.
3278 (ASSERT_MAYBE_NE): Likewise.
3279 (ASSERT_MAYBE_NE_AT): Likewise.
3280 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
3281 the assertion to pass.
3282 (ASSERT_STREQ_AT): Likewise.
3284 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
3286 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
3287 interaction with -pie.
3289 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3291 * selftest.h: Fix alphabetization of per-source-file selftest
3294 2018-04-30 Jason Merrill <jason@redhat.com>
3296 PR c++/61982 - dead stores to destroyed objects.
3297 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
3300 2018-04-30 Jason Merrill <jason@redhat.com>
3302 * tree.c (build_clobber): New.
3303 * tree.h: Declare it.
3304 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
3306 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3308 * diagnostic-show-locus.c (layout::layout): Update for
3309 location_get_source_line returning a char_span.
3310 (struct char_span): Move to input.h.
3311 (struct correction): Update for fields in char_span becoming
3313 (struct source_line): Update for location_get_source_line
3314 returning a char_span.
3315 (layout::print_line): Likewise.
3316 * edit-context.c (edited_file::print_content): Likewise.
3317 (edited_file::print_diff_hunk): Likewise.
3318 (edited_file::print_run_of_changed_lines): Likewise.
3319 (edited_file::get_num_lines): Likewise.
3320 (edited_line::edited_line): Likewise.
3321 * final.c (asm_show_source): Likewise.
3322 * input.c (location_get_source_line): Convert return type
3323 from const char * to char_span, losing the final "line_len"
3325 (dump_location_info): Update for the above.
3326 (get_substring_ranges_for_loc): Likewise. Use a char_span
3327 when handling the literal within the line.
3328 (test_reading_source_line): Update for location_get_source_line
3329 returning a char_span.
3330 * input.h (class char_span): Move here from
3331 diagnostic-show-locus.c, converting from a struct to a class.
3332 Make data members private.
3333 (char_span::operator bool): New.
3334 (char_span::length): New.
3335 (char_span::get_buffer): New.
3336 (char_span::operator[]): New.
3337 (char_span::subspan): Make const.
3338 (char_span::xstrdup): New.
3339 (location_get_source_line): Convert return type from const char *
3340 to char_span, losing the final "line_size" param.
3342 2018-04-30 Jan Hubicka <jh@suse.cz>
3344 * lto-wrapper.c (ltrans_priorities): New static var.
3345 (cmp_priority): New.
3346 (run_gcc): Read priorities and if doing parallel build order
3347 the Makefile by them.
3349 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3351 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
3353 2018-04-30 Richard Biener <rguenther@suse.de>
3355 * tree-cfg.c (verify_address): Remove base argument, add
3356 flag whether to check TREE_ADDRESSABLE and do that.
3357 (verify_expr): Remove.
3358 (verify_types_in_gimple_reference): Add pieces from verify_expr.
3359 (verify_gimple_assign_single): Likewise.
3360 (verify_gimple_switch): Likewise.
3361 (verify_expr_location_1): Dereference tp once. Add (disabled)
3362 piece from verify_expr.
3363 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
3365 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3367 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
3369 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3371 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
3372 (small_data_pattern): Likewise.
3373 (arc_rewrite_small_data): Likewise.
3374 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
3375 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
3376 (get_symbol_alignment): New function.
3377 (legitimate_small_data_address_p): Likewise.
3378 (legitimate_scaled_address): Update, call
3379 legitimate_small_data_address_p.
3380 (output_sdata): New static variable.
3381 (arc_print_operand): Update how we handle small data operands.
3382 (arc_print_operand_address): Likewise.
3383 (arc_legitimate_address_p): Update, use
3384 legitimate_small_data_address_p.
3385 (arc_rewrite_small_data_p): Remove.
3386 (arc_rewrite_small_data_1): Likewise.
3387 (arc_rewrite_small_data): Likewise.
3388 (small_data_pattern): Likewise.
3389 (compact_sda_memory_operand): Update to use
3390 legitimate_small_data_address_p and get_symbol_alignment.
3391 (prepare_move_operands): Don't rewite sdata pattern.
3392 (prepare_extend_operands): Remove.
3393 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
3395 (zero_extendqisi2): Likewise.
3396 (zero_extendhisi2): Likewise.
3397 (extendqihi2): Likewise.
3398 (extendqisi2): Likewise.
3399 (extendhisi2): Likewise.
3403 * config/arc/constraints.md (Usd): Change it to memory constraint.
3405 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3407 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
3408 as source of std instructions.
3409 * config/arc/arc.md (movsi_insn): Update pattern predicate to
3410 allow 6-bit constants as source for store instructions.
3411 (movdi_insn): Update instruction pattern to allow 6-bit constants
3412 as source for store instructions.
3414 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
3416 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
3418 2018-04-30 Nathan Sidwell <nathan@acm.org>
3419 Sandra Loosemore <sandra@codesourcery.com>
3421 * dumpfile.c (dump_open): Allow '-' for stdout.
3422 * doc/invoke.texi (Developer Options): Document dump filename
3423 determination early. Document stdin/stdout selection.
3425 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
3427 Microblaze Target: PIC data text relative
3429 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
3430 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
3432 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
3433 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
3434 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
3435 New addressing mode for data-text relative position indepenedent code.
3436 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
3437 'ADDRESS_SYMBOLIC_TXT_REL'.
3438 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
3439 (microblaze_legitimate_pic_operand): Exclude function calls from
3440 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
3441 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
3443 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
3444 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
3445 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
3446 for 'address + offset'.
3447 (microblaze_expand_prologue): Add new function prologue call for
3449 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
3450 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
3451 table in case of TARGET_PIC_DATA_TEXT_REL.
3452 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
3453 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
3454 Add new macros 'UNSPEC_TEXT',
3455 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
3456 + exclude function calls from 'UNSPEC_PLT' in case of data text
3458 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
3459 new target hook for generating address diff vector tables in case of
3461 * doc/tm.texi : Regenerate.
3462 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
3463 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
3464 of addr diff vector generation.
3465 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
3466 target hook definition.
3467 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
3468 Add default function for generate_pic_addr_diff_vec -> flag_pic.
3469 * doc/invoke.texi (Add new pic option): Add new microblaze pic
3470 option for data text relative.
3472 2018-04-30 Richard Biener <rguenther@suse.de>
3474 * tree-chrec.h (evolution_function_is_constant_p): Remove
3476 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
3478 2018-04-30 Richard Biener <rguenther@suse.de>
3481 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
3483 2018-04-30 Richard Biener <rguenther@suse.de>
3485 PR tree-optimization/28364
3486 PR tree-optimization/85275
3487 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
3488 copying first exit test.
3490 2018-04-28 Mark Wielaard <mark@klomp.org>
3492 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
3494 (dwarf_AT): Handle DW_AT_addr_base.
3495 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
3497 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
3500 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
3501 (*ashl<dwi>3_doubleword_mask_1): Ditto.
3502 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
3503 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
3505 2018-04-28 Richard Biener <rguenther@suse.de>
3507 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
3508 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
3509 to reflect use. Only add interesting stmts.
3511 2018-04-27 Martin Jambor <mjambor@suse.cz>
3514 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
3515 the jump function allows for passing through aggregate values.
3517 2018-04-27 David Malcolm <dmalcolm@redhat.com>
3519 * input.h (in_system_header_at): Convert from macro to inline
3521 (from_macro_expansion_at): Likewise.
3522 (from_macro_definition_at): Likewise.
3524 2018-04-27 Jeff Law <law@redhat.com>
3526 * config.gcc: Mark tile* targets as deprecated/obsolete.
3528 2018-04-27 Richard Biener <rguenther@suse.de>
3530 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
3533 2018-04-27 Richard Biener <rguenther@suse.de>
3535 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
3537 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
3539 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
3540 with Yd constraint. Set "preferred_for_speed" attribute from
3541 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
3543 (*movdi_internal): Ditto.
3544 (movti_interunit splitters): Remove
3545 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
3546 (movdi_interunit splitters): Ditto.
3547 * config/i386/constraints.md (Ye): Remove.
3548 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
3550 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3553 * config/aarch64/constraints.md (Usg): Limit to 31.
3556 2018-04-27 Jakub Jelinek <jakub@redhat.com>
3558 PR tree-optimization/85529
3559 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
3560 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
3561 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
3562 zero extension or masking of the MSB bit.
3563 (optimize_range_tests): Add FIRST_BB argument, pass it through
3564 to optimize_range_tests_var_bound.
3565 (maybe_optimize_range_tests, reassociate_bb): Adjust
3566 optimize_range_tests callers.
3568 2018-04-26 Richard Biener <rguenther@suse.de>
3569 Jakub Jelinek <jakub@redhat.com>
3571 * cgraph.h (symbol_table): Just declare debug method here.
3572 * symtab.c (symbol_table::debug): Define.
3574 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
3576 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
3578 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
3580 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
3581 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
3582 (*movdi_internal): Substitute Yi and Yj constraint with x
3583 and Ym and Yn constraint with y constraint. Update "isa"
3584 attribute and set "preferred_for_speed" attribute from
3585 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
3586 (*movsi_internal): Ditto.
3587 (*movdf_internal): Ditto.
3588 (*movsf_internal): Ditto.
3589 (*zero_extendsidi2): Ditto.
3590 * config/i386/sse.md (vec_set<mode>_0): Ditto.
3591 (sse2_loadld): Ditto.
3592 (*vec_extract<ssevecmodelower>_0): Ditto.
3593 (*vec_extractv4si_0_zext_sse4): Ditto.
3594 (vec_concatv2di): Ditto.
3595 (*vec_dup<mode>): Ditto.
3596 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
3597 * config/i386/constraints.md (Yi): Remove.
3602 2018-04-26 Nathan Sidwell <nathan@acm.org>
3604 * dumpfile.c (dump_open): New.
3605 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
3606 (dump_finish): Detect stdio/stderr by value not name.
3608 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
3610 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
3612 2018-04-26 Tom de Vries <tom@codesourcery.com>
3615 * config/nvptx/nvptx.c (verify_neutering_jumps)
3616 (verify_neutering_labels): New function
3617 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
3619 2018-04-26 Tom de Vries <tom@codesourcery.com>
3622 * config/nvptx/nvptx.c (needs_neutering_p): New function.
3623 (nvptx_single): Use needs_neutering_p to skip over insns that do not
3626 2018-04-26 Richard Biener <rguenther@suse.de>
3627 Tom de Vries <tom@codesourcery.com>
3630 * lto-streamer-out.c (output_function): Fixup loops if required to match
3631 discovery done in the reader.
3633 2018-04-26 Richard Biener <rguenther@suse.de>
3635 PR tree-optimization/85116
3636 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
3637 have a loop exit from the single latch predecessor. Remove
3638 case of header with just condition.
3639 (ch_base::copy_headers): Exclude infinite loops from any
3641 (pass_ch::execute): Record exits.
3643 2018-04-26 Richard Biener <rguenther@suse.de>
3645 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
3646 prologue cost vector and pass it to vect_get_load_cost.
3647 (vect_get_peeling_costs_all_drs): Likewise.
3648 (vect_peeling_hash_get_lowest_cost): Likewise.
3649 (vect_enhance_data_refs_alignment): Likewise.
3651 2018-04-26 Richard Biener <rguenther@suse.de>
3654 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
3655 checking of integer<->pointer conversions.
3656 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
3657 sign-/zero-extending pointer types.
3658 (expand_omp_for_static_chunk): Likewise.
3660 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
3661 Jean Lee <xiaoyur347@gmail.com>
3663 * config/mips/mips.c (mips_asan_shadow_offset): New function.
3664 (TARGET_ASAN_SHADOW_OFFSET): Define.
3665 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
3666 true for -fsanitize=address.
3668 2018-04-25 Mark Wielaard <mark@klomp.org>
3670 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
3673 2018-04-25 Jakub Jelinek <jakub@redhat.com>
3675 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
3676 than "alu", remove explicit "memory" and "imm_disp" attributes.
3677 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
3680 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
3681 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
3682 gen_lowpart_no_emit.
3684 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
3687 * config/i386/i386.c (ix86_expand_builtin): Change memory
3688 operand to XI, extend p0 to Pmode.
3689 * config/i386/i386.md: Change unspec volatile and operand
3690 1 mode to XI, change operand 0 mode to P.
3692 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3694 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
3695 GET_MODE_MASK before any checking.
3696 (nds32_can_use_bset_p): Likewise.
3697 (nds32_can_use_btgl_p): Likewise.
3699 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3701 * config/nds32/nds32-doubleword.md: New define_split pattern for
3702 illegal register number.
3704 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3706 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
3708 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3710 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
3712 2018-04-25 Richard Biener <rguenther@suse.de>
3714 * lto-streamer.h (LTO_major_version): Bump to 8.
3716 2018-04-25 Jakub Jelinek <jakub@redhat.com>
3718 * BASE-VER: Set to 9.0.0.
3720 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
3722 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
3723 in __abskf2 and __powikf2.
3725 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3728 * config/aarch64/constraints.md (Usg, Usj): New constraints.
3729 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
3730 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
3731 Use the above on operand 2. Reindent.
3732 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
3734 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
3737 * common/config/i386/i386-common.c (ix86_handle_option): Don't
3739 * config/i386/i386.opt (mcet): Removed.
3740 * doc/install.texi: Remove -mcet documentation.
3741 * doc/invoke.texi: Likewise.
3743 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
3746 * doc/install.texi: Remove -mcet from bootstrap-cet.
3748 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3751 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
3752 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
3756 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
3757 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
3758 containing a CONST_VECTOR.
3760 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
3762 * doc/install.texi: Update newlib dependency for nvptx.
3764 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3767 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
3768 instead of INTVAL when shifting x left.
3770 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
3772 PR tree-optimization/85478
3773 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
3774 vect_grouped_store_supported for single element vectors.
3776 2018-04-24 Richard Biener <rguenther@suse.de>
3779 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
3780 load cost increase to the case of non-constant step.
3782 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3785 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
3786 destination if any_malformed_asm.
3788 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
3791 * expr.c (store_field): In the bitfield case, if the value comes from
3792 a function call and is returned in registers by means of a PARALLEL,
3793 do not change the mode of the temporary unless BLKmode and VOIDmode.
3795 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
3797 PR rtl-optimization/85423
3798 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
3799 dependencies to debug insns when the previous insn is non-debug.
3801 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
3803 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
3804 enums into a single definition.
3805 (fls): Fix predicates and printing.
3808 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
3810 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
3811 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
3812 and short u6 immediate.
3813 (check_if_valid_sleep_operand): Remove.
3814 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
3816 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3818 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
3819 flag_always_save_lp condition.
3820 * config/nds32/nds32.opt (malways-save-lp): New option.
3822 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3824 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
3825 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
3826 * config/nds32/nds32.h
3827 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
3828 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
3830 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3832 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
3833 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
3835 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3836 Chung-Ju Wu <jasonwucj@gmail.com>
3838 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
3840 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
3841 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
3843 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3845 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
3847 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3849 * config/nds32/nds32-protos.h (nds32_data_alignment,
3850 nds32_local_alignment): Declare.
3851 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
3852 nds32_local_alignment): New functions.
3853 (TARGET_CONSTANT_ALIGNMENT): Define.
3854 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
3856 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3858 * config/nds32/nds32.c
3859 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
3860 (TARGET_MODES_TIEABLE_P): Likewise.
3862 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3864 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
3867 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
3868 Chung-Ju Wu <jasonwucj@gmail.com>
3870 * config/nds32/constants.md (unspec_volatile_element): Add enum values
3871 for unaligned access.
3872 * config/nds32/nds32-intrinsic.c: Implementation of expanding
3874 * config/nds32/nds32-intrinsic.md: Likewise.
3875 * config/nds32/nds32_intrinsic.h: Likewise.
3876 * config/nds32/nds32.h (nds32_builtins): Likewise.
3877 * config/nds32/nds32.opt (munaligned-access): New option.
3878 * config/nds32/nds32.c (nds32_asm_file_start): Display
3879 flag_unaligned_access status.
3881 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
3883 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
3884 -mno-relax is present.
3885 * config/riscv/linux.h (LINK_SPEC): Ditto.
3887 2018-04-20 Martin Sebor <msebor@redhat.com>
3890 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
3892 (gimple_fold_builtin_stxcpy_chk): Same.
3893 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
3895 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
3898 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
3899 __powikf2 when long double is IEEE 128-bit.
3901 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
3903 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
3904 step to make sure stack always aligned.
3906 2018-04-20 Carl Love <cel@us.ibm.com>
3909 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
3910 size check for arg0.
3912 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
3913 Tom de Vries <tom@codesourcery.com>
3916 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
3917 Emit insns for calls too.
3918 (nvptx_find_par): Always look for worker-level predecessor insn.
3919 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
3921 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
3922 (nvptx_process_pars): Propagate frames for calls.
3924 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
3927 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
3929 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
3930 (ix86_handle_option): Don't handle OPT_mibt.
3931 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
3933 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
3935 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
3936 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
3937 (ix86_target_macros): Define __CET__ with flag_cf_protection
3938 for -fcf-protection.
3939 * config/i386/i386.c (isa2_opts): Remove -mibt.
3940 * config/i386/i386.h (TARGET_IBT): Removed.
3941 (TARGET_IBT_P): Likewise.
3942 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
3943 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
3944 * config/i386/i386.opt (mcet): Update help message.
3947 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
3948 -mcet as an alias for -mshstk.
3950 2018-04-20 Richard Biener <rguenther@suse.de>
3953 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
3954 complexity by forcing a single use of the multiply operand.
3956 2018-04-20 Martin Jambor <mjambor@suse.cz>
3959 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
3960 recursion dependency to only apply to non-clones.
3962 2018-04-20 Martin Jambor <mjambor@suse.cz>
3965 * ipa-cp.c (create_specialized_node): Check that clones of
3966 self-recursive edges exist during IPA-CP.
3968 2018-04-19 Toon Moene <toon@moene.org>
3970 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
3973 2018-04-19 Jakub Jelinek <jakub@redhat.com>
3975 PR tree-optimization/85467
3976 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
3977 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
3978 VECTOR_CST element to type.
3980 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
3983 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
3984 * config/i386/i386.md (builtin_setjmp_setup): Removed.
3985 (builtin_longjmp): Likewise.
3986 (save_stack_nonlocal): New pattern.
3987 (restore_stack_nonlocal): Likewise.
3989 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
3992 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
3993 Replace ASM_OUTPUT_LABEL with fprintf.
3995 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
3998 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
3999 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
4000 * config/i386/i386-c.c (ix86_target_macros_internal): Also
4001 define __IBT__ and __SHSTK__ for -fcf-protection.
4002 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
4004 (ix86_trampoline_init): Likewise.
4005 (x86_output_mi_thunk): Likewise.
4006 (ix86_notrack_prefixed_insn_p): Likewise.
4007 (ix86_option_override_internal): Don't disallow -fcf-protection.
4008 * config/i386/i386.md (rdssp<mode>): Also enable for
4010 (incssp<mode>): Likewise.
4011 (nop_endbr): Likewise.
4012 * config/i386/i386.opt (mcet): Change help message to built-in
4016 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
4017 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
4018 enable CET built-in functions.
4020 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
4022 * common/config/i386/i386-common.c
4023 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
4024 OPTION_MASK_ISA_MOVDIRI_UNSET,
4025 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
4026 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
4027 * config.gcc (movdirintrin.h): New header.
4028 * config/i386/cpuid.h (bit_MOVDIRI,
4029 bit_MOVDIR64B): New bits.
4030 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
4032 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
4033 (VOID, PVOID, PCVOID)): New function types.
4034 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
4035 __builtin_ia32_directstoreu_u64,
4036 __builtin_ia32_movdir64b): New builtins.
4037 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
4038 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
4040 (ix86_valid_target_attribute_inner_p): Ditto.
4041 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
4042 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
4043 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
4044 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
4045 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
4046 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
4047 (movdiri<mode>, movdir64b_<mode>): New.
4048 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
4049 * config/i386/immintrin.h: Include movdirintrin.h.
4050 * config/i386/movdirintrin.h: New file.
4051 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
4053 2018-04-19 Richard Biener <rguenther@suse.de>
4056 * cfg.c (clear_bb_flags): When loop state says we have
4057 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
4059 2018-04-19 Richard Biener <rguenther@suse.de>
4061 PR tree-optimization/84737
4062 * tree-vect-data-refs.c (vect_copy_ref_info): New function
4063 copying restrict info.
4064 (vect_setup_realignment): Use it.
4065 * tree-vectorizer.h (vect_copy_ref_info): Declare.
4066 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
4067 the first DR to all generated stores.
4068 (vectorizable_load): Likewise for loads.
4070 2018-04-19 Jakub Jelinek <jakub@redhat.com>
4072 PR tree-optimization/85446
4073 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
4074 the integral and pointer types to have the same precision.
4076 * doc/install.texi: Document --disable-cet being the default and
4079 2018-04-18 Martin Liska <mliska@suse.cz>
4081 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
4084 2018-04-18 Martin Liska <mliska@suse.cz>
4087 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
4090 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
4091 arguments if they are comparable.
4093 2018-04-18 Martin Liska <mliska@suse.cz>
4096 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
4099 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
4102 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
4105 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
4106 ENDBR after calling __morestack.
4108 2018-04-18 David Malcolm <dmalcolm@redhat.com>
4111 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
4112 by using gcc_base_ver to generate a gcc_driver_version, and use
4113 it when generating GCC_DRIVER_NAME.
4114 * configure: Regenerate.
4116 2018-04-18 Jakub Jelinek <jakub@redhat.com>
4119 * config.gcc: Obsolete powerpc*-*-*spe*.
4121 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4124 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
4125 (stabstr_D): Change type of unum from unsigned int to
4126 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
4129 2018-04-17 Jim Wilson <jimw@sifive.com>
4132 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
4133 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
4134 Set arg_pointer_offset after using pretend_args_size.
4136 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4138 PR rtl-optimization/85431
4139 * dse.c (record_store): Ignore zero width stores.
4142 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
4143 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
4144 __builtin_stack_restore rather than after it.
4145 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
4146 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
4147 argument instead of virtual_dynamic_stack_rtx.
4149 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
4151 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
4153 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
4154 Add note to error message to explain internal mapping of overloaded
4155 built-in function name to non-overloaded built-in function name.
4156 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
4159 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
4162 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
4163 where the inputs overlap with the output.
4165 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4167 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
4168 (=v, v) alternative and explicit "memory" attribute.
4169 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
4170 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4172 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
4173 "sselog1" type instead of "sselog".
4174 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
4175 "sselog". Remove explicit "memory" attribute.
4176 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
4177 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4179 (vec_extract_hi_v32hi): Merge all alternatives into one, use
4180 "sselog1" type instead of "sselog". Remove explicit "memory"
4182 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
4183 use "sselog1" type instead of "sselog". Remove explicit "memory"
4185 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
4186 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4188 (vec_extract_hi_v64qi): Merge all alternatives into one, use
4189 "sselog1" type instead of "sselog". Remove explicit "memory"
4191 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
4192 use "sselog1" type instead of "sselog". Remove explicit "memory"
4196 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
4199 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
4202 2018-04-17 Martin Jambor <mjambor@suse.cz>
4205 * ipa-cp.c (create_specialized_node): Call
4206 expand_all_artificial_thunks if necessary.
4208 2018-04-17 Martin Liska <mliska@suse.cz>
4211 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
4212 in message, remote space in between '_G' and '('.
4214 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4217 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
4218 avx512f_vmcmp<mode>3<round_saeonly_name>,
4219 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
4220 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
4221 avx512f_rndscale<mode><round_saeonly_name>,
4222 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
4223 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
4224 Use %<iptr>2 instead of %2 for -masm=intel.
4225 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
4226 avx512f_vcvttss2usi<round_saeonly_name>,
4227 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
4229 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
4230 avx512f_vcvttsd2usi<round_saeonly_name>,
4231 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
4232 Use %q1 instead of %1 for -masm=intel.
4233 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
4234 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
4235 of %3 for -masm=intel.
4236 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
4238 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
4240 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
4242 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
4243 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
4245 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
4247 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
4248 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
4249 %g1 and one with %0 and %1.
4250 (avx512er_vmrcp28<mode><round_saeonly_name>,
4251 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
4253 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
4254 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
4255 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
4256 of %0 and %{%4%} for -masm=intel.
4257 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
4258 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
4259 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
4260 order of %0 and %{%5%}%{z%} for -masm=intel.
4262 2018-04-17 Jan Hubicka <jh@suse.cz>
4265 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
4267 2018-04-17 Martin Liska <mliska@suse.cz>
4270 * multiple_target.c (create_dispatcher_calls): Set apostrophes
4271 for target_clone error message. Make default implementation
4272 clone to be a local declaration.
4273 (separate_attrs): Add new argument and check for an empty
4275 (expand_target_clones): Handle it.
4276 (ipa_target_clone): Make redirection just for target_clones
4279 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
4280 Tom de Vries <tom@codesourcery.com>
4283 * omp-expand.c (expand_oacc_for): Add dummy false branch for
4284 tiled basic blocks without omp continue statements.
4286 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
4289 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
4290 vec_extract expression as having side effects to make sure it gets
4293 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
4296 * config/i386/i386.c (get_builtin_code_for_version): Check
4299 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
4302 * gcc/config.gcc: Support "skylake".
4303 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
4305 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
4306 (processor_target_table): Add "skylake".
4307 (ix86_option_override_internal): Add "skylake".
4308 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
4309 PROCESSOR_CANNONLAKE.
4310 (get_builtin_code_for_version): Fix priority for
4311 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
4312 PROCESSOR_SKYLAKE-AVX512.
4313 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
4314 (processor_type): Add PROCESSOR_SKYLAKE.
4316 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
4317 Jason Merrill <jason@redhat.com>
4320 * convert.c (convert_to_integer_1): Use direct recursion for
4321 enumeral types and types with a precision less than the number
4322 of bits in their mode.
4324 2018-04-16 Julia Koval <julia.koval@intel.com>
4327 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
4328 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
4330 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
4333 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
4334 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
4335 and -mno-direct-move.
4337 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
4340 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
4341 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
4342 Ensure negative shifts result in {0}.
4344 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
4346 PR rtl-optimization/79916
4347 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
4348 regs (if any) to define how to gnerate SD moves when LRA is in
4351 2018-04-13 Jakub Jelinek <jakub@redhat.com>
4353 PR rtl-optimization/85393
4354 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
4355 * except.c (expand_dw2_landing_pad_for_region): Make static.
4356 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
4357 a label and unconditional jump to old_bb, rather than
4358 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
4361 PR rtl-optimization/85376
4362 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
4363 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
4364 instead of a specific value.
4366 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
4367 Bin Cheng <bin.cheng@arm.com>
4369 PR tree-optimization/82965
4370 PR tree-optimization/83991
4371 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
4372 by_profile_only parameter.
4373 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
4374 information if the loop was predicted to iterate too many times.
4375 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
4377 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
4380 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
4383 2018-04-13 Martin Liska <mliska@suse.cz>
4384 Jakub Jelinek <jakub@redhat.com>
4387 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
4388 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
4389 * builtins.c (expand_builtin_memory_copy_args): Use
4390 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
4391 handle dest_addr == pc_rtx.
4393 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
4396 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
4397 asked to not generate direct moves.
4398 (fix_trunc<mode>si2_stfiwx): Similar.
4399 (fix_trunc<mode>si2_internal): Similar.
4401 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4404 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
4405 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
4406 lookup if dest in some wider mode is known to be const0_rtx and
4407 if so, record permanent equivalence for it to be ZERO_EXTEND of
4408 the narrower mode destination.
4410 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
4412 * lto-streamer-out.c (output_function): Revert 259346.
4413 * omp-expand.c (expand_oacc_for): Likewise.
4415 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
4417 PR rtl-optimization/85354
4418 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
4419 * sel-sched.c (sel_global_init): ... here.
4421 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
4424 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
4425 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
4426 mode for PE-COFF targets.
4427 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
4428 (i386_pe_asm_lto_end): Likewise.
4429 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
4430 (TARGET_ASM_LTO_END): Likewise.
4431 * config/i386/winnt.c (saved_debug_info_level): New static variable.
4432 (i386_pe_asm_lto_start): New function.
4433 (i386_pe_asm_lto_end): Likewise.
4435 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
4436 Richard Biener <rguenther@suse.de>
4439 * lto-streamer-out.c (output_function): Fix CFG loop state before
4441 * omp-expand.c (expand_oacc_for): Handle calls to internal
4442 functions like regular functions.
4444 2018-04-12 Richard Biener <rguenther@suse.de>
4447 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
4448 for the early LTO debug to properly generate references to it
4449 during DIE emission. Do not re-use that for the skeleton for
4451 (dwarf2out_early_finish): Likewise.
4453 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4456 * config/i386/sse.md
4457 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
4458 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
4459 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
4460 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
4461 and output is a reg, avoid creating invalid lowpart subreg, but
4462 instead split into a 512-bit move. Don't split if not AVX512VL,
4463 input is xmm16+ reg and output is a mem.
4464 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
4465 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
4466 xmm16+ reg and output is a mem.
4468 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4470 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
4471 also for flag_dwarf2_cfi_asm.
4473 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4475 PR rtl-optimization/85342
4476 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
4477 a bool scalar var inside of the loop instead. Don't try to update
4478 recog_data.operand after failed apply_change_group.
4480 2018-04-12 Tom de Vries <tom@codesourcery.com>
4483 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
4484 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
4485 array with flexible array member as array without given dimension.
4486 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
4487 argument for undefined param to true.
4489 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
4492 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
4493 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
4494 from PowerPC section.
4495 * config/rs6000/sysv4.opt (mcall-): Improve help text.
4496 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
4497 help text that is too long.
4498 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
4499 help text that is too long.
4500 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
4501 help text that is too long.
4503 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
4505 * config/alpha/alpha.md (stack_probe_internal): Rename
4506 from "probe_stack". Update all callers.
4508 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4510 PR rtl-optimization/84566
4511 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
4512 sched_macro_fuse_insns.
4514 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4517 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
4518 (compute_block_dependences): ... from here.
4520 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4522 PR tree-optimization/85331
4523 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
4524 from int to HOST_WIDE_INT.
4526 2018-04-11 Martin Jambor <mjambor@suse.cz>
4529 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
4530 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
4531 not the same as the source val.
4532 (cgraph_edge_brings_value_p): New parameter.
4533 (gather_edges_for_value): Pass destination value to
4534 cgraph_edge_brings_value_p.
4535 (perhaps_add_new_callers): Likewise.
4536 (get_info_about_necessary_edges): Likewise and exclude values brought
4537 only by self-recursive edges.
4538 (create_specialized_node): Redirect only clones of self-calling edges.
4539 (+self_recursive_pass_through_p): New function.
4540 (find_more_scalar_values_for_callers_subset): Use it.
4541 (find_aggregate_values_for_callers_subset): Likewise.
4542 (known_aggs_to_agg_replacement_list): Removed.
4543 (decide_whether_version_node): Re-calculate known constants for all
4544 remaining context clones.
4546 2018-04-11 Richard Biener <rguenther@suse.de>
4549 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
4550 from early DWARF output.
4551 (dwarf2out_early_finish): Output line info unconditionally into
4552 early DWARF and add reference to it.
4554 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4557 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
4558 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
4559 other than V2DFmode using iptr mode attribute.
4560 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
4562 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4564 PR rtl-optimization/84659
4565 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
4567 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4570 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
4572 (output_loc_list): Pass address of a dummy size variable even in the
4573 locview handling loop.
4574 (index_location_lists): Add comment on why skip_loc_list_entry can't
4577 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
4580 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
4583 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
4586 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
4587 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
4588 and -mstring-compare-inline-limit.
4590 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4593 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
4594 for stack clash protection in a register whenever we need it to be in
4597 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4599 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
4600 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
4602 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4605 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
4607 (mlong-double-): Ditto.
4608 * config/rs6000/sysv4.opt (msdata=): Ditto.
4609 (mtls-size=): Ditto.
4611 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
4613 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
4614 erroneous entries for
4615 "vector int vec_ldl (int, long int *)", and
4616 "vector unsigned int vec_ldl (int, unsigned long int *)".
4617 Add comments and entries for
4618 "vector bool char vec_ldl (int, bool char *)",
4619 "vector bool short vec_ldl (int, bool short *)",
4620 "vector bool int vec_ldl (int, bool int *)",
4621 "vector bool long long vec_ldl (int, bool long long *)",
4622 "vector pixel vec_ldl (int, pixel *)",
4623 "vector long long vec_ldl (int, long long *)",
4624 "vector unsigned long long vec_ldl (int, unsigned long long *)".
4625 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
4626 type tree bool_long_long_type_node and correct definition of
4627 bool_V2DI_type_node to make reference to this new type tree.
4628 (rs6000_mangle_type): Replace erroneous reference to
4629 bool_long_type_node with bool_long_long_type_node.
4630 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
4631 comments to emphasize sign distinctions for char and int types and
4632 replace RS6000_BTI_bool_long constant with
4633 RS6000_BTI_bool_long_long constant. Also add comment to restrict
4634 use of RS6000_BTI_pixel.
4635 (bool_long_type_node): Remove this macro definition.
4636 (bool_long_long_type_node): New macro definition
4638 2018-04-10 Jakub Jelinek <jakub@redhat.com>
4640 PR rtl-optimization/85300
4641 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
4642 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
4643 simplify_unary_operation fails.
4645 2018-04-10 Martin Liska <mliska@suse.cz>
4647 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
4648 cgraph_edge and ipa_ref.
4650 2018-04-10 Jakub Jelinek <jakub@redhat.com>
4654 * config/i386/sse.md
4655 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
4656 computation of the VEC_MERGE selector from mask.
4657 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
4658 Fix decoding of the VEC_MERGE selector into mask.
4660 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
4662 PR tree-optimization/85286
4663 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
4665 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
4667 * final.c (final_1): Set insn_last_address as well as
4668 insn_current_address.
4670 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4673 * explow.c (emit_stack_probe): Call validize_mem on memory location
4674 before passing it to gen_probe_stack. Create address operand and
4675 legitimize it for the probe_stack_address case.
4677 2018-04-09 Jan Hubicka <jh@suse.cz>
4680 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
4681 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
4682 * tree.c (free_lang_data_in_type): Fix handling of binfos;
4684 (free_lang_data): Rebuild type inheritance graph.
4686 2018-04-09 Martin Sebor <msebor@redhat.com>
4688 * invoke.texi (-finline-small-functions): Mention other optimization
4690 (-findirect-inlining, -fpartial-inlining): Same.
4691 (-finline-functions-called-once): Same.
4692 (-freorder-blocks-and-partition): Same.
4694 2018-04-09 Jan Hubicka <jh@suse.cz>
4697 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
4698 jumps; choose last target that matches the criteria (i.e.
4699 no partition changes for non-crossing jumps).
4700 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
4701 support for redirecting crossing jumps to non-crossing.
4703 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
4705 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
4706 also for naked functions.
4708 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
4710 * config/arc/arc.md (add_shift): New pattern.
4711 (add_shift2): Likewise.
4712 (sub_shift): Likewise.
4713 (sub_shift_cmp0_noout): Likewise.
4714 (compare_si_ashiftsi): Likewise.
4715 (xbfu_cmp0_noout): New combine pattern.
4716 (xbfu_cmp0"): Likewise.
4717 (movsi_set_cc_insn): Place the predicable variant first.
4718 (commutative_binary_cmp0_noout): Remove clobber.
4719 (commutative_binary_cmp0): New pattern.
4720 (noncommutative_binary_cmp0): Likewise.
4721 (noncommutative_binary_cmp0_noout): Likewise.
4722 (noncommutative_binary_comparison_result_used): Removed.
4723 (rsub_cmp0): New pattern.
4724 (rsub_cmp0_noout): Likewise.
4725 (extzvsi): Changed, keep only meaningful variants.
4726 (SQH, SEZ): New iterators.
4727 (SQH_postfix): New mode attribute.
4728 (SEZ_prefix): New code attribute.
4729 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
4730 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
4731 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
4733 (noncommutative_operator): Check the availability of barrel
4736 2018-04-09 Richard Biener <rguenther@suse.de>
4738 PR tree-optimization/85284
4739 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
4740 Only use the niter constraining form of simple_iv when the exit
4743 2018-04-09 Tom de Vries <tom@codesourcery.com>
4746 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
4747 (define_expand "*memory_barrier"): New define_expand.
4748 (define_insn "memory_barrier"): New insn.
4750 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4752 PR rtl-optimization/80463
4753 PR rtl-optimization/83972
4754 PR rtl-optimization/83480
4756 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
4757 correct producer for the insn.
4758 (tidy_control_flow): Fixup seqnos in case of debug insns.
4760 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4762 PR rtl-optimization/83913
4764 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
4765 different sched-times when merging exprs.
4767 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4769 PR rtl-optimization/83962
4771 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
4772 tidy_fallthru_edge and tidy_control_flow.
4774 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4776 PR rtl-optimization/83530
4778 * sel-sched.c (force_next_insn): New global variable.
4779 (remove_insn_for_debug): When force_next_insn is true, also leave only
4780 next insn in the ready list.
4781 (sel_sched_region): When the region wasn't scheduled, make another pass
4782 over it with force_next_insn set to 1.
4784 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
4786 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
4788 * config/nds32/constants.md (unspec_volatile_element): Add enum values
4789 for interrupt control.
4790 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
4791 functions for interrupt control.
4792 * config/nds32/nds32-intrinsic.md: Likewise.
4793 * config/nds32/nds32_intrinsic.h: Likewise.
4794 * config/nds32/nds32.h (nds32_builtins): Likewise.
4796 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
4798 * config/nds32/nds32.c (nds32_init_machine_status,
4799 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
4800 strict_aligned_p field.
4801 (nds32_expand_to_rtl_hook): New function.
4802 (TARGET_EXPAND_TO_RTL_HOOK): Define.
4803 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
4805 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4806 Chung-Ju Wu <jasonwucj@gmail.com>
4808 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
4809 * config/nds32/nds32-n7.md: New file.
4810 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
4811 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
4813 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
4814 * config/nds32/nds32.md (pipeline_model): Add n7.
4815 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
4816 * config/nds32/pipelines.md: Include n7 settings.
4818 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4819 Chung-Ju Wu <jasonwucj@gmail.com>
4821 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
4822 * config/nds32/nds32-e8.md: New file.
4823 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
4824 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
4826 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
4827 * config/nds32/nds32.md (pipeline_model): Add e8.
4828 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
4829 * config/nds32/pipelines.md: Include e8 settings.
4831 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4832 Chung-Ju Wu <jasonwucj@gmail.com>
4834 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
4835 * config/nds32/nds32-n8.md: New file.
4836 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
4837 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
4839 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
4840 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
4841 * config/nds32/nds32.md (pipeline_model): Add n8.
4842 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
4843 * config/nds32/pipelines.md: Include n8 settings.
4845 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4846 Chung-Ju Wu <jasonwucj@gmail.com>
4848 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
4849 * config/nds32/nds32-n9-2r1w.md: New file.
4850 * config/nds32/nds32-n9-3r2w.md: New file.
4851 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
4852 nds32_register_ports): New or modify for cpu n9.
4853 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
4855 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
4856 * config/nds32/nds32-utils.c: New file.
4857 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
4858 TARGET_MUL_SLOW): Define.
4859 * config/nds32/nds32.md (pipeline_model): New attribute.
4860 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
4861 New options that support cpu n9.
4862 * config/nds32/pipelines.md: Include n9 settings.
4863 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
4865 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
4867 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
4868 information if necessary.
4869 (output_cond_branch_compare_zero): Likewise.
4870 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
4871 (nds32_target_alignment): Refine for alignment.
4872 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
4873 (FUNCTION_BOUNDARY): Modify.
4874 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
4876 * config/nds32/nds32.opt (malways-align, malign-functions): New.
4878 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
4880 * config/nds32/constants.md (unspec_volatile_element): Add values for
4881 TLB operation and data prefetch.
4882 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
4883 functions for TLB operation and data prefetch.
4884 * config/nds32/nds32-intrinsic.md: Likewise.
4885 * config/nds32/nds32_intrinsic.h: Likewise.
4886 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
4887 (nds32_print_operand): Likewise.
4888 * config/nds32/nds32.h (nds32_builtins): Likewise.
4890 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
4891 Andrew Pinski <pinsika@gcc.gnu.org>
4894 * match.pd: Use constant_boolean_node of correct type instead of
4895 boolean_true_node or boolean_false_node for simplifying
4896 pointer comparisons to zero.
4898 2018-04-07 Jakub Jelinek <jakub@redhat.com>
4900 PR tree-optimization/80021
4901 * tree.c (verify_type_variant): Make error call in verify_variant_match
4902 translatable and remove final full stop.
4904 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4906 * config/nds32/constants.md (unspec_volatile_element): Add
4907 UNSPEC_VOLATILE_EH_RETURN.
4908 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
4909 nds32_output_stack_pop): Support dwarf exception handling process.
4910 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
4911 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
4912 exception handling process.
4913 (nds32_compute_stack_frame): Likewise.
4914 (nds32_return_addr_rtx): Likewise.
4915 (nds32_initial_elimination_offset): Likewise.
4916 (nds32_expand_prologue): Likewise.
4917 (nds32_expand_epilogue): Likewise.
4918 (nds32_dynamic_chain_address): New function.
4919 * config/nds32/nds32.h (machine_function): Add fields for dwarf
4921 (DYNAMIC_CHAIN_ADDRESS): Define.
4922 (EH_RETURN_DATA_REGNO): Define.
4923 (EH_RETURN_STACKADJ_RTX): Define.
4924 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
4925 patterns for dwarf exception handling.
4927 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4929 * config/nds32/nds32.h: Clean up obsolete macros.
4931 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4933 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4934 Add enum values for particular instructions.
4935 * config/nds32/nds32-intrinsic.c: Implementation of expanding
4936 particular intrinsic functions.
4937 * config/nds32/nds32-intrinsic.md: Likewise.
4938 * config/nds32/nds32_intrinsic.h: Likewise.
4939 * config/nds32/nds32.h (nds32_builtins): Likewise.
4940 * config/nds32/nds32.md (type): Add pbsad and pbsada.
4941 (btst, ave): New patterns for particular instructions.
4943 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4945 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4946 Add enum values for atomic load/store and memory sync.
4947 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
4949 * config/nds32/nds32-intrinsic.md: Likewise.
4950 * config/nds32/nds32_intrinsic.h: Likewise.
4951 * config/nds32/nds32.h (nds32_builtins): Likewise.
4953 2018-04-07 Jakub Jelinek <jakub@redhat.com>
4955 PR tree-optimization/85257
4956 * fold-const.c (native_encode_vector): If not all elts could fit
4957 and off is -1, return 0 rather than offset.
4958 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
4959 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
4960 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
4961 adjust buffer in native_interpret_expr call.
4963 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4965 * config/nds32/constants.md (unspec_volatile_element): Add cache
4966 control enum values.
4967 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
4968 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
4969 * config/nds32/nds32.c (nds32_cctl_names): New.
4970 (nds32_print_operand): Handle cache control register names.
4971 * config/nds32/nds32.h (nds32_builtins): New enum values.
4972 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
4974 * config/nds32/nds32.md (type): Add mmu.
4975 * config/nds32/pipelines.md (simple_insn): Add mmu.
4977 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4979 * config/nds32/nds32.md (type): Remove call.
4980 * config/nds32/pipelines.md (simple_insn): Likewise.
4982 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4984 * config/nds32/constants.md (unspec_volatile_element): Add
4985 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
4986 UNSPEC_VOLATILE_FMFCFG.
4987 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
4988 description for fmfcfg and fmfcsr.
4989 (bdesc_1arg): Add fmtcsr.
4990 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
4991 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
4992 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
4993 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
4994 unspec_fmfcfg): New patterns.
4995 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
4996 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
4997 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
4998 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
4999 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
5000 __nds32__fmfcfg): Define.
5002 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5004 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
5005 intrinsic register names.
5006 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
5007 intrinsic register enum values and macros.
5009 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5011 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
5012 for load/store addressing form.
5013 (nds32_print_operand_address): Likewise.
5015 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
5018 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
5019 based on LABEL_REF. Remove useless assertion.
5020 (pic_address_needs_scratch): Fix formatting.
5021 (sparc_legitimize_pic_address): Minor tweaks.
5022 (sparc_delegitimize_address): Adjust assertion accordingly.
5023 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
5024 into symbolic_operand.
5025 (movsi_high_pic_label_ref): Likewise.
5026 (movsi_lo_sum_pic_label_ref): Likewise.
5027 (movdi_pic_label_ref): Likewise.
5028 (movdi_high_pic_label_ref): Likewise.
5029 (movdi_lo_sum_pic_label_ref): Likewise.
5031 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
5033 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
5034 custom LIB_SPEC setup.
5036 2018-04-06 Ruslan Bukin <br@bsdpad.com>
5037 Kito Cheng <kito.cheng@gmail.com>
5039 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
5040 * config/riscv/freebsd.h: New.
5042 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
5044 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
5045 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
5048 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
5049 Kito Cheng <kito.cheng@gmail.com>
5051 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
5052 nds32_output_call, nds32_symbol_binds_local_p): New functions.
5053 * config/nds32/nds32-protos.h (nds32_output_call,
5054 nds32_output_return): Declare.
5055 * config/nds32/nds32.md: Refine all the call and return patterns.
5057 2018-04-06 Jakub Jelinek <jakub@redhat.com>
5060 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
5061 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
5063 PR rtl-optimization/84872
5064 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
5065 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
5068 2018-04-06 Tamar Christina <tamar.christina@arm.com>
5070 * expr.c (copy_blkmode_to_reg): Revert 254862.
5071 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
5073 2018-04-06 Richard Biener <rguenther@suse.de>
5076 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
5077 after seeing a component reference with an adjacent field. Treat
5078 refs to arrays at struct end of external decls similar to
5079 refs to unconstrained commons.
5081 2018-04-06 Jakub Jelinek <jakub@redhat.com>
5084 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
5085 look through SAVE_EXPRs with non-side-effects argument. Adjust
5087 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
5090 2018-04-06 Richard Biener <rguenther@suse.de>
5093 * alias.c (find_base_term): New wrapper around find_base_term
5094 unwinding CSELIB_VAL_PTR changes.
5095 (find_base_term): Do not restore CSELIB_VAL_PTR during the
5098 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5100 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
5102 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
5103 constant definitions.
5104 ("nop"): lr 0,0 -> nopr r0
5105 ("nop_lr0", "nop_lr1"): New insn definitions.
5107 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
5109 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
5110 NDS32_V3PUSH_AVAILABLE_P macro.
5112 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
5113 Chung-Ju Wu <jasonwucj@gmail.com>
5115 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
5116 (nds32*-*-*): Add float and fpu_config into supported_defaults.
5117 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
5118 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
5119 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
5120 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
5121 * config/nds32/constraints.md: New constraints and checking for hard
5122 float configuration.
5123 * config/nds32/iterators.md: New mode iterator and attribute for hard
5124 float configuration.
5125 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
5127 * config/nds32/nds32-fpu.md: New file.
5128 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
5129 deal with hard float code generation.
5130 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
5132 (abi_type, float_reg_number): New enum type.
5133 * config/nds32/nds32-predicates.c: New predicates for hard float.
5134 * config/nds32/nds32-protos.h: Declare functions for hard float.
5135 * config/nds32/nds32.c: Implementation for hard float configuration.
5136 * config/nds32/nds32.h: Definitions for hard float configuration.
5137 * config/nds32/nds32.md: Include hard float machine description and
5138 modify patterns for hard float configuration.
5139 * config/nds32/nds32.opt: New options for hard float configuration.
5140 * config/nds32/predicates.md: New predicates for hard float
5143 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5145 * common/config/nds32/nds32-common.c
5146 (nds32_option_optimization_table): Enable -mreleax-hint by default.
5148 2018-04-05 Jakub Jelinek <jakub@redhat.com>
5151 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
5152 CONSTRUCTOR_ELT (ctor, ...)->value.
5154 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
5157 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
5159 2018-04-05 Tom de Vries <tom@codesourcery.com>
5162 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
5165 2018-04-05 Shiva Chen <shiva0217@gmail.com>
5166 Kito Cheng <kito.cheng@gmail.com>
5168 * config/nds32/constraints.md (U33): Fine-tune checking condition.
5169 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
5170 * config/nds32/nds32.h (nds32_16bit_address_type): Add
5171 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
5173 2018-04-05 Shiva Chen <shiva0217@gmail.com>
5174 Kito Cheng <kito.cheng@gmail.com>
5176 * config/nds32/constraints.md (Ufe): New memory constraint.
5177 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
5178 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
5179 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
5181 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
5182 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
5184 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5186 * config/nds32/nds32.md: Use optimize_size in the condition for
5187 alu-shift instructions.
5189 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5191 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
5193 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5195 * config/nds32/nds32.md (negsi2): Refine pattern.
5197 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
5198 Chung-Ju Wu <jasonwucj@gmail.com>
5200 * config/nds32/iterators.md (shift_rotate): New code iterator.
5201 (shift): New code attribute.
5202 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
5203 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
5204 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
5205 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
5206 bit-wise operations.
5207 (andsi3, *andsi3): Ditto.
5208 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
5209 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
5210 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
5211 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
5212 nds32_ior_operand, nds32_xor_operand): New predicates.
5214 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5216 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
5217 (addsi3, subsi3): ... this.
5219 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5221 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
5223 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5225 * config/nds32/nds32.md: Adjust indention.
5227 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
5229 * config/nds32/nds32.md (feature): New attribute.
5231 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5233 * config/nds32/nds32.md (subtype): New attribute.
5235 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
5238 * config/arm/arm-builtins.c (arm_expand_builtin): Change
5239 expansion to perform a bitwise AND of the argument followed by a
5240 boolean negation of the result.
5242 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
5244 PR rtl-optimization/84878
5245 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
5246 the basic block. Assert the use reference is not artificial and that
5247 it has an associated insn.
5249 2018-04-04 Michael Matz <matz@suse.de>
5251 * builtins.c (compute_objsize): Pass correct operand
5252 to array_at_struct_end_p.
5254 2018-04-04 Richard Biener <rguenther@suse.de>
5257 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
5258 from contexts for DINFO_LEVEL_TERSE and below.
5260 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
5262 * config/nds32/nds32-doubleword.md (move_<mode>): Require
5263 resiter_operand condition.
5264 * config/nds32/nds32.md (*move<mode>): Ditto.
5266 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
5267 Monk Chiang <sh.chiang04@gmail.com>
5269 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
5271 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5273 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
5275 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5276 Kito Cheng <kito.cheng@gmail.com>
5278 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
5279 nds32_cond_code_str, output_cond_branch,
5280 output_cond_branch_compare_zero, nds32_expand_cbranch,
5281 nds32_expand_cstore, nds32_expand_movcc,
5282 nds32_output_cbranchsi4_equality_zero,
5283 nds32_output_cbranchsi4_equality_reg,
5284 nds32_output_cbranchsi4_equality_reg_or_const_int,
5285 nds32_output_cbranchsi4_greater_less_zero: New functions.
5286 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
5287 nds32_expand_cstore, nds32_expand_movcc,
5288 nds32_output_cbranchsi4_equality_zero,
5289 nds32_output_cbranchsi4_equality_reg,
5290 nds32_output_cbranchsi4_equality_reg_or_const_int,
5291 nds32_output_cbranchsi4_greater_less_zero): Declare.
5292 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
5293 nds32_rimm11s_operand): New predicates.
5294 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
5295 * config/nds32/nds32.md: Rewrite all the branch and conditional move
5298 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
5300 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
5301 * config/nds32/nds32.md: Ditto.
5302 * config/nds32/pipelines.md: Ditto.
5304 2018-04-04 Richard Biener <rguenther@suse.de>
5306 PR tree-optimization/85168
5307 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
5308 propagating abnormals.
5310 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5312 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
5314 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5315 Kito Cheng <kito.cheng@gmail.com>
5317 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
5318 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
5319 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
5320 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
5321 * config/nds32/nds32.md (sibcall_internal): New.
5322 (sibcall_register): Remove.
5323 (sibcall_immediate): Remove.
5324 (sibcall_value_internal): New.
5325 (sibcall_value_register): Remove.
5326 (sibcall_value_immediate): Remove.
5327 * config/nds32/predicates.md (nds32_general_register_operand): New.
5328 (nds32_call_address_operand): New.
5330 2018-04-03 Jakub Jelinek <jakub@redhat.com>
5332 PR rtl-optimization/85167
5333 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
5334 bb_defs if *split_p, instead preinitialize it to NULL.
5336 PR tree-optimization/85156
5337 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
5338 evaluating the argument multiple times.
5340 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
5342 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
5344 (_mm_cvtpd_ps): Likewise.
5345 (_mm_cvttpd_epi32): Likewise.
5346 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
5347 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
5348 vector, pixel, and bool following altivec.h include.
5350 2018-04-03 Martin Sebor <msebor@redhat.com>
5352 * doc/extend.texi (Common Function Attributes): Clarify.
5353 (const attribute): Likewise.
5354 (pure attribute): Likewise.
5356 2018-04-03 Jakub Jelinek <jakub@redhat.com>
5359 * config/i386/i386.c (ix86_expand_vector_set): Use
5360 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
5362 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
5364 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
5365 instructions when changing rounding bits to preserve precision bits
5366 in the x87 control word.
5368 2018-04-03 Martin Liska <mliska@suse.cz>
5370 PR tree-optimization/82491
5371 * rtl.h (strip_offset_and_add): Replace += suboffset with
5372 poly_uint64 () + suboffset.
5374 2018-03-29 Martin Liska <mliska@suse.cz>
5375 Martin Jambor <mjambor@suse.cz>
5378 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
5379 param_type is not an integral or pointer type.
5381 2018-04-03 Richard Biener <rguenther@suse.de>
5383 * sese.h (recompute_all_dominators): Remove.
5385 2018-04-02 Martin Sebor <msebor@redhat.com>
5387 * doc/invoke.texi (-Wrestrict): Fix typos.
5389 2018-04-02 Jim Wilson <jimw@sifive.com>
5391 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
5392 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
5393 (<optab>di3, <optab>si3_extend): Likewise.
5394 (<optab>si3_mask, <optab>si3_mask_1): New.
5395 (<optab>di3_mask, <optab>di3_mask_1): New.
5396 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
5397 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
5398 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
5400 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
5402 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
5405 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
5407 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
5408 (nds32_canonicalize_comparison): New function.
5410 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5411 Kito Cheng <kito.cheng@gmail.com>
5412 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5414 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
5415 * config/nds32/constants.md (unspec_volatile_element): Add
5416 UNSPEC_VOLATILE_RELAX_GROUP.
5417 * config/nds32/nds32-relax-opt.c: New file.
5418 * config/nds32/nds32-predicates.c
5419 (nds32_symbol_load_store_p): New function.
5420 * config/nds32/nds32-protos.h
5421 (nds32_symbol_load_store_p): Declare function.
5422 (make_pass_nds32_relax_opt): Declare new rtl pass function.
5423 * config/nds32/nds32.c
5424 (nds32_register_pass): New function to register pass.
5425 (nds32_register_passes): New function to register passes.
5426 * config/nds32/nds32.md (relax_group): New pattern.
5427 * config/nds32/nds32.opt (mrelax-hint): New option.
5428 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
5430 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
5432 * config/nds32/t-nds32: Modify files dependency.
5434 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5436 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
5437 (PROFILE_HOOK): Define its implementation.
5439 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5441 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
5442 type and 32-bit size.
5444 2018-04-01 Jakub Jelinek <jakub@redhat.com>
5447 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
5448 (V_128_256): New mode iterator.
5449 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
5450 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
5451 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
5453 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
5454 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
5456 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
5459 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
5460 NaN inputs correctly.
5462 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
5465 * config/rs6000/vsx.md (??r): New mode attribute.
5466 (*vsx_mov<mode>_64bit): Use it.
5467 (*vsx_mov<mode>_32bit): Likewise.
5469 2018-03-30 Martin Sebor <msebor@redhat.com>
5471 PR tree-optimization/84818
5472 * builtins.c (check_access): Use warning_n.
5474 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5477 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
5479 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
5482 2018-03-30 Julia Koval <julia.koval@intel.com>
5485 * x86-tune.def (movx, partial_reg_dependency): Enable for
5488 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
5491 * lra-constraints.c (process_alt_operands): Move setting
5492 this_alternative_matches below.
5494 2018-03-29 Martin Liska <mliska@suse.cz>
5497 * doc/invoke.texi: Document how LTO works with debug info.
5498 Describe auto-load support of binutils. Mention 'x86-64'
5499 as valid option value of -march option.
5501 2018-03-29 Jakub Jelinek <jakub@redhat.com>
5503 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
5506 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
5507 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
5508 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
5511 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
5514 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
5515 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
5516 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
5517 for RS6000_BTM_POWERPC64.
5518 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
5519 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
5520 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
5525 2018-03-28 Carl Love <cel@us.ibm.com>
5528 2017-09-27 Carl Love <cel@us.ibm.com>
5530 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
5531 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
5532 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
5535 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5537 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
5538 instead of __vector bool.
5539 (_mm_max_pu8): Likewise.
5540 (_mm_min_pi16): Likewise.
5542 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
5545 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
5546 (DIVWEUO): Likewise.
5548 (DIVDEUO): Likewise.
5549 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
5550 DIVWEUO and DIVDEUO.
5551 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
5552 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
5553 (div_extend): Likewise.
5554 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
5556 (__builtin_divweuo): Likewise.
5557 (__builtin_divdeo): Likewise.
5558 (__builtin_divdeuo): Likewise.
5560 2018-03-28 Jakub Jelinek <jakub@redhat.com>
5563 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
5564 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
5566 PR tree-optimization/82004
5567 * gimple-match-head.c (optimize_pow_to_exp): New function.
5568 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
5569 Don't fold to exp if optimize_pow_to_exp is false.
5571 2018-03-28 Martin Liska <mliska@suse.cz>
5574 * calls.c (initialize_argument_information): Fix trailing space.
5575 * common.opt: Fix typo and provide better explanation for
5576 -fsanitize-coverage option.
5577 * config/i386/i386.opt: Fix typo.
5579 2018-03-28 Jakub Jelinek <jakub@redhat.com>
5580 Martin Liska <mliska@suse.cz>
5583 * gimplify.c (asan_poison_variable): Don't do the check for
5584 gimplify_omp_ctxp here.
5585 (gimplify_decl_expr): Do it here.
5586 (gimplify_target_expr): Likewise.
5588 2018-03-28 Martin Liska <mliska@suse.cz>
5591 * config/i386/i386.c (ix86_function_arg_advance): Do not call
5592 chkp_type_bounds_count if MPX is not enabled.
5594 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
5596 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
5598 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
5601 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
5602 function to create the function decl for complex long double
5603 multiply and divide for -mabi=ieeelongdouble.
5604 (init_float128_ieee): Call it.
5606 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
5609 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
5610 -fcf-protection=branch -mibt.
5611 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
5613 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5616 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
5618 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
5621 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
5622 extern array declarations.
5624 2018-03-27 Richard Biener <rguenther@suse.de>
5627 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
5628 explicit single_use checks.
5630 2018-03-27 Richard Biener <rguenther@suse.de>
5632 PR tree-optimization/85082
5633 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
5636 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5638 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
5639 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
5640 Turn on fasynchronous-unwind-tables and funwind-tables.
5642 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
5645 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
5646 (*bmi_blsr_<mode>_ccz): Ditto.
5648 2018-03-26 Tom de Vries <tom@codesourcery.com>
5650 PR tree-optimization/85063
5651 * omp-general.c (offloading_function_p): New function. Factor out
5653 * omp-offload.c (pass_omp_target_link::gate): ... here.
5654 * omp-general.h (offloading_function_p): Declare.
5655 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
5656 with attribute omp declare target for offloading functions.
5658 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
5660 PR tree-optimization/84005
5661 * tree-data-ref.h (get_base_for_alignment): Declare.
5662 * tree-data-ref.c (get_base_for_alignment_1): New function.
5663 (get_base_for_alignment): Likewise.
5664 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
5665 get_base_for_alignment to find a suitable base object, instead
5666 of always using drb->base_address.
5668 2018-03-23 Jakub Jelinek <jakub@redhat.com>
5671 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
5672 known size by default.
5674 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
5677 * lra-constraints.c (process_alt_operands): Don't match BLKmode
5678 and non BLKmode operands.
5680 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5683 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
5684 Clean up attributes.
5686 2018-03-23 Richard Biener <rguenther@suse.de>
5689 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
5690 we are going to emit early debug for LTO.
5692 2018-03-23 Jakub Jelinek <jakub@redhat.com>
5695 * function.c (match_asm_constraints_1): Don't optimize if input
5696 doesn't satisfy general_operand predicate for output's mode.
5699 * alias.c (write_dependence_p): Don't require for x_canonicalized
5700 non-VOIDmode if x has VOIDmode.
5703 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
5704 just don't try to optimize it rather than assert it never happens.
5706 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
5708 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
5709 macro expansions for definition of ST_INTERNAL_<mode> and
5710 LD_INTERNAL_<mode> builtins.
5711 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
5713 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
5715 (altivec_expand_st_builtin): Likewise.
5716 (altivec_expand_builtin): Remove calls to deleted functions.
5717 (rs6000_address_for_altivec): Delete this function.
5718 * config/rs6000/vector.md: Remove expands for
5719 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
5721 2018-03-22 Sudakshina Das <sudi.das@arm.com>
5724 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
5725 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
5726 re-computing once computed.
5727 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
5728 (arm_init_machine_status): Initialize
5729 machine->static_chain_stack_bytes.
5731 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
5734 * doc/extend.texi: Add four new prototypes for vec_ld.
5735 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
5736 definitions for more logical presentation.
5737 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
5738 entries for V1TI variants of __builtin_altivec_ld builtin.
5739 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
5740 handling of V1TI variant of LVX icode pattern.
5741 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
5742 (rs6000_gimple_fold_builtin): Likewise.
5743 (altivec_init_builtins): Add code to define
5744 __builtin_altivec_lvx_v1ti function.
5746 2018-03-22 Jakub Jelinek <jakub@redhat.com>
5749 * function.c (match_asm_constraints_1): Don't do the optimization
5750 if input isn't a REG, SUBREG, MEM or constant.
5752 2018-03-22 Tom de Vries <tom@codesourcery.com>
5754 PR tree-optimization/84956
5755 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
5756 bb_has_abnormal_pred.
5758 2018-03-22 Jakub Jelinek <jakub@redhat.com>
5761 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
5762 DECL_INITIAL (decl) to decl at the end.
5763 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
5766 2018-03-21 Joseph Myers <joseph@codesourcery.com>
5768 * doc/extend.texi (__builtin_tgmath): Document when complex
5769 integer types are treated as _Complex _Float64.
5771 2018-03-21 Tom de Vries <tom@codesourcery.com>
5773 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
5775 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5777 PR tree-optimization/84960
5778 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
5779 if it is ENTRY block, move them into single succ of ENTRY in that case.
5781 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
5783 PR tree-optimization/84811
5784 * poly-int.h (poly_span_traits): Remove the T3 parameter and
5785 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
5786 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
5787 (known_subrange_p): Update accordingly. Cast each value involved
5788 in the size comparison, rather than casting the result of the
5791 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5793 PR tree-optimization/84982
5794 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
5795 by flipping the least significant bit rather than all bits from
5796 bitpos to bitpos + bitsize - 1.
5798 2018-03-21 Nathan Sidwell <nathan@acm.org>
5800 * doc/extend.texi (Deprecated Features): Remove mention of
5801 long-deleted deprecations.
5803 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5806 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
5807 * configure: Regenerate.
5809 2018-03-21 Tom de Vries <tom@codesourcery.com>
5811 PR tree-optimization/83126
5812 * tree-parloops.c (num_phis): New function.
5813 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
5815 2018-03-21 Nathan Sidwell <nathan@acm.org>
5817 * doc/extend.texi (Deprecated Features): Update deprecated flags,
5818 mention anon-struct/union members and trailing attributes.
5820 2018-03-21 Bin Cheng <bin.cheng@arm.com>
5822 PR tree-optimization/84969
5823 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
5824 builtin memset partitions if they set different rhs values.
5826 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5828 PR rtl-optimization/84989
5829 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
5830 VEC_DUPLICATE with scalar result mode.
5832 2018-03-21 Martin Liska <mliska@suse.cz>
5835 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
5836 not intended return statement.
5838 2018-03-21 Martin Liska <mliska@suse.cz>
5841 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
5842 (chkp_find_bound_slots_1): Limit number of iterations.
5844 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
5847 * Minor grammar fixes for x86 options.
5849 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5852 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
5853 holding REG_CFA_RESTORE notes, instead turn them into a USE.
5855 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
5858 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
5859 (altivec_lvx_<mode>_1op): Likewise.
5860 (altivec_stvx_<mode>_2op): Likewise.
5861 (altivec_stvx_<mode>_1op): Likewise.
5862 (altivec_lvx_<VM2:mode>): New define_expand.
5863 (altivec_stvx_<VM2:mode>): Likewise.
5864 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
5865 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
5866 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
5867 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
5868 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
5869 (rs6000_gen_lvx): Likewise.
5870 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
5871 (altivec_expand_stv_builtin): Likewise.
5872 (altivec_expand_builtin): Likewise.
5873 * config/rs6000/vector.md: Likewise.
5875 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5878 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
5881 2018-03-20 Richard Biener <rguenther@suse.de>
5884 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
5885 sign-conversions as zero, fall back to standard scalar_stmt
5888 2018-03-20 Martin Liska <mliska@suse.cz>
5891 * predict.c (rebuild_frequencies): Handle case when we have
5892 PROFILE_ABSENT, but flag_guess_branch_prob is false.
5894 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5897 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
5898 flag_section_anchors.
5899 * varasm.c (use_blocks_for_decl_p): Remove hack for
5900 dw2_force_const_mem.
5903 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
5905 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
5906 be created, use lowpart_subreg of operands[0] rather than operands[0]
5908 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
5909 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
5910 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
5911 and n constraint instead of aarch64_shift_imm_di and Usd.
5912 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
5913 (*aarch64_<optab>_reg_minus<mode>3): ... this.
5915 2018-03-20 Sudakshina Das <sudi.das@arm.com>
5918 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
5919 to favor GPR over NEON registers.
5920 (<shift>di3_neon): Likewise.
5922 2018-03-20 Tom de Vries <tom@codesourcery.com>
5925 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
5926 (nvptx_process_pars): Emit bar.sync asap and alap.
5928 2018-03-20 Tom de Vries <tom@codesourcery.com>
5931 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
5932 seen_label if seen_label is already set.
5934 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5937 * config/i386/i386.c (fold_builtin_cpu): For features above 31
5938 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
5939 Use 1U instead of 1. Formatting fixes.
5942 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
5943 instead of TREE_TYPE (s1) for the return value.
5945 2018-03-19 Jakub Jelinek <jakub@redhat.com>
5947 PR tree-optimization/84946
5948 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
5949 bitsize + bitsize in poly_uint64 rather than poly_int64.
5952 * dwarf2asm.c: Include fold-const.c.
5953 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
5954 of decl rather than decl itself.
5956 PR rtl-optimization/84643
5957 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
5959 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
5962 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
5963 calling assemble_variable.
5965 2018-03-19 Sudakshina Das <sudi.das@arm.com>
5968 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
5969 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
5971 2018-03-19 Jim Wilson <jimw@sifive.com>
5974 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
5975 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
5976 (riscv_first_stack_step): Likewise.
5977 (riscv_option_override): Use STACK_BOUNDARY instead of
5979 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
5981 (BIGGEST_ALIGNMENT): Set to 128.
5982 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
5983 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
5986 2018-03-19 Richard Biener <rguenther@suse.de>
5988 PR tree-optimization/84933
5989 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
5990 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
5992 2018-03-19 Richard Biener <rguenther@suse.de>
5994 PR tree-optimization/84859
5995 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
5996 (cond_if_else_store_replacement): Perform sinking operation on
5997 single-store BBs regardless of MAX_STORES_TO_SINK setting.
5998 Generalize what a BB with a single eligible store is.
6000 2018-03-19 Richard Biener <rguenther@suse.de>
6002 PR tree-optimization/84929
6003 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
6004 chrec_is_positive against non-chrec arg.
6006 2018-03-19 Tamar Christina <tamar.christina@arm.com>
6009 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
6011 2018-03-18 Martin Liska <mliska@suse.cz>
6013 PR rtl-optimization/84635
6014 * regrename.c (build_def_use): Use matches_mode only when
6017 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
6019 PR tree-optimization/84913
6020 * tree-vect-loop.c (vectorizable_reduction): Don't try to
6021 vectorize chains of COND_EXPRs.
6023 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
6025 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
6027 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
6029 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
6031 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
6033 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
6035 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
6036 Kito Cheng <kito.cheng@gmail.com>
6038 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
6039 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
6040 (nds32_adjust_reg_alloc_order): New function.
6041 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
6043 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
6045 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
6046 nds32_print_operand, nds32_print_operand_address): Use
6047 HOST_WIDE_INT_PRINT_DEC instead.
6049 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
6051 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
6053 2018-03-17 Jakub Jelinek <jakub@redhat.com>
6056 * config/i386/i386.c (initial_ix86_tune_features,
6057 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
6059 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
6060 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
6061 rather than 1u << ix86_tune. Formatting fix.
6062 (ix86_option_override_internal): Change ix86_arch_mask from
6063 unsigned int to unsigned HOST_WIDE_INT, initialize to
6064 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
6065 (ix86_function_specific_restore): Likewise.
6067 2018-03-16 Jakub Jelinek <jakub@redhat.com>
6070 * postreload.c (reload_combine_recognize_pattern): Perform
6071 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
6072 truncate_int_for_mode the result for the destination's mode.
6075 * hsa-gen.c (mem_type_for_type): Fix comment typo.
6076 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
6078 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
6081 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
6084 * lra-assigns.c (lra_split_hard_reg_for): Don't use
6085 regno_allocno_class_array and sorted_pseudos.
6086 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
6087 insns where regno is used.
6089 2018-03-16 Martin Liska <mliska@suse.cz>
6092 * multiple_target.c (create_dispatcher_calls): Redirect
6093 reference in the symbol table.
6095 2018-03-16 Martin Liska <mliska@suse.cz>
6098 * multiple_target.c (create_dispatcher_calls): Redirect also
6101 2018-03-16 Jakub Jelinek <jakub@redhat.com>
6105 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
6106 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
6107 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
6109 2018-03-16 Julia Koval <julia.koval@intel.com>
6111 * doc/invoke.texi (Skylake Server): Add CLWB.
6112 Cannonlake): Remove CLWB.
6114 2018-03-16 Jakub Jelinek <jakub@redhat.com>
6116 PR tree-optimization/84841
6117 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
6119 (FLOAT_ONE_CONST_TYPE): Define.
6120 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
6121 (sort_by_operand_rank): Put entries with higher constant_type last
6122 rather than first to match comments.
6124 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
6126 * config/nios2/nios2.md (movsi_internal): Fix thinko in
6129 2018-03-15 Jakub Jelinek <jakub@redhat.com>
6132 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
6133 check and use address of target always.
6135 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
6138 * config/i386/i386.c (indirect_thunk_needed): Update comments.
6139 (indirect_thunk_bnd_needed): Likewise.
6140 (indirect_thunks_used): Likewise.
6141 (indirect_thunks_bnd_used): Likewise.
6142 (indirect_return_needed): New.
6143 (indirect_return_bnd_needed): Likewise.
6144 (output_indirect_thunk_function): Add a bool argument for
6146 (output_indirect_thunk_function): Don't generate alias for
6147 function return thunk.
6148 (ix86_code_end): Call output_indirect_thunk_function to generate
6149 function return thunks.
6150 (ix86_output_function_return): Set indirect_return_bnd_needed
6151 and indirect_return_needed instead of indirect_thunk_bnd_needed
6152 and indirect_thunk_needed.
6154 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
6156 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
6157 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
6158 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
6160 2018-03-15 David Malcolm <dmalcolm@redhat.com>
6161 Paul Hua <paul.hua.gm@gmail.com>
6164 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
6166 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
6168 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
6169 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
6172 2018-03-15 Tamar Christina <tamar.christina@arm.com>
6175 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
6176 instead of GET_MODE_SIZE when comparing Units.
6178 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
6181 * varasm.c (hash_section): Return an unchangeble hash value
6182 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
6183 Return !aarch64_can_use_per_function_literal_pools_p ().
6185 2018-03-15 Jakub Jelinek <jakub@redhat.com>
6188 * optabs.c (emit_conditional_move): Pass address of cmode's copy
6189 rather than address of cmode as last argument to prepare_cmp_insn.
6191 2018-03-15 Julia Koval <julia.koval@intel.com>
6193 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
6194 F_AVX512VNNI, F_AVX512BITALG): New.
6196 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
6199 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
6200 insn for floating-point loads and stores.
6202 2018-03-14 Carl Love <cel@us.ibm.com>
6204 * config/rs6000/rs6000-c.c: Add macro definitions for
6205 ALTIVEC_BUILTIN_VEC_PERMXOR.
6206 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
6207 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
6208 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
6209 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
6211 * config/doc/extend.texi: Add prototypes for vec_permxor.
6213 2018-03-14 David Malcolm <dmalcolm@redhat.com>
6216 * diagnostic-show-locus.c (class layout_point): Convert m_line
6217 from int to linenum_type.
6218 (line_span::comparator): Use linenum "compare" function when
6219 comparing line numbers.
6220 (test_line_span): New function.
6221 (layout_range::contains_point): Convert param "row" from int to
6223 (layout_range::intersects_line_p): Likewise.
6224 (layout::will_show_line_p): Likewise.
6225 (layout::print_source_line): Likewise.
6226 (layout::should_print_annotation_line_p): Likewise.
6227 (layout::print_annotation_line): Likewise.
6228 (layout::print_leading_fixits): Likewise.
6229 (layout::annotation_line_showed_range_p): Likewise.
6230 (struct line_corrections): Likewise for field m_row.
6231 (line_corrections::line_corrections): Likewise for param "row".
6232 (layout::print_trailing_fixits): Likewise.
6233 (layout::get_state_at_point): Likewise.
6234 (layout::get_x_bound_for_row): Likewise.
6235 (layout::print_line): Likewise.
6236 (diagnostic_show_locus): Likewise for locals "last_line" and
6238 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
6239 * input.c (selftest::test_linenum_comparisons): New function.
6240 (selftest::input_c_tests): Call it.
6241 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
6242 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
6243 * selftest.h (ASSERT_GT): New macro.
6244 (ASSERT_GT_AT): New macro.
6245 (ASSERT_LT): New macro.
6246 (ASSERT_LT_AT): New macro.
6248 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
6250 PR rtl-optimization/84780
6251 * combine.c (distribute_links): Don't make a link based on pc_rtx.
6253 2018-03-14 Martin Liska <mliska@suse.cz>
6255 * tree.c (record_node_allocation_statistics): Use
6256 get_stats_node_kind.
6257 (get_stats_node_kind): New function extracted from
6258 record_node_allocation_statistics.
6259 (free_node): Use get_stats_node_kind.
6261 2018-03-14 Richard Biener <rguenther@suse.de>
6263 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
6264 that the value-set of ANTIC_IN doesn't grow.
6267 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
6269 (BB_VISITED_WITH_VISITED_SUCCS): New define.
6270 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
6272 2018-03-14 Julia Koval <julia.koval@intel.com>
6274 * config.gcc (icelake-client, icelake-server): New.
6276 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
6277 (initial_ix86_arch_features): Ditto.
6278 (PTA_SKYLAKE): Add SGX.
6279 (PTA_ICELAKE): Remove.
6280 (PTA_ICELAKE_CLIENT): New.
6281 (PTA_ICELAKE_SERVER): New.
6282 (ix86_option_override_internal): Split up icelake on icelake client and
6284 (get_builtin_code_for_version): Ditto.
6285 (fold_builtin_cpu): Ditto.
6286 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
6287 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
6288 * config/i386/i386.h (processor_type): Ditto.
6289 * doc/invoke.texi: Ditto.
6291 2018-03-14 Jakub Jelinek <jakub@redhat.com>
6294 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
6295 INTEGER_CST offset, add it together with bitpos / 8 and
6296 sign extend based on POINTER_SIZE.
6300 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
6303 * config/i386/constraints.md (Yc): New register constraint.
6304 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
6305 Use Yc constraint for alternative 2 of operand 0. Remove
6306 preferred_for_speed attribute.
6308 2018-03-14 Richard Biener <rguenther@suse.de>
6310 PR tree-optimization/84830
6311 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
6312 with the old one to avoid oscillations.
6314 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
6317 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
6319 (assign_by_spills): Return a flag of reload assignment failure.
6320 Do not process the reload assignment failures. Do not spill other
6321 reload pseudos if they has the same reg class. Update n if
6323 (lra_assign): Add a return arg. Set up from the result of
6324 assign_by_spills call.
6325 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6326 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6327 usage_insns if it is not NULL.
6328 (spill_hard_reg_in_range): New function.
6329 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6330 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6331 function prototypes.
6332 (lra_assign): Change prototype.
6333 * lra.c (lra): Add code to deal with fails by splitting hard reg
6336 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
6338 * config/riscv/riscv.opt (mrelax): New option.
6339 * config/riscv/riscv.c (riscv_file_start): Emit ".option
6340 "norelax" when riscv_mrelax is disabled.
6341 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
6343 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6346 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
6347 reassociation for int modes.
6349 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6351 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
6352 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
6354 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
6355 * config/aarch64/aarch64-sve.md
6356 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
6357 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
6358 (*extend<mode><Vwide>2): Rename to...
6359 (aarch64_sve_extend<mode><Vwide>2): ...this.
6360 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
6361 renaming the old pattern to...
6362 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
6364 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
6365 define_expand, renaming the old pattern to...
6366 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
6367 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
6368 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
6369 account when deciding which SVE instruction the optab should use.
6370 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
6372 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6374 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
6375 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
6376 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
6377 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
6378 (tlsdesc_small_<mode>): Turn a define_expand and use
6379 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
6380 (tlsdesc_small_advsimd_<mode>): ...this.
6381 (tlsdesc_small_sve_<mode>): New pattern.
6383 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6385 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
6386 (UNSPEC_UMUL_HIGHPART): New constants.
6387 (MUL_HIGHPART): New int iteraor.
6388 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
6389 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
6391 (*<su>mul<mode>3_highpart): New define_insn.
6393 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
6396 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
6399 2018-03-13 Martin Liska <mliska@suse.cz>
6402 * (sem_item_optimizer::sem_item_optimizer): Initialize new
6404 (sem_item_optimizer::~sem_item_optimizer): Release it.
6405 (sem_item_optimizer::merge_classes): Register variable aliases.
6406 (sem_item_optimizer::fixup_pt_set): New function.
6407 (sem_item_optimizer::fixup_points_to_sets): Likewise.
6408 * ipa-icf.h: Declare new variables and functions.
6410 2018-03-13 Jakub Jelinek <jakub@redhat.com>
6413 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
6414 integer_pow2p@2 and test integer_pow2p in condition.
6415 (A < 0 ? C : 0): Similarly for @1.
6418 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
6419 characters starting at p contain '\0' character, don't look beyond
6423 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
6424 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
6427 * reg-stack.c (change_stack): Change update_end var from int to
6428 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
6429 also call set_block_for_insn on the newly added insns and rescan.
6432 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
6433 on the last operand.
6436 * tree.c (stabilize_reference_1): Return save_expr (e) for
6437 STATEMENT_LIST even if it doesn't have side-effects.
6439 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
6441 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
6443 2018-03-12 Renlin Li <renlin.li@arm.com>
6445 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
6446 aarch64_output_scalar_simd_mov_immediate.
6448 2018-03-12 Martin Sebor <msebor@redhat.com>
6450 PR tree-optimization/83456
6451 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
6452 for perfectly overlapping calls to memcpy.
6453 (gimple_fold_builtin_memory_chk): Same.
6454 (gimple_fold_builtin_strcpy): Handle no-warning.
6455 (gimple_fold_builtin_stxcpy_chk): Same.
6456 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
6458 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
6460 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
6461 parameter. Use it for SFmode.
6462 (rs6000_function_arg_advance_1): Adjust.
6463 (rs6000_function_arg): Adjust.
6464 (rs6000_gimplify_va_arg): Pass false for that new parameter.
6466 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
6468 PR rtl-optimization/84169
6469 PR rtl-optimization/84780
6470 * combine.c (can_combine_p): Check for a 2-insn combination whether
6471 the destination register is used between the two insns, too.
6473 2018-03-12 Richard Biener <rguenther@suse.de>
6475 PR tree-optimization/84803
6476 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
6477 for refs DR analysis didn't process.
6479 2018-03-12 Richard Biener <rguenther@suse.de>
6481 PR tree-optimization/84777
6482 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
6483 force-vectorize loops ignore whether we are optimizing for size.
6485 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
6487 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
6488 (TARGET_MD_ASM_ADJUST): Define.
6490 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
6491 Kito Cheng <kito.cheng@gmail.com>
6492 Chung-Ju Wu <jasonwucj@gmail.com>
6494 * config/nds32/nds32.c (nds32_compute_stack_frame,
6495 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
6496 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
6497 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
6498 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
6499 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
6500 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
6501 * config/nds32/nds32.md (prologue, epilogue): Use macro
6502 NDS32_V3PUSH_AVAILABLE_P to do checking.
6504 2018-03-11 Jakub Jelinek <jakub@redhat.com>
6507 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
6508 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
6509 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
6510 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
6511 addition of most attributes on !orig_type_die or the attribute not
6512 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
6514 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6515 Chung-Ju Wu <jasonwucj@gmail.com>
6517 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
6519 * config/nds32/nds32.opt (mvh): New option.
6521 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6522 Chung-Ju Wu <jasonwucj@gmail.com>
6524 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
6526 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
6527 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
6530 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6531 Chung-Ju Wu <jasonwucj@gmail.com>
6533 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
6535 * config/nds32/nds32-multiple.md (strlensi): New pattern.
6536 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
6538 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
6539 Kito Cheng <kito.cheng@gmail.com>
6540 Chung-Ju Wu <jasonwucj@gmail.com>
6542 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
6543 UNSPEC_FFMISM and UNSPEC_FLMISM.
6544 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
6545 for ffb, ffmism and flmism.
6546 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
6547 (unspec_ffmism): Ditto.
6548 (unspec_flmism): Ditto.
6549 (nds32_expand_builtin_impl): Check if string extension is available.
6550 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
6551 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
6553 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
6556 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
6559 * lra-assigns.c (assign_by_spills): Return a flag of reload
6560 assignment failure. Do not process the reload assignment
6561 failures. Do not spill other reload pseudos if they has the same
6563 (lra_assign): Add a return arg. Set up from the result of
6564 assign_by_spills call.
6565 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6566 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6567 usage_insns if it is not NULL.
6568 (spill_hard_reg_in_range): New function.
6569 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6570 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6571 function prototypes.
6572 (lra_assign): Change prototype.
6573 * lra.c (lra): Add code to deal with fails by splitting hard reg
6576 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
6579 * config/i386/i386.opt: Replace Enforcment with Enforcement.
6581 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
6584 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
6585 (dw_val_node): Add val_symbolic_view.
6586 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
6587 (symview_upper_bound): New.
6588 (new_line_info_table): Initialize symviews_since_reset.
6589 (dwarf2out_source_line): Count symviews_since_reset and set
6590 symview_upper_bound.
6591 (dw_val_equal_p): Handle symview.
6592 (add_AT_symview): New.
6593 (print_dw_val): Handle symview.
6594 (attr_checksum, attr_checksum_ordered): Likewise.
6595 (same_dw_val_p, size_of_die): Likewise.
6596 (value_format, output_die): Likewise.
6597 (add_high_low_attributes): Use add_AT_symview for entry_view.
6598 (dwarf2out_finish): Reset symview_upper_bound, clear
6601 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
6604 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
6605 Add strict argument and use it.
6606 (rs6000_split_multireg_move): Update for new strict argument.
6607 (mem_operand_gpr): Disallow all non-offsettable addresses.
6608 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
6610 2018-03-09 Jakub Jelinek <jakub@redhat.com>
6613 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
6614 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
6615 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
6618 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
6619 decl, use remap_type if we want to use the type.
6621 2018-03-09 Martin Sebor <msebor@redhat.com>
6623 PR tree-optimization/84526
6624 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
6626 (builtin_access::generic_overlap): Be prepared to handle non-array
6629 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
6631 PR rtl-optimization/84682
6632 * lra-constraints.c (process_address_1): Check is_address flag
6633 for address constraints.
6634 (process_alt_operands): Likewise.
6635 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
6636 preprocess_constraints.
6637 * recog.h (preprocess_constraints): Add oploc parameter.
6639 * recog.c (preprocess_constraints): Test address_operand for
6640 CT_ADDRESS constraints.
6642 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
6645 * lra-assigns.c (assign_by_spills): Return a flag of reload
6646 assignment failure. Do not process the reload assignment
6647 failures. Do not spill other reload pseudos if they has the same
6649 (lra_assign): Add a return arg. Set up from the result of
6650 assign_by_spills call.
6651 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6652 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6653 usage_insns if it is not NULL.
6654 (spill_hard_reg_in_range): New function.
6655 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6656 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6657 function prototypes.
6658 (lra_assign): Change prototype.
6659 * lra.c (lra): Add code to deal with fails by splitting hard reg
6662 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6665 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
6666 Accept complain bool parameter. Only emit errors if it is true.
6667 (arm_parse_cpu_option_name): Likewise.
6668 (arm_target_thumb_only): Adjust callers of the above.
6669 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
6670 prototype to take a default true bool parameter.
6671 (arm_parse_arch_option_name): Likewise.
6673 2018-03-09 David Malcolm <dmalcolm@redhat.com>
6674 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
6678 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
6679 * configure: Regenerate.
6680 * configure.ac ("linker --version-script option"): New.
6681 ("linker soname option"): New.
6683 2018-03-09 Richard Biener <rguenther@suse.de>
6685 PR tree-optimization/84775
6686 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
6687 immediate uses of predicate stmts and mark them modified.
6690 PR tree-optimization/84178
6691 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
6693 (version_loop_for_if_conversion): Delay update_ssa call.
6694 (tree_if_conversion): Delay update_ssa until after predicate
6697 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
6700 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
6701 when the function accesses prior frames.
6703 2018-03-08 Jakub Jelinek <jakub@redhat.com>
6706 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
6707 gen_llsym, otherwise call maybe_gen_llsym.
6710 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
6711 has ',' character inside of it.
6713 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6716 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
6717 as clobbering CC_REGNUM.
6719 2018-03-08 Richard Biener <rguenther@suse.de>
6722 * tree-scalar-evolution.c: Include tree-into-ssa.h.
6723 (follow_copies_to_constant): Do not follow SSA names registered
6726 2018-03-08 Richard Biener <rguenther@suse.de>
6728 PR tree-optimization/84178
6729 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
6731 (version_loop_for_if_conversion): Delay update_ssa call.
6732 (tree_if_conversion): Delay update_ssa until after predicate
6735 2018-03-08 David Malcolm <dmalcolm@redhat.com>
6737 PR tree-optimization/84178
6738 * tree-if-conv.c (release_bb_predicate): Remove the
6739 the assertion that the stmts have NULL use_ops.
6740 Discard the statements, asserting that they haven't
6741 yet been added to a BB.
6743 2018-03-08 Richard Biener <rguenther@suse.de>
6745 PR tree-optimization/84746
6746 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
6747 (phi_translate): Pass in destination ANTIC_OUT set.
6748 (phi_translate_1): Likewise. For a simplified result lookup
6749 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
6750 (phi_translate_set): Adjust.
6751 (do_pre_regular_insertion): Likewise.
6752 (do_pre_partial_partial_insertion): Likewise.
6754 2018-03-08 Martin Liska <mliska@suse.cz>
6756 PR gcov-profile/84735
6757 * doc/gcov.texi: Document usage of profile files.
6758 * gcov-io.h: Document changes in the format.
6760 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
6764 * dwarf2out.c (struct dw_line_info_table): Update comments for
6766 (FORCE_RESET_NEXT_VIEW): New.
6767 (FORCE_RESETTING_VIEW_P): New.
6768 (RESETTING_VIEW_P): Check for -1 too.
6769 (ZERO_VIEW_P): Likewise.
6770 (new_line_info_table): Force-reset next view.
6771 (dwarf2out_begin_function): Likewise.
6772 (dwarf2out_source_line): Simplify zero_view_p initialization.
6773 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
6774 view directly. Omit view when omitting .loc at line 0.
6776 2018-03-08 Jakub Jelinek <jakub@redhat.com>
6778 PR tree-optimization/84740
6779 * tree-switch-conversion.c (process_switch): Call build_constructors
6780 only if info.phi_count is non-zero.
6782 PR tree-optimization/84739
6783 * tree-tailcall.c (find_tail_calls): Check call arguments against
6784 DECL_ARGUMENTS (current_function_decl) rather than
6785 DECL_ARGUMENTS (func) when checking for tail recursion.
6787 2018-03-07 Jakub Jelinek <jakub@redhat.com>
6789 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
6790 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
6791 Volker Reichelt's entry and add entries for people that perform
6792 GCC fuzzy testing and report numerous bugs.
6794 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
6797 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
6798 readonly data in sdata, if that is disabled.
6799 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
6800 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
6801 -mreadonly-in-sdata option.
6803 2018-03-07 Martin Sebor <msebor@redhat.com>
6805 PR tree-optimization/84468
6806 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
6807 basic block when looking for nul assignment.
6809 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
6812 * except.h (output_function_exception_table): Adjust prototype.
6813 * except.c (output_function_exception_table): Remove FNNAME parameter
6814 and add SECTION parameter. Ouput one part of the table at a time.
6815 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
6816 the first part of the exception table and emit unwind directives.
6817 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
6818 (i386_pe_seh_cold_init): Likewise.
6819 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
6820 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
6821 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
6822 (ix86_output_call_insn): Emit a nop in one more case for SEH.
6823 * config/i386/winnt.c: Include except.h.
6824 (struct seh_frame_state): Add reg_offset, after_prologue and
6825 in_cold_section fields.
6826 (i386_pe_seh_end_prologue): Set seh->after_prologue.
6827 (i386_pe_seh_cold_init): New function.
6828 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
6829 to seh->in_cold_section.
6830 (seh_emit_push): Record the offset of the push.
6831 (seh_emit_save): Record the offet of the save.
6832 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
6833 Test seh->after_prologue to disregard the epilogue.
6834 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
6835 (i386_pe_end_cold_function): New function.
6837 2018-03-07 Jakub Jelinek <jakub@redhat.com>
6840 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
6841 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
6844 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
6846 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
6847 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
6850 * multiple_target.c: Include tree-inline.h and intl.h.
6851 (expand_target_clones): Diagnose and fail if node->definition and
6852 !tree_versionable_function_p (node->decl).
6854 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
6856 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
6858 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
6859 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
6860 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
6862 2018-03-06 Jakub Jelinek <jakub@redhat.com>
6865 * combine.c (try_combine): Use reg_or_subregno instead of handling
6866 just paradoxical SUBREGs and REGs.
6868 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
6870 * config/arc/arc.c (arc_finalize_pic): Remove function.
6871 (arc_must_save_register): We use single base PIC register, remove
6872 checks to save/restore the PIC register.
6873 (arc_expand_prologue): Likewise.
6874 * config/arc/arc-protos.h (arc_set_default_type_attributes):
6876 (arc_verify_short): Likewise.
6877 (arc_attr_type): Likewise.
6878 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
6879 (walk_stores): Likewise.
6880 (arc_address_cost): Make it static.
6881 (arc_verify_short): Likewise.
6882 (branch_dest): Likewise.
6883 (arc_attr_type): Likewise.
6884 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
6885 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
6886 (arc_final_prescan_insn): Remove inserting the nops due to
6887 hardware hazards. It is done in reorg step.
6888 (insn_length_variant_t): Remove.
6889 (insn_length_parameters_t): Likewise.
6890 (arc_insn_length_parameters): Likewise.
6891 (arc_get_insn_variants): Likewise.
6892 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
6894 2018-03-06 Jakub Jelinek <jakub@redhat.com>
6897 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
6900 PR tree-optimization/84687
6901 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
6903 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
6905 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6907 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
6908 Rename to ppc_speculation_barrier.
6909 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
6910 __builtin_ppc_speculation_barrier.
6912 2018-03-05 Jakub Jelinek <jakub@redhat.com>
6915 * combine.c (combine_simplify_rtx): Don't try to simplify if
6916 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
6919 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
6921 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
6922 to 32 bytes when compiling for POWER9.
6924 2018-03-05 Jakub Jelinek <jakub@redhat.com>
6927 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
6928 regparm >= 3 with no arg reg available also for calls with
6929 flag_force_indirect_call. Pass decl to ix86_function_regparm.
6932 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
6934 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
6936 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
6939 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
6941 2018-03-05 Richard Biener <rguenther@suse.de>
6943 PR tree-optimization/84486
6944 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
6945 When inserting a __builtin_assume_aligned call set the LHS
6946 SSA name alignment info accordingly.
6948 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
6950 PR tree-optimization/84114
6951 * config/aarch64/aarch64.c (aarch64_reassociation_width)
6952 Avoid reassociation of FLOAT_MODE addition.
6954 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
6956 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
6957 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
6958 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
6959 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
6960 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
6961 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
6962 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
6964 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
6965 __builtin_ia32_wbinvd): New builtins.
6966 (SPECIAL_ARGS2): New.
6967 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
6968 (SPECIAL_ARGS2): New.
6969 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
6970 (ix86_valid_target_attribute_inner_p): Ditto.
6971 (ix86_init_mmx_sse_builtins): Add special_args2.
6972 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
6973 TARGET_WBNOINVD_P): New.
6974 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
6975 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
6976 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
6977 * config/i386/immintrin.h (_wbinvd): New intrinsic.
6978 * config/i386/pconfigintrin.h: New file.
6979 * config/i386/wbnoinvdintrin.h: Ditto.
6980 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
6981 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
6983 2018-03-05 Richard Biener <rguenther@suse.de>
6985 PR tree-optimization/84670
6986 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
6988 (BB_VISITED_WITH_VISITED_SUCCS): New define.
6989 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
6990 (compute_antic_aux): Only assert the number of values in ANTIC_IN
6991 doesn't grow if all successors (recursively) were visited at least
6994 2018-03-05 Richard Biener <rguenther@suse.de>
6996 PR tree-optimization/84650
6997 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
6998 if executed in the loop pipeline.
7000 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
7002 * doc/configfiles.texi (Configuration Files): Move info about
7003 conditionalizing $target-protos.h to...
7004 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
7005 differs from $target-protos.h.
7007 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
7008 Chung-Ju Wu <jasonwucj@gmail.com>
7010 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
7011 * config/nds32/nds32-multiple.md (setmemsi): Define.
7012 * config/nds32/nds32-memory-manipulation.c
7013 (nds32_gen_dup_4_byte_to_word_value): New.
7014 (emit_setmem_word_loop): New.
7015 (emit_setmem_byte_loop): New.
7016 (nds32_expand_setmem_loop): New.
7017 (nds32_expand_setmem_loop_v3m): New.
7018 (nds32_expand_setmem_unroll): New.
7019 (nds32_expand_setmem): New.
7021 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
7022 Chung-Ju Wu <jasonwucj@gmail.com>
7024 * config/nds32/nds32-memory-manipulation.c
7025 (nds32_emit_load_store): New.
7026 (nds32_emit_post_inc_load_store): New.
7027 (nds32_emit_mem_move): New.
7028 (nds32_emit_mem_move_block): New.
7029 (nds32_expand_movmemsi_loop_unknown_size): New.
7030 (nds32_expand_movmemsi_loop_known_size): New.
7031 (nds32_expand_movmemsi_loop): New.
7032 (nds32_expand_movmemsi_unroll): New.
7033 (nds32_expand_movmemqi): Rename ...
7034 (nds32_expand_movmemsi): ... to this.
7035 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
7036 (movmemsi): ... to this.
7037 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
7038 (nds32_expand_movmemsi): ... to this.
7040 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
7041 Monk Chiang <sh.chiang04@gmail.com>
7042 Chung-Ju Wu <jasonwucj@gmail.com>
7044 * config/nds32/nds32-protos.h
7045 (nds32_expand_load_multiple): New arguments.
7046 (nds32_expand_store_multiple): Ditto.
7047 (nds32_valid_multiple_load_store): Rename ...
7048 (nds32_valid_multiple_load_store_p): ... to this.
7049 * config/nds32/nds32-memory-manipulation.c
7050 (nds32_expand_load_multiple): Refine implementation.
7051 (nds32_expand_store_multiple): Ditto.
7052 * config/nds32/nds32-multiple.md
7053 (load_multiple): Update nds32_expand_load_multiple interface.
7054 (store_multiple): Update nds32_expand_store_multiple interface.
7055 * config/nds32/nds32-predicates.c
7056 (nds32_valid_multiple_load_store): Rename ...
7057 (nds32_valid_multiple_load_store_p): ... to this and refine
7059 * config/nds32/predicates.md
7060 (nds32_load_multiple_and_update_address_operation): New predicate.
7061 (nds32_store_multiple_and_update_address_operation): New predicate.
7063 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
7064 Chung-Ju Wu <jasonwucj@gmail.com>
7066 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
7067 (combo): New attribute.
7068 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
7070 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
7072 * config/nds32/nds32.opt: Change -mcmodel= default value.
7074 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
7075 Monk Chiang <sh.chiang04@gmail.com>
7076 Chung-Ju Wu <jasonwucj@gmail.com>
7078 * config/nds32/constants.md (unspec_element): New enum.
7079 * config/nds32/constraints.md (Umw): New constraint.
7080 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
7081 * config/nds32/nds32-intrinsic.md: Likewise.
7082 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
7083 (nds32_valid_smw_lwm_base_p): New.
7084 (nds32_output_smw_single_word): New.
7085 (nds32_output_lmw_single_word): New.
7086 (nds32_expand_unaligned_load): New.
7087 (nds32_expand_unaligned_store): New.
7088 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
7089 (nds32_output_smw_single_word): Declare.
7090 (nds32_output_lmw_single_word): Declare.
7091 (nds32_expand_unaligned_load): Declare.
7092 (nds32_expand_unaligned_store): Declare.
7093 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
7094 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
7095 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
7096 NDS32_BUILTIN_UASTORE_DW.
7097 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
7100 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
7101 Kito Cheng <kito.cheng@gmail.com>
7102 Chung-Ju Wu <jasonwucj@gmail.com>
7104 * config/nds32/nds32-intrinsic.c
7105 (nds32_expand_builtin_null_ftype_reg): Delete.
7106 (nds32_expand_builtin_reg_ftype_imm): Ditto.
7107 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
7108 (nds32_read_argument): New.
7109 (nds32_legitimize_target): Ditto.
7110 (nds32_legitimize_argument): Ditto.
7111 (nds32_check_constant_argument): Ditto.
7112 (nds32_expand_unop_builtin): Ditto.
7113 (nds32_expand_unopimm_builtin): Ditto.
7114 (nds32_expand_binop_builtin): Ditto.
7115 (nds32_builtin_decl_impl): Ditto.
7116 (builtin_description): Ditto.
7117 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
7118 (nds32_init_builtins_impl): Ditto.
7119 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
7120 (nds32_builtin_decl): New.
7121 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
7122 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
7124 2018-03-02 Jeff Law <law@redhat.com>
7126 * reorg.c (stop_search_p): Handle DEBUG_INSN.
7127 (redundant_insn, fill_simple_delay_slots): Likewise.
7128 (fill_slots_from_thread): Likewise.
7129 * resource.c (mark_referenced_resources): Likewise.
7130 (mark_set_resources, find_dead_or_set_registers): Likewise.
7132 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7134 * substring-locations.h (format_warning_va): Formatting fix for
7136 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
7138 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
7139 * substring-locations.c: Include intl.h.
7140 (format_warning_va): Turned into small wrapper around
7141 format_warning_n_va, renamed to ...
7142 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
7143 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
7145 (format_warning_at_substring_n): New function.
7146 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
7147 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
7148 format_warning_at_substring with just a shorter name instead of
7149 const function pointer.
7150 (fmtwarn_n): New function.
7151 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
7152 appropriate, get rid of all the fmtstr temporaries, move conditionals
7153 with G_() wrapped string literals directly into fmtwarn arguments,
7154 cast dir.len to (int), formatting fixes.
7156 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
7158 * doc/invoke.texi: Remove "Cilk Plus" references.
7160 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7161 Richard Biener <rguenther@suse.de>
7164 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
7165 for error or warning attributes if CALL_FROM_THUNK_P is set.
7168 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7171 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
7172 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
7175 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
7176 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
7178 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
7179 instead of -1U in last predictors element's probability member.
7181 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
7184 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
7185 arguments if they are comparable.
7187 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
7189 PR tree-optimization/84634
7190 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
7191 masks and masked_loop_p with a single loop_masks, making sure it's
7192 null for bb vectorization.
7194 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
7196 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
7197 (vect_analyze_data_ref_access): Use loop->safe_len rather than
7198 loop->force_vectorize to check whether there is no alias.
7200 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7203 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
7205 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
7207 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
7208 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
7209 instead of a loop around prev_real_insn.
7210 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
7214 * config/i386/i386.c (ix86_print_operand): Use conditional
7215 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
7218 2018-03-02 Richard Biener <rguenther@suse.de>
7220 PR tree-optimization/84427
7221 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
7222 (bitmap_set_subtract_values): Rewrite to handle multiple
7225 (prune_clobbered_mems): Likewise.
7226 (phi_translate): Take edge instead of pred/phiblock.
7227 (phi_translate_1): Likewise.
7228 (phi_translate_set): Likewise. Insert all translated
7229 exprs for a value into the set, keeping possibly multiple
7230 expressions per value.
7231 (compute_antic_aux): Adjust for phi_translate changes.
7232 When intersecting union the expressions and prune those
7233 not in the final value set, keeping possibly multiple
7234 expressions per value. Do not use value-insertion
7235 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
7236 all expressions. Add verification that the value-sets
7237 only shrink during iteration.
7238 (compute_partial_antic_aux): Adjust for the phi_translate changes.
7239 (do_pre_regular_insertion): Likewise.
7240 (do_pre_partial_partial_insertion): Likewise.
7242 2018-03-02 Richard Biener <rguenther@suse.de>
7245 * config/darwin.c (saved_debug_info_level): New static global.
7246 (darwin_asm_lto_start): Disable debug info generation for LTO out.
7247 (darwin_asm_lto_end): Restore debug info generation settings.
7249 2018-03-01 Martin Liska <mliska@suse.cz>
7252 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
7255 2018-03-01 Richard Biener <rguenther@suse.de>
7258 * dwarf2out.c (gen_variable_die): Properly handle late VLA
7259 type annotation with LTO when debug was disabled at compile-time.
7261 2018-03-01 Matthew Fortune <mfortune@gmail.com>
7263 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
7265 (mips_final_postscan_insn): Likewise.
7267 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
7269 PR rtl-optimization/84528
7270 * alias.c (init_alias_target): Add commentary.
7271 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
7272 a unique base value if the frame pointer is not eliminated
7273 to the stack pointer.
7275 2018-03-01 Tom de Vries <tom@codesourcery.com>
7277 PR rtl-optimization/83327
7278 * lra-int.h (hard_regs_spilled_into): Declare.
7279 * lra.c (hard_regs_spilled_into): Define.
7280 (init_reg_info): Init hard_regs_spilled_into.
7281 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
7282 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
7283 (process_bb_lives): Handle hard_regs_spilled_into.
7284 (lra_create_live_ranges_1): Before doing liveness propagation, clear
7285 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
7287 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
7289 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
7290 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
7291 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
7292 * config/rs6000/aix72.h: New file.
7294 2018-02-28 Jakub Jelinek <jakub@redhat.com>
7296 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
7297 instead of warning_at with conditional singular and plural messages
7301 * stor-layout.c (update_alignment_for_field): For
7302 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
7303 && !DECL_PACKED (field), do the alignment update, just use
7304 only desired_align instead of MAX (type_align, desired_align)
7306 (place_field): Don't do known_align < desired_align handling
7307 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
7308 is non-NULL, instead do it after rli->prev_field handling and
7309 only if not within a bitfield word. For DECL_PACKED (field)
7310 use type_align of BITS_PER_UNIT.
7312 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
7314 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
7315 superfluous parentheses and trailing spaces.
7317 2018-02-28 Richard Biener <rguenther@suse.de>
7319 PR tree-optimization/84584
7320 * graphite-scop-detection.c (scop_detection::add_scop): Discard
7321 SCoPs with fake exit edge.
7323 2018-02-28 Martin Liska <mliska@suse.cz>
7326 * timevar.c (timer::print): Fix format to properly print 100%
7329 2018-02-28 Richard Biener <rguenther@suse.de>
7332 * genmatch.c (capture_info::walk_match): Do not mark
7333 captured expressions without operands as expr_p given
7334 they act more like predicates and should be subject to
7335 "lost tail" side-effect preserving.
7337 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
7339 PR rtl-optimization/81611
7340 * auto-inc-dec.c (attempt_change): Move dead note from
7341 mem_insn if it's the next use of regno
7342 (find_address): Take address use of reg holding
7343 non-incremented value. Add parm to limit search to the named
7345 (merge_in_block): Attempt to use a mem insn that is the next
7346 use of the original regno.
7348 2018-02-27 Martin Sebor <msebor@redhat.com>
7351 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
7352 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
7354 2018-02-27 Martin Sebor <msebor@redhat.com>
7356 PR translation/84207
7357 * diagnostic-core.h (warning_n, error_n, inform_n): Change
7358 n argument to unsigned HOST_WIDE_INT.
7359 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
7360 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
7361 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
7362 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
7364 2018-02-27 Richard Biener <rguenther@suse.de>
7366 PR tree-optimization/84512
7367 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
7368 Do not use the estimate returned from record_stmt_cost for
7369 the scalar iteration cost but sum properly using add_stmt_cost.
7371 2018-02-27 Richard Biener <rguenther@suse.de>
7373 PR tree-optimization/84466
7374 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
7375 Adjust last change to less strictly validate use operands.
7377 2018-02-27 Martin Liska <mliska@suse.cz>
7379 PR gcov-profile/84548
7380 * gcov.c (process_file): Allow partial overlap and consider it
7381 also as group functions.
7382 (output_lines): Properly calculate range of lines for a group.
7384 2018-02-27 Martin Liska <mliska@suse.cz>
7386 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
7387 'ggc' suffixes. Change first column width.
7388 (timer::print): Fix formatting of the column.
7390 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
7392 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
7393 preserve inline entry blocks for the sake of debug inline
7394 entry point markers alone.
7395 (remove_unused_locals): Suggest in comments a better place to
7396 force the preservation of inline entry blocks that are
7397 otherwise unused, but do not preserve them.
7399 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7401 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
7403 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7406 * config/i386/constraints.md (Bs): Replace
7407 ix86_indirect_branch_register with
7408 TARGET_INDIRECT_BRANCH_REGISTER.
7410 * config/i386/i386.md (indirect_jump): Likewise.
7411 (tablejump): Likewise.
7412 (*sibcall_memory): Likewise.
7413 (*sibcall_value_memory): Likewise.
7414 Peepholes of indirect call and jump via memory: Likewise.
7415 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
7416 (*sibcall_value_GOT_32): Likewise.
7417 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7418 (GOT_memory_operand): Likewise.
7419 (call_insn_operand): Likewise.
7420 (sibcall_insn_operand): Likewise.
7421 (GOT32_symbol_operand): Likewise.
7422 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
7424 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
7426 PR rtl-optimization/83496
7427 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
7428 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
7429 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
7430 redundant insn, if any.
7431 (relax_delay_slots): Likewise.
7432 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
7434 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
7436 PR tree-optimization/83965
7437 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
7438 that grouped statements are part of a reduction chain. Return
7439 true if the statement is not marked as a reduction itself but
7441 (vect_recog_dot_prod_pattern): Don't check whether the statement
7442 is part of a group here.
7443 (vect_recog_sad_pattern): Likewise.
7444 (vect_recog_widen_sum_pattern): Likewise.
7446 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
7449 * final.c (rest_of_clean_state): Also look for calls inside sequences.
7451 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7454 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
7456 (ix86_output_indirect_function_return): New prototype.
7457 (ix86_split_simple_return_pop_internal): Likewise.
7458 * config/i386/i386.c (indirect_return_via_cx): New.
7459 (indirect_return_via_cx_bnd): Likewise.
7460 (indirect_thunk_name): Handle return va CX_REG.
7461 (output_indirect_thunk_function): Create alias for
7462 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
7463 (ix86_output_indirect_jmp): Remove the bool argument.
7464 (ix86_output_indirect_function_return): New function.
7465 (ix86_split_simple_return_pop_internal): Likewise.
7466 * config/i386/i386.md (*indirect_jump): Don't pass false
7467 to ix86_output_indirect_jmp.
7468 (*tablejump_1): Likewise.
7469 (simple_return_pop_internal): Change it to define_insn_and_split.
7470 Call ix86_split_simple_return_pop_internal to split it for
7472 (simple_return_indirect_internal): Call
7473 ix86_output_indirect_function_return instead of
7474 ix86_output_indirect_jmp.
7476 2018-02-26 Jakub Jelinek <jakub@redhat.com>
7479 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
7480 memset and value initialization afterwards.
7482 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
7484 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
7486 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
7489 * common/config/aarch64/aarch64-common.c
7490 (aarch_option_optimization_table[]): Switch
7491 off fomit-frame-pointer
7493 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
7494 Chung-Ju Wu <jasonwucj@gmail.com>
7496 * config/nds32/nds32-multiple.md (load_multiple): Disallow
7498 (store_multiple): Ditto.
7500 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
7502 * config.gcc: Add --with-cpu support for nds32 target.
7503 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
7504 * config/nds32/nds32.opt: Add -mcpu= option.
7506 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
7508 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
7509 isel=yes): Warn for these deprecated options.
7511 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
7513 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
7514 ISA_2_5_MASKS_EMBEDDED.
7516 2018-02-23 Jakub Jelinek <jakub@redhat.com>
7518 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
7519 p->max as pointers rather than using iterative_hash_expr.
7521 2018-02-23 Carl Love <cel@us.ibm.com>
7523 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
7524 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
7526 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
7527 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
7528 P8V_BUILTIN_VEC_VUNSIGNED2.
7530 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
7533 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
7534 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
7536 * lra-constraints.c (curr_insn_transform): Set up
7537 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
7539 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
7540 * lra-eliminations.c (spill_pseudos): Ditto.
7541 (process_insn_for_elimination): Ditto.
7542 * lra-lives.c (reg_early_clobber_p): Use the new macros.
7543 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
7544 LRA_NON_CLOBBERED_ALT.
7546 2018-02-22 Martin Sebor <msebor@redhat.com>
7548 PR tree-optimization/84480
7549 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
7550 to maybe_diag_stxncpy_trunc. Call it.
7551 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
7552 from gimple_fold_builtin_strcpy. Print inlining stack.
7553 (handle_builtin_stxncpy): Print inlining stack.
7554 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
7556 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
7559 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
7560 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
7561 and -fcheck-pointer-bounds are used together.
7562 (indirect_thunk_prefix): New enum.
7563 (indirect_thunk_need_prefix): New function.
7564 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
7565 "_nt" instead of "_bnd" for NOTRACK prefix.
7566 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
7567 (output_indirect_thunk_function): Likewise.
7569 (ix86_code_end): Update output_indirect_thunk_function calls.
7570 (ix86_output_indirect_branch_via_reg): Replace
7571 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
7572 (ix86_output_indirect_branch_via_push): Likewise.
7573 (ix86_output_function_return): Likewise.
7574 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
7575 incompatible with -fcf-protection=branch and
7576 -fcheck-pointer-bounds.
7578 2018-02-22 Steve Ellcey <sellcey@cavium.com>
7581 * config/aarch64/aarch64.c (aarch64_print_address_internal):
7582 Change gcc_assert call to output_operand_lossage.
7584 2018-02-22 Steve Ellcey <sellcey@cavium.com>
7586 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
7588 2018-02-22 DJ Delorie <dj@redhat.com>
7589 Sebastian Perta <sebastian.perta@renesas.com>
7590 Oleg Endo <olegendo@gcc.gnu.org>
7592 * config/rx/rx.c (rx_rtx_costs): New function.
7593 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
7595 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
7597 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
7599 2018-02-22 Martin Liska <mliska@suse.cz>
7602 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
7603 Add "native" as a possible value.
7605 2018-02-22 Martin Liska <mliska@suse.cz>
7608 * config/i386/i386.c (ix86_option_override_internal):
7609 Add "native" as a possible value for -march and -mtune.
7611 2018-02-22 Jakub Jelinek <jakub@redhat.com>
7614 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
7615 to all type variants.
7617 PR tree-optimization/84503
7618 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
7619 width as info->bitpos + info->bitsize - start.
7620 (merged_store_group::merge_overlapping): Simplify width computation.
7621 (check_no_overlap): New function.
7622 (imm_store_chain_info::try_coalesce_bswap): Compute expected
7623 start + width and last_order of the group, fail if check_no_overlap
7625 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
7626 to group if check_no_overlap fails.
7628 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
7630 * config/rs6000/altivec.md: Delete contraint arguments to
7631 define_expand, define_split, and define_peephole2, and in
7632 define_insn_and_split if always unused.
7633 * config/rs6000/darwin.md: Ditto.
7634 * config/rs6000/dfp.md: Ditto.
7635 * config/rs6000/rs6000.md: Ditto.
7636 * config/rs6000/sync.md: Ditto.
7637 * config/rs6000/vector.md: Ditto.
7638 * config/rs6000/vsx.md: Ditto.
7640 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
7642 * config/rs6000/altivec.md: Write output control strings as braced
7643 blocks instead of double-quoted strings.
7644 * config/rs6000/darwin.md: Ditto.
7645 * config/rs6000/rs6000.md: Ditto.
7646 * config/rs6000/vector.md: Ditto.
7647 * config/rs6000/vsx.md: Ditto.
7649 2018-02-21 Jason Merrill <jason@redhat.com>
7651 PR c++/84314 - ICE with templates and fastcall attribute.
7652 * attribs.c (build_type_attribute_qual_variant): Remove assert.
7654 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
7656 * ipa-cp.c (determine_versionability): Fix comment typos.
7658 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
7661 * ipa-cp.c (determine_versionability): Do not version functions caling
7664 2018-02-21 Martin Liska <mliska@suse.cz>
7667 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
7668 Add "native" as a possible value.
7669 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
7670 the macro when native cpu detection is available.
7672 2018-02-21 Martin Liska <mliska@suse.cz>
7675 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
7676 Add "native" as a possible value.
7677 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
7678 when native cpu detection is available.
7680 2018-02-21 Jakub Jelinek <jakub@redhat.com>
7681 Martin Sebor <msebor@redhat.com>
7683 PR tree-optimization/84478
7684 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
7686 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
7687 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
7688 support which is conservatively correct, for 2 only stay conservative
7689 for maxlen. Formatting and comment capitalization fixes. Add STRICT
7690 argument to the 2 argument get_range_strlen, adjust 6 arg
7691 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
7693 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
7694 (gimple_fold_builtin_strlen): Pass true as last argument to
7697 2018-02-20 Martin Sebor <msebor@redhat.com>
7700 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
7701 (builtin_memref::set_base_and_offset): Same. Handle inner references.
7702 (builtin_memref::builtin_memref): Factor out parts into
7703 set_base_and_offset and call it.
7705 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
7708 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
7709 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
7710 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
7711 search at the associated MODE_INT.
7713 2018-02-20 Jeff Law <law@redhat.com>
7716 PR tree-optimization/81592
7718 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
7719 for range data rather than using global data.
7720 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
7721 range data rather than using global data.
7722 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
7723 pass it to children as needed.
7724 (struct directive::fmtresult): Similarly.
7725 (struct directive::set_width): Similarly.
7726 (struct directive::set_precision): Similarly.
7727 (format_integer, format_directive, parse_directive): Similarly.
7728 (format_none): Accept unnamed vr_values parameter.
7729 (format_percent, format_floating, format_character): Similarly.
7730 (format_string, format_plain): Similarly.
7731 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
7732 the EVRP range analyzer for range data rather than using global data.
7733 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
7734 gimple-ssa-evrp-analyze.h
7735 (class sprintf_dom_walker): Add after_dom_children member function.
7736 Add evrp_range_analyzer member.
7737 (sprintf_dom_walker::before_dom_children): Call into the EVRP
7738 range analyzer as needed.
7739 (sprintf_dom_walker::after_dom_children): New member function.
7740 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
7742 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
7743 (evrp_range_analyzer::pop_to_marker): Likewise.
7745 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
7747 PR tree-optimization/84419
7748 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
7749 with the required type if its current type is compatible but
7752 2018-02-20 Jakub Jelinek <jakub@redhat.com>
7755 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
7756 after vectorization.
7758 2018-02-20 Martin Liska <mliska@suse.cz>
7761 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
7762 possible values if we don't have a hint.
7764 2018-02-20 Martin Liska <mliska@suse.cz>
7768 * final.c (shorten_branches): Build align_tab array with one
7770 * opts.c (finish_options): Add alignment option limit check.
7771 (MAX_CODE_ALIGN): Likewise.
7772 (MAX_CODE_ALIGN_VALUE): Likewise.
7773 * doc/invoke.texi: Document maximum allowed option value for
7774 all -falign-* options.
7776 2018-02-19 Jakub Jelinek <jakub@redhat.com>
7779 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
7780 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
7781 * var-tracking.c (emit_note_insn_var_location): Remove all references
7782 to NOTE_INSN_CALL_ARG_LOCATION.
7783 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
7784 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
7785 Use copy_rtx_if_shared.
7786 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
7787 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
7788 (dwarf2out_var_location): Remove handling of
7789 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
7791 * final.c (final_scan_insn): Remove all references to
7792 NOTE_INSN_CALL_ARG_LOCATION.
7793 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
7794 before dumping final insns.
7795 * except.c (emit_note_eh_region_end): Remove all references to
7796 NOTE_INSN_CALL_ARG_LOCATION.
7797 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
7798 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
7799 * config/arc/arc.c (hwloop_optimize): Likewise.
7800 * config/arm/arm.c (create_fix_barrier): Likewise.
7801 * config/s390/s390.c (s390_chunkify_start): Likewise.
7802 * config/sh/sh.c (find_barrier): Likewise.
7803 * config/i386/i386.c (rest_of_insert_endbranch,
7804 ix86_seh_fixup_eh_fallthru): Likewise.
7805 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
7806 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
7807 * config/frv/frv.c (frv_function_prologue): Likewise.
7808 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
7810 (note_outside_basic_block_p): Remove all references to
7811 NOTE_INSN_CALL_ARG_LOCATION.
7812 * gengtype.c (adjust_field_rtx_def): Likewise.
7813 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
7815 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
7816 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
7819 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
7822 PR tree-optimization/84452
7823 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
7824 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
7827 2018-02-19 Martin Liska <mliska@suse.cz>
7830 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
7832 2018-02-19 Martin Liska <mliska@suse.cz>
7833 Richard Sandiford <richard.sandiford@linaro.org>
7835 PR tree-optimization/82491
7836 * gimple-fold.c (get_base_constructor): Make earlier bail out
7839 2018-02-19 Carl Love <cel@us.ibm.com>
7841 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
7842 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
7844 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
7845 P8V_BUILTIN_VEC_NEG.
7847 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
7849 * config/rl78/rl78.md (movdf): New define expand.
7851 2018-02-19 Martin Liska <mliska@suse.cz>
7854 * doc/invoke.texi: Fix typo.
7855 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
7857 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
7859 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
7860 handle rs6000_single_float and rs6000_double_float specially for
7863 2018-02-16 Jeff Law <law@redhat.com>
7865 * config/rx/rx.c (add_pop_cfi_notes): New function.;
7868 2018-02-16 Jakub Jelinek <jakub@redhat.com>
7871 * ipa-inline.c (inline_small_functions): Fix a typo.
7873 2018-02-16 Nathan Sidwell <nathan@acm.org>
7875 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
7877 2018-02-16 Carl Love <cel@us.ibm.com>
7879 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
7880 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
7881 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
7882 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
7883 expansion to P8V_BUILTIN_VEC_FLOAT2.
7885 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
7887 PR rtl-optimization/70023
7888 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
7889 src_regno into account.
7891 2018-02-16 Carl Love <cel@us.ibm.com>
7893 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
7894 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
7895 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
7896 * config/rs6000/rs6000.c: Remove case statements for
7897 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
7898 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
7899 and P9V_BUILTIN_VEC_VINSERT4B.
7900 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
7901 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
7902 * config/rs6000/vsx.md:
7903 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
7906 2018-02-16 Carl Love <cel@us.ibm.com>
7908 * config/rs6000/altivec.h: Add builtin names vec_extract4b
7910 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
7912 * config/rs6000/rs6000-c.c: Add the definitions for
7913 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
7914 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
7915 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
7916 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
7917 definition for insert4b and define insn *insert3b_internal.
7918 * doc/extend.texi: Add documentation for vec_extract4b.
7920 2018-02-16 Nathan Sidwell <nathan@acm.org>
7922 * doc/extend.texi (Backwards Compatibility): Mention friend
7923 injection. Note for-scope is deprecated.
7924 * doc/invoke.texi (-ffriend-injection): Deprecate.
7926 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
7928 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
7929 that moved to I2, also allow destinations that are a paradoxical
7930 subreg (instead of a normal reg).
7932 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
7935 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
7938 2018-02-16 Richard Biener <rguenther@suse.de>
7940 PR tree-optimization/84037
7941 PR tree-optimization/84016
7943 * config/i386/i386.c (ix86_builtin_vectorization_cost):
7944 Adjust vec_construct for the fact we need additional higher latency
7945 128bit inserts for AVX256 and AVX512 vector builds.
7946 (ix86_add_stmt_cost): Scale vector construction cost for
7949 2018-02-16 Richard Biener <rguenther@suse.de>
7951 PR tree-optimization/84417
7952 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
7953 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
7954 (non_rewritable_lvalue_p): Likewise, use poly-ints.
7956 2018-02-16 Martin Liska <mliska@suse.cz>
7959 * internal-fn.def (ASAN_CHECK): Set proper flags.
7960 (ASAN_MARK): Likewise.
7962 2018-02-16 Julia Koval <julia.koval@intel.com>
7964 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
7965 from PTA_CANNONLAKE.
7967 2018-02-16 Jakub Jelinek <jakub@redhat.com>
7970 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
7971 Use ++iter rather than iter++ for std::list iterators.
7972 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
7973 defer deleting them until all nodes in the forest are processed. Do
7974 free even leaf nodes. Change to_process into auto_vec.
7977 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
7978 * vec.h (vec_default_construct): Use memset instead of placement new
7979 if BROKEN_VALUE_INITIALIZATION is defined.
7980 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
7981 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
7984 PR rtl-optimization/83723
7985 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
7986 * lra.c (lra_substitute_pseudo): Likewise. If true, use
7987 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
7989 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
7991 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
7993 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
7995 PR rtl-optimization/81443
7996 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
7997 from inner REGs to paradoxical SUBREGs.
7999 2018-02-16 Richard Biener <rguenther@suse.de>
8001 PR tree-optimization/84399
8002 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
8003 For operands we can analyze at their definition make sure we can
8004 analyze them at each use as well.
8006 2018-02-16 Richard Biener <rguenther@suse.de>
8008 PR tree-optimization/84190
8009 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
8010 volatile accesses if the decl isn't volatile.
8012 2018-02-15 Jason Merrill <jason@redhat.com>
8014 PR c++/84314 - ICE with templates and fastcall attribute.
8015 * attribs.c (build_type_attribute_qual_variant): Don't clobber
8016 TYPE_CANONICAL on an existing type.
8018 2018-02-15 Jakub Jelinek <jakub@redhat.com>
8020 PR tree-optimization/84383
8021 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
8022 dstoff nor call operand_equal_p if dstbase is NULL.
8024 PR tree-optimization/84334
8025 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
8026 also a CONSTANT_CLASS_P, punt.
8028 2018-02-14 Jim Wilson <jimw@sifive.com>
8030 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
8031 first SMALL_OPERAND check. New local min_second_step. Move assert
8032 to where locals are set. Add TARGET_RVC support.
8033 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
8035 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
8037 * doc/invoke.texi: Correct -Wformat-overflow code sample.
8039 2018-02-14 Martin Sebor <msebor@redhat.com>
8041 PR tree-optimization/83698
8042 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
8043 arrays constrain the offset range to their bounds.
8044 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
8045 (builtin_access::overlap): Avoid setting the size of overlap if it's
8047 (maybe_diag_overlap): Also consider arrays when deciding what values
8048 of offsets to include in diagnostics.
8050 2018-02-14 Martin Sebor <msebor@redhat.com>
8053 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
8054 that correspond to the kind of a declaration.
8056 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
8059 * config/pa/pa.md: Load address of PIC label using the linkage table
8060 if the label is nonlocal.
8062 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
8064 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
8065 warning message if user requests -maltivec=be.
8066 * doc/invoke.texi: Document deprecation of -maltivec=be.
8068 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
8071 * config/rs6000/rs6000-c.c: Update definitions for
8072 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
8073 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
8075 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
8078 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
8079 add _get_ssp intrinsics. Remove argument from
8080 __builtin_ia32_rdssp[d|q].
8081 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
8082 * config/i386/i386-builtin.def: Remove argument from
8083 __builtin_ia32_rdssp[d|q].
8084 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
8085 ix86_expand_special_args_builtin for _rdssp[d|q].
8086 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
8087 Clear register before usage.
8088 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
8089 Add documentation for new _get_ssp and _inc_ssp intrinsics.
8091 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
8093 PR tree-optimization/84357
8094 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
8095 operand 1 of an ARRAY_REF too.
8097 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
8100 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
8101 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
8103 (set_of_reg): New struct.
8104 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
8105 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
8106 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
8108 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
8109 Split into bitclr, bitset, bitinvert patterns if appropriate.
8110 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
8111 use rx_fuse_in_memory_bitop.
8112 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
8113 to named insn, correct maximum insn length.
8115 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
8118 * machmode.def: Define a complex mode for PARTIAL_INT.
8119 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
8121 * doc/rtl.texi: Document CSPImode.
8122 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
8124 (msp430_hard_regno_nregs_with_padding): Likewise.
8126 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
8129 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
8131 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
8133 PR rtl-optimization/84169
8134 * combine.c (try_combine): New variable split_i2i3. Set it to true if
8135 we generated a parallel as new i3 and we split that to new i2 and i3
8136 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
8137 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
8138 those to i2, not i1. Partially rewrite this scan code.
8140 2018-02-13 Jakub Jelinek <jakub@redhat.com>
8143 * stor-layout.c (place_field): For variable length fields, adjust
8144 offset_align afterwards not just based on the field's alignment,
8145 but also on the size.
8148 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
8149 of exps and logs in the use_exp2 case.
8151 2018-02-13 Jeff Law <law@redhat.com>
8153 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
8156 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
8159 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
8163 * common.opt (gas-loc-support, gas-locview-support): New.
8164 (ginline-points, ginternal-reset-location-views): New.
8165 * doc/invoke.texi: Document them. Use @itemx where intended.
8166 (gvariable-location-views): Adjust.
8167 * target.def (reset_location_view): New.
8168 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
8169 (TARGET_RESET_LOCATION_VIEW): New.
8170 * doc/tm.texi: Rebuilt.
8171 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
8172 (dwarf2out_default_as_locview_support): New.
8173 (output_asm_line_debug_info): Use option variables.
8174 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
8175 (output_loc_list): Likewise.
8176 (add_high_low_attributes): Check option variables.
8177 Don't output entry view attribute in strict mode.
8178 (gen_inlined_subroutine_die): Check option variables.
8179 (dwarf2out_inline_entry): Likewise.
8180 (init_sections_and_labels): Likewise.
8181 (dwarf2out_early_finish): Likewise.
8182 (maybe_reset_location_view): New, from...
8183 (dwarf2out_var_location): ... here. Call it.
8184 * debug.h (dwarf2out_default_as_loc_support): Declare.
8185 (dwarf2out_default_as_locview_support): Declare.
8186 * hooks.c (hook_int_rtx_insn_0): New.
8187 * hooks.h (hook_int_rtx_insn_0): Declare.
8188 * toplev.c (process_options): Take -gas-loc-support and
8189 -gas-locview-support from dwarf2out. Enable
8190 -gvariable-location-views by default only with locview
8191 assembler support. Enable -ginternal-reset-location-views by
8192 default only if the target defines the corresponding hook.
8193 Enable -ginline-points by default if location views are
8194 enabled; force it disabled if statement frontiers are
8196 * tree-inline.c (expand_call_inline): Check option variables.
8197 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
8199 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
8201 PR tree-optimization/84321
8202 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
8203 handling. Also check whether the anti-range contains any values
8204 that satisfy the mask; switch to a VR_RANGE if not.
8206 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
8209 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
8211 2018-02-13 Martin Jambor <mjambor@suse.cz>
8214 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
8215 of call statements, also set location of a load to a temporary.
8217 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
8219 * config/rl78/rl78.c (add_vector_labels): New function.
8220 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
8221 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
8222 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
8223 which checks that no arguments are passed.
8224 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
8225 * doc/extend.texi: Documentation for the new attribute.
8227 2018-02-13 Andreas Schwab <schwab@suse.de>
8229 * config/riscv/linux.h (CPP_SPEC): Define.
8231 2018-02-13 Jakub Jelinek <jakub@redhat.com>
8234 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
8235 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
8236 OPTION_MASK_ISA_AES as first argument to def_builtin_const
8237 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
8238 instead of OPTION_MASK_ISA_PCLMUL as first argument to
8239 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
8240 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
8241 temporarily for AES and PCLMUL builtins.
8243 PR tree-optimization/84339
8244 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
8245 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
8249 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
8250 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
8251 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
8253 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
8255 * omp-simd-clone.h: New file.
8256 * omp-simd-clone.c: Include omp-simd-clone.h.
8257 (expand_simd_clones): No longer static.
8258 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
8259 cgraph.h and omp-simd-clone.h.
8260 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
8261 (vect_recog_widen_shift_pattern): Formatting fix.
8262 (vect_pattern_recog_1): Don't check optab for calls.
8265 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
8266 operands[2] into a REG before using gen_lowpart on it.
8268 2018-02-12 Jeff Law <law@redhat.com>
8271 * config/sh/sh.c (find_barrier): Consider a sibling call
8274 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
8275 successfully back substituting a reg.
8277 2018-02-12 Richard Biener <rguenther@suse.de>
8279 PR tree-optimization/84037
8280 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
8281 parameter, move visited init to caller.
8282 (vect_slp_analyze_operations): Separate cost from validity
8283 check, initialize visited once for all instances.
8284 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
8286 * tree-vect-stmts.c (vect_model_simple_cost): Make early
8288 (vect_model_promotion_demotion_cost): Likewise.
8289 (vectorizable_bswap): Guard cost modeling with !slp_node
8290 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
8292 (vectorizable_call): Likewise.
8293 (vectorizable_conversion): Likewise.
8294 (vectorizable_assignment): Likewise.
8295 (vectorizable_shift): Likewise.
8296 (vectorizable_operation): Likewise.
8297 (vectorizable_store): Likewise.
8298 (vectorizable_load): Likewise.
8299 (vectorizable_condition): Likewise.
8300 (vectorizable_comparison): Likewise.
8302 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
8305 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
8306 (ASAN_MARK): Fix fnspec to account for return value, change pointer
8307 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
8309 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
8312 * params.def (inline-min-speedup): Increase from 8 to 15.
8313 (max-inline-insns-auto): Decrease from 40 to 30.
8314 * ipa-split.c (consider_split): Add some buffer for function to
8315 be considered inlining candidate.
8316 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
8319 2018-02-12 Richard Biener <rguenther@suse.de>
8321 PR tree-optimization/84037
8322 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
8323 matched stmts if we cannot swap the non-matched ones.
8325 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
8327 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
8328 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
8329 _mm_maskz_scalef_round_ss): New intrinsics.
8330 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
8331 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
8332 __builtin_ia32_scalefss_round): Remove.
8333 (__builtin_ia32_scalefsd_mask_round,
8334 __builtin_ia32_scalefss_mask_round): New intrinsics.
8335 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
8336 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
8337 ((match_operand:VF_128 2 "<round_nimm_predicate>"
8338 "<round_constraint>")): Changed to ...
8339 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
8340 "<round_scalar_constraint>")): ... this.
8341 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
8342 %0, %1, %2<round_op3>}"): Changed to ...
8343 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
8344 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
8345 %2<round_scalar_mask_op3>}"): ... this.
8346 * config/i386/subst.md (round_scalar_nimm_predicate): New.
8348 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
8350 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
8351 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
8352 (_mm_maskz_sqrt_round_ss): New intrinsics.
8353 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
8354 (__builtin_ia32_sqrtsd_mask_round)
8355 (__builtin_ia32_sqrtss_mask_round): New builtins.
8356 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
8357 (__builtin_ia32_sqrtss_round): Remove.
8358 (__builtin_ia32_sqrtsd_mask_round)
8359 (__builtin_ia32_sqrtss_mask_round): New builtins.
8360 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
8361 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
8362 ((match_operand:VF_128 1 "vector_operand"
8363 "xBm,<round_constraint>")): Changed to ...
8364 ((match_operand:VF_128 1 "vector_operand"
8365 "xBm,<round_scalar_constraint>")): ... this.
8366 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
8367 %0, %2, %<iptr>1<round_op3>}): Changed to ...
8368 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
8369 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
8370 %<iptr>1<round_scalar_mask_op3>}): ... this.
8371 ((set_attr "prefix" "<round_prefix>")): Changed to ...
8372 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
8374 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
8377 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
8378 Cast vec_cmpeq result to correct type.
8379 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
8380 Cast vec_cmpgt result to correct type.
8382 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
8384 * final.c (final_scan_insn_1): Renamed from...
8385 (final_scan_insn): ... this. New wrapper, to recover
8386 seen from the outermost call in recursive ones.
8387 * config/sparc/sparc.c (output_return): Drop seen from call.
8388 (output_sibcall): Likewise.
8389 * config/visium/visium.c (output_branch): Likewise.
8391 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
8393 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
8396 2018-02-10 Alan Modra <amodra@gmail.com>
8399 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
8400 Specify LR as an input.
8402 2018-02-10 Jakub Jelinek <jakub@redhat.com>
8405 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
8406 remove_member_access_dummy_vars): New functions.
8407 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
8408 lower_omp_1, execute_lower_omp): Use them.
8410 PR rtl-optimization/84308
8411 * shrink-wrap.c (spread_components): Release todo vector.
8413 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
8415 PR rtl-optimization/57193
8416 * ira-color.c (struct allocno_color_data): Add member
8417 conflict_allocno_hard_prefs.
8418 (update_conflict_allocno_hard_prefs): New.
8419 (bucket_allocno_compare_func): Add a preference based on
8420 conflict_allocno_hard_prefs.
8421 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
8422 (color_allocnos): Remove a dead code. Initiate
8423 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
8425 2018-02-09 Jakub Jelinek <jakub@redhat.com>
8428 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
8429 constraint from =wa to wa. Avoid a subreg on the output operand,
8430 instead use a pseudo and subreg it in a move.
8431 (p9_xxbrd_<mode>): Changed to ...
8432 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
8433 (p9_xxbrd_v2df): New expander.
8434 (p9_xxbrw_<mode>): Changed to ...
8435 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
8436 (p9_xxbrw_v4sf): New expander.
8438 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
8440 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
8442 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
8445 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
8446 multiply in 32-bit mode.
8447 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
8448 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
8451 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
8453 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
8454 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
8455 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
8456 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
8458 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
8461 * dwarf2out.c (is_trivial_indirect_ref): New function.
8462 (dwarf2out_late_global_decl): Do not generate a location
8463 attribute for variables that have a non-trivial DECL_VALUE_EXPR
8464 and that are not defined in the current unit.
8466 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
8468 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
8469 instead of a libcall for UNORDERED.
8471 2018-02-09 Tamar Christina <tamar.christina@arm.com>
8474 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
8475 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
8477 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8480 * config/s390/s390.c (s390_set_current_function): Invoke
8481 s390_indirect_branch_settings also if fndecl didn't change.
8483 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
8485 * config/rs6000/rs6000.md (blockage): Set length to zero.
8487 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
8489 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
8491 2018-02-09 Jakub Jelinek <jakub@redhat.com>
8494 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
8495 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
8499 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
8500 PARALLEL incoming that failed vt_get_decl_and_offset check.
8503 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
8504 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
8506 (get_variable_section): For decls in named .bss* sections pass true as
8507 second argument to bss_initializer_p.
8509 2018-02-09 Marek Polacek <polacek@redhat.com>
8510 Jakub Jelinek <jakub@redhat.com>
8513 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
8514 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
8515 Sync some changes from cxx_fold_indirect_ref.
8517 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
8519 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
8521 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
8522 (BLOCK_INLINE_ENTRY_LABEL): New.
8523 (dwarf2out_var_location): Disregard inline entry markers.
8524 (inline_entry_data): New struct.
8525 (inline_entry_data_hasher): New hashtable type.
8526 (inline_entry_data_hasher::hash): New.
8527 (inline_entry_data_hasher::equal): New.
8528 (inline_entry_data_table): New variable.
8529 (add_high_low_attributes): Add DW_AT_entry_pc and
8530 DW_AT_GNU_entry_view attributes if a pending entry is found
8531 in inline_entry_data_table. Add old entry_pc attribute only
8532 if debug nonbinding markers are disabled.
8533 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
8534 markers are enabled.
8535 (block_within_block_p, dwarf2out_inline_entry): New.
8536 (dwarf2out_finish): Check that no entries remained in
8537 inline_entry_data_table.
8538 * final.c (reemit_insn_block_notes): Handle inline entry notes.
8539 (final_scan_insn, notice_source_line): Likewise.
8540 (rest_of_clean_state): Skip inline entry markers.
8541 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
8543 * gimple.c (gimple_build_debug_inline_entry): New.
8544 * gimple.h (enum gimple_debug_subcode): Add
8545 GIMPLE_DEBUG_INLINE_ENTRY.
8546 (gimple_build_debug_inline_entry): Declare.
8547 (gimple_debug_inline_entry_p): New.
8548 (gimple_debug_nonbind_marker_p): Adjust.
8549 * insn-notes.def (INLINE_ENTRY): New.
8550 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
8551 inline entry marker notes.
8552 (print_insn): Likewise.
8553 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
8554 (INSN_DEBUG_MARKER_KIND): Likewise.
8555 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
8556 * tree-inline.c (expand_call_inline): Build and insert
8557 debug_inline_entry stmt.
8558 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
8559 inline entry blocks early, if nonbind markers are enabled.
8560 (dump_scope_block): Dump fragment info.
8561 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
8562 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
8563 (gimple_build_debug_inline_entry): New.
8564 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
8565 Enable/disable inline entry points too.
8566 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
8567 (DEBUG_INSN): Describe inline entry markers.
8569 * common.opt (gvariable-location-views): New.
8570 (gvariable-location-views=incompat5): New.
8571 * config.in: Rebuilt.
8572 * configure: Rebuilt.
8573 * configure.ac: Test assembler for view support.
8574 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
8575 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
8576 * dwarf2out.c (var_loc_view): New typedef.
8577 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
8578 (dwarf2out_locviews_in_attribute): New.
8579 (dwarf2out_locviews_in_loclist): New.
8580 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
8581 (enum dw_line_info_opcode): Add LI_adv_address.
8582 (struct dw_line_info_table): Add view.
8583 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
8584 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
8585 (zero_view_p): New variable.
8586 (ZERO_VIEW_P): New macro.
8587 (output_asm_line_debug_info): New.
8588 (struct var_loc_node): Add view.
8589 (add_AT_view_list, AT_loc_list): New.
8590 (add_var_loc_to_decl): Add view param. Test it against last.
8591 (new_loc_list): Add view params. Record them.
8592 (AT_loc_list_ptr): Handle loc and view lists.
8593 (view_list_to_loc_list_val_node): New.
8594 (print_dw_val): Handle dw_val_class_view_list.
8595 (size_of_die): Likewise.
8596 (value_format): Likewise.
8597 (loc_list_has_views): New.
8598 (gen_llsym): Set vl_symbol too.
8599 (maybe_gen_llsym, skip_loc_list_entry): New.
8600 (dwarf2out_maybe_output_loclist_view_pair): New.
8601 (output_loc_list): Output view list or entries too.
8602 (output_view_list_offset): New.
8603 (output_die): Handle dw_val_class_view_list.
8604 (output_dwarf_version): New.
8605 (output_compilation_unit_header): Use it.
8606 (output_skeleton_debug_sections): Likewise.
8607 (output_rnglists, output_line_info): Likewise.
8608 (output_pubnames, output_aranges): Update version comments.
8609 (output_one_line_info_table): Output view numbers in asm comments.
8610 (dw_loc_list): Determine current endview, pass it to new_loc_list.
8611 Call maybe_gen_llsym.
8612 (loc_list_from_tree_1): Adjust.
8613 (add_AT_location_description): Create view list attribute if
8614 needed, check it's absent otherwise.
8615 (convert_cfa_to_fb_loc_list): Adjust.
8616 (maybe_emit_file): Call output_asm_line_debug_info for test.
8617 (dwarf2out_var_location): Reset views as needed. Precompute
8618 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
8619 attribute. Set view.
8620 (new_line_info_table): Reset next view.
8621 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
8622 (dwarf2out_source_line): Likewise. Output view resets and labels to
8623 the assembler, or select appropriate line info opcodes.
8624 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
8625 (optimize_string_length): Catch it. Adjust.
8626 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
8627 dw_val_class_view_list, and remove it if no longer needed.
8628 (hash_loc_list): Hash view numbers.
8629 (loc_list_hasher::equal): Compare them.
8630 (optimize_location_lists): Check whether a view list symbol is
8631 needed, and whether the locview attribute is present, and
8632 whether they match. Remove the locview attribute if no longer
8634 (index_location_lists): Call skip_loc_list_entry for test.
8635 (dwarf2out_finish): Call output_asm_line_debug_info for test.
8636 Use output_dwarf_version.
8637 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
8638 (struct dw_val_node): Add val_view_list.
8639 * final.c (SEEN_NEXT_VIEW): New.
8640 (set_next_view_needed): New.
8641 (clear_next_view_needed): New.
8642 (maybe_output_next_view): New.
8643 (final_start_function): Rename to...
8644 (final_start_function_1): ... this. Take pointer to FIRST,
8645 add SEEN parameter. Emit param bindings in the initial view.
8646 (final_start_function): Reintroduce SEEN-less interface.
8647 (final): Rename to...
8648 (final_1): ... this. Take SEEN parameter. Output final pending
8649 next view at the end.
8650 (final): Reintroduce seen-less interface.
8651 (final_scan_insn): Output pending next view before switching
8652 sections or ending a block. Mark the next view as needed when
8653 outputting variable locations. Notify debug backend of section
8654 changes, and of location view changes.
8655 (rest_of_handle_final): Adjust.
8656 * toplev.c (process_options): Autodetect value for debug variable
8657 location views option. Warn on incompat5 without -gdwarf-5.
8658 * doc/invoke.texi (gvariable-location-views): New.
8659 (gvariable-location-views=incompat5): New.
8660 (gno-variable-location-views): New.
8662 2018-02-08 David Malcolm <dmalcolm@redhat.com>
8664 PR tree-optimization/84136
8665 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
8666 that the result of find_edge is non-NULL.
8668 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
8671 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
8672 storing integer register in SImode. Fix cost of 256 and 512
8673 byte aligned SSE register store.
8675 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
8677 * config/i386/i386.c (ix86_multiplication_cost): Fix
8678 multiplication cost for TARGET_AVX512DQ.
8680 2018-02-08 Marek Polacek <polacek@redhat.com>
8682 PR tree-optimization/84238
8683 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
8686 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8688 PR tree-optimization/84265
8689 * tree-vect-stmts.c (vectorizable_store): Don't treat
8690 VMAT_CONTIGUOUS accesses as grouped.
8691 (vectorizable_load): Likewise.
8693 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8695 PR tree-optimization/81635
8696 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
8697 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
8698 (test_round_for_mask): New functions.
8699 (wide_int_cc_tests): Call test_round_for_mask.
8700 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
8701 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
8702 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
8703 range returned by get_range_info.
8705 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
8708 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
8709 * symtab.c: Include builtins.h
8710 (symtab_node::output_to_lto_symbol_table_p): Move here
8711 from lto-streamer-out.c:output_symbol_p.
8712 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
8713 (output_symbol_p): Move all logic to symtab.c
8714 (produce_symtab): Update.
8716 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8718 * config/s390/s390-opts.h (enum indirect_branch): Define.
8719 * config/s390/s390-protos.h (s390_return_addr_from_memory)
8720 (s390_indirect_branch_via_thunk)
8721 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
8722 (enum s390_indirect_branch_type): Define.
8723 * config/s390/s390.c (struct s390_frame_layout, struct
8724 machine_function): Remove.
8725 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
8726 (indirect_branch_table_label_no, indirect_branch_table_name):
8728 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
8729 (enum s390_indirect_branch_option): Define.
8730 (s390_return_addr_from_memory): New function.
8731 (s390_handle_string_attribute): New function.
8732 (s390_attribute_table): Add new attribute handler.
8733 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
8734 (s390_indirect_branch_via_thunk): New function.
8735 (s390_indirect_branch_via_inline_thunk): New function.
8736 (s390_function_ok_for_sibcall): When jumping via thunk disallow
8737 sibling call optimization for non z10 compiles.
8738 (s390_emit_call): Force indirect branch target to be a single
8739 register. Add r1 clobber for non-z10 compiles.
8740 (s390_emit_epilogue): Emit return jump via return_use expander.
8741 (s390_reorg): Handle JUMP_INSNs as execute targets.
8742 (s390_option_override_internal): Perform validity checks for the
8743 new command line options.
8744 (s390_indirect_branch_attrvalue): New function.
8745 (s390_indirect_branch_settings): New function.
8746 (s390_set_current_function): Invoke s390_indirect_branch_settings.
8747 (s390_output_indirect_thunk_function): New function.
8748 (s390_code_end): Implement target hook.
8749 (s390_case_values_threshold): Implement target hook.
8750 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
8752 * config/s390/s390.h (struct s390_frame_layout)
8753 (struct machine_function): Move here from s390.c.
8754 (TARGET_INDIRECT_BRANCH_NOBP_RET)
8755 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
8756 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
8757 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
8758 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
8759 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
8760 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
8761 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
8762 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
8763 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
8764 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
8765 (mnemonic attribute): Add values which aren't recognized
8767 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
8768 pattern for branch conversion. Fix mnemonic attribute.
8769 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
8770 indirect branch via thunk if requested.
8771 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
8772 ("*indirect_jump"): Disable for branch conversion using out of
8774 ("indirect_jump_via_thunk<mode>_z10")
8775 ("indirect_jump_via_thunk<mode>")
8776 ("indirect_jump_via_inlinethunk<mode>_z10")
8777 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
8778 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
8779 ("casesi_jump_via_inlinethunk<mode>_z10")
8780 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
8781 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
8782 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
8783 ("*indirect2_jump"): Disable for branch conversion.
8784 ("casesi_jump"): Turn into expander and expand patterns for branch
8786 ("return_use"): New expander.
8787 ("*return"): Emit return via thunk and rename it to ...
8788 ("*return<mode>"): ... this one.
8789 * config/s390/s390.opt: Add new options and and enum for the
8792 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8794 * lra-constraints.c (match_reload): Unconditionally use
8795 gen_lowpart_SUBREG, rather than selecting between that
8796 and equivalent gen_rtx_SUBREG code.
8798 2018-02-08 Richard Biener <rguenther@suse.de>
8800 PR tree-optimization/84233
8801 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
8802 changed flag instead of boguously re-using phi_inserted.
8804 2018-02-08 Martin Jambor <mjambor@suse.cz>
8806 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
8807 static local variables.
8809 2018-02-08 Richard Biener <rguenther@suse.de>
8811 PR tree-optimization/84278
8812 * tree-vect-stmts.c (vectorizable_store): When looking for
8813 smaller vector types to perform grouped strided loads/stores
8814 make sure the mode is supported by the target.
8815 (vectorizable_load): Likewise.
8817 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
8819 * config/aarch64/aarch64.c (aarch64_components_for_bb):
8820 Increase LDP/STP opportunities by adding adjacent callee-saves.
8822 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
8824 PR rtl-optimization/84068
8825 PR rtl-optimization/83459
8826 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
8828 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
8830 PR tree-optimization/84224
8831 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
8832 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
8835 2018-02-07 Iain Sandoe <iain@codesourcery.com>
8838 * config/rs6000/altivec.md (*restore_world): Remove LR use.
8839 * config/rs6000/predicates.md (restore_world_operation): Adjust op
8840 count, remove one USE.
8842 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
8844 * doc/install.texi (Configuration): Document the
8845 --with-long-double-format={ibm,ieee} PowerPC configuration
8849 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
8850 Convert from define_expand to be define_insn_and_split. Rework
8851 float/double/_Float128 conversions to QI/HI/SImode to work with
8852 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
8853 conversions to QI/HImode types did a store and then a load to
8854 truncate the value. For conversions to VSX registers, don't split
8855 the insn, instead emit the code directly. Use the code iterator
8856 any_fix to combine signed and unsigned conversions.
8857 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
8858 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
8859 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
8860 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
8861 (fix_<mode>di2_hw): Likewise.
8862 (fixuns_<mode>di2_hw): Likewise.
8863 (fix_<mode>si2_hw): Likewise.
8864 (fixuns_<mode>si2_hw): Likewise.
8865 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
8866 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
8867 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
8868 fix<uns>_trunc<SFDF:mode>si2_p8.
8869 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
8871 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
8872 (fix<uns>_<mode>_mem): Likewise.
8873 (fctiw<u>z_<mode>_mem): Likewise.
8874 (fix<uns>_<mode>_mem): Likewise.
8875 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
8876 the register allocator from doing a direct move to the GPRs to do
8877 a store, and instead use the ISA 3.0 store byte/half-word from
8878 vector register instruction. For IEEE 128-bit floating point,
8879 also optimize stores of 32-bit ints.
8880 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
8882 2018-02-07 Alan Hayward <alan.hayward@arm.com>
8884 * genextract.c (push_pathstr_operand): New function to support
8886 (walk_rtx): Call push_pathstr_operand.
8887 (print_path): Support [a-zA-Z].
8889 2018-02-07 Richard Biener <rguenther@suse.de>
8891 PR tree-optimization/84037
8892 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
8893 (cse_and_gimplify_to_preheader): Declare.
8894 (vect_get_place_in_interleaving_chain): Likewise.
8895 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
8897 (_loop_vec_info::~_loop_vec_info): Delete it.
8898 (cse_and_gimplify_to_preheader): New function.
8899 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
8900 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
8901 (vectorizable_load): Likewise. For grouped stores always base
8902 the IV on the first element.
8903 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
8904 condition before gimplifying.
8906 2018-02-07 Jakub Jelinek <jakub@redhat.com>
8908 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
8909 *DIV_EXPR and *MOD_EXPR.
8911 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
8914 * config/i386/i386.c (ix86_option_override_internal): Mask out
8915 the CF_SET bit when checking -fcf-protection.
8917 2018-02-07 Tom de Vries <tom@codesourcery.com>
8920 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
8923 2018-02-07 Richard Biener <rguenther@suse.de>
8925 PR tree-optimization/84204
8926 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
8929 PR tree-optimization/84205
8930 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
8931 special-case isl_ast_op_zdiv_r.
8933 PR tree-optimization/84223
8934 * graphite-scop-detection.c (gather_bbs::before_dom_children):
8935 Only add conditions from within the region.
8936 (gather_bbs::after_dom_children): Adjust.
8938 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
8941 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
8942 * config/avr/avr.md: Only post-reload split REG-REG moves if
8943 either register is GENERAL_REG_P.
8945 2018-02-07 Jakub Jelinek <jakub@redhat.com>
8947 PR tree-optimization/84235
8948 * tree-ssa-scopedtables.c
8949 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
8950 if the subtraction is performed in floating point type where NaNs are
8951 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
8952 build 1. Formatting fix.
8954 2018-02-06 Jakub Jelinek <jakub@redhat.com>
8957 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
8958 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
8959 and skip it regardless of bb boundaries. Use CALL_P macro,
8960 don't test INSN_P (insn) together with CALL_P or JUMP_P check
8961 unnecessarily, formatting fix.
8963 2018-02-06 Michael Collison <michael.collison@arm.com>
8965 * config/arm/thumb2.md:
8966 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
8967 (*thumb_mov_notscc): Ditto.
8969 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
8972 * config/rs6000/rs6000.md (su code attribute): Use "u" for
8973 unsigned_fix, not "s".
8975 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8977 * configure.ac (gcc_fn_eh_frame_ro): New function.
8978 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
8979 correct .eh_frame permissions.
8980 * configure: Regenerate.
8982 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
8984 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
8987 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8989 * config/rs6000/rs6000.c (rs6000_option_override_internal):
8990 Display warning message for -mno-speculate-indirect-jumps.
8992 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
8994 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
8996 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
8998 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
9000 PR tree-optimization/84225
9001 * tree-eh.c (find_trapping_overflow): Only call
9002 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
9004 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
9007 * config/i386/i386.c: Reimplement the check of possible options
9008 -mibt/-mshstk conbination. Change error messages.
9009 * doc/invoke.texi: Fix a typo: remove extra '='.
9011 2018-02-06 Marek Polacek <polacek@redhat.com>
9013 PR tree-optimization/84228
9014 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
9016 2018-02-06 Tamar Christina <tamar.christina@arm.com>
9019 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
9020 emitted arch directives.
9021 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
9022 __ARM_FEATURE_COPROC before changing architectures.
9024 2018-02-06 Richard Biener <rguenther@suse.de>
9026 * config/i386/i386.c (print_reg): Fix typo.
9027 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
9029 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
9031 * configure: Regenerate.
9033 2018-02-05 Martin Sebor <msebor@redhat.com>
9035 PR tree-optimization/83369
9036 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
9039 2018-02-05 Martin Liska <mliska@suse.cz>
9041 * doc/invoke.texi: Cherry-pick upstream r323995.
9043 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
9045 * ira.c (ira_init_register_move_cost): Adjust comment.
9047 2018-02-05 Martin Liska <mliska@suse.cz>
9049 PR gcov-profile/84137
9050 * doc/gcov.texi: Fix typo in documentation.
9052 2018-02-05 Martin Liska <mliska@suse.cz>
9054 PR gcov-profile/83879
9055 * doc/gcov.texi: Document necessity of --dynamic-list-data when
9056 using dlopen functionality.
9058 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
9060 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
9061 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
9062 _mm_maskz_range_ss, _mm_mask_range_round_ss,
9063 _mm_maskz_range_round_ss): New intrinsics.
9064 (__builtin_ia32_rangesd128_round)
9065 (__builtin_ia32_rangess128_round): Remove.
9066 (__builtin_ia32_rangesd128_mask_round,
9067 __builtin_ia32_rangess128_mask_round): New builtins.
9068 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
9069 __builtin_ia32_rangess128_round): Remove.
9070 (__builtin_ia32_rangesd128_mask_round,
9071 __builtin_ia32_rangess128_mask_round): New builtins.
9072 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
9073 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
9074 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
9075 "<round_saeonly_constraint>")): Changed to ...
9076 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
9077 "<round_saeonly_scalar_constraint>")): ... this.
9078 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
9079 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
9080 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
9081 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
9082 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
9084 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
9086 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
9088 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
9089 Remove all values except native, 8540 and 8548.
9091 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
9093 * config/i386/i386.c (ix86_output_function_return): Pass
9094 INVALID_REGNUM, instead of -1, as invalid register number to
9095 indirect_thunk_name and output_indirect_thunk.
9097 2018-02-02 Julia Koval <julia.koval@intel.com>
9099 * config.gcc: Add -march=icelake.
9100 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
9101 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
9102 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
9103 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
9104 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
9105 (processor_target_table): Add icelake.
9106 (ix86_option_override_internal): Handle new PTAs.
9107 (get_builtin_code_for_version): Handle icelake.
9108 (M_INTEL_COREI7_ICELAKE): New.
9109 (fold_builtin_cpu): Handle icelake.
9110 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
9111 * doc/invoke.texi: Add -march=icelake.
9113 2018-02-02 Julia Koval <julia.koval@intel.com>
9115 * config/i386/i386.c (ix86_option_override_internal): Change flags type
9116 to wide_int_bitmask.
9117 * wide-int-bitmask.h: New.
9119 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
9122 * config/i386/i386.md: Replace Pmode with word_mode in
9123 builtin_setjmp_setup and builtin_longjmp to support x32.
9125 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
9129 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
9131 (rs6000_supported_cpu_names): New static variable.
9132 (linux_cpu_translation_table): Likewise.
9133 (elf_platform) <cpu>: Define new static variable and use it.
9134 Translate kernel AT_PLATFORM name to canonical name if needed.
9135 Error if platform name is unknown.
9137 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
9140 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
9142 2018-02-01 Jeff Law <law@redhat.com>
9145 * config/i386/i386.c (release_scratch_register_on_entry): Add new
9146 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
9147 the scratch if RELEASE_VIA_POP is false.
9148 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
9149 If we have to save a temporary register, decrement SIZE appropriately.
9150 Pass new arguments to release_scratch_register_on_entry.
9151 (ix86_adjust_stack_and_probe): Likewise.
9152 (ix86_emit_probe_stack_range): Pass new arguments to
9153 release_scratch_register_on_entry.
9155 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
9157 PR rtl-optimization/84157
9158 * combine.c (change_zero_ext): Use REG_P predicate in
9159 front of HARD_REGISTER_P predicate.
9161 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
9163 * config/avr/avr.c (avr_option_override): Move disabling of
9164 -fdelete-null-pointer-checks to...
9165 * common/config/avr/avr-common.c (avr_option_optimization_table):
9168 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9170 PR tree-optimization/81635
9171 * tree-data-ref.c (split_constant_offset_1): For types that
9172 wrap on overflow, try to use range info to prove that wrapping
9175 2018-02-01 Renlin Li <renlin.li@arm.com>
9178 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
9180 (aarch64_register_move_cost): Likewise.
9181 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
9183 (REG_CLASS_NAMES): Likewise.
9184 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
9185 TAILCALL_ADDR_REGS. Remove IP registers.
9186 * config/aarch64/aarch64.md (Ucs): Update register constraint.
9188 2018-02-01 Richard Biener <rguenther@suse.de>
9190 * domwalk.h (dom_walker::dom_walker): Add additional constructor
9191 for specifying RPO order and allow NULL for that.
9192 * domwalk.c (dom_walker::dom_walker): Likewise.
9193 (dom_walker::walk): Handle NULL RPO order.
9194 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
9196 (rewrite_update_dom_walker): Likewise.
9197 (mark_def_dom_walker): Likewise.
9199 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9201 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
9202 (aarch64_maybe_expand_sve_subreg_move): Declare.
9203 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
9204 * config/aarch64/predicates.md (aarch64_any_register_operand): New
9206 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
9207 that are semantically a reverse operation.
9208 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
9209 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
9210 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
9212 (aarch64_can_change_mode_class): For big-endian, forbid changes
9213 between two SVE modes if they have different element sizes.
9215 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9217 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
9218 the TImode handling for big-endian targets.
9220 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9222 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
9223 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
9225 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
9226 Remove BSWAP handing for big-endian targets and use the form of
9227 LD1RQ appropariate for the mode.
9229 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9231 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
9232 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
9235 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9238 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
9239 check for operands that need to go through aarch64_sve_reload_be.
9241 2018-02-01 Jakub Jelinek <jakub@redhat.com>
9243 PR tree-optimization/81661
9244 PR tree-optimization/84117
9245 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
9246 * tree-eh.c: Include gimplify.h.
9247 (find_trapping_overflow, replace_trapping_overflow,
9248 rewrite_to_non_trapping_overflow): New functions.
9249 * tree-vect-loop.c: Include tree-eh.h.
9250 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
9251 * tree-data-ref.c: Include tree-eh.h.
9252 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
9254 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
9256 PR rtl-optimization/84123
9257 * combine.c (change_zero_ext): Check if hard register satisfies
9258 can_change_dest_mode before calling gen_lowpart_SUBREG.
9260 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
9263 * ira.c (ira_init_register_move_cost): Remove assert.
9265 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
9267 PR rtl-optimization/84071
9268 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
9269 * doc/tm.texi: Regenerate.
9271 2018-01-31 Richard Biener <rguenther@suse.de>
9273 PR tree-optimization/84132
9274 * tree-data-ref.c (analyze_miv_subscript): Properly
9275 check whether evolution_function_is_affine_multivariate_p
9276 before calling gcd_of_steps_may_divide_p.
9278 2018-01-31 Julia Koval <julia.koval@intel.com>
9281 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
9282 * config/i386/i386.md (rdpid_rex64) New.
9283 (rdpid): Make 32bit only.
9285 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
9288 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
9289 an IDENTIFIER_NODE for FUNCTION_TYPE's.
9291 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
9294 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
9296 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
9298 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
9300 PR rtl-optimization/84071
9301 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
9302 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
9304 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
9306 * config/arc/arc.c (arc_handle_aux_attribute): New function.
9307 (arc_attribute_table): Add 'aux' attribute.
9308 (arc_in_small_data_p): Consider aux like variables.
9309 (arc_is_aux_reg_p): New function.
9310 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
9311 (arc_get_aux_arg): New function.
9312 (prepare_move_operands): Handle aux-register access.
9313 (arc_handle_aux_attribute): New function.
9314 * doc/extend.texi (ARC Variable attributes): Add subsection.
9316 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
9318 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
9319 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
9320 (arc_attribute_table): Add 'uncached' attribute.
9321 (arc_print_operand): Print '.di' flag for uncached memory
9323 (arc_in_small_data_p): Do not consider for small data the uncached
9325 (arc_is_uncached_mem_p): New function.
9326 * config/arc/predicates.md (compact_store_memory_operand): Check
9327 for uncached memory accesses.
9328 (nonvol_nonimm_operand): Likewise.
9329 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
9331 2018-01-31 Jakub Jelinek <jakub@redhat.com>
9334 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
9335 falign-loops=): Add Optimization flag.
9337 2018-01-30 Jeff Law <law@redhat.com>
9340 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
9341 INT_REGISTERS_SAVED. Check it prior to calling
9342 get_scratch_register_on_entry.
9343 (ix86_adjust_stack_and_probe): Similarly.
9344 (ix86_emit_probe_stack_range): Similarly.
9345 (ix86_expand_prologue): Corresponding changes.
9347 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9350 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
9351 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
9353 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
9356 * lra-constraints.c (curr_insn_transform): Process AND in the
9359 2018-01-30 Jakub Jelinek <jakub@redhat.com>
9361 PR rtl-optimization/83986
9362 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
9363 dependence against last_pending_memory_flush in addition to
9366 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
9368 PR tree-optimization/81611
9369 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
9372 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9375 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
9378 2018-01-30 Richard Biener <rguenther@suse.de>
9379 Jakub Jelinek <jakub@redhat.com>
9381 PR tree-optimization/84111
9382 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
9383 inner loops added during recursion, as they don't have up-to-date
9386 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
9389 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
9390 (can_inline_edge_by_limits_p): ... here.
9391 (can_early_inline_edge_p, check_callers,
9392 update_caller_keys, update_callee_keys, recursive_inlining,
9393 add_new_edges_to_heap, speculation_useful_p,
9394 inline_small_functions,
9395 inline_small_functions, flatten_function,
9396 inline_to_all_callers_1): Update.
9398 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
9400 * profile-count.c (profile_count::combine_with_ipa_count): Handle
9403 2018-01-30 Richard Biener <rguenther@suse.de>
9405 PR tree-optimization/83008
9406 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
9407 invariant and constant vector uses in stmts when they need
9410 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9413 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
9414 * configure: Regenerate.
9416 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9418 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
9420 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
9421 Use gen_rtx_REG rather than gen_lowpart.
9423 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9425 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
9426 rather than 0 when creating partial subregs.
9428 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9430 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
9433 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
9436 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
9437 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
9438 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
9439 flags. This restores the settings used before the 2017-07-24.
9440 Turning off pre increment/decrement/modify allows IVOPTS to
9441 optimize DF/SF loops where the index is an int.
9443 2018-01-29 Richard Biener <rguenther@suse.de>
9444 Kelvin Nilsen <kelvin@gcc.gnu.org>
9447 * tree-vect-stmts.c (vectorizable_call): Don't call
9448 targetm.vectorize_builtin_md_vectorized_function if callee is
9451 2018-01-22 Carl Love <cel@us.ibm.com>
9453 * doc/extend.tex: Fix typo in second arg in
9454 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
9456 2018-01-29 Richard Biener <rguenther@suse.de>
9458 PR tree-optimization/84086
9459 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
9460 (flush_ssaname_freelist): When SSA names were released reset
9461 the SCEV hash table.
9463 2018-01-29 Richard Biener <rguenther@suse.de>
9465 PR tree-optimization/84057
9466 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
9467 removed paths when removing edges.
9469 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
9471 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
9472 -mfunction-return=@var{choice}.
9474 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
9477 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
9479 (layout::print_source_line): Likewise.
9480 (test_get_line_width_without_trailing_whitespace): Add test cases.
9482 2018-01-27 Jakub Jelinek <jakub@redhat.com>
9485 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
9488 2018-01-26 Jim Wilson <jimw@sifive.com>
9490 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
9492 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
9495 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9497 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
9498 and CMP + SUB-immediate -> SUBS.
9500 2018-01-26 Martin Sebor <msebor@redhat.com>
9502 PR tree-optimization/83896
9503 * tree-ssa-strlen.c (get_string_len): Rename...
9504 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
9505 Avoid assuming length is constant.
9506 (handle_char_store): Use HOST_WIDE_INT for string length.
9508 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
9511 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
9512 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
9514 2018-01-26 Richard Biener <rguenther@suse.de>
9516 PR rtl-optimization/84003
9517 * dse.c (record_store): Only record redundant stores when
9518 the earlier store aliases at least all accesses the later one does.
9520 2018-01-26 Jakub Jelinek <jakub@redhat.com>
9522 PR rtl-optimization/83985
9523 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
9524 REG_CFA_RESTORE insns.
9525 (delete_unmarked_insns): Don't ignore separate shrink wrapping
9526 REG_CFA_RESTORE insns here.
9529 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
9530 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
9532 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9534 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
9535 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
9536 (arc_init): Likewise.
9537 (arc_override_options): Likewise.
9538 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
9540 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
9542 * config/arc/arc.h (TARGET_DBNZ): Define.
9543 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
9544 properly set the tune attribute.
9545 (dbnz): Use TARGET_DBNZ guard.
9546 * config/arc/arc.opt (mtune): Add core3 option.
9548 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9550 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
9551 recognize new pic like addresses.
9552 (arc_delegitimize_address): Clean up.
9554 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9556 * config/arc/arc-arches.def: Option mrf16 valid for all
9558 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
9559 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
9560 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
9561 * config/arc/arc-tables.opt: Regenerate.
9562 * config/arc/arc.c (arc_conditional_register_usage): Handle
9563 reduced register file case.
9564 (arc_file_start): Set must have build attributes.
9565 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
9567 * config/arc/arc.opt (mrf16): Add new option.
9568 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
9569 * config/arc/genmultilib.awk: Handle new mrf16 option.
9570 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
9571 * config/arc/t-multilib: Regenerate.
9572 * doc/invoke.texi (ARC Options): Document mrf16 option.
9574 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9576 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
9577 * config/arc/arc.c (arc_handle_secure_attribute): New function.
9578 (arc_attribute_table): Add 'secure_call' attribute.
9579 (arc_print_operand): Print secure call operand.
9580 (arc_function_ok_for_sibcall): Don't optimize tail calls when
9582 (arc_is_secure_call_p): New function. * config/arc/arc.md
9583 (call_i): Add support for sjli instruction.
9584 (call_value_i): Likewise.
9585 * config/arc/constraints.md (Csc): New constraint.
9587 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9588 John Eric Martin <John.Martin@emmicro-us.com>
9590 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
9591 * config/arc/arc.c (_arc_jli_section): New struct.
9592 (arc_jli_section): New type.
9593 (rc_jli_sections): New static variable.
9594 (arc_handle_jli_attribute): New function.
9595 (arc_attribute_table): Add jli_always and jli_fixed attribute.
9596 (arc_file_end): New function.
9597 (TARGET_ASM_FILE_END): Define.
9598 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
9599 (arc_add_jli_section): New function.
9600 (jli_call_scan): Likewise.
9601 (arc_reorg): Call jli_call_scan.
9602 (arc_output_addsi): Remove 'S' from printing asm operand.
9603 (arc_is_jli_call_p): New function.
9604 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
9606 (movhi_insn): Likewise.
9607 (movsi_insn): Likewise.
9608 (movsi_set_cc_insn): Likewise.
9609 (loadqi_update): Likewise.
9610 (load_zeroextendqisi_update): Likewise.
9611 (load_signextendqisi_update): Likewise.
9612 (loadhi_update): Likewise.
9613 (load_zeroextendhisi_update): Likewise.
9614 (load_signextendhisi_update): Likewise.
9615 (loadsi_update): Likewise.
9616 (loadsf_update): Likewise.
9617 (movsicc_insn): Likewise.
9618 (bset_insn): Likewise.
9619 (bxor_insn): Likewise.
9620 (bclr_insn): Likewise.
9621 (bmsk_insn): Likewise.
9622 (bicsi3_insn): Likewise.
9623 (cmpsi_cc_c_insn): Likewise.
9624 (movsi_ne): Likewise.
9625 (movsi_cond_exec): Likewise.
9626 (clrsbsi2): Likewise.
9637 (call_i): Remove 'S' asm letter, add jli instruction.
9638 (call_value_i): Likewise.
9639 * config/arc/arc.op (mjli-always): New option.
9640 * config/arc/constraints.md (Cji): New constraint.
9641 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
9643 (subsf3_fpx): Likewise.
9644 (mulsf3_fpx): Likewise.
9645 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
9647 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
9649 * doc/invoke.texi (ARC): Document mjli-always option.
9651 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
9653 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
9654 avoid addition with 0 and use incw and decw where possible.
9656 2018-01-26 Richard Biener <rguenther@suse.de>
9658 PR tree-optimization/81082
9659 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
9660 association if it requires casting to unsigned.
9661 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
9662 from fold_plusminus_mult_expr to catch important cases late when
9663 range info is available.
9665 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9667 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
9668 * configure.ac (hidden_linkonce): New test.
9669 * configure: Regenerate.
9670 * config.in: Regenerate.
9672 2018-01-26 Julia Koval <julia.koval@intel.com>
9674 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
9675 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
9676 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
9677 _mm_mask_bitshuffle_epi64_mask): Fix type.
9678 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
9679 USI_FTYPE_V4DI_V4DI_USI): Remove.
9680 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
9681 __builtin_ia32_vpshufbitqmb256_mask,
9682 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
9683 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
9684 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
9686 2018-01-26 Alan Modra <amodra@gmail.com>
9689 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
9690 UNSPEC_VBPERMQ. Sort other unspecs.
9692 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
9694 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
9696 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
9699 * predict.c (drop_profile): Do not push/pop cfun; update also
9701 (handle_missing_profiles): Fix logic looking for zero profiles.
9703 2018-01-25 Jakub Jelinek <jakub@redhat.com>
9706 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
9707 on functions with #pragma omp declare simd or functions with simd
9709 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
9710 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
9711 Remove trailing \n from warning_at calls.
9713 2018-01-25 Tom de Vries <tom@codesourcery.com>
9716 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
9717 for neutered workers.
9719 2018-01-24 Joseph Myers <joseph@codesourcery.com>
9722 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
9723 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
9725 2018-01-24 Jeff Law <law@redhat.com>
9728 * i386.c (get_probe_interval): Move to earlier point.
9729 (ix86_compute_frame_layout): If -fstack-clash-protection and
9730 the frame is larger than the probe interval, then use pushes
9731 to save registers rather than reg->mem moves.
9732 (ix86_expand_prologue): Remove conditional for int_registers_saved
9735 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
9738 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
9739 min/max for never referenced object.
9741 2018-01-24 Jakub Jelinek <jakub@redhat.com>
9744 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
9746 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
9747 attributes from DECL_ATTRIBUTES (decl) without affecting
9748 DECL_ATTRIBUTES (current_function_decl).
9749 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
9750 functions with non-NULL DECL_ABSTRACT_ORIGIN.
9752 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
9754 PR tree-optimization/83979
9755 * fold-const.c (fold_comparison): Use constant_boolean_node
9756 instead of boolean_{true,false}_node.
9758 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
9760 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
9763 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9765 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9766 Simplify the clause that sets the length attribute.
9767 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9768 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
9769 clause that sets the length attribute.
9770 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9772 2018-01-24 Tom de Vries <tom@codesourcery.com>
9775 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
9776 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
9777 Add strict parameter.
9778 (prevent_branch_around_nothing): Insert dummy insn between branch to
9779 label and label with no ptx insn inbetween.
9780 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
9782 2018-01-24 Tom de Vries <tom@codesourcery.com>
9785 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
9786 for neutered threads in warp.
9787 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
9789 2018-01-24 Richard Biener <rguenther@suse.de>
9791 PR tree-optimization/83176
9792 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
9795 2018-01-24 Richard Biener <rguenther@suse.de>
9797 PR tree-optimization/82819
9798 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
9799 code generating pluses that are no-ops in the target precision.
9801 2018-01-24 Richard Biener <rguenther@suse.de>
9804 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
9806 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9808 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
9809 to merge probabilities.
9810 * predict.c (probably_never_executed): Also mark as cold functions
9811 with global 0 profile and guessed local profile.
9812 * profile-count.c (profile_probability::combine_with_count): New
9814 * profile-count.h (profile_probability::operator*,
9815 profile_probability::operator*=, profile_probability::operator/,
9816 profile_probability::operator/=): Reduce precision to adjusted
9817 and set value to guessed on contradictory divisions.
9818 (profile_probability::combine_with_freq): Remove.
9819 (profile_probability::combine_wiht_count): Declare.
9820 (profile_count::force_nonzero):: Set to adjusted.
9821 (profile_count::probability_in):: Set quality to adjusted.
9822 * tree-ssa-tail-merge.c (replace_block_by): Use
9825 2018-01-23 Andrew Waterman <andrew@sifive.com>
9826 Jim Wilson <jimw@sifive.com>
9828 * config/riscv/riscv.c (riscv_stack_boundary): New.
9829 (riscv_option_override): Set riscv_stack_boundary. Handle
9830 riscv_preferred_stack_boundary_arg.
9831 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
9832 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
9833 (STACK_BOUNDARY): Set to riscv_stack_boundary.
9834 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
9835 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
9836 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
9838 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
9841 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
9842 of struct ix86_frame.
9843 (ix86_expand_epilogue): Likewise. Add a local variable for
9844 the reg_save_offset field in struct ix86_frame.
9846 2018-01-23 Bin Cheng <bin.cheng@arm.com>
9848 PR tree-optimization/82604
9849 * tree-loop-distribution.c (enum partition_kind): New enum item
9850 PKIND_PARTIAL_MEMSET.
9851 (partition_builtin_p): Support above new enum item.
9852 (generate_code_for_partition): Ditto.
9853 (compute_access_range): Differentiate cases that equality can be
9854 proven at all loops, the innermost loops or no loops.
9855 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
9856 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
9857 (finalize_partitions, distribute_loop): Don't fuse partition of
9858 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
9859 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
9862 2018-01-23 Martin Liska <mliska@suse.cz>
9864 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
9865 order to ignore the predictor.
9866 (PRED_POLYMORPHIC_CALL): Likewise.
9867 (PRED_RECURSIVE_CALL): Likewise.
9869 2018-01-23 Martin Liska <mliska@suse.cz>
9871 * tree-profile.c (tree_profiling): Print function header to
9872 aware reader which function we are working on.
9873 * value-prof.c (gimple_find_values_to_profile): Do not print
9874 not interesting value histograms.
9876 2018-01-23 Martin Liska <mliska@suse.cz>
9878 * profile-count.h (enum profile_quality): Add
9879 profile_uninitialized as the first value. Do not number values
9880 as they are zero based.
9881 (profile_count::verify): Update sanity check.
9882 (profile_probability::verify): Likewise.
9884 2018-01-23 Nathan Sidwell <nathan@acm.org>
9886 * doc/invoke.texi (ffor-scope): Deprecate.
9888 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9890 PR tree-optimization/83510
9891 * domwalk.c (set_all_edges_as_executable): New function.
9892 (dom_walker::dom_walker): Convert bool param
9893 "skip_unreachable_blocks" to enum reachability. Move setup of
9894 edge flags to set_all_edges_as_executable and only do it when
9895 reachability is REACHABLE_BLOCKS.
9896 * domwalk.h (enum dom_walker::reachability): New enum.
9897 (dom_walker::dom_walker): Convert bool param
9898 "skip_unreachable_blocks" to enum reachability.
9899 (set_all_edges_as_executable): New decl.
9900 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
9901 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
9903 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
9904 but converting true to REACHABLE_BLOCKS.
9905 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
9907 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
9908 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
9909 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
9911 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
9912 if check_all_array_refs will be called.
9914 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9916 * tree.c (selftest::test_location_wrappers): Add more test
9919 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9921 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
9922 (selftest::test_bit_in_range): Likewise.
9924 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
9927 * doc/sourcebuild.texi (vect_float): Say that the selector
9928 only describes the situation when -funsafe-math-optimizations is on.
9929 (vect_float_strict): Document.
9931 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
9933 PR tree-optimization/83965
9934 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
9935 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
9936 instead of checking only for a reduction.
9937 (vect_recog_widen_sum_pattern): Likewise.
9939 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9941 * predict.c (probably_never_executed): Only use precise profile info.
9942 (compute_function_frequency): Skip after inlining hack since we now
9943 have quality checking.
9945 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9947 * profile-count.h (profile_probability::very_unlikely,
9948 profile_probability::unlikely, profile_probability::even): Set
9949 precision to guessed.
9951 2018-01-23 Richard Biener <rguenther@suse.de>
9953 PR tree-optimization/83963
9954 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
9955 Properly terminate dominator walk when crossing the exit edge not
9956 when visiting its source block.
9958 2018-01-23 Jakub Jelinek <jakub@redhat.com>
9961 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
9962 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
9964 2018-01-22 Jakub Jelinek <jakub@redhat.com>
9966 PR tree-optimization/83957
9967 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
9968 semicolon after for body surrounded by braces.
9970 PR tree-optimization/83081
9971 * profile-count.h (profile_probability::split): New method.
9972 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
9973 Use profile_probability::split.
9974 (do_compare_rtx_and_jump): Fix adjustment of probabilities
9975 when splitting a single conditional jump into 2.
9977 2018-01-22 David Malcolm <dmalcolm@redhat.com>
9979 PR tree-optimization/69452
9980 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
9983 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9985 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
9986 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
9987 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
9989 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9991 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
9992 * config/rl78/rl78.md (movdi): New define_expand.
9993 * config/rl78/rl78.c (rl78_split_movdi): New function.
9995 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
9998 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
10000 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
10001 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
10002 128-bit to produce an UNSPEC move to get the double word with the
10003 signbit and then a shift directly to do signbit.
10004 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
10005 implementation with a new version that just does either a direct
10006 move or a regular move. Move memory interface to separate insns.
10007 Move insns so they are next to the expander.
10008 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
10009 with signbit move. Split big and little endian case.
10010 (signbit<mode>2_dm_mem_le): Likewise.
10011 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
10012 (signbit<mode>2_dm2): Likewise.
10014 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10016 * config/rl78/rl78.md (anddi3): New define_expand.
10018 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10020 * config/rl78/rl78.md (umindi3): New define_expand.
10022 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10024 * config/rl78/rl78.md (smindi3): New define_expand.
10026 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10028 * config/rl78/rl78.md (smaxdi3): New define_expand.
10030 2018-01-22 Carl Love <cel@us.ibm.com>
10032 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
10033 LVX_V1TI): Add macro expansion.
10034 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
10035 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
10036 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
10037 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
10038 Change check to determine if the instruction is a byte reversing
10039 entry. Fix typo in comment.
10040 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
10041 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
10042 Add def_builtin calls for new builtins.
10043 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
10044 Add define_insn expansion.
10046 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10048 * config/rl78/rl78.md (umaxdi3): New define_expand.
10050 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10052 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
10053 for non-QImode registers.
10055 2018-01-22 Richard Biener <rguenther@suse.de>
10057 PR tree-optimization/83963
10058 * graphite-scop-detection.c (scop_detection::get_sese): Delay
10059 including the loop exit block.
10060 (scop_detection::merge_sese): Likewise.
10061 (scop_detection::add_scop): Do it here instead.
10063 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10065 * doc/sourcebuild.texi (arm_softfloat): Document.
10067 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
10070 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
10071 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
10072 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
10074 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10075 David Edelsohn <dje.gcc@gmail.com>
10078 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
10079 Change "crset eq" to "crset 2".
10080 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
10081 (*call_indirect_aix<mode>_nospec): Likewise.
10082 (*call_value_indirect_aix<mode>_nospec): Likewise.
10083 (*call_indirect_elfv2<mode>_nospec): Likewise.
10084 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
10085 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
10086 change assembly output from . to $.
10087 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
10088 (indirect_jump<mode>_nospec): Change assembly output from . to $.
10089 (*tablejump<mode>_internal1_nospec): Likewise.
10091 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
10094 * config/sh/sh_optimize_sett_clrt.cc:
10095 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
10097 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
10099 PR tree-optimization/83940
10100 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
10101 offset_dt to vect_constant_def rather than vect_unknown_def_type.
10102 (vect_check_load_store_mask): Add a mask_dt_out parameter and
10103 use it to pass back the definition type.
10104 (vect_check_store_rhs): Likewise rhs_dt_out.
10105 (vect_build_gather_load_calls): Add a mask_dt argument and use
10106 it instead of a call to vect_is_simple_use.
10107 (vectorizable_store): Update calls to vect_check_load_store_mask
10108 and vect_check_store_rhs. Use the dt returned by the latter instead
10109 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
10110 instead of calls to vect_is_simple_use. Pass the scalar rather
10111 than the vector operand to vect_is_simple_use when handling
10112 second and subsequent copies of an rhs value.
10113 (vectorizable_load): Update calls to vect_check_load_store_mask
10114 and vect_build_gather_load_calls. Use the cached mask_dt and
10115 gs_info.offset_dt instead of calls to vect_is_simple_use.
10117 2018-01-20 Jakub Jelinek <jakub@redhat.com>
10119 PR middle-end/83945
10120 * tree-emutls.c: Include gimplify.h.
10121 (lower_emutls_2): New function.
10122 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
10123 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
10124 it before further processing.
10127 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
10128 UINTVAL (trueop1) instead of INTVAL (op1).
10130 2018-01-19 Jakub Jelinek <jakub@redhat.com>
10134 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
10135 INCOMING_FRAME_SP_OFFSET if not defined.
10136 (scan_trace): Add ENTRY argument. If true and
10137 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
10138 emit a note to adjust the CFA offset.
10139 (create_cfi_notes): Adjust scan_trace callers.
10140 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
10141 INCOMING_FRAME_SP_OFFSET in the CIE.
10142 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
10143 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
10145 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
10146 * doc/tm.texi: Regenerated.
10148 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
10150 PR rtl-optimization/83147
10151 * lra-constraints.c (remove_inheritance_pseudos): Use
10152 lra_substitute_pseudo_within_insn.
10154 2018-01-19 Tom de Vries <tom@codesourcery.com>
10155 Cesar Philippidis <cesar@codesourcery.com>
10158 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
10160 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
10163 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
10164 spaces for function labels.
10166 2018-01-19 Martin Liska <mliska@suse.cz>
10168 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
10169 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
10170 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
10171 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
10172 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
10173 (PRED_CONST_RETURN): Change from 69 to 65.
10174 (PRED_NULL_RETURN): Change from 91 to 71.
10175 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
10176 (PRED_LOOP_GUARD): Change from 66 to 73.
10178 2018-01-19 Martin Liska <mliska@suse.cz>
10180 * predict.c (predict_insn_def): Add new assert.
10181 (struct branch_predictor): Change type to signed integer.
10182 (test_prediction_value_range): Amend test to cover
10183 PROB_UNINITIALIZED.
10184 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
10185 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
10186 (PRED_LOOP_ITERATIONS_MAX): Likewise.
10187 (PRED_LOOP_IV_COMPARE): Likewise.
10188 * predict.h (PROB_UNINITIALIZED): Define new constant.
10190 2018-01-19 Martin Liska <mliska@suse.cz>
10192 * predict.c (dump_prediction): Add new format for
10193 analyze_brprob.py script which is enabled with -details
10195 * profile-count.h (precise_p): New function.
10197 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
10199 PR tree-optimization/83922
10200 * tree-vect-loop.c (vect_verify_full_masking): Return false if
10201 there are no statements that need masking.
10202 (vect_active_double_reduction_p): New function.
10203 (vect_analyze_loop_operations): Use it when handling phis that
10204 are not in the loop header.
10206 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
10208 PR tree-optimization/83914
10209 * tree-vect-loop.c (vectorizable_induction): Don't convert
10210 init_expr or apply the peeling adjustment for inductions
10211 that are nested within the vectorized loop.
10213 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10215 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
10218 2018-01-18 Jakub Jelinek <jakub@redhat.com>
10222 * function.h (gimplify_parameters): Add gimple_seq * argument.
10223 * function.c: Include gimple.h and options.h.
10224 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
10225 for the added local temporaries if needed.
10226 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
10227 if there are any parameter cleanups, wrap whole body into a
10228 try/finally with the cleanups.
10230 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
10233 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
10234 Use GET_MODE_CLASS for scalar floating point.
10236 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
10240 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
10241 Fix call of call_cgraph_insertion_hooks.
10243 2018-01-18 Martin Sebor <msebor@redhat.com>
10245 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
10247 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
10250 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
10253 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
10256 * common.opt: (-ffile-prefix-map): New option.
10257 * opts.c (common_handle_option): Defer it.
10258 * opts-global.c (handle_common_deferred_options): Handle it.
10259 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
10260 * file-prefix-map.h: New file.
10261 (remap_debug_filename, add_debug_prefix_map): ...here.
10262 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
10263 * final.c (debug_prefix_map, add_debug_prefix_map
10264 remap_debug_filename): Move to...
10265 * file-prefix-map.c: New file.
10266 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
10267 generalize, get rid of alloca(), use strrchr() instead of strchr().
10268 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
10269 Implement in terms of add_prefix_map().
10270 (remap_macro_filename, remap_debug_filename): Implement in term of
10272 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
10273 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
10274 * dbxout.c: Include file-prefix-map.h.
10275 * varasm.c: Likewise.
10276 * vmsdbgout.c: Likewise.
10277 * xcoffout.c: Likewise.
10278 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
10279 * doc/cppopts.texi (-fmacro-prefix-map): Document.
10280 * doc/invoke.texi (-ffile-prefix-map): Document.
10281 (-fdebug-prefix-map): Update description.
10283 2018-01-18 Martin Liska <mliska@suse.cz>
10285 * config/i386/i386.c (indirect_thunk_name): Document that also
10287 (output_indirect_thunk): Document why both instructions
10288 (pause and lfence) are generated.
10290 2018-01-18 Richard Biener <rguenther@suse.de>
10292 PR tree-optimization/83887
10293 * graphite-scop-detection.c
10294 (scop_detection::get_nearest_dom_with_single_entry): Remove.
10295 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
10296 (scop_detection::merge_sese): Re-implement with a flood-fill
10297 algorithm that properly finds a SESE region if it exists.
10299 2018-01-18 Jakub Jelinek <jakub@redhat.com>
10302 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
10303 pointer_diff optimizations use view_convert instead of convert.
10305 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10307 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
10308 Generate different code for -mno-speculate-indirect-jumps.
10309 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
10310 (*call_indirect_aix<mode>): Disable for
10311 -mno-speculate-indirect-jumps.
10312 (*call_indirect_aix<mode>_nospec): New define_insn.
10313 (*call_value_indirect_aix<mode>): Disable for
10314 -mno-speculate-indirect-jumps.
10315 (*call_value_indirect_aix<mode>_nospec): New define_insn.
10316 (*sibcall_nonlocal_sysv<mode>): Generate different code for
10317 -mno-speculate-indirect-jumps.
10318 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
10320 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
10322 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
10323 long double type, set the flags for noting the default long double
10324 type, even if we don't pass or return a long double type.
10326 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
10329 * ipa-inline.c (flatten_function): Do not overwrite final inlining
10332 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
10334 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
10335 support for merge[hl].
10336 (fold_mergehl_helper): New helper function.
10337 (tree-vector-builder.h): New #include for tree_vector_builder usage.
10338 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
10339 (altivec_vmrglw_direct): Add xxmrglw insn.
10341 2018-01-17 Andrew Waterman <andrew@sifive.com>
10343 * config/riscv/riscv.c (riscv_conditional_register_usage): If
10344 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
10346 2018-01-17 David Malcolm <dmalcolm@redhat.com>
10349 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
10350 call the lto_location_cache before reading the
10351 DECL_SOURCE_LOCATION of the types.
10353 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
10354 Richard Sandiford <richard.sandiford@linaro.org>
10356 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
10357 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
10358 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
10359 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
10360 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
10362 * config/aarch64/constraints.md (aarch64_movti_operand):
10364 * config/aarch64/predicates.md (Uti): Add new constraint.
10366 2018-01-17 Carl Love <cel@us.ibm.com>
10368 * config/rs6000/vsx.md (define_expand xl_len_r,
10369 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
10370 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
10372 (define_expand, define_insn): Move the shift left from the
10373 define_insn to the define_expand for lxvl and stxvl instructions.
10374 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
10375 and XL_LEN_R definitions to PURE.
10377 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
10379 * config/i386/i386.c (indirect_thunk_name): Declare regno
10380 as unsigned int. Compare regno with INVALID_REGNUM.
10381 (output_indirect_thunk): Ditto.
10382 (output_indirect_thunk_function): Ditto.
10383 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
10384 in the call to output_indirect_thunk_function.
10386 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
10388 PR middle-end/83884
10389 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
10390 rather than the size of inner_type to determine the stack slot size
10391 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
10393 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
10396 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
10399 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
10401 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
10402 endian Linux systems to optionally enable multilibs for selecting
10403 the long double type if the user configured an explicit type.
10404 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
10405 have no long double multilibs if not defined.
10406 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
10407 warn if the user used -mabi={ieee,ibm}longdouble and we built
10408 multilibs for long double.
10409 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
10410 appropriate multilib option.
10411 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
10413 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
10414 for building long double multilibs.
10415 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
10417 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
10419 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
10422 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
10424 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
10427 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
10430 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
10433 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
10435 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
10436 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
10438 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
10440 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
10441 different rtl trees depending on TARGET_64BIT.
10442 (rs6000_gen_lvx): Likewise.
10444 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
10446 * config/visium/visium.md (nop): Tweak comment.
10447 (hazard_nop): Likewise.
10449 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10451 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
10452 -mspeculate-indirect-jumps.
10453 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
10454 for -mno-speculate-indirect-jumps.
10455 (*call_indirect_elfv2<mode>_nospec): New define_insn.
10456 (*call_value_indirect_elfv2<mode>): Disable for
10457 -mno-speculate-indirect-jumps.
10458 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
10459 (indirect_jump): Emit different RTL for
10460 -mno-speculate-indirect-jumps.
10461 (*indirect_jump<mode>): Disable for
10462 -mno-speculate-indirect-jumps.
10463 (*indirect_jump<mode>_nospec): New define_insn.
10464 (tablejump): Emit different RTL for
10465 -mno-speculate-indirect-jumps.
10466 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
10467 (tablejumpsi_nospec): New define_expand.
10468 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
10469 (tablejumpdi_nospec): New define_expand.
10470 (*tablejump<mode>_internal1): Disable for
10471 -mno-speculate-indirect-jumps.
10472 (*tablejump<mode>_internal1_nospec): New define_insn.
10473 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
10476 2018-01-16 Artyom Skrobov tyomitch@gmail.com
10478 * caller-save.c (insert_save): Drop unnecessary parameter. All
10481 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10482 Richard Biener <rguenth@suse.de>
10485 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
10486 return early, inline manually is_gimple_sizepos. Make sure if we
10487 call gimplify_expr we don't end up with a gimple constant.
10488 * tree.c (variably_modified_type_p): Don't return true for
10489 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
10490 * gimplify.h (is_gimple_sizepos): Remove.
10492 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10494 PR tree-optimization/83857
10495 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
10496 vectorizable_live_operation for pure SLP statements.
10497 (vectorizable_live_operation): Handle PHIs.
10499 2018-01-16 Richard Biener <rguenther@suse.de>
10501 PR tree-optimization/83867
10502 * tree-vect-stmts.c (vect_transform_stmt): Precompute
10503 nested_in_vect_loop_p since the scalar stmt may get invalidated.
10505 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10508 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
10509 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
10510 If off is not INTEGER_CST, issue a may not be aligned warning
10511 rather than isn't aligned. Use isn%'t rather than isn't.
10512 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
10514 <case MULT_EXPR>: Improve the case when bottom and one of the
10515 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
10516 operand, in that case check if the other operand is multiple of
10517 bottom divided by the INTEGER_CST operand.
10519 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10522 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
10523 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
10524 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
10525 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
10526 * config/pa/pa.c (pa_function_arg_advance): Likewise.
10527 (pa_function_arg, pa_arg_partial_bytes): Likewise.
10528 (pa_function_arg_size): New function.
10530 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10532 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
10533 in a separate statement.
10535 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10537 PR tree-optimization/83847
10538 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
10539 group gathers and scatters.
10541 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10543 PR rtl-optimization/86620
10544 * params.def (max-sched-ready-insns): Bump minimum value to 1.
10546 PR rtl-optimization/83213
10547 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
10548 to last if both are JUMP_INSNs.
10550 PR tree-optimization/83843
10551 * gimple-ssa-store-merging.c
10552 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
10553 store_immediate_info for bswap/nop orig_stores.
10555 2018-01-15 Andrew Waterman <andrew@sifive.com>
10557 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
10559 <UDIV>: Increase cost if !TARGET_DIV.
10561 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
10563 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
10564 (define_attr "cr_logical_3op"): New.
10565 (cceq_ior_compare): Adjust.
10566 (cceq_ior_compare_complement): Adjust.
10567 (*cceq_rev_compare): Adjust.
10568 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
10569 (is_cracked_insn): Adjust.
10570 (insn_must_be_first_in_group): Adjust.
10571 * config/rs6000/40x.md: Adjust.
10572 * config/rs6000/440.md: Adjust.
10573 * config/rs6000/476.md: Adjust.
10574 * config/rs6000/601.md: Adjust.
10575 * config/rs6000/603.md: Adjust.
10576 * config/rs6000/6xx.md: Adjust.
10577 * config/rs6000/7450.md: Adjust.
10578 * config/rs6000/7xx.md: Adjust.
10579 * config/rs6000/8540.md: Adjust.
10580 * config/rs6000/cell.md: Adjust.
10581 * config/rs6000/e300c2c3.md: Adjust.
10582 * config/rs6000/e500mc.md: Adjust.
10583 * config/rs6000/e500mc64.md: Adjust.
10584 * config/rs6000/e5500.md: Adjust.
10585 * config/rs6000/e6500.md: Adjust.
10586 * config/rs6000/mpc.md: Adjust.
10587 * config/rs6000/power4.md: Adjust.
10588 * config/rs6000/power5.md: Adjust.
10589 * config/rs6000/power6.md: Adjust.
10590 * config/rs6000/power7.md: Adjust.
10591 * config/rs6000/power8.md: Adjust.
10592 * config/rs6000/power9.md: Adjust.
10593 * config/rs6000/rs64.md: Adjust.
10594 * config/rs6000/titan.md: Adjust.
10596 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10598 * config/i386/predicates.md (indirect_branch_operand): Rewrite
10599 ix86_indirect_branch_register logic.
10601 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10603 * config/i386/constraints.md (Bs): Update
10604 ix86_indirect_branch_register check. Don't check
10605 ix86_indirect_branch_register with GOT_memory_operand.
10607 * config/i386/predicates.md (GOT_memory_operand): Don't check
10608 ix86_indirect_branch_register here.
10609 (GOT32_symbol_operand): Likewise.
10611 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10613 * config/i386/predicates.md (constant_call_address_operand):
10614 Rewrite ix86_indirect_branch_register logic.
10615 (sibcall_insn_operand): Likewise.
10617 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10619 * config/i386/constraints.md (Bs): Replace
10620 ix86_indirect_branch_thunk_register with
10621 ix86_indirect_branch_register.
10623 * config/i386/i386.md (indirect_jump): Likewise.
10624 (tablejump): Likewise.
10625 (*sibcall_memory): Likewise.
10626 (*sibcall_value_memory): Likewise.
10627 Peepholes of indirect call and jump via memory: Likewise.
10628 * config/i386/i386.opt: Likewise.
10629 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10630 (GOT_memory_operand): Likewise.
10631 (call_insn_operand): Likewise.
10632 (sibcall_insn_operand): Likewise.
10633 (GOT32_symbol_operand): Likewise.
10635 2018-01-15 Jakub Jelinek <jakub@redhat.com>
10637 PR middle-end/83837
10638 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
10639 type rather than type addr's type points to.
10640 (expand_omp_atomic_mutex): Likewise.
10641 (expand_omp_atomic): Likewise.
10643 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10646 * config/i386/i386.c (output_indirect_thunk_function): Use
10647 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
10648 for __x86_return_thunk.
10650 2018-01-15 Richard Biener <rguenther@suse.de>
10652 PR middle-end/83850
10653 * expmed.c (extract_bit_field_1): Fix typo.
10655 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10658 * config/arm/iterators.md (VF): New mode iterator.
10659 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
10660 Remove integer-related logic from pattern.
10661 (neon_vabd<mode>_3): Likewise.
10663 2018-01-15 Jakub Jelinek <jakub@redhat.com>
10665 PR middle-end/82694
10666 * common.opt (fstrict-overflow): No longer an alias.
10667 (fwrapv-pointer): New option.
10668 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
10669 also for pointer types based on flag_wrapv_pointer.
10670 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
10671 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
10672 opts->x_flag_wrapv got set.
10673 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
10674 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
10675 POINTER_TYPE_OVERFLOW_UNDEFINED.
10676 * match.pd: Likewise in address comparison pattern.
10677 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
10679 2018-01-15 Richard Biener <rguenther@suse.de>
10682 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
10683 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
10684 Reset type names to their identifier if their TYPE_DECL doesn't
10685 have linkage (and thus is used for ODR and devirt).
10686 (save_debug_info_for_decl): Remove.
10687 (save_debug_info_for_type): Likewise.
10688 (add_tree_to_fld_list): Adjust.
10689 * tree-pretty-print.c (dump_generic_node): Make dumping of
10690 type names more robust.
10692 2018-01-15 Richard Biener <rguenther@suse.de>
10694 * BASE-VER: Bump to 8.0.1.
10696 2018-01-14 Martin Sebor <msebor@redhat.com>
10699 * builtins.c (check_access): Avoid warning when the no-warning bit
10702 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
10704 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
10705 * ira-color (allocno_hard_regs_compare): Likewise.
10707 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
10710 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
10711 Use .pushsection/.popsection.
10713 2018-01-14 Martin Sebor <msebor@redhat.com>
10716 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
10718 2018-01-14 Jakub Jelinek <jakub@redhat.com>
10720 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
10721 entry from extra_headers.
10722 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
10723 extra_headers, make the list bitwise identical to the i?86-*-* one.
10725 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10727 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
10728 -mcmodel=large with -mindirect-branch=thunk,
10729 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
10730 -mfunction-return=thunk-extern.
10731 * doc/invoke.texi: Document -mcmodel=large is incompatible with
10732 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
10733 -mfunction-return=thunk and -mfunction-return=thunk-extern.
10735 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10737 * config/i386/i386.c (print_reg): Print the name of the full
10738 integer register without '%'.
10739 (ix86_print_operand): Handle 'V'.
10740 * doc/extend.texi: Document 'V' modifier.
10742 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10744 * config/i386/constraints.md (Bs): Disallow memory operand for
10745 -mindirect-branch-register.
10747 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10748 (GOT_memory_operand): Likewise.
10749 (call_insn_operand): Likewise.
10750 (sibcall_insn_operand): Likewise.
10751 (GOT32_symbol_operand): Likewise.
10752 * config/i386/i386.md (indirect_jump): Call convert_memory_address
10753 for -mindirect-branch-register.
10754 (tablejump): Likewise.
10755 (*sibcall_memory): Likewise.
10756 (*sibcall_value_memory): Likewise.
10757 Disallow peepholes of indirect call and jump via memory for
10758 -mindirect-branch-register.
10759 (*call_pop): Replace m with Bw.
10760 (*call_value_pop): Likewise.
10761 (*sibcall_pop_memory): Replace m with Bs.
10762 * config/i386/i386.opt (mindirect-branch-register): New option.
10763 * doc/invoke.texi: Document -mindirect-branch-register option.
10765 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10767 * config/i386/i386-protos.h (ix86_output_function_return): New.
10768 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
10769 set function_return_type.
10770 (indirect_thunk_name): Add ret_p to indicate thunk for function
10772 (output_indirect_thunk_function): Pass false to
10773 indirect_thunk_name.
10774 (ix86_output_indirect_branch_via_reg): Likewise.
10775 (ix86_output_indirect_branch_via_push): Likewise.
10776 (output_indirect_thunk_function): Create alias for function
10777 return thunk if regno < 0.
10778 (ix86_output_function_return): New function.
10779 (ix86_handle_fndecl_attribute): Handle function_return.
10780 (ix86_attribute_table): Add function_return.
10781 * config/i386/i386.h (machine_function): Add
10782 function_return_type.
10783 * config/i386/i386.md (simple_return_internal): Use
10784 ix86_output_function_return.
10785 (simple_return_internal_long): Likewise.
10786 * config/i386/i386.opt (mfunction-return=): New option.
10787 (indirect_branch): Mention -mfunction-return=.
10788 * doc/extend.texi: Document function_return function attribute.
10789 * doc/invoke.texi: Document -mfunction-return= option.
10791 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10793 * config/i386/i386-opts.h (indirect_branch): New.
10794 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
10795 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
10796 with local indirect jump when converting indirect call and jump.
10797 (ix86_set_indirect_branch_type): New.
10798 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
10799 (indirectlabelno): New.
10800 (indirect_thunk_needed): Likewise.
10801 (indirect_thunk_bnd_needed): Likewise.
10802 (indirect_thunks_used): Likewise.
10803 (indirect_thunks_bnd_used): Likewise.
10804 (INDIRECT_LABEL): Likewise.
10805 (indirect_thunk_name): Likewise.
10806 (output_indirect_thunk): Likewise.
10807 (output_indirect_thunk_function): Likewise.
10808 (ix86_output_indirect_branch_via_reg): Likewise.
10809 (ix86_output_indirect_branch_via_push): Likewise.
10810 (ix86_output_indirect_branch): Likewise.
10811 (ix86_output_indirect_jmp): Likewise.
10812 (ix86_code_end): Call output_indirect_thunk_function if needed.
10813 (ix86_output_call_insn): Call ix86_output_indirect_branch if
10815 (ix86_handle_fndecl_attribute): Handle indirect_branch.
10816 (ix86_attribute_table): Add indirect_branch.
10817 * config/i386/i386.h (machine_function): Add indirect_branch_type
10818 and has_local_indirect_jump.
10819 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
10821 (tablejump): Likewise.
10822 (*indirect_jump): Use ix86_output_indirect_jmp.
10823 (*tablejump_1): Likewise.
10824 (simple_return_indirect_internal): Likewise.
10825 * config/i386/i386.opt (mindirect-branch=): New option.
10826 (indirect_branch): New.
10829 (thunk-inline): Likewise.
10830 (thunk-extern): Likewise.
10831 * doc/extend.texi: Document indirect_branch function attribute.
10832 * doc/invoke.texi: Document -mindirect-branch= option.
10834 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
10837 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
10839 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
10841 * ipa-inline.c (want_inline_small_function_p): Return false if
10842 inlining has already failed with CIF_FINAL_ERROR.
10843 (update_caller_keys): Call want_inline_small_function_p before
10845 (update_callee_keys): Likewise.
10847 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
10849 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
10851 (rs6000_quadword_masked_address_p): Likewise.
10852 (quad_aligned_load_p): Likewise.
10853 (quad_aligned_store_p): Likewise.
10854 (const_load_sequence_p): Add comment to describe the outer-most loop.
10855 (mimic_memory_attributes_and_flags): New function.
10856 (rs6000_gen_stvx): Likewise.
10857 (replace_swapped_aligned_store): Likewise.
10858 (rs6000_gen_lvx): Likewise.
10859 (replace_swapped_aligned_load): Likewise.
10860 (replace_swapped_load_constant): Capitalize argument name in
10861 comment describing this function.
10862 (rs6000_analyze_swaps): Add a third pass to search for vector loads
10863 and stores that access quad-word aligned addresses and replace
10864 with stvx or lvx instructions when appropriate.
10865 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
10866 New function prototype.
10867 (rs6000_quadword_masked_address_p): Likewise.
10868 (rs6000_gen_lvx): Likewise.
10869 (rs6000_gen_stvx): Likewise.
10870 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
10871 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
10872 when memory address is aligned.
10873 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
10874 this split to select lvx instruction when memory address is aligned.
10875 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
10876 instruction when memory address is aligned.
10877 (*vsx_le_perm_load_v16qi): Likewise.
10878 (four unnamed splitters): Modify to select the stvx instruction
10879 when memory is aligned.
10881 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
10883 * predict.c (determine_unlikely_bbs): Handle correctly BBs
10884 which appears in the queue multiple times.
10886 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10887 Alan Hayward <alan.hayward@arm.com>
10888 David Sherwood <david.sherwood@arm.com>
10890 * tree-vectorizer.h (vec_lower_bound): New structure.
10891 (_loop_vec_info): Add check_nonzero and lower_bounds.
10892 (LOOP_VINFO_CHECK_NONZERO): New macro.
10893 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
10894 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
10895 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
10896 fields. Make seg_len the distance travelled, not including the
10898 (dr_direction_indicator): Declare.
10899 (dr_zero_step_indicator): Likewise.
10900 (dr_known_forward_stride_p): Likewise.
10901 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
10903 (runtime_alias_check_p): Allow runtime alias checks with
10905 (operator ==): Compare access_size and align.
10906 (prune_runtime_alias_test_list): Rework for new distinction between
10907 the access_size and seg_len.
10908 (create_intersect_range_checks_index): Likewise. Cope with polynomial
10910 (get_segment_min_max): New function.
10911 (create_intersect_range_checks): Use it.
10912 (dr_step_indicator): New function.
10913 (dr_direction_indicator): Likewise.
10914 (dr_zero_step_indicator): Likewise.
10915 (dr_known_forward_stride_p): Likewise.
10916 * tree-loop-distribution.c (data_ref_segment_size): Return
10917 DR_STEP * (niters - 1).
10918 (compute_alias_check_pairs): Update call to the dr_with_seg_len
10920 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
10921 (vect_preserves_scalar_order_p): New function, split out from...
10922 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
10923 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
10924 (vect_vfa_access_size): New function.
10925 (vect_vfa_align): Likewise.
10926 (vect_compile_time_alias): Take access_size_a and access_b arguments.
10927 (dump_lower_bound): New function.
10928 (vect_check_lower_bound): Likewise.
10929 (vect_small_gap_p): Likewise.
10930 (vectorizable_with_step_bound_p): Likewise.
10931 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
10932 depencies if the vectorization factor is 1. Convert the checks
10933 for nonzero steps into checks on the bounds of DR_STEP. Try using
10934 a bunds check for variable steps if the minimum required step is
10935 relatively small. Update calls to the dr_with_seg_len
10936 constructor and to vect_compile_time_alias.
10937 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
10939 (vect_loop_versioning): Call it.
10940 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
10942 (vect_estimate_min_profitable_iters): Account for any bounds checks.
10944 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10945 Alan Hayward <alan.hayward@arm.com>
10946 David Sherwood <david.sherwood@arm.com>
10948 * doc/sourcebuild.texi (vect_scatter_store): Document.
10949 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
10951 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
10953 * genopinit.c (main): Add supports_vec_scatter_store and
10954 supports_vec_scatter_store_cached to target_optabs.
10955 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
10956 IFN_MASK_SCATTER_STORE.
10957 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
10959 * internal-fn.h (internal_store_fn_p): Declare.
10960 (internal_fn_stored_value_index): Likewise.
10961 * internal-fn.c (scatter_store_direct): New macro.
10962 (expand_scatter_store_optab_fn): New function.
10963 (direct_scatter_store_optab_supported_p): New macro.
10964 (internal_store_fn_p): New function.
10965 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
10966 IFN_MASK_SCATTER_STORE.
10967 (internal_fn_mask_index): Likewise.
10968 (internal_fn_stored_value_index): New function.
10969 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
10970 for scatter stores.
10971 * optabs-query.h (supports_vec_scatter_store_p): Declare.
10972 * optabs-query.c (supports_vec_scatter_store_p): New function.
10973 * tree-vectorizer.h (vect_get_store_rhs): Declare.
10974 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
10975 true for scatter stores.
10976 (vect_gather_scatter_fn_p): Handle scatter stores too.
10977 (vect_check_gather_scatter): Consider using scatter stores if
10978 supports_vec_scatter_store_p.
10979 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
10980 scatter stores too.
10981 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
10982 internal_fn_stored_value_index.
10983 (check_load_store_masking): Handle scatter stores too.
10984 (vect_get_store_rhs): Make public.
10985 (vectorizable_call): Use internal_store_fn_p.
10986 (vectorizable_store): Handle scatter store internal functions.
10987 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
10988 when deciding whether the end of the group has been reached.
10989 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
10990 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
10991 (mask_scatter_store<mode>): New insns.
10993 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10994 Alan Hayward <alan.hayward@arm.com>
10995 David Sherwood <david.sherwood@arm.com>
10997 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
10998 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
10999 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
11001 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
11002 Use vect_truncate_gather_scatter_offset if we can't treat the
11003 operation as a normal gather load or scatter store.
11004 (get_group_load_store_type): Take the gather_scatter_info
11005 as argument. Try using a gather load or scatter store for
11006 single-element groups.
11007 (get_load_store_type): Update calls to get_group_load_store_type
11008 and vect_use_strided_gather_scatters_p.
11010 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11011 Alan Hayward <alan.hayward@arm.com>
11012 David Sherwood <david.sherwood@arm.com>
11014 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
11015 optional tree argument.
11016 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
11018 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
11019 but continue to use the current value as a fallback.
11020 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
11021 to compare the updates.
11022 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
11023 (get_load_store_type): Use it when handling a strided access.
11024 (vect_get_strided_load_store_ops): New function.
11025 (vect_get_data_ptr_increment): Likewise.
11026 (vectorizable_load): Handle strided gather loads. Always pass
11027 a step to vect_create_data_ref_ptr and bump_vector_ptr.
11029 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11030 Alan Hayward <alan.hayward@arm.com>
11031 David Sherwood <david.sherwood@arm.com>
11033 * doc/md.texi (gather_load@var{m}): Document.
11034 (mask_gather_load@var{m}): Likewise.
11035 * genopinit.c (main): Add supports_vec_gather_load and
11036 supports_vec_gather_load_cached to target_optabs.
11037 * optabs-tree.c (init_tree_optimization_optabs): Use
11038 ggc_cleared_alloc to allocate target_optabs.
11039 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
11040 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
11042 * internal-fn.h (internal_load_fn_p): Declare.
11043 (internal_gather_scatter_fn_p): Likewise.
11044 (internal_fn_mask_index): Likewise.
11045 (internal_gather_scatter_fn_supported_p): Likewise.
11046 * internal-fn.c (gather_load_direct): New macro.
11047 (expand_gather_load_optab_fn): New function.
11048 (direct_gather_load_optab_supported_p): New macro.
11049 (direct_internal_fn_optab): New function.
11050 (internal_load_fn_p): Likewise.
11051 (internal_gather_scatter_fn_p): Likewise.
11052 (internal_fn_mask_index): Likewise.
11053 (internal_gather_scatter_fn_supported_p): Likewise.
11054 * optabs-query.c (supports_at_least_one_mode_p): New function.
11055 (supports_vec_gather_load_p): Likewise.
11056 * optabs-query.h (supports_vec_gather_load_p): Declare.
11057 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
11058 and memory_type field.
11059 (NUM_PATTERNS): Bump to 15.
11060 * tree-vect-data-refs.c: Include internal-fn.h.
11061 (vect_gather_scatter_fn_p): New function.
11062 (vect_describe_gather_scatter_call): Likewise.
11063 (vect_check_gather_scatter): Try using internal functions for
11064 gather loads. Recognize existing calls to a gather load function.
11065 (vect_analyze_data_refs): Consider using gather loads if
11066 supports_vec_gather_load_p.
11067 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
11068 (vect_get_gather_scatter_offset_type): Likewise.
11069 (vect_convert_mask_for_vectype): Likewise.
11070 (vect_add_conversion_to_patterm): Likewise.
11071 (vect_try_gather_scatter_pattern): Likewise.
11072 (vect_recog_gather_scatter_pattern): New pattern recognizer.
11073 (vect_vect_recog_func_ptrs): Add it.
11074 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
11075 internal_fn_mask_index and internal_gather_scatter_fn_p.
11076 (check_load_store_masking): Take the gather_scatter_info as an
11077 argument and handle gather loads.
11078 (vect_get_gather_scatter_ops): New function.
11079 (vectorizable_call): Check internal_load_fn_p.
11080 (vectorizable_load): Likewise. Handle gather load internal
11082 (vectorizable_store): Update call to check_load_store_masking.
11083 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
11084 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
11085 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
11086 (aarch64_gather_scale_operand_d): New predicates.
11087 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
11088 (mask_gather_load<mode>): New insns.
11090 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11091 Alan Hayward <alan.hayward@arm.com>
11092 David Sherwood <david.sherwood@arm.com>
11094 * optabs.def (fold_left_plus_optab): New optab.
11095 * doc/md.texi (fold_left_plus_@var{m}): Document.
11096 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
11097 * internal-fn.c (fold_left_direct): Define.
11098 (expand_fold_left_optab_fn): Likewise.
11099 (direct_fold_left_optab_supported_p): Likewise.
11100 * fold-const-call.c (fold_const_fold_left): New function.
11101 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
11102 * tree-parloops.c (valid_reduction_p): New function.
11103 (gather_scalar_reductions): Use it.
11104 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
11105 (vect_finish_replace_stmt): Declare.
11106 * tree-vect-loop.c (fold_left_reduction_fn): New function.
11107 (needs_fold_left_reduction_p): New function, split out from...
11108 (vect_is_simple_reduction): ...here. Accept reductions that
11109 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
11110 (vect_force_simple_reduction): Also store the reduction type in
11111 the assignment's STMT_VINFO_REDUC_TYPE.
11112 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
11113 (merge_with_identity): New function.
11114 (vect_expand_fold_left): Likewise.
11115 (vectorize_fold_left_reduction): Likewise.
11116 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
11117 scalar phi in place for it. Check for target support and reject
11118 cases that would reassociate the operation. Defer the transform
11119 phase to vectorize_fold_left_reduction.
11120 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
11121 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
11122 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
11124 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11126 * tree-if-conv.c (predicate_mem_writes): Remove redundant
11127 call to ifc_temp_var.
11129 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11130 Alan Hayward <alan.hayward@arm.com>
11131 David Sherwood <david.sherwood@arm.com>
11133 * target.def (legitimize_address_displacement): Take the original
11134 offset as a poly_int.
11135 * targhooks.h (default_legitimize_address_displacement): Update
11137 * targhooks.c (default_legitimize_address_displacement): Likewise.
11138 * doc/tm.texi: Regenerate.
11139 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
11140 as an argument, moving assert of ad->disp == ad->disp_term to...
11141 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
11142 Try calling targetm.legitimize_address_displacement before expanding
11143 the address rather than afterwards, and adjust for the new interface.
11144 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
11145 Match the new hook interface. Handle SVE addresses.
11146 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
11147 new hook interface.
11149 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11151 * Makefile.in (OBJS): Add early-remat.o.
11152 * target.def (select_early_remat_modes): New hook.
11153 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
11154 * doc/tm.texi: Regenerate.
11155 * targhooks.h (default_select_early_remat_modes): Declare.
11156 * targhooks.c (default_select_early_remat_modes): New function.
11157 * timevar.def (TV_EARLY_REMAT): New timevar.
11158 * passes.def (pass_early_remat): New pass.
11159 * tree-pass.h (make_pass_early_remat): Declare.
11160 * early-remat.c: New file.
11161 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
11163 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
11165 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11166 Alan Hayward <alan.hayward@arm.com>
11167 David Sherwood <david.sherwood@arm.com>
11169 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
11170 vfm1 with a bound_epilog parameter.
11171 (vect_do_peeling): Update calls accordingly, and move the prologue
11172 call earlier in the function. Treat the base bound_epilog as 0 for
11173 fully-masked loops and retain vf - 1 for other loops. Add 1 to
11174 this base when peeling for gaps.
11175 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
11176 with fully-masked loops.
11177 (vect_estimate_min_profitable_iters): Handle the single peeled
11178 iteration in that case.
11180 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11181 Alan Hayward <alan.hayward@arm.com>
11182 David Sherwood <david.sherwood@arm.com>
11184 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
11185 single-element interleaving even if the size is not a power of 2.
11186 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
11187 accesses for single-element interleaving if the group size is
11190 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11191 Alan Hayward <alan.hayward@arm.com>
11192 David Sherwood <david.sherwood@arm.com>
11194 * doc/md.texi (fold_extract_last_@var{m}): Document.
11195 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
11196 * optabs.def (fold_extract_last_optab): New optab.
11197 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
11198 * internal-fn.c (fold_extract_direct): New macro.
11199 (expand_fold_extract_optab_fn): Likewise.
11200 (direct_fold_extract_optab_supported_p): Likewise.
11201 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
11202 * tree-vect-loop.c (vect_model_reduction_cost): Handle
11203 EXTRACT_LAST_REDUCTION.
11204 (get_initial_def_for_reduction): Do not create an initial vector
11205 for EXTRACT_LAST_REDUCTION reductions.
11206 (vectorizable_reduction): Leave the scalar phi in place for
11207 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
11208 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
11209 epilogue code for EXTRACT_LAST_REDUCTION and defer the
11210 transform phase to vectorizable_condition.
11211 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
11213 (vect_finish_stmt_generation): ...here.
11214 (vect_finish_replace_stmt): New function.
11215 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
11216 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
11218 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
11220 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11221 Alan Hayward <alan.hayward@arm.com>
11222 David Sherwood <david.sherwood@arm.com>
11224 * doc/md.texi (extract_last_@var{m}): Document.
11225 * optabs.def (extract_last_optab): New optab.
11226 * internal-fn.def (EXTRACT_LAST): New internal function.
11227 * internal-fn.c (cond_unary_direct): New macro.
11228 (expand_cond_unary_optab_fn): Likewise.
11229 (direct_cond_unary_optab_supported_p): Likewise.
11230 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
11231 loops using EXTRACT_LAST.
11232 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
11233 (extract_last_<mode>): ...this optab.
11234 (vec_extract<mode><Vel>): Update accordingly.
11236 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11237 Alan Hayward <alan.hayward@arm.com>
11238 David Sherwood <david.sherwood@arm.com>
11240 * target.def (empty_mask_is_expensive): New hook.
11241 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
11242 * doc/tm.texi: Regenerate.
11243 * targhooks.h (default_empty_mask_is_expensive): Declare.
11244 * targhooks.c (default_empty_mask_is_expensive): New function.
11245 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
11246 if the target says that empty masks are expensive.
11247 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
11249 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
11251 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11252 Alan Hayward <alan.hayward@arm.com>
11253 David Sherwood <david.sherwood@arm.com>
11255 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
11256 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
11257 (vect_use_loop_mask_for_alignment_p): New function.
11258 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
11259 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
11260 niters_skip argument. Make sure that the first niters_skip elements
11261 of the first iteration are inactive.
11262 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
11263 Update call to vect_set_loop_masks_directly.
11264 (get_misalign_in_elems): New function, split out from...
11265 (vect_gen_prolog_loop_niters): ...here.
11266 (vect_update_init_of_dr): Take a code argument that specifies whether
11267 the adjustment should be added or subtracted.
11268 (vect_update_init_of_drs): Likewise.
11269 (vect_prepare_for_masked_peels): New function.
11270 (vect_do_peeling): Skip prologue peeling if we're using a mask
11271 instead. Update call to vect_update_inits_of_drs.
11272 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11274 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
11275 alignment. Do not include the number of peeled iterations in
11276 the minimum threshold in that case.
11277 (vectorizable_induction): Adjust the start value down by
11278 LOOP_VINFO_MASK_SKIP_NITERS iterations.
11279 (vect_transform_loop): Call vect_prepare_for_masked_peels.
11280 Take the number of skipped iterations into account when calculating
11282 * tree-vect-stmts.c (vect_gen_while_not): New function.
11284 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11285 Alan Hayward <alan.hayward@arm.com>
11286 David Sherwood <david.sherwood@arm.com>
11288 * doc/sourcebuild.texi (vect_fully_masked): Document.
11289 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
11290 default value to 0.
11291 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
11293 (vect_analyze_loop_2): ...here. Don't check the vectorization
11294 factor against the number of loop iterations if the loop is
11297 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11298 Alan Hayward <alan.hayward@arm.com>
11299 David Sherwood <david.sherwood@arm.com>
11301 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
11302 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
11303 (dump_groups): Update accordingly.
11304 (iv_use::mem_type): New member variable.
11305 (address_p): New function.
11306 (record_use): Add a mem_type argument and initialize the new
11308 (record_group_use): Add a mem_type argument. Use address_p.
11309 Remove obsolete null checks of base_object. Update call to record_use.
11310 (find_interesting_uses_op): Update call to record_group_use.
11311 (find_interesting_uses_cond): Likewise.
11312 (find_interesting_uses_address): Likewise.
11313 (get_mem_type_for_internal_fn): New function.
11314 (find_address_like_use): Likewise.
11315 (find_interesting_uses_stmt): Try find_address_like_use before
11316 calling find_interesting_uses_op.
11317 (addr_offset_valid_p): Use the iv mem_type field as the type
11318 of the addressed memory.
11319 (add_autoinc_candidates): Likewise.
11320 (get_address_cost): Likewise.
11321 (split_small_address_groups_p): Use address_p.
11322 (split_address_groups): Likewise.
11323 (add_iv_candidate_for_use): Likewise.
11324 (autoinc_possible_for_pair): Likewise.
11325 (rewrite_groups): Likewise.
11326 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
11327 (determine_group_iv_cost): Update after split of USE_ADDRESS.
11328 (get_alias_ptr_type_for_ptr_address): New function.
11329 (rewrite_use_address): Rewrite address uses in calls that were
11330 identified by find_address_like_use.
11332 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11333 Alan Hayward <alan.hayward@arm.com>
11334 David Sherwood <david.sherwood@arm.com>
11336 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
11338 * gimple-expr.h (is_gimple_addressable: Likewise.
11339 * gimple-expr.c (is_gimple_address): Likewise.
11340 * internal-fn.c (expand_call_mem_ref): New function.
11341 (expand_mask_load_optab_fn): Use it.
11342 (expand_mask_store_optab_fn): Likewise.
11344 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11345 Alan Hayward <alan.hayward@arm.com>
11346 David Sherwood <david.sherwood@arm.com>
11348 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
11349 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
11350 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
11351 (cond_umax@var{mode}): Document.
11352 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
11353 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
11354 (cond_umin_optab, cond_umax_optab): New optabs.
11355 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
11356 (COND_IOR, COND_XOR): New internal functions.
11357 * internal-fn.h (get_conditional_internal_fn): Declare.
11358 * internal-fn.c (cond_binary_direct): New macro.
11359 (expand_cond_binary_optab_fn): Likewise.
11360 (direct_cond_binary_optab_supported_p): Likewise.
11361 (get_conditional_internal_fn): New function.
11362 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
11363 Cope with reduction statements that are vectorized as calls rather
11365 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
11366 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
11367 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
11368 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
11369 (UNSPEC_COND_EOR): New unspecs.
11370 (optab): Add mappings for them.
11371 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
11372 (sve_int_op, sve_fp_op): New int attributes.
11374 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11375 Alan Hayward <alan.hayward@arm.com>
11376 David Sherwood <david.sherwood@arm.com>
11378 * optabs.def (while_ult_optab): New optab.
11379 * doc/md.texi (while_ult@var{m}@var{n}): Document.
11380 * internal-fn.def (WHILE_ULT): New internal function.
11381 * internal-fn.h (direct_internal_fn_supported_p): New override
11382 that takes two types as argument.
11383 * internal-fn.c (while_direct): New macro.
11384 (expand_while_optab_fn): New function.
11385 (convert_optab_supported_p): Likewise.
11386 (direct_while_optab_supported_p): New macro.
11387 * wide-int.h (wi::udiv_ceil): New function.
11388 * tree-vectorizer.h (rgroup_masks): New structure.
11389 (vec_loop_masks): New typedef.
11390 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
11391 and fully_masked_p.
11392 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
11393 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
11394 (vect_max_vf): New function.
11395 (slpeel_make_loop_iterate_ntimes): Delete.
11396 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
11397 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
11398 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
11399 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
11400 internal-fn.h, stor-layout.h and optabs-query.h.
11401 (vect_set_loop_mask): New function.
11402 (add_preheader_seq): Likewise.
11403 (add_header_seq): Likewise.
11404 (interleave_supported_p): Likewise.
11405 (vect_maybe_permute_loop_masks): Likewise.
11406 (vect_set_loop_masks_directly): Likewise.
11407 (vect_set_loop_condition_masked): Likewise.
11408 (vect_set_loop_condition_unmasked): New function, split out from
11409 slpeel_make_loop_iterate_ntimes.
11410 (slpeel_make_loop_iterate_ntimes): Rename to..
11411 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
11412 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
11413 (vect_do_peeling): Update call accordingly.
11414 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
11416 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11417 mask_compare_type, can_fully_mask_p and fully_masked_p.
11418 (release_vec_loop_masks): New function.
11419 (_loop_vec_info): Use it to free the loop masks.
11420 (can_produce_all_loop_masks_p): New function.
11421 (vect_get_max_nscalars_per_iter): Likewise.
11422 (vect_verify_full_masking): Likewise.
11423 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
11424 retries, and free the mask rgroups before retrying. Check loop-wide
11425 reasons for disallowing fully-masked loops. Make the final decision
11426 about whether use a fully-masked loop or not.
11427 (vect_estimate_min_profitable_iters): Do not assume that peeling
11428 for the number of iterations will be needed for fully-masked loops.
11429 (vectorizable_reduction): Disable fully-masked loops.
11430 (vectorizable_live_operation): Likewise.
11431 (vect_halve_mask_nunits): New function.
11432 (vect_double_mask_nunits): Likewise.
11433 (vect_record_loop_mask): Likewise.
11434 (vect_get_loop_mask): Likewise.
11435 (vect_transform_loop): Handle the case in which the final loop
11436 iteration might handle a partial vector. Call vect_set_loop_condition
11437 instead of slpeel_make_loop_iterate_ntimes.
11438 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
11439 (check_load_store_masking): New function.
11440 (prepare_load_store_mask): Likewise.
11441 (vectorizable_store): Handle fully-masked loops.
11442 (vectorizable_load): Likewise.
11443 (supportable_widening_operation): Use vect_halve_mask_nunits for
11445 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
11446 (vect_gen_while): New function.
11447 * config/aarch64/aarch64.md (umax<mode>3): New expander.
11448 (aarch64_uqdec<mode>): New insn.
11450 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11451 Alan Hayward <alan.hayward@arm.com>
11452 David Sherwood <david.sherwood@arm.com>
11454 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
11455 (reduc_xor_scal_optab): New optabs.
11456 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
11457 (reduc_xor_scal_@var{m}): Document.
11458 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
11459 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
11460 internal functions.
11461 * fold-const-call.c (fold_const_call): Handle them.
11462 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
11463 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
11464 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
11465 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
11466 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
11467 (UNSPEC_XORV): New unspecs.
11468 (optab): Add entries for them.
11469 (BITWISEV): New int iterator.
11470 (bit_reduc_op): New int attributes.
11472 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11473 Alan Hayward <alan.hayward@arm.com>
11474 David Sherwood <david.sherwood@arm.com>
11476 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
11477 * internal-fn.def (VEC_SHL_INSERT): New internal function.
11478 * optabs.def (vec_shl_insert_optab): New optab.
11479 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
11480 (duplicate_and_interleave): Likewise.
11481 * tree-vect-loop.c: Include internal-fn.h.
11482 (neutral_op_for_slp_reduction): New function, split out from
11483 get_initial_defs_for_reduction.
11484 (get_initial_def_for_reduction): Handle option 2 for variable-length
11485 vectors by loading the neutral value into a vector and then shifting
11486 the initial value into element 0.
11487 (get_initial_defs_for_reduction): Replace the code argument with
11488 the neutral value calculated by neutral_op_for_slp_reduction.
11489 Use gimple_build_vector for constant-length vectors.
11490 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
11491 but the first group_size elements have a neutral value.
11492 Use duplicate_and_interleave otherwise.
11493 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
11494 Update call to get_initial_defs_for_reduction. Handle SLP
11495 reductions for variable-length vectors by creating one vector
11496 result for each scalar result, with the elements associated
11497 with other scalar results stubbed out with the neutral value.
11498 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
11499 Require IFN_VEC_SHL_INSERT for double reductions on
11500 variable-length vectors, or SLP reductions that have
11501 a neutral value. Require can_duplicate_and_interleave_p
11502 support for variable-length unchained SLP reductions if there
11503 is no neutral value, such as for MIN/MAX reductions. Also require
11504 the number of vector elements to be a multiple of the number of
11505 SLP statements when doing variable-length unchained SLP reductions.
11506 Update call to vect_create_epilog_for_reduction.
11507 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
11508 and remove initial values.
11509 (duplicate_and_interleave): Make public.
11510 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
11511 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
11513 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11514 Alan Hayward <alan.hayward@arm.com>
11515 David Sherwood <david.sherwood@arm.com>
11517 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
11518 (can_duplicate_and_interleave_p): New function.
11519 (vect_get_and_check_slp_defs): Take the vector of statements
11520 rather than just the current one. Remove excess parentheses.
11521 Restriction rejectinon of vect_constant_def and vect_external_def
11522 for variable-length vectors to boolean types, or types for which
11523 can_duplicate_and_interleave_p is false.
11524 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
11525 (duplicate_and_interleave): New function.
11526 (vect_get_constant_vectors): Use gimple_build_vector for
11527 constant-length vectors and suitable variable-length constant
11528 vectors. Use duplicate_and_interleave for other variable-length
11529 vectors. Don't defer the update when inserting new statements.
11531 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11532 Alan Hayward <alan.hayward@arm.com>
11533 David Sherwood <david.sherwood@arm.com>
11535 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
11536 min_profitable_iters doesn't go negative.
11538 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11539 Alan Hayward <alan.hayward@arm.com>
11540 David Sherwood <david.sherwood@arm.com>
11542 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
11543 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
11544 * optabs.def (vec_mask_load_lanes_optab): New optab.
11545 (vec_mask_store_lanes_optab): Likewise.
11546 * internal-fn.def (MASK_LOAD_LANES): New internal function.
11547 (MASK_STORE_LANES): Likewise.
11548 * internal-fn.c (mask_load_lanes_direct): New macro.
11549 (mask_store_lanes_direct): Likewise.
11550 (expand_mask_load_optab_fn): Handle masked operations.
11551 (expand_mask_load_lanes_optab_fn): New macro.
11552 (expand_mask_store_optab_fn): Handle masked operations.
11553 (expand_mask_store_lanes_optab_fn): New macro.
11554 (direct_mask_load_lanes_optab_supported_p): Likewise.
11555 (direct_mask_store_lanes_optab_supported_p): Likewise.
11556 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
11558 (vect_load_lanes_supported): Likewise.
11559 * tree-vect-data-refs.c (strip_conversion): New function.
11560 (can_group_stmts_p): Likewise.
11561 (vect_analyze_data_ref_accesses): Use it instead of checking
11562 for a pair of assignments.
11563 (vect_store_lanes_supported): Take a masked_p parameter.
11564 (vect_load_lanes_supported): Likewise.
11565 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
11566 vect_store_lanes_supported and vect_load_lanes_supported.
11567 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
11568 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
11569 parameter. Don't allow gaps for masked accesses.
11570 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
11571 and vect_load_lanes_supported.
11572 (get_load_store_type): Take a masked_p parameter and update
11573 call to get_group_load_store_type.
11574 (vectorizable_store): Update call to get_load_store_type.
11575 Handle IFN_MASK_STORE_LANES.
11576 (vectorizable_load): Update call to get_load_store_type.
11577 Handle IFN_MASK_LOAD_LANES.
11579 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11580 Alan Hayward <alan.hayward@arm.com>
11581 David Sherwood <david.sherwood@arm.com>
11583 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
11585 * config/aarch64/aarch64-protos.h
11586 (aarch64_sve_struct_memory_operand_p): Declare.
11587 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
11588 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
11589 (VPRED, vpred): Handle SVE structure modes.
11590 * config/aarch64/constraints.md (Utx): New constraint.
11591 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
11592 (aarch64_sve_struct_nonimmediate_operand): New predicates.
11593 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
11594 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
11595 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
11596 structure modes. Split into pieces after RA.
11597 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
11598 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
11600 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
11601 SVE structure modes.
11602 (aarch64_classify_address): Likewise.
11603 (sizetochar): Move earlier in file.
11604 (aarch64_print_operand): Handle SVE register lists.
11605 (aarch64_array_mode): New function.
11606 (aarch64_sve_struct_memory_operand_p): Likewise.
11607 (TARGET_ARRAY_MODE): Redefine.
11609 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11610 Alan Hayward <alan.hayward@arm.com>
11611 David Sherwood <david.sherwood@arm.com>
11613 * target.def (array_mode): New target hook.
11614 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
11615 * doc/tm.texi: Regenerate.
11616 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
11617 * hooks.c (hook_optmode_mode_uhwi_none): New function.
11618 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
11619 targetm.array_mode.
11620 * stor-layout.c (mode_for_array): Likewise. Support polynomial
11623 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11624 Alan Hayward <alan.hayward@arm.com>
11625 David Sherwood <david.sherwood@arm.com>
11627 * fold-const.c (fold_binary_loc): Check the argument types
11628 rather than the result type when testing for a vector operation.
11630 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11632 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
11633 * doc/tm.texi: Regenerate.
11635 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11636 Alan Hayward <alan.hayward@arm.com>
11637 David Sherwood <david.sherwood@arm.com>
11639 * doc/invoke.texi (-msve-vector-bits=): Document new option.
11640 (sve): Document new AArch64 extension.
11641 * doc/md.texi (w): Extend the description of the AArch64
11642 constraint to include SVE vectors.
11643 (Upl, Upa): Document new AArch64 predicate constraints.
11644 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
11646 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
11647 (msve-vector-bits=): New option.
11648 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
11649 SVE when these are disabled.
11650 (sve): New extension.
11651 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
11652 modes. Adjust their number of units based on aarch64_sve_vg.
11653 (MAX_BITSIZE_MODE_ANY_MODE): Define.
11654 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
11655 aarch64_addr_query_type.
11656 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
11657 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
11658 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
11659 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
11660 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
11661 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
11662 (aarch64_simd_imm_zero_p): Delete.
11663 (aarch64_check_zero_based_sve_index_immediate): Declare.
11664 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
11665 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
11666 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
11667 (aarch64_sve_float_mul_immediate_p): Likewise.
11668 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
11669 rather than an rtx.
11670 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
11671 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
11672 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
11673 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
11674 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
11675 (aarch64_regmode_natural_size): Likewise.
11676 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
11677 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
11679 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
11680 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
11681 for VG and the SVE predicate registers.
11682 (V_ALIASES): Add a "z"-prefixed alias.
11683 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
11684 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
11685 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
11686 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
11687 (REG_CLASS_NAMES): Add entries for them.
11688 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
11689 and the predicate registers.
11690 (aarch64_sve_vg): Declare.
11691 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
11692 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
11693 (REGMODE_NATURAL_SIZE): Define.
11694 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
11696 * config/aarch64/aarch64.c: Include cfgrtl.h.
11697 (simd_immediate_info): Add a constructor for series vectors,
11698 and an associated step field.
11699 (aarch64_sve_vg): New variable.
11700 (aarch64_dbx_register_number): Handle VG and the predicate registers.
11701 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
11702 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
11703 (VEC_ANY_DATA, VEC_STRUCT): New constants.
11704 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
11705 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
11706 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
11707 (aarch64_get_mask_mode): New functions.
11708 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
11709 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
11710 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
11711 predicate modes and predicate registers. Explicitly restrict
11712 GPRs to modes of 16 bytes or smaller. Only allow FP registers
11713 to store a vector mode if it is recognized by
11714 aarch64_classify_vector_mode.
11715 (aarch64_regmode_natural_size): New function.
11716 (aarch64_hard_regno_caller_save_mode): Return the original mode
11718 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
11719 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
11720 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
11721 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
11723 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
11724 does not overlap dest if the function is frame-related. Handle
11726 (aarch64_split_add_offset): New function.
11727 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
11728 them aarch64_add_offset.
11729 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
11730 and update call to aarch64_sub_sp.
11731 (aarch64_add_cfa_expression): New function.
11732 (aarch64_expand_prologue): Pass extra temporary registers to the
11733 functions above. Handle the case in which we need to emit new
11734 DW_CFA_expressions for registers that were originally saved
11735 relative to the stack pointer, but now have to be expressed
11736 relative to the frame pointer.
11737 (aarch64_output_mi_thunk): Pass extra temporary registers to the
11739 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
11740 IP0 and IP1 values for SVE frames.
11741 (aarch64_expand_vec_series): New function.
11742 (aarch64_expand_sve_widened_duplicate): Likewise.
11743 (aarch64_expand_sve_const_vector): Likewise.
11744 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
11745 Handle SVE constants. Use emit_move_insn to move a force_const_mem
11746 into the register, rather than emitting a SET directly.
11747 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
11748 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
11749 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
11750 (offset_9bit_signed_scaled_p): New functions.
11751 (aarch64_replicate_bitmask_imm): New function.
11752 (aarch64_bitmask_imm): Use it.
11753 (aarch64_cannot_force_const_mem): Reject expressions involving
11754 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
11755 (aarch64_classify_index): Handle SVE indices, by requiring
11756 a plain register index with a scale that matches the element size.
11757 (aarch64_classify_address): Handle SVE addresses. Assert that
11758 the mode of the address is VOIDmode or an integer mode.
11759 Update call to aarch64_classify_symbol.
11760 (aarch64_classify_symbolic_expression): Update call to
11761 aarch64_classify_symbol.
11762 (aarch64_const_vec_all_in_range_p): New function.
11763 (aarch64_print_vector_float_operand): Likewise.
11764 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
11765 "vN" for FP registers with SVE modes. Handle (const ...) vectors
11766 and the FP immediates 1.0 and 0.5.
11767 (aarch64_print_address_internal): Handle SVE addresses.
11768 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
11769 (aarch64_regno_regclass): Handle predicate registers.
11770 (aarch64_secondary_reload): Handle big-endian reloads of SVE
11772 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
11773 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
11774 (aarch64_convert_sve_vector_bits): New function.
11775 (aarch64_override_options): Use it to handle -msve-vector-bits=.
11776 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
11777 rather than an rtx.
11778 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
11779 Handle SVE vector and predicate modes. Accept VL-based constants
11780 that need only one temporary register, and VL offsets that require
11781 no temporary registers.
11782 (aarch64_conditional_register_usage): Mark the predicate registers
11783 as fixed if SVE isn't available.
11784 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
11785 Return true for SVE vector and predicate modes.
11786 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
11787 rather than an unsigned int. Handle SVE modes.
11788 (aarch64_preferred_simd_mode): Update call accordingly. Handle
11790 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
11792 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
11793 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
11794 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
11795 (aarch64_sve_float_mul_immediate_p): New functions.
11796 (aarch64_sve_valid_immediate): New function.
11797 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
11798 Explicitly reject structure modes. Check for INDEX constants.
11799 Handle PTRUE and PFALSE constants.
11800 (aarch64_check_zero_based_sve_index_immediate): New function.
11801 (aarch64_simd_imm_zero_p): Delete.
11802 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
11803 vector modes. Accept constants in the range of CNT[BHWD].
11804 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
11805 ask for an Advanced SIMD mode.
11806 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
11807 (aarch64_simd_vector_alignment): Handle SVE predicates.
11808 (aarch64_vectorize_preferred_vector_alignment): New function.
11809 (aarch64_simd_vector_alignment_reachable): Use it instead of
11811 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
11812 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
11814 (MAX_VECT_LEN): Delete.
11815 (expand_vec_perm_d): Add a vec_flags field.
11816 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
11817 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
11818 (aarch64_evpc_ext): Don't apply a big-endian lane correction
11820 (aarch64_evpc_rev): Rename to...
11821 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
11822 (aarch64_evpc_rev_global): New function.
11823 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
11824 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
11826 (aarch64_evpc_sve_tbl): New function.
11827 (aarch64_expand_vec_perm_const_1): Update after rename of
11828 aarch64_evpc_rev. Handle SVE permutes too, trying
11829 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
11830 than aarch64_evpc_tbl.
11831 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
11832 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
11833 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
11834 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
11835 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
11836 (aarch64_expand_sve_vcond): New functions.
11837 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
11838 of aarch64_vector_mode_p.
11839 (aarch64_dwarf_poly_indeterminate_value): New function.
11840 (aarch64_compute_pressure_classes): Likewise.
11841 (aarch64_can_change_mode_class): Likewise.
11842 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
11843 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
11844 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
11845 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
11846 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
11847 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
11848 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
11849 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
11851 (Dn, Dl, Dr): Accept const as well as const_vector.
11852 (Dz): Likewise. Compare against CONST0_RTX.
11853 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
11854 of "vector" where appropriate.
11855 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
11856 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
11857 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
11858 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
11859 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
11860 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
11861 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
11862 (v_int_equiv): Extend to SVE modes.
11863 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
11865 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
11866 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
11867 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
11868 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
11869 (SVE_COND_FP_CMP): New int iterators.
11870 (perm_hilo): Handle the new unpack unspecs.
11871 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
11873 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
11874 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
11875 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
11876 (aarch64_equality_operator, aarch64_constant_vector_operand)
11877 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
11878 (aarch64_sve_nonimmediate_operand): Likewise.
11879 (aarch64_sve_general_operand): Likewise.
11880 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
11881 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
11882 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
11883 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
11884 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
11885 (aarch64_sve_float_arith_immediate): Likewise.
11886 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
11887 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
11888 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
11889 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
11890 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
11891 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
11892 (aarch64_sve_float_arith_operand): Likewise.
11893 (aarch64_sve_float_arith_with_sub_operand): Likewise.
11894 (aarch64_sve_float_mul_operand): Likewise.
11895 (aarch64_sve_vec_perm_operand): Likewise.
11896 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
11897 (aarch64_mov_operand): Accept const_poly_int and const_vector.
11898 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
11899 as well as const_vector.
11900 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
11901 in file. Use CONST0_RTX and CONSTM1_RTX.
11902 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
11903 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
11904 Use aarch64_simd_imm_zero.
11905 * config/aarch64/aarch64-sve.md: New file.
11906 * config/aarch64/aarch64.md: Include it.
11907 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
11908 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
11909 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
11910 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
11911 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
11912 (sve): New attribute.
11913 (enabled): Disable instructions with the sve attribute unless
11915 (movqi, movhi): Pass CONST_POLY_INT operaneds through
11916 aarch64_expand_mov_immediate.
11917 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
11918 CNT[BHSD] immediates.
11919 (movti): Split CONST_POLY_INT moves into two halves.
11920 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
11921 Split additions that need a temporary here if the destination
11922 is the stack pointer.
11923 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
11924 (*add<mode>3_poly_1): New instruction.
11925 (set_clobber_cc): New expander.
11927 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11929 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
11930 parameter and use it instead of GET_MODE_SIZE (innermode). Use
11931 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
11932 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
11933 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
11934 Change innermode from fixed_mode_size to machine_mode.
11935 (simplify_subreg): Update call accordingly. Handle a constant-sized
11936 subreg of a variable-length CONST_VECTOR.
11938 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11939 Alan Hayward <alan.hayward@arm.com>
11940 David Sherwood <david.sherwood@arm.com>
11942 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
11943 (add_offset_to_base): New function, split out from...
11944 (create_mem_ref): ...here. When handling a scale other than 1,
11945 check first whether the address is valid without the offset.
11946 Add it into the base if so, leaving the index and scale as-is.
11948 2018-01-12 Jakub Jelinek <jakub@redhat.com>
11951 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
11952 fold_for_warn before checking if arg2 is INTEGER_CST.
11954 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
11956 * config/rs6000/predicates.md (load_multiple_operation): Delete.
11957 (store_multiple_operation): Delete.
11958 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
11959 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
11960 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
11961 guarded by TARGET_STRING.
11962 (rs6000_output_load_multiple): Delete.
11963 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
11964 OPTION_MASK_STRING / TARGET_STRING handling.
11965 (print_operand) <'N', 'O'>: Add comment that these are unused now.
11966 (const rs6000_opt_masks) <"string">: Change mask to 0.
11967 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
11968 (MASK_STRING): Delete.
11969 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
11971 (load_multiple): Delete.
11978 (store_multiple): Delete.
11985 (movmemsi_8reg): Delete.
11986 (corresponding unnamed define_insn): Delete.
11987 (movmemsi_6reg): Delete.
11988 (corresponding unnamed define_insn): Delete.
11989 (movmemsi_4reg): Delete.
11990 (corresponding unnamed define_insn): Delete.
11991 (movmemsi_2reg): Delete.
11992 (corresponding unnamed define_insn): Delete.
11993 (movmemsi_1reg): Delete.
11994 (corresponding unnamed define_insn): Delete.
11995 * config/rs6000/rs6000.opt (mno-string): New.
11996 (mstring): Replace by deprecation warning stub.
11997 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
11999 2018-01-12 Jakub Jelinek <jakub@redhat.com>
12001 * regrename.c (regrename_do_replace): If replacing the same
12002 reg multiple times, try to reuse last created gen_raw_REG.
12005 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
12006 main to workaround a bug in GDB.
12008 2018-01-12 Tom de Vries <tom@codesourcery.com>
12011 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
12013 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
12015 PR rtl-optimization/80481
12016 * ira-color.c (get_cap_member): New function.
12017 (allocnos_conflict_by_live_ranges_p): Use it.
12018 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
12019 (setup_slot_coalesced_allocno_live_ranges): Ditto.
12021 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
12024 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
12025 (*saddl_se_1): Ditto.
12026 (*ssubsi_1): Ditto.
12027 (*ssubl_se_1): Ditto.
12029 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
12031 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
12032 rather than wi::to_widest for DR_INITs.
12033 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
12034 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
12035 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
12037 (vect_analyze_group_access_1): Note that here.
12039 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
12041 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
12042 polynomial type sizes.
12044 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
12046 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
12047 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
12048 (gimple_add_tmp_var): Likewise.
12050 2018-01-12 Martin Liska <mliska@suse.cz>
12052 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
12053 (gimple_alloc_sizes): Likewise.
12054 (dump_gimple_statistics): Use PRIu64 in printf format.
12055 * gimple.h: Change uint64_t to int.
12057 2018-01-12 Martin Liska <mliska@suse.cz>
12059 * tree-core.h: Use uint64_t instead of int.
12060 * tree.c (tree_node_counts): Likewise.
12061 (tree_node_sizes): Likewise.
12062 (dump_tree_statistics): Use PRIu64 in printf format.
12064 2018-01-12 Martin Liska <mliska@suse.cz>
12066 * Makefile.in: As qsort_chk is implemented in vec.c, add
12067 vec.o to linkage of gencfn-macros.
12068 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
12069 passing the info to record_node_allocation_statistics.
12070 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
12072 * ggc-common.c (struct ggc_usage): Add operator== and use
12073 it in operator< and compare function.
12074 * mem-stats.h (struct mem_usage): Likewise.
12075 * vec.c (struct vec_usage): Remove operator< and compare
12076 function. Can be simply inherited.
12078 2018-01-12 Martin Jambor <mjambor@suse.cz>
12081 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
12082 * tree-ssa-math-opts.c: Include domwalk.h.
12083 (convert_mult_to_fma_1): New function.
12084 (fma_transformation_info): New type.
12085 (fma_deferring_state): Likewise.
12086 (cancel_fma_deferring): New function.
12087 (result_of_phi): Likewise.
12088 (last_fma_candidate_feeds_initial_phi): Likewise.
12089 (convert_mult_to_fma): Added deferring logic, split actual
12090 transformation to convert_mult_to_fma_1.
12091 (math_opts_dom_walker): New type.
12092 (math_opts_dom_walker::after_dom_children): New method, body moved
12093 here from pass_optimize_widening_mul::execute, added deferring logic
12095 (pass_optimize_widening_mul::execute): Moved most of code to
12096 math_opts_dom_walker::after_dom_children.
12097 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
12098 * config/i386/i386.c (ix86_option_override_internal): Added
12099 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
12101 2018-01-12 Richard Biener <rguenther@suse.de>
12104 * dwarf2out.c (gen_variable_die): Do not reset old_die for
12105 inline instance vars.
12107 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
12110 * config/rx/rx.c (rx_is_restricted_memory_address):
12111 Handle SUBREG case.
12113 2018-01-12 Richard Biener <rguenther@suse.de>
12115 PR tree-optimization/80846
12116 * target.def (split_reduction): New target hook.
12117 * targhooks.c (default_split_reduction): New function.
12118 * targhooks.h (default_split_reduction): Declare.
12119 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
12120 target requests first reduce vectors by combining low and high
12122 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
12123 (get_vectype_for_scalar_type_and_size): Export.
12124 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
12125 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
12126 * doc/tm.texi: Regenerate.
12127 * config/i386/i386.c (ix86_split_reduction): Implement
12128 TARGET_VECTORIZE_SPLIT_REDUCTION.
12130 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
12133 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
12134 in PIC mode except for TARGET_VXWORKS_RTP.
12135 * config/sparc/sparc.c: Include cfgrtl.h.
12136 (TARGET_INIT_PIC_REG): Define.
12137 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
12138 (sparc_pic_register_p): New predicate.
12139 (sparc_legitimate_address_p): Use it.
12140 (sparc_legitimize_pic_address): Likewise.
12141 (sparc_delegitimize_address): Likewise.
12142 (sparc_mode_dependent_address_p): Likewise.
12143 (gen_load_pcrel_sym): Remove 4th parameter.
12144 (load_got_register): Adjust call to above. Remove obsolete stuff.
12145 (sparc_expand_prologue): Do not call load_got_register here.
12146 (sparc_flat_expand_prologue): Likewise.
12147 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
12148 (sparc_use_pseudo_pic_reg): New function.
12149 (sparc_init_pic_reg): Likewise.
12150 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
12151 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
12153 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
12155 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
12156 Add item for branch_cost.
12158 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
12160 PR rtl-optimization/83565
12161 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
12162 not extend the result to a larger mode for rotate operations.
12163 (num_sign_bit_copies1): Likewise.
12165 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12168 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
12170 Use values-Xc.o for -pedantic.
12171 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
12173 2018-01-12 Martin Liska <mliska@suse.cz>
12176 * ipa-devirt.c (final_warning_record::grow_type_warnings):
12178 (possible_polymorphic_call_targets): Use it.
12179 (ipa_devirt): Likewise.
12181 2018-01-12 Martin Liska <mliska@suse.cz>
12183 * profile-count.h (enum profile_quality): Use 0 as invalid
12184 enum value of profile_quality.
12186 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
12188 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
12189 -mext-string options.
12191 2018-01-12 Richard Biener <rguenther@suse.de>
12193 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
12194 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
12195 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
12197 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
12199 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
12201 * configure.ac (--with-long-double-format): Add support for the
12202 configuration option to change the default long double format on
12204 * config.gcc (powerpc*-linux*-*): Likewise.
12205 * configure: Regenerate.
12206 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
12207 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
12208 used without modification.
12210 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12212 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
12213 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
12214 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
12215 MISC_BUILTIN_SPEC_BARRIER.
12216 (rs6000_init_builtins): Likewise.
12217 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
12219 (speculation_barrier): New define_insn.
12220 * doc/extend.texi: Document __builtin_speculation_barrier.
12222 2018-01-11 Jakub Jelinek <jakub@redhat.com>
12225 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
12226 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
12227 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
12229 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
12230 integral modes instead of "ss" and "sd".
12231 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
12232 vectors with 32-bit and 64-bit elements.
12233 (vecdupssescalarmodesuffix): New mode attribute.
12234 (vec_dup<mode>): Use it.
12236 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
12239 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
12240 frame if argument is passed on stack.
12242 2018-01-11 Jakub Jelinek <jakub@redhat.com>
12245 * ree.c (combine_reaching_defs): Optimize also
12246 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
12247 reg2=any_extend(exp); reg1=reg2;, formatting fix.
12249 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
12251 PR middle-end/83189
12252 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
12254 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
12256 PR middle-end/83718
12257 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
12258 after they are computed.
12260 2018-01-11 Bin Cheng <bin.cheng@arm.com>
12262 PR tree-optimization/83695
12263 * gimple-loop-linterchange.cc
12264 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
12265 reset cached scev information after interchange.
12266 (pass_linterchange::execute): Remove call to scev_reset_htab.
12268 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12270 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
12271 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
12272 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
12273 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
12274 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
12275 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
12276 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
12277 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
12278 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
12279 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
12280 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
12281 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
12282 (V_lane_reg): Likewise.
12283 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
12285 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
12286 (vfmal_lane_low<mode>_intrinsic,
12287 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
12288 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
12289 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
12290 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
12291 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
12292 vfmsl_lane_high<mode>_intrinsic): New define_insns.
12294 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12296 * config/arm/arm-cpus.in (fp16fml): New feature.
12297 (ALL_SIMD): Add fp16fml.
12298 (armv8.2-a): Add fp16fml as an option.
12299 (armv8.3-a): Likewise.
12300 (armv8.4-a): Add fp16fml as part of fp16.
12301 * config/arm/arm.h (TARGET_FP16FML): Define.
12302 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
12304 * config/arm/arm-modes.def (V2HF): Define.
12305 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
12306 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
12307 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
12308 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
12309 vfmsl_low, vfmsl_high): New set of builtins.
12310 * config/arm/iterators.md (PLUSMINUS): New code iterator.
12311 (vfml_op): New code attribute.
12312 (VFMLHALVES): New int iterator.
12313 (VFML, VFMLSEL): New mode attributes.
12314 (V_reg): Define mapping for V2HF.
12315 (V_hi, V_lo): New mode attributes.
12316 (VF_constraint): Likewise.
12317 (vfml_half, vfml_half_selector): New int attributes.
12318 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
12320 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
12321 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
12323 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
12324 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
12325 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
12326 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
12328 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
12329 Document new effective target and option set.
12331 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12333 * config/arm/arm-cpus.in (armv8_4): New feature.
12334 (ARMv8_4a): New fgroup.
12335 (armv8.4-a): New arch.
12336 * config/arm/arm-tables.opt: Regenerate.
12337 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
12338 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
12339 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
12340 Add matching rules for -march=armv8.4-a and extensions.
12341 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
12343 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
12346 * config/rx/rx.md (BW): New mode attribute.
12347 (sync_lock_test_and_setsi): Add mode suffix to insn output.
12349 2018-01-11 Richard Biener <rguenther@suse.de>
12351 PR tree-optimization/83435
12352 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
12353 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
12354 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
12356 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12357 Alan Hayward <alan.hayward@arm.com>
12358 David Sherwood <david.sherwood@arm.com>
12360 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
12362 (aarch64_classify_address): Initialize it. Track polynomial offsets.
12363 (aarch64_print_address_internal): Use it to check for a zero offset.
12365 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12366 Alan Hayward <alan.hayward@arm.com>
12367 David Sherwood <david.sherwood@arm.com>
12369 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
12370 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
12371 Return a poly_int64 rather than a HOST_WIDE_INT.
12372 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
12373 rather than a HOST_WIDE_INT.
12374 * config/aarch64/aarch64.h (aarch64_frame): Protect with
12375 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
12376 hard_fp_offset, frame_size, initial_adjust, callee_offset and
12377 final_offset from HOST_WIDE_INT to poly_int64.
12378 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
12379 to_constant when getting the number of units in an Advanced SIMD
12381 (aarch64_builtin_vectorized_function): Check for a constant number
12383 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
12385 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
12386 attribute instead of GET_MODE_NUNITS.
12387 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
12388 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
12389 GET_MODE_SIZE for fixed-size registers.
12390 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
12391 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
12392 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
12393 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
12394 (aarch64_print_operand, aarch64_print_address_internal)
12395 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
12396 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
12397 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
12398 Handle polynomial GET_MODE_SIZE.
12399 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
12400 wider than SImode without modification.
12401 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
12402 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
12403 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
12404 passing and returning SVE modes.
12405 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
12406 rather than GEN_INT.
12407 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
12408 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
12409 (aarch64_allocate_and_probe_stack_space): Likewise.
12410 (aarch64_layout_frame): Cope with polynomial offsets.
12411 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
12412 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
12413 polynomial offsets.
12414 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
12415 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
12416 poly_int64 rather than a HOST_WIDE_INT.
12417 (aarch64_get_separate_components, aarch64_process_components)
12418 (aarch64_expand_prologue, aarch64_expand_epilogue)
12419 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
12420 (aarch64_anchor_offset): New function, split out from...
12421 (aarch64_legitimize_address): ...here.
12422 (aarch64_builtin_vectorization_cost): Handle polynomial
12423 TYPE_VECTOR_SUBPARTS.
12424 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
12426 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
12427 number of elements from the PARALLEL rather than the mode.
12428 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
12429 rather than GET_MODE_BITSIZE.
12430 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
12431 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
12432 (aarch64_expand_vec_perm_const_1): Handle polynomial
12433 d->perm.length () and d->perm elements.
12434 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
12435 Apply to_constant to d->perm elements.
12436 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
12437 polynomial CONST_VECTOR_NUNITS.
12438 (aarch64_move_pointer): Take amount as a poly_int64 rather
12440 (aarch64_progress_pointer): Avoid temporary variable.
12441 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
12442 the mode attribute instead of GET_MODE.
12444 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12445 Alan Hayward <alan.hayward@arm.com>
12446 David Sherwood <david.sherwood@arm.com>
12448 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
12449 x exists before using it.
12450 (aarch64_add_constant_internal): Rename to...
12451 (aarch64_add_offset_1): ...this. Replace regnum with separate
12452 src and dest rtxes. Handle the case in which they're different,
12453 including when the offset is zero. Replace scratchreg with an rtx.
12454 Use 2 additions if there is no spare register into which we can
12455 move a 16-bit constant.
12456 (aarch64_add_constant): Delete.
12457 (aarch64_add_offset): Replace reg with separate src and dest
12458 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
12459 Use aarch64_add_offset_1.
12460 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
12461 an rtx rather than an int. Take the delta as a poly_int64
12462 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
12463 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
12464 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
12465 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
12466 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
12467 and aarch64_add_sp.
12468 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
12469 aarch64_add_constant.
12471 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12473 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
12474 Use scalar_float_mode.
12476 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12478 * config/aarch64/aarch64-simd.md
12479 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
12480 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
12481 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
12482 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
12483 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
12484 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
12485 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
12486 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
12487 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
12488 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
12490 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12493 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
12494 targ_options->x_arm_arch_string is non NULL.
12496 2018-01-11 Tamar Christina <tamar.christina@arm.com>
12498 * config/aarch64/aarch64.h
12499 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
12501 2018-01-11 Sudakshina Das <sudi.das@arm.com>
12504 * expmed.c (emit_store_flag_force): Swap if const op0
12505 and change VOIDmode to mode of op0.
12507 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12509 PR rtl-optimization/83761
12510 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
12511 than bytes to mode_for_size.
12513 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12515 PR middle-end/83189
12516 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
12517 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
12520 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12522 PR middle-end/83575
12523 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
12524 when in layout mode.
12525 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
12526 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
12529 2018-01-10 Michael Collison <michael.collison@arm.com>
12531 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
12532 * config/aarch64/aarch64-option-extension.def: Add
12533 AARCH64_OPT_EXTENSION of 'fp16fml'.
12534 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12535 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
12536 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
12537 * config/aarch64/constraints.md (Ui7): New constraint.
12538 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
12539 (VFMLA_SEL_W): Ditto.
12542 (VFMLA16_LOW): New int iterator.
12543 (VFMLA16_HIGH): Ditto.
12544 (UNSPEC_FMLAL): New unspec.
12545 (UNSPEC_FMLSL): Ditto.
12546 (UNSPEC_FMLAL2): Ditto.
12547 (UNSPEC_FMLSL2): Ditto.
12548 (f16mac): New code attribute.
12549 * config/aarch64/aarch64-simd-builtins.def
12550 (aarch64_fmlal_lowv2sf): Ditto.
12551 (aarch64_fmlsl_lowv2sf): Ditto.
12552 (aarch64_fmlalq_lowv4sf): Ditto.
12553 (aarch64_fmlslq_lowv4sf): Ditto.
12554 (aarch64_fmlal_highv2sf): Ditto.
12555 (aarch64_fmlsl_highv2sf): Ditto.
12556 (aarch64_fmlalq_highv4sf): Ditto.
12557 (aarch64_fmlslq_highv4sf): Ditto.
12558 (aarch64_fmlal_lane_lowv2sf): Ditto.
12559 (aarch64_fmlsl_lane_lowv2sf): Ditto.
12560 (aarch64_fmlal_laneq_lowv2sf): Ditto.
12561 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
12562 (aarch64_fmlalq_lane_lowv4sf): Ditto.
12563 (aarch64_fmlsl_lane_lowv4sf): Ditto.
12564 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
12565 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
12566 (aarch64_fmlal_lane_highv2sf): Ditto.
12567 (aarch64_fmlsl_lane_highv2sf): Ditto.
12568 (aarch64_fmlal_laneq_highv2sf): Ditto.
12569 (aarch64_fmlsl_laneq_highv2sf): Ditto.
12570 (aarch64_fmlalq_lane_highv4sf): Ditto.
12571 (aarch64_fmlsl_lane_highv4sf): Ditto.
12572 (aarch64_fmlalq_laneq_highv4sf): Ditto.
12573 (aarch64_fmlsl_laneq_highv4sf): Ditto.
12574 * config/aarch64/aarch64-simd.md:
12575 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
12576 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
12577 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
12578 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
12579 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
12580 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
12581 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
12582 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
12583 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
12584 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
12585 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
12586 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
12587 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
12588 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
12589 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
12590 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
12591 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
12592 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
12593 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
12594 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
12595 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
12596 (vfmlsl_low_u32): Ditto.
12597 (vfmlalq_low_u32): Ditto.
12598 (vfmlslq_low_u32): Ditto.
12599 (vfmlal_high_u32): Ditto.
12600 (vfmlsl_high_u32): Ditto.
12601 (vfmlalq_high_u32): Ditto.
12602 (vfmlslq_high_u32): Ditto.
12603 (vfmlal_lane_low_u32): Ditto.
12604 (vfmlsl_lane_low_u32): Ditto.
12605 (vfmlal_laneq_low_u32): Ditto.
12606 (vfmlsl_laneq_low_u32): Ditto.
12607 (vfmlalq_lane_low_u32): Ditto.
12608 (vfmlslq_lane_low_u32): Ditto.
12609 (vfmlalq_laneq_low_u32): Ditto.
12610 (vfmlslq_laneq_low_u32): Ditto.
12611 (vfmlal_lane_high_u32): Ditto.
12612 (vfmlsl_lane_high_u32): Ditto.
12613 (vfmlal_laneq_high_u32): Ditto.
12614 (vfmlsl_laneq_high_u32): Ditto.
12615 (vfmlalq_lane_high_u32): Ditto.
12616 (vfmlslq_lane_high_u32): Ditto.
12617 (vfmlalq_laneq_high_u32): Ditto.
12618 (vfmlslq_laneq_high_u32): Ditto.
12619 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
12620 (AARCH64_FL_FOR_ARCH8_4): New.
12621 (AARCH64_ISA_F16FML): New ISA flag.
12622 (TARGET_F16FML): New feature flag for fp16fml.
12623 (doc/invoke.texi): Document new fp16fml option.
12625 2018-01-10 Michael Collison <michael.collison@arm.com>
12627 * config/aarch64/aarch64-builtins.c:
12628 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
12629 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12630 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
12631 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
12632 (AARCH64_ISA_SHA3): New ISA flag.
12633 (TARGET_SHA3): New feature flag for sha3.
12634 * config/aarch64/iterators.md (sha512_op): New int attribute.
12635 (CRYPTO_SHA512): New int iterator.
12636 (UNSPEC_SHA512H): New unspec.
12637 (UNSPEC_SHA512H2): Ditto.
12638 (UNSPEC_SHA512SU0): Ditto.
12639 (UNSPEC_SHA512SU1): Ditto.
12640 * config/aarch64/aarch64-simd-builtins.def
12641 (aarch64_crypto_sha512hqv2di): New builtin.
12642 (aarch64_crypto_sha512h2qv2di): Ditto.
12643 (aarch64_crypto_sha512su0qv2di): Ditto.
12644 (aarch64_crypto_sha512su1qv2di): Ditto.
12645 (aarch64_eor3qv8hi): Ditto.
12646 (aarch64_rax1qv2di): Ditto.
12647 (aarch64_xarqv2di): Ditto.
12648 (aarch64_bcaxqv8hi): Ditto.
12649 * config/aarch64/aarch64-simd.md:
12650 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
12651 (aarch64_crypto_sha512su0qv2di): Ditto.
12652 (aarch64_crypto_sha512su1qv2di): Ditto.
12653 (aarch64_eor3qv8hi): Ditto.
12654 (aarch64_rax1qv2di): Ditto.
12655 (aarch64_xarqv2di): Ditto.
12656 (aarch64_bcaxqv8hi): Ditto.
12657 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
12658 (vsha512h2q_u64): Ditto.
12659 (vsha512su0q_u64): Ditto.
12660 (vsha512su1q_u64): Ditto.
12661 (veor3q_u16): Ditto.
12662 (vrax1q_u64): Ditto.
12663 (vxarq_u64): Ditto.
12664 (vbcaxq_u16): Ditto.
12665 * config/arm/types.md (crypto_sha512): New type attribute.
12666 (crypto_sha3): Ditto.
12667 (doc/invoke.texi): Document new sha3 option.
12669 2018-01-10 Michael Collison <michael.collison@arm.com>
12671 * config/aarch64/aarch64-builtins.c:
12672 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
12673 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12674 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
12675 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
12676 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
12677 (AARCH64_ISA_SM4): New ISA flag.
12678 (TARGET_SM4): New feature flag for sm4.
12679 * config/aarch64/aarch64-simd-builtins.def
12680 (aarch64_sm3ss1qv4si): Ditto.
12681 (aarch64_sm3tt1aq4si): Ditto.
12682 (aarch64_sm3tt1bq4si): Ditto.
12683 (aarch64_sm3tt2aq4si): Ditto.
12684 (aarch64_sm3tt2bq4si): Ditto.
12685 (aarch64_sm3partw1qv4si): Ditto.
12686 (aarch64_sm3partw2qv4si): Ditto.
12687 (aarch64_sm4eqv4si): Ditto.
12688 (aarch64_sm4ekeyqv4si): Ditto.
12689 * config/aarch64/aarch64-simd.md:
12690 (aarch64_sm3ss1qv4si): Ditto.
12691 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
12692 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
12693 (aarch64_sm4eqv4si): Ditto.
12694 (aarch64_sm4ekeyqv4si): Ditto.
12695 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
12696 (sm3part_op): Ditto.
12697 (CRYPTO_SM3TT): Ditto.
12698 (CRYPTO_SM3PART): Ditto.
12699 (UNSPEC_SM3SS1): New unspec.
12700 (UNSPEC_SM3TT1A): Ditto.
12701 (UNSPEC_SM3TT1B): Ditto.
12702 (UNSPEC_SM3TT2A): Ditto.
12703 (UNSPEC_SM3TT2B): Ditto.
12704 (UNSPEC_SM3PARTW1): Ditto.
12705 (UNSPEC_SM3PARTW2): Ditto.
12706 (UNSPEC_SM4E): Ditto.
12707 (UNSPEC_SM4EKEY): Ditto.
12708 * config/aarch64/constraints.md (Ui2): New constraint.
12709 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
12710 * config/arm/types.md (crypto_sm3): New type attribute.
12711 (crypto_sm4): Ditto.
12712 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
12713 (vsm3tt1aq_u32): Ditto.
12714 (vsm3tt1bq_u32): Ditto.
12715 (vsm3tt2aq_u32): Ditto.
12716 (vsm3tt2bq_u32): Ditto.
12717 (vsm3partw1q_u32): Ditto.
12718 (vsm3partw2q_u32): Ditto.
12719 (vsm4eq_u32): Ditto.
12720 (vsm4ekeyq_u32): Ditto.
12721 (doc/invoke.texi): Document new sm4 option.
12723 2018-01-10 Michael Collison <michael.collison@arm.com>
12725 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
12726 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
12727 (AARCH64_FL_FOR_ARCH8_4): New.
12728 (AARCH64_FL_V8_4): New flag.
12729 (doc/invoke.texi): Document new armv8.4-a option.
12731 2018-01-10 Michael Collison <michael.collison@arm.com>
12733 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12734 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
12735 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
12736 * config/aarch64/aarch64-option-extension.def: Add
12737 AARCH64_OPT_EXTENSION of 'sha2'.
12738 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
12739 (crypto): Disable sha2 and aes if crypto disabled.
12740 (crypto): Enable aes and sha2 if enabled.
12741 (simd): Disable sha2 and aes if simd disabled.
12742 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
12744 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
12745 (TARGET_SHA2): New feature flag for sha2.
12746 (TARGET_AES): New feature flag for aes.
12747 * config/aarch64/aarch64-simd.md:
12748 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
12749 conditional on TARGET_AES.
12750 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
12751 (aarch64_crypto_sha1hsi): Make pattern conditional
12753 (aarch64_crypto_sha1hv4si): Ditto.
12754 (aarch64_be_crypto_sha1hv4si): Ditto.
12755 (aarch64_crypto_sha1su1v4si): Ditto.
12756 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
12757 (aarch64_crypto_sha1su0v4si): Ditto.
12758 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
12759 (aarch64_crypto_sha256su0v4si): Ditto.
12760 (aarch64_crypto_sha256su1v4si): Ditto.
12761 (doc/invoke.texi): Document new aes and sha2 options.
12763 2018-01-10 Martin Sebor <msebor@redhat.com>
12765 PR tree-optimization/83781
12766 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
12769 2018-01-11 Martin Sebor <msebor@gmail.com>
12770 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12772 PR tree-optimization/83501
12773 PR tree-optimization/81703
12775 * tree-ssa-strlen.c (get_string_cst): Rename...
12776 (get_string_len): ...to this. Handle global constants.
12777 (handle_char_store): Adjust.
12779 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
12780 Jim Wilson <jimw@sifive.com>
12782 * config/riscv/riscv-protos.h (riscv_output_return): New.
12783 * config/riscv/riscv.c (struct machine_function): New naked_p field.
12784 (riscv_attribute_table, riscv_output_return),
12785 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
12786 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
12787 (riscv_compute_frame_info): Only compute frame->mask if not a naked
12789 (riscv_expand_prologue): Add early return for naked function.
12790 (riscv_expand_epilogue): Likewise.
12791 (riscv_function_ok_for_sibcall): Return false for naked function.
12792 (riscv_set_current_function): New.
12793 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
12794 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
12795 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
12796 * doc/extend.texi (RISC-V Function Attributes): New.
12798 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
12800 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
12801 check for 128-bit long double before checking TCmode.
12802 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
12803 128-bit long doubles before checking TFmode or TCmode.
12804 (FLOAT128_IBM_P): Likewise.
12806 2018-01-10 Martin Sebor <msebor@redhat.com>
12808 PR tree-optimization/83671
12809 * builtins.c (c_strlen): Unconditionally return zero for the empty
12811 Use -Warray-bounds for warnings.
12812 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
12813 for non-constant array indices with COMPONENT_REF, arrays of
12814 arrays, and pointers to arrays.
12815 (gimple_fold_builtin_strlen): Determine and set length range for
12816 non-constant character arrays.
12818 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
12820 PR middle-end/81897
12821 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
12824 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
12826 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
12828 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
12831 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
12832 VECTOR_MEM_ALTIVEC_OR_VSX_P.
12833 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
12834 indexed_or_indirect_operand predicate.
12835 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
12836 (*vsx_le_perm_load_v8hi): Likewise.
12837 (*vsx_le_perm_load_v16qi): Likewise.
12838 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
12839 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
12840 (*vsx_le_perm_store_v8hi): Likewise.
12841 (*vsx_le_perm_store_v16qi): Likewise.
12842 (eight unnamed splitters): Likewise.
12844 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
12846 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
12847 * config/rs6000/emmintrin.h: Likewise.
12848 * config/rs6000/mmintrin.h: Likewise.
12849 * config/rs6000/xmmintrin.h: Likewise.
12851 2018-01-10 David Malcolm <dmalcolm@redhat.com>
12854 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
12856 * tree.c (tree_nop_conversion): Return true for location wrapper
12858 (maybe_wrap_with_location): New function.
12859 (selftest::check_strip_nops): New function.
12860 (selftest::test_location_wrappers): New function.
12861 (selftest::tree_c_tests): Call it.
12862 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
12863 (maybe_wrap_with_location): New decl.
12864 (EXPR_LOCATION_WRAPPER_P): New macro.
12865 (location_wrapper_p): New inline function.
12866 (tree_strip_any_location_wrapper): New inline function.
12868 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
12871 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
12872 stack_realign_offset for the largest alignment of stack slot
12874 (ix86_find_max_used_stack_alignment): New function.
12875 (ix86_finalize_stack_frame_flags): Use it. Set
12876 max_used_stack_alignment if we don't realign stack.
12877 * config/i386/i386.h (machine_function): Add
12878 max_used_stack_alignment.
12880 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
12882 * config/arm/arm.opt (-mbranch-cost): New option.
12883 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
12886 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
12889 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
12890 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
12892 2018-01-10 Richard Biener <rguenther@suse.de>
12895 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
12896 early out so it also covers the case where we have a non-NULL
12899 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
12901 PR tree-optimization/83753
12902 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
12903 for non-strided grouped accesses if the number of elements is 1.
12905 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12908 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
12909 * i386.h (TARGET_USE_GATHER): Define.
12910 * x86-tune.def (X86_TUNE_USE_GATHER): New.
12912 2018-01-10 Martin Liska <mliska@suse.cz>
12915 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
12916 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
12918 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
12919 CLEANUP_NO_PARTITIONING is not set.
12921 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
12923 * doc/rtl.texi: Remove documentation of (const ...) wrappers
12924 for vectors, as a partial revert of r254296.
12925 * rtl.h (const_vec_p): Delete.
12926 (const_vec_duplicate_p): Don't test for vector CONSTs.
12927 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
12928 * expmed.c (make_tree): Likewise.
12931 * common.md (E, F): Use CONSTANT_P instead of checking for
12933 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
12934 checking for CONST_VECTOR.
12936 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
12938 PR middle-end/83575
12939 * predict.c (force_edge_cold): Handle in more sane way edges
12940 with no prediction.
12942 2018-01-09 Carl Love <cel@us.ibm.com>
12944 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
12946 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
12947 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
12948 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
12949 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
12950 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
12951 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
12952 * config/rs6000/rs6000-protos.h: Add extern defition for
12953 rs6000_generate_float2_double_code.
12954 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
12956 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
12957 (float2_v2df): Add define_expand.
12959 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
12962 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
12963 op_mode in the force_to_mode call.
12965 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
12967 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
12968 instead of checking each element individually.
12969 (aarch64_evpc_uzp): Likewise.
12970 (aarch64_evpc_zip): Likewise.
12971 (aarch64_evpc_ext): Likewise.
12972 (aarch64_evpc_rev): Likewise.
12973 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
12974 instead of checking each element individually. Return true without
12976 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
12977 whether all selected elements come from the same input, instead of
12978 checking each element individually. Remove calls to gen_rtx_REG,
12979 start_sequence and end_sequence and instead assert that no rtl is
12982 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
12984 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
12985 order of HIGH and CONST checks.
12987 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
12989 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
12990 if the destination isn't an SSA_NAME.
12992 2018-01-09 Richard Biener <rguenther@suse.de>
12994 PR tree-optimization/83668
12995 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
12997 (canonicalize_loop_form): ... here, renamed from ...
12998 (canonicalize_loop_closed_ssa_form): ... this and amended to
12999 swap successor edges for loop exit blocks to make us use
13000 the RPO order we need for initial schedule generation.
13002 2018-01-09 Joseph Myers <joseph@codesourcery.com>
13004 PR tree-optimization/64811
13005 * match.pd: When optimizing comparisons with Inf, avoid
13006 introducing or losing exceptions from comparisons with NaN.
13008 2018-01-09 Martin Liska <mliska@suse.cz>
13011 * asan.c (shadow_mem_size): Add gcc_assert.
13013 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
13015 Don't save registers in main().
13018 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
13019 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
13020 * config/avr/avr.c (avr_set_current_function): Don't error if
13021 naked, OS_task or OS_main are specified at the same time.
13022 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
13024 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
13026 * common/config/avr/avr-common.c (avr_option_optimization_table):
13027 Switch on -mmain-is-OS_task for optimizing compilations.
13029 2018-01-09 Richard Biener <rguenther@suse.de>
13031 PR tree-optimization/83572
13032 * graphite.c: Include cfganal.h.
13033 (graphite_transform_loops): Connect infinite loops to exit
13034 and remove fake edges at the end.
13036 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
13038 * ipa-inline.c (edge_badness): Revert accidental checkin.
13040 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
13043 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
13044 symbols; not inline clones.
13046 2018-01-09 Jakub Jelinek <jakub@redhat.com>
13049 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
13050 hard registers. Formatting fixes.
13052 PR preprocessor/83722
13053 * gcc.c (try_generate_repro): Pass
13054 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
13055 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
13058 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
13059 Kito Cheng <kito.cheng@gmail.com>
13061 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
13062 (riscv_leaf_function_p): Delete.
13063 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
13065 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13067 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
13069 (do_ifelse): New function.
13070 (do_isel): New function.
13071 (do_sub3): New function.
13072 (do_add3): New function.
13073 (do_load_mask_compare): New function.
13074 (do_overlap_load_compare): New function.
13075 (expand_compare_loop): New function.
13076 (expand_block_compare): Call expand_compare_loop() when appropriate.
13077 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
13078 option description.
13079 (-mblock-compare-inline-loop-limit): New option.
13081 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13084 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
13085 Reverse order of second and third operands in first alternative.
13086 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
13087 of first and second elements in UNSPEC_VPERMR vector.
13088 (altivec_expand_vec_perm_le): Likewise.
13090 2018-01-08 Jeff Law <law@redhat.com>
13092 PR rtl-optimizatin/81308
13093 * tree-switch-conversion.c (cfg_altered): New file scoped static.
13094 (process_switch): If group_case_labels makes a change, then set
13096 (pass_convert_switch::execute): If a switch is converted, then
13097 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
13099 PR rtl-optimization/81308
13100 * recog.c (split_all_insns): Conditionally cleanup the CFG after
13103 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
13105 PR target/83663 - Revert r255946
13106 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
13107 generation for cases where splatting a value is not useful.
13108 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
13109 across a vec_duplicate and a paradoxical subreg forming a vector
13110 mode to a vec_concat.
13112 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13114 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
13115 -march=armv8.3-a variants.
13116 * config/arm/t-multilib: Likewise.
13117 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
13119 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13121 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
13123 (cceq_ior_compare_complement): Give it a name so I can use it, and
13124 change boolean_or_operator predicate to boolean_operator so it can
13125 be used to generate a crand.
13126 (eqne): New code iterator.
13127 (bd/bd_neg): New code_attrs.
13128 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
13129 a single define_insn.
13130 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
13131 decrement (bdnzt/bdnzf/bdzt/bdzf).
13132 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
13133 with the new names of the branch decrement patterns, and added the
13134 names of the branch decrement conditional patterns.
13136 2018-01-08 Richard Biener <rguenther@suse.de>
13138 PR tree-optimization/83563
13139 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
13142 2018-01-08 Richard Biener <rguenther@suse.de>
13144 PR middle-end/83713
13145 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
13147 2018-01-08 Richard Biener <rguenther@suse.de>
13149 PR tree-optimization/83685
13150 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
13151 references to abnormals.
13153 2018-01-08 Richard Biener <rguenther@suse.de>
13156 * dwarf2out.c (output_indirect_strings): Handle empty
13157 skeleton_debug_str_hash.
13158 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
13160 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
13162 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
13163 (emit_store_direct): Likewise.
13164 (arc_trampoline_adjust_address): Likewise.
13165 (arc_asm_trampoline_template): New function.
13166 (arc_initialize_trampoline): Use asm_trampoline_template.
13167 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
13168 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
13169 * config/arc/arc.md (flush_icache): Delete pattern.
13171 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
13173 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
13174 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
13177 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
13180 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
13181 by not USED_FOR_TARGET.
13182 (make_pass_resolve_sw_modes): Likewise.
13184 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
13186 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
13189 2018-01-08 Richard Biener <rguenther@suse.de>
13191 PR middle-end/83580
13192 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
13194 2018-01-08 Richard Biener <rguenther@suse.de>
13196 PR middle-end/83517
13197 * match.pd ((t * 2) / 2) -> t): Add missing :c.
13199 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
13201 PR middle-end/81897
13202 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
13203 basic blocks with a small number of successors.
13204 (convert_control_dep_chain_into_preds): Improve handling of
13206 (dump_predicates): Split apart into...
13207 (dump_pred_chain): ...here...
13208 (dump_pred_info): ...and here.
13209 (can_one_predicate_be_invalidated_p): Add debugging printfs.
13210 (can_chain_union_be_invalidated_p): Improve check for invalidation
13212 (uninit_uses_cannot_happen): Avoid unnecessary if
13213 convert_control_dep_chain_into_preds yielded nothing.
13215 2018-01-06 Martin Sebor <msebor@redhat.com>
13217 PR tree-optimization/83640
13218 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
13219 subtracting negative offset from size.
13220 (builtin_access::overlap): Adjust offset bounds of the access to fall
13221 within the size of the object if possible.
13223 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
13225 PR rtl-optimization/83699
13226 * expmed.c (extract_bit_field_1): Restrict the vector usage of
13227 extract_bit_field_as_subreg to cases in which the extracted
13228 value is also a vector.
13230 * lra-constraints.c (process_alt_operands): Test for the equivalence
13231 substitutions when detecting a possible reload cycle.
13233 2018-01-06 Jakub Jelinek <jakub@redhat.com>
13236 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
13237 by default if flag_selective_schedling{,2}. Formatting fixes.
13239 PR rtl-optimization/83682
13240 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
13241 if it has non-VECTOR_MODE element mode.
13242 (vec_duplicate_p): Likewise.
13244 PR middle-end/83694
13245 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
13246 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
13248 2018-01-05 Jakub Jelinek <jakub@redhat.com>
13251 * config/i386/i386-builtin.def
13252 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
13253 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
13254 Require also OPTION_MASK_ISA_AVX512F in addition to
13255 OPTION_MASK_ISA_GFNI.
13256 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
13257 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
13258 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
13259 to OPTION_MASK_ISA_GFNI.
13260 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
13261 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
13262 OPTION_MASK_ISA_AVX512BW.
13263 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
13264 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
13265 addition to OPTION_MASK_ISA_GFNI.
13266 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
13267 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
13268 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
13269 to OPTION_MASK_ISA_GFNI.
13270 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
13271 a requirement for all ISAs rather than any of them with a few
13273 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
13275 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
13276 bitmasks to be enabled with 3 exceptions, instead of requiring any
13277 enabled ISA with lots of exceptions.
13278 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
13279 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
13280 Change avx512bw in isa attribute to avx512f.
13281 * config/i386/sgxintrin.h: Add license boilerplate.
13282 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
13283 to __AVX512F__ and __AVX512VL to __AVX512VL__.
13284 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
13285 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
13287 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
13288 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
13289 temporarily sse2 rather than sse if not enabled already.
13292 * config/i386/sse.md (VI248_VLBW): Rename to ...
13293 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
13294 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
13295 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
13296 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
13297 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
13298 mode iterator instead of VI248_VLBW.
13300 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
13302 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
13303 (record_modified): Skip clobbers; add debug output.
13304 (param_change_prob): Use sreal frequencies.
13306 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
13308 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
13309 punt for user-aligned variables.
13311 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
13313 * tree-chrec.c (chrec_contains_symbols): Return true for
13316 2018-01-05 Sudakshina Das <sudi.das@arm.com>
13319 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
13320 of (x|y) == x for BICS pattern.
13322 2018-01-05 Jakub Jelinek <jakub@redhat.com>
13324 PR tree-optimization/83605
13325 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
13326 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
13329 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
13331 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
13332 * config/epiphany/rtems.h: New file.
13334 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13335 Uros Bizjak <ubizjak@gmail.com>
13338 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
13339 QIreg_operand instead of register_operand predicate.
13340 * config/i386/i386.c (ix86_rop_should_change_byte_p,
13341 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
13342 comments instead of -fmitigate[-_]rop.
13344 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13347 * cgraphunit.c (symbol_table::compile): Switch to text_section
13348 before calling assembly_start debug hook.
13349 * run-rtl-passes.c (run_rtl_passes): Likewise.
13352 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13354 * tree-vrp.c (extract_range_from_binary_expr_1): Check
13355 range_int_cst_p rather than !symbolic_range_p before calling
13356 extract_range_from_multiplicative_op_1.
13358 2018-01-04 Jeff Law <law@redhat.com>
13360 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
13361 redundant test in assertion.
13363 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13365 * doc/rtl.texi: Document machine_mode wrapper classes.
13367 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13369 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
13370 using tree_to_uhwi.
13372 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13374 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
13375 the VEC_PERM_EXPR fold to fail.
13377 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13380 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
13381 to switched_sections.
13383 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13386 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
13387 test for d.testing.
13389 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
13392 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
13393 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
13395 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13398 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
13399 is BLKmode and bitpos not zero or mode change is needed.
13401 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13404 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
13407 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
13410 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
13411 instead of MULT rtx. Update all corresponding splitters.
13412 (*saddl_se): Ditto.
13413 (*ssub<modesuffix>): Ditto.
13414 (*ssubl_se): Ditto.
13415 (*cmp_sadd_di): Update split patterns.
13416 (*cmp_sadd_si): Ditto.
13417 (*cmp_sadd_sidi): Ditto.
13418 (*cmp_ssub_di): Ditto.
13419 (*cmp_ssub_si): Ditto.
13420 (*cmp_ssub_sidi): Ditto.
13421 * config/alpha/predicates.md (const23_operand): New predicate.
13422 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
13423 Look for ASHIFT, not MULT inner operand.
13424 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
13426 2018-01-04 Martin Liska <mliska@suse.cz>
13428 PR gcov-profile/83669
13429 * gcov.c (output_intermediate_file): Add version to intermediate
13431 * doc/gcov.texi: Document new field 'version' in intermediate
13432 file format. Fix location of '-k' option of gcov command.
13434 2018-01-04 Martin Liska <mliska@suse.cz>
13437 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
13439 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13441 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
13443 2018-01-03 Martin Sebor <msebor@redhat.com>
13445 PR tree-optimization/83655
13446 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
13447 checking calls with invalid arguments.
13449 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13451 * tree-vect-stmts.c (vect_get_store_rhs): New function.
13452 (vectorizable_mask_load_store): Delete.
13453 (vectorizable_call): Return false for masked loads and stores.
13454 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
13455 instead of gimple_assign_rhs1.
13456 (vectorizable_load): Handle IFN_MASK_LOAD.
13457 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
13459 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13461 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
13463 (vectorizable_mask_load_store): ...here.
13464 (vectorizable_load): ...and here.
13466 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13468 * tree-vect-stmts.c (vect_build_all_ones_mask)
13469 (vect_build_zero_merge_argument): New functions, split out from...
13470 (vectorizable_load): ...here.
13472 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13474 * tree-vect-stmts.c (vect_check_store_rhs): New function,
13476 (vectorizable_mask_load_store): ...here.
13477 (vectorizable_store): ...and here.
13479 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13481 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
13483 (vectorizable_mask_load_store): ...here.
13485 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13487 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
13488 (vect_model_store_cost): Take a vec_load_store_type instead of a
13490 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
13491 (vect_model_store_cost): Take a vec_load_store_type instead of a
13493 (vectorizable_mask_load_store): Update accordingly.
13494 (vectorizable_store): Likewise.
13495 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
13497 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13499 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
13500 IFN_MASK_LOAD calls here rather than...
13501 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
13503 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13504 Alan Hayward <alan.hayward@arm.com>
13505 David Sherwood <david.sherwood@arm.com>
13507 * expmed.c (extract_bit_field_1): For vector extracts,
13508 fall back to extract_bit_field_as_subreg if vec_extract
13511 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13512 Alan Hayward <alan.hayward@arm.com>
13513 David Sherwood <david.sherwood@arm.com>
13515 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
13516 they are variable or constant sized.
13517 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
13518 slots for constant-sized data.
13520 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13521 Alan Hayward <alan.hayward@arm.com>
13522 David Sherwood <david.sherwood@arm.com>
13524 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
13525 handling COND_EXPRs with boolean comparisons, try to find a better
13526 basis for the mask type than the boolean itself.
13528 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13530 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
13531 is calculated and how it can be overridden.
13532 * genmodes.c (max_bitsize_mode_any_mode): New variable.
13533 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
13535 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
13538 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13539 Alan Hayward <alan.hayward@arm.com>
13540 David Sherwood <david.sherwood@arm.com>
13542 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
13543 Remove the mode argument.
13544 (aarch64_simd_valid_immediate): Remove the mode and inverse
13546 * config/aarch64/iterators.md (bitsize): New iterator.
13547 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
13548 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
13549 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
13550 aarch64_simd_valid_immediate.
13551 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
13552 (aarch64_reg_or_bic_imm): Likewise.
13553 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
13554 with an insn_type enum and msl with a modifier_type enum.
13555 Replace element_width with a scalar_mode. Change the shift
13556 to unsigned int. Add constructors for scalar_float_mode and
13557 scalar_int_mode elements.
13558 (aarch64_vect_float_const_representable_p): Delete.
13559 (aarch64_can_const_movi_rtx_p)
13560 (aarch64_simd_scalar_immediate_valid_for_move)
13561 (aarch64_simd_make_constant): Update call to
13562 aarch64_simd_valid_immediate.
13563 (aarch64_advsimd_valid_immediate_hs): New function.
13564 (aarch64_advsimd_valid_immediate): Likewise.
13565 (aarch64_simd_valid_immediate): Remove mode and inverse
13566 arguments. Rewrite to use the above. Use const_vec_duplicate_p
13567 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
13568 and aarch64_float_const_representable_p on the result.
13569 (aarch64_output_simd_mov_immediate): Remove mode argument.
13570 Update call to aarch64_simd_valid_immediate and use of
13571 simd_immediate_info.
13572 (aarch64_output_scalar_simd_mov_immediate): Update call
13575 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13576 Alan Hayward <alan.hayward@arm.com>
13577 David Sherwood <david.sherwood@arm.com>
13579 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
13580 (mode_nunits): Likewise CONST_MODE_NUNITS.
13581 * machmode.def (ADJUST_NUNITS): Document.
13582 * genmodes.c (mode_data::need_nunits_adj): New field.
13583 (blank_mode): Update accordingly.
13584 (adj_nunits): New variable.
13585 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
13587 (emit_mode_size_inline): Set need_bytesize_adj for all modes
13588 listed in adj_nunits.
13589 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
13590 listed in adj_nunits. Don't emit case statements for such modes.
13591 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
13592 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
13593 nothing if adj_nunits is nonnull.
13594 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
13595 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
13596 (emit_mode_fbit): Update use of print_maybe_const_decl.
13597 (emit_move_size): Likewise. Treat the array as non-const
13599 (emit_mode_adjustments): Handle adj_nunits.
13601 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13603 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
13604 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
13605 (VECTOR_MODES): Use it.
13606 (make_vector_modes): Take the prefix as an argument.
13608 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13609 Alan Hayward <alan.hayward@arm.com>
13610 David Sherwood <david.sherwood@arm.com>
13612 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
13613 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
13614 for MODE_VECTOR_BOOL.
13615 * machmode.def (VECTOR_BOOL_MODE): Document.
13616 * genmodes.c (VECTOR_BOOL_MODE): New macro.
13617 (make_vector_bool_mode): New function.
13618 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
13620 * lto-streamer-in.c (lto_input_mode_table): Likewise.
13621 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
13623 * stor-layout.c (int_mode_for_mode): Likewise.
13624 * tree.c (build_vector_type_for_mode): Likewise.
13625 * varasm.c (output_constant_pool_2): Likewise.
13626 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
13627 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
13628 for MODE_VECTOR_BOOL.
13629 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
13630 of mode class checks.
13631 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
13632 instead of a list of mode class checks.
13633 (expand_vector_scalar_condition): Likewise.
13634 (type_for_widest_vector_mode): Handle BImode as an inner mode.
13636 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13637 Alan Hayward <alan.hayward@arm.com>
13638 David Sherwood <david.sherwood@arm.com>
13640 * machmode.h (mode_size): Change from unsigned short to
13642 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
13643 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
13644 or if measurement_type is not polynomial.
13645 (fixed_size_mode::includes_p): Check for constant-sized modes.
13646 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
13647 return a poly_uint16 rather than an unsigned short.
13648 (emit_mode_size): Change the type of mode_size from unsigned short
13649 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
13650 (emit_mode_adjustments): Cope with polynomial vector sizes.
13651 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13653 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13655 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
13656 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
13657 * caller-save.c (setup_save_areas): Likewise.
13658 (replace_reg_with_saved_mem): Likewise.
13659 * calls.c (emit_library_call_value_1): Likewise.
13660 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
13661 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
13662 (gen_lowpart_for_combine): Likewise.
13663 * convert.c (convert_to_integer_1): Likewise.
13664 * cse.c (equiv_constant, cse_insn): Likewise.
13665 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
13666 (cselib_subst_to_values): Likewise.
13667 * dce.c (word_dce_process_block): Likewise.
13668 * df-problems.c (df_word_lr_mark_ref): Likewise.
13669 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
13670 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
13671 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
13672 (rtl_for_decl_location): Likewise.
13673 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
13674 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
13675 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
13676 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
13677 (expand_expr_real_1): Likewise.
13678 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
13679 (pad_below): Likewise.
13680 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
13681 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
13682 * ira.c (get_subreg_tracking_sizes): Likewise.
13683 * ira-build.c (ira_create_allocno_objects): Likewise.
13684 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
13685 (ira_sort_regnos_for_alter_reg): Likewise.
13686 * ira-costs.c (record_operand_costs): Likewise.
13687 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
13688 (resolve_simple_move): Likewise.
13689 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
13690 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
13691 (lra_constraints): Likewise.
13692 (CONST_POOL_OK_P): Reject variable-sized modes.
13693 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
13694 (add_pseudo_to_slot, lra_spill): Likewise.
13695 * omp-low.c (omp_clause_aligned_alignment): Likewise.
13696 * optabs-query.c (get_best_extraction_insn): Likewise.
13697 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
13698 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
13699 (expand_mult_highpart, valid_multiword_target_p): Likewise.
13700 * recog.c (offsettable_address_addr_space_p): Likewise.
13701 * regcprop.c (maybe_mode_change): Likewise.
13702 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
13703 * regrename.c (build_def_use): Likewise.
13704 * regstat.c (dump_reg_info): Likewise.
13705 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
13706 (find_reloads, find_reloads_subreg_address): Likewise.
13707 * reload1.c (eliminate_regs_1): Likewise.
13708 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
13709 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
13710 (simplify_binary_operation_1, simplify_subreg): Likewise.
13711 * targhooks.c (default_function_arg_padding): Likewise.
13712 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
13713 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
13714 (verify_gimple_assign_ternary): Likewise.
13715 * tree-inline.c (estimate_move_cost): Likewise.
13716 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13717 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
13718 (get_address_cost_ainc): Likewise.
13719 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
13720 (vect_supportable_dr_alignment): Likewise.
13721 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
13722 (vectorizable_reduction): Likewise.
13723 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
13724 (vectorizable_operation, vectorizable_load): Likewise.
13725 * tree.c (build_same_sized_truth_vector_type): Likewise.
13726 * valtrack.c (cleanup_auto_inc_dec): Likewise.
13727 * var-tracking.c (emit_note_insn_var_location): Likewise.
13728 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
13729 (ADDR_VEC_ALIGN): Likewise.
13731 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13732 Alan Hayward <alan.hayward@arm.com>
13733 David Sherwood <david.sherwood@arm.com>
13735 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
13737 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
13738 or if measurement_type is polynomial.
13739 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
13740 * combine.c (make_extraction): Likewise.
13741 * dse.c (find_shift_sequence): Likewise.
13742 * dwarf2out.c (mem_loc_descriptor): Likewise.
13743 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
13744 (extract_bit_field, extract_low_bits): Likewise.
13745 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
13746 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
13747 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
13748 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
13749 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
13750 * reload.c (find_reloads): Likewise.
13751 * reload1.c (alter_reg): Likewise.
13752 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
13753 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
13754 * tree-if-conv.c (predicate_mem_writes): Likewise.
13755 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
13756 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
13757 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
13758 * valtrack.c (dead_debug_insert_temp): Likewise.
13759 * varasm.c (mergeable_constant_section): Likewise.
13760 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
13762 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13763 Alan Hayward <alan.hayward@arm.com>
13764 David Sherwood <david.sherwood@arm.com>
13766 * expr.c (expand_assignment): Cope with polynomial mode sizes
13767 when assigning to a CONCAT.
13769 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13770 Alan Hayward <alan.hayward@arm.com>
13771 David Sherwood <david.sherwood@arm.com>
13773 * machmode.h (mode_precision): Change from unsigned short to
13775 (mode_to_precision): Return a poly_uint16 rather than an unsigned
13777 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
13778 or if measurement_type is not polynomial.
13779 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
13780 in which the mode is already known to be a scalar_int_mode.
13781 * genmodes.c (emit_mode_precision): Change the type of mode_precision
13782 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
13784 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13785 for GET_MODE_PRECISION.
13786 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13787 for GET_MODE_PRECISION.
13788 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
13790 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
13791 (expand_field_assignment, make_extraction): Likewise.
13792 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
13793 (get_last_value): Likewise.
13794 * convert.c (convert_to_integer_1): Likewise.
13795 * cse.c (cse_insn): Likewise.
13796 * expr.c (expand_expr_real_1): Likewise.
13797 * lra-constraints.c (simplify_operand_subreg): Likewise.
13798 * optabs-query.c (can_atomic_load_p): Likewise.
13799 * optabs.c (expand_atomic_load): Likewise.
13800 (expand_atomic_store): Likewise.
13801 * ree.c (combine_reaching_defs): Likewise.
13802 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
13803 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
13804 * tree.h (type_has_mode_precision_p): Likewise.
13805 * ubsan.c (instrument_si_overflow): Likewise.
13807 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13808 Alan Hayward <alan.hayward@arm.com>
13809 David Sherwood <david.sherwood@arm.com>
13811 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
13812 polynomial numbers of units.
13813 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
13814 (valid_vector_subparts_p): New function.
13815 (build_vector_type): Remove temporary shim and take the number
13816 of units as a poly_uint64 rather than an int.
13817 (build_opaque_vector_type): Take the number of units as a
13818 poly_uint64 rather than an int.
13819 * tree.c (build_vector_from_ctor): Handle polynomial
13820 TYPE_VECTOR_SUBPARTS.
13821 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
13822 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
13823 (build_vector_from_val): If the number of units is variable,
13824 use build_vec_duplicate_cst for constant operands and
13825 VEC_DUPLICATE_EXPR otherwise.
13826 (make_vector_type): Remove temporary is_constant ().
13827 (build_vector_type, build_opaque_vector_type): Take the number of
13828 units as a poly_uint64 rather than an int.
13829 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
13831 * cfgexpand.c (expand_debug_expr): Likewise.
13832 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
13833 (store_constructor, expand_expr_real_1): Likewise.
13834 (const_scalar_mask_from_tree): Likewise.
13835 * fold-const-call.c (fold_const_reduction): Likewise.
13836 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
13837 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
13838 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
13839 (fold_relational_const): Likewise.
13840 (native_interpret_vector): Likewise. Change the size from an
13841 int to an unsigned int.
13842 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
13843 TYPE_VECTOR_SUBPARTS.
13844 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
13845 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
13846 duplicating a non-constant operand into a variable-length vector.
13847 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
13848 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
13849 * ipa-icf.c (sem_variable::equals): Likewise.
13850 * match.pd: Likewise.
13851 * omp-simd-clone.c (simd_clone_subparts): Likewise.
13852 * print-tree.c (print_node): Likewise.
13853 * stor-layout.c (layout_type): Likewise.
13854 * targhooks.c (default_builtin_vectorization_cost): Likewise.
13855 * tree-cfg.c (verify_gimple_comparison): Likewise.
13856 (verify_gimple_assign_binary): Likewise.
13857 (verify_gimple_assign_ternary): Likewise.
13858 (verify_gimple_assign_single): Likewise.
13859 * tree-pretty-print.c (dump_generic_node): Likewise.
13860 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13861 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
13862 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
13863 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
13864 (vect_shift_permute_load_chain): Likewise.
13865 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
13866 (expand_vector_condition, optimize_vector_constructor): Likewise.
13867 (lower_vec_perm, get_compute_type): Likewise.
13868 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
13869 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
13870 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
13871 (vect_recog_mask_conversion_pattern): Likewise.
13872 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
13873 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
13874 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13875 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
13876 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
13877 (vectorizable_shift, vectorizable_operation, vectorizable_store)
13878 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
13879 (supportable_widening_operation): Likewise.
13880 (supportable_narrowing_operation): Likewise.
13881 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
13883 * varasm.c (output_constant): Likewise.
13885 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13886 Alan Hayward <alan.hayward@arm.com>
13887 David Sherwood <david.sherwood@arm.com>
13889 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
13890 so that both the length == 3 and length != 3 cases set up their
13891 own permute vectors. Add comments explaining why we know the
13892 number of elements is constant.
13893 (vect_permute_load_chain): Likewise.
13895 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13896 Alan Hayward <alan.hayward@arm.com>
13897 David Sherwood <david.sherwood@arm.com>
13899 * machmode.h (mode_nunits): Change from unsigned char to
13901 (ONLY_FIXED_SIZE_MODES): New macro.
13902 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
13903 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
13904 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
13906 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
13907 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
13908 or if measurement_type is not polynomial.
13909 * genmodes.c (ZERO_COEFFS): New macro.
13910 (emit_mode_nunits_inline): Make mode_nunits_inline return a
13912 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
13913 Use ZERO_COEFFS when emitting initializers.
13914 * data-streamer.h (bp_pack_poly_value): New function.
13915 (bp_unpack_poly_value): Likewise.
13916 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13917 for GET_MODE_NUNITS.
13918 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13919 for GET_MODE_NUNITS.
13920 * tree.c (make_vector_type): Remove temporary shim and make
13921 the real function take the number of units as a poly_uint64
13922 rather than an int.
13923 (build_vector_type_for_mode): Handle polynomial nunits.
13924 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
13925 * emit-rtl.c (const_vec_series_p_1): Likewise.
13926 (gen_rtx_CONST_VECTOR): Likewise.
13927 * fold-const.c (test_vec_duplicate_folding): Likewise.
13928 * genrecog.c (validate_pattern): Likewise.
13929 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
13930 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
13931 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
13932 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
13933 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
13934 * rtlanal.c (subreg_get_info): Likewise.
13935 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13936 (vect_grouped_load_supported): Likewise.
13937 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
13938 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
13939 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
13940 (simplify_const_unary_operation, simplify_binary_operation_1)
13941 (simplify_const_binary_operation, simplify_ternary_operation)
13942 (test_vector_ops_duplicate, test_vector_ops): Likewise.
13943 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
13944 instead of CONST_VECTOR_NUNITS.
13945 * varasm.c (output_constant_pool_2): Likewise.
13946 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
13947 explicit-encoded elements in the XVEC for variable-length vectors.
13949 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13951 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
13953 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13954 Alan Hayward <alan.hayward@arm.com>
13955 David Sherwood <david.sherwood@arm.com>
13957 * coretypes.h (fixed_size_mode): Declare.
13958 (fixed_size_mode_pod): New typedef.
13959 * builtins.h (target_builtins::x_apply_args_mode)
13960 (target_builtins::x_apply_result_mode): Change type to
13961 fixed_size_mode_pod.
13962 * builtins.c (apply_args_size, apply_result_size, result_vector)
13963 (expand_builtin_apply_args_1, expand_builtin_apply)
13964 (expand_builtin_return): Update accordingly.
13966 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13968 * cse.c (hash_rtx_cb): Hash only the encoded elements.
13969 * cselib.c (cselib_hash_rtx): Likewise.
13970 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
13971 CONST_VECTOR encoding.
13973 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13974 Jeff Law <law@redhat.com>
13977 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
13978 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
13979 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
13980 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
13983 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
13984 explicitly probe *sp in a noreturn function if there were any callee
13985 register saves or frame pointer is needed.
13987 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13990 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
13991 BLKmode for ternary, binary or unary expressions.
13994 * var-tracking.c (delete_vta_debug_insn): New inline function.
13995 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
13996 insns from get_insns () to NULL instead of each bb separately.
13997 Use delete_vta_debug_insn. No longer static.
13998 (vt_debug_insns_local, variable_tracking_main_1): Adjust
13999 delete_vta_debug_insns callers.
14000 * rtl.h (delete_vta_debug_insns): Declare.
14001 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
14002 instead of variable_tracking_main.
14004 2018-01-03 Martin Sebor <msebor@redhat.com>
14006 PR tree-optimization/83603
14007 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
14008 arguments past the endof the argument list in functions declared
14009 without a prototype.
14010 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
14011 Avoid checking when arguments are null.
14013 2018-01-03 Martin Sebor <msebor@redhat.com>
14016 * doc/extend.texi (attribute const): Fix a typo.
14017 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
14018 issuing -Wsuggest-attribute for void functions.
14020 2018-01-03 Martin Sebor <msebor@redhat.com>
14022 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
14023 offset_int::from instead of wide_int::to_shwi.
14024 (maybe_diag_overlap): Remove assertion.
14025 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
14026 * gimple-ssa-sprintf.c (format_directive): Same.
14027 (parse_directive): Same.
14028 (sprintf_dom_walker::compute_format_length): Same.
14029 (try_substitute_return_value): Same.
14031 2018-01-03 Jeff Law <law@redhat.com>
14033 PR middle-end/83654
14034 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
14035 non-constant residual for zero at runtime and avoid probing in
14036 that case. Reorganize code for trailing problem to mirror handling
14039 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14041 PR tree-optimization/83501
14042 * tree-ssa-strlen.c (get_string_cst): New.
14043 (handle_char_store): Call get_string_cst.
14045 2018-01-03 Martin Liska <mliska@suse.cz>
14047 PR tree-optimization/83593
14048 * tree-ssa-strlen.c: Include tree-cfg.h.
14049 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
14050 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
14051 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
14053 (strlen_dom_walker::before_dom_children): Call
14054 gimple_purge_dead_eh_edges. Dump tranformation with details
14056 (strlen_dom_walker::before_dom_children): Update call by adding
14057 new argument cleanup_eh.
14058 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
14060 2018-01-03 Martin Liska <mliska@suse.cz>
14063 * cif-code.def (VARIADIC_THUNK): New enum value.
14064 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
14067 2018-01-03 Jan Beulich <jbeulich@suse.com>
14069 * sse.md (mov<mode>_internal): Tighten condition for when to use
14070 vmovdqu<ssescalarsize> for TI and OI modes.
14072 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14074 Update copyright years.
14076 2018-01-03 Martin Liska <mliska@suse.cz>
14079 * ipa-visibility.c (function_and_variable_visibility): Skip
14080 functions with noipa attribure.
14082 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14084 * gcc.c (process_command): Update copyright notice dates.
14085 * gcov-dump.c (print_version): Ditto.
14086 * gcov.c (print_version): Ditto.
14087 * gcov-tool.c (print_version): Ditto.
14088 * gengtype.c (create_file): Ditto.
14089 * doc/cpp.texi: Bump @copying's copyright year.
14090 * doc/cppinternals.texi: Ditto.
14091 * doc/gcc.texi: Ditto.
14092 * doc/gccint.texi: Ditto.
14093 * doc/gcov.texi: Ditto.
14094 * doc/install.texi: Ditto.
14095 * doc/invoke.texi: Ditto.
14097 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14099 * vector-builder.h (vector_builder::m_full_nelts): Change from
14100 unsigned int to poly_uint64.
14101 (vector_builder::full_nelts): Update prototype accordingly.
14102 (vector_builder::new_vector): Likewise.
14103 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
14104 (vector_builder::operator ==): Likewise.
14105 (vector_builder::finalize): Likewise.
14106 * int-vector-builder.h (int_vector_builder::int_vector_builder):
14107 Take the number of elements as a poly_uint64 rather than an
14109 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
14110 from unsigned int to poly_uint64.
14111 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
14112 (vec_perm_indices::new_vector): Likewise.
14113 (vec_perm_indices::length): Likewise.
14114 (vec_perm_indices::nelts_per_input): Likewise.
14115 (vec_perm_indices::input_nelts): Likewise.
14116 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
14117 number of elements per input as a poly_uint64 rather than an
14118 unsigned int. Use the original encoding for variable-length
14119 vectors, rather than clamping each individual element.
14120 For the second and subsequent elements in each pattern,
14121 clamp the step and base before clamping their sum.
14122 (vec_perm_indices::series_p): Handle polynomial element counts.
14123 (vec_perm_indices::all_in_range_p): Likewise.
14124 (vec_perm_indices_to_tree): Likewise.
14125 (vec_perm_indices_to_rtx): Likewise.
14126 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
14127 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
14128 (tree_vector_builder::new_binary_operation): Handle polynomial
14129 element counts. Return false if we need to know the number
14130 of elements at compile time.
14131 * fold-const.c (fold_vec_perm): Punt if the number of elements
14132 isn't known at compile time.
14134 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14136 * vec-perm-indices.h (vec_perm_builder): Change element type
14137 from HOST_WIDE_INT to poly_int64.
14138 (vec_perm_indices::element_type): Update accordingly.
14139 (vec_perm_indices::clamp): Handle polynomial element_types.
14140 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
14141 (vec_perm_indices::all_in_range_p): Likewise.
14142 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
14144 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
14145 polynomial vec_perm_indices element types.
14146 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
14147 * fold-const.c (fold_vec_perm): Likewise.
14148 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
14149 * tree-vect-generic.c (lower_vec_perm): Likewise.
14150 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
14151 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
14152 element type to HOST_WIDE_INT.
14154 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14155 Alan Hayward <alan.hayward@arm.com>
14156 David Sherwood <david.sherwood@arm.com>
14158 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
14159 rather than an int. Use plus_constant.
14160 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
14161 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
14163 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14164 Alan Hayward <alan.hayward@arm.com>
14165 David Sherwood <david.sherwood@arm.com>
14167 * calls.c (emit_call_1, expand_call): Change struct_value_size from
14168 a HOST_WIDE_INT to a poly_int64.
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 * calls.c (load_register_parameters): Cope with polynomial
14175 mode sizes. Require a constant size for BLKmode parameters
14176 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
14177 forces a parameter to be padded at the lsb end in order to
14178 fill a complete number of words, require the parameter size
14179 to be ordered wrt UNITS_PER_WORD.
14181 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14182 Alan Hayward <alan.hayward@arm.com>
14183 David Sherwood <david.sherwood@arm.com>
14185 * reload1.c (spill_stack_slot_width): Change element type
14186 from unsigned int to poly_uint64_pod.
14187 (alter_reg): Treat mode sizes as polynomial.
14189 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14190 Alan Hayward <alan.hayward@arm.com>
14191 David Sherwood <david.sherwood@arm.com>
14193 * reload.c (complex_word_subreg_p): New function.
14194 (reload_inner_reg_of_subreg, push_reload): Use it.
14196 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14197 Alan Hayward <alan.hayward@arm.com>
14198 David Sherwood <david.sherwood@arm.com>
14200 * lra-constraints.c (process_alt_operands): Reject matched
14201 operands whose sizes aren't ordered.
14202 (match_reload): Refer to this check here.
14204 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14205 Alan Hayward <alan.hayward@arm.com>
14206 David Sherwood <david.sherwood@arm.com>
14208 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
14209 that the mode size is in the set {1, 2, 4, 8, 16}.
14211 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14212 Alan Hayward <alan.hayward@arm.com>
14213 David Sherwood <david.sherwood@arm.com>
14215 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
14216 Use plus_constant instead of gen_rtx_PLUS.
14218 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14219 Alan Hayward <alan.hayward@arm.com>
14220 David Sherwood <david.sherwood@arm.com>
14222 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
14223 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
14224 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
14225 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
14226 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
14227 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
14228 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
14229 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
14230 * config/i386/i386.c (ix86_push_rounding): ...this new function.
14231 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
14233 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
14234 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
14235 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
14236 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
14237 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
14238 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
14239 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
14240 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
14241 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
14242 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
14244 * expr.c (emit_move_resolve_push): Treat the input and result
14245 of PUSH_ROUNDING as a poly_int64.
14246 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
14247 (emit_push_insn): Likewise.
14248 * lra-eliminations.c (mark_not_eliminable): Likewise.
14249 * recog.c (push_operand): Likewise.
14250 * reload1.c (elimination_effects): Likewise.
14251 * rtlanal.c (nonzero_bits1): Likewise.
14252 * calls.c (store_one_arg): Likewise. Require the padding to be
14253 known at compile time.
14255 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14256 Alan Hayward <alan.hayward@arm.com>
14257 David Sherwood <david.sherwood@arm.com>
14259 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
14260 Use plus_constant instead of gen_rtx_PLUS.
14262 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14263 Alan Hayward <alan.hayward@arm.com>
14264 David Sherwood <david.sherwood@arm.com>
14266 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
14267 rather than an int.
14269 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14270 Alan Hayward <alan.hayward@arm.com>
14271 David Sherwood <david.sherwood@arm.com>
14273 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
14274 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
14275 via stack temporaries. Treat the mode size as polynomial too.
14277 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14278 Alan Hayward <alan.hayward@arm.com>
14279 David Sherwood <david.sherwood@arm.com>
14281 * expr.c (expand_expr_real_2): When handling conversions involving
14282 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
14283 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
14284 as a poly_uint64 too.
14286 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14287 Alan Hayward <alan.hayward@arm.com>
14288 David Sherwood <david.sherwood@arm.com>
14290 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
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 * combine.c (can_change_dest_mode): Handle polynomial
14297 REGMODE_NATURAL_SIZE.
14298 * expmed.c (store_bit_field_1): Likewise.
14299 * expr.c (store_constructor): Likewise.
14300 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
14301 and polynomial REGMODE_NATURAL_SIZE.
14302 (gen_lowpart_common): Likewise.
14303 * reginfo.c (record_subregs_of_mode): Likewise.
14304 * rtlanal.c (read_modify_subreg_p): Likewise.
14306 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14307 Alan Hayward <alan.hayward@arm.com>
14308 David Sherwood <david.sherwood@arm.com>
14310 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
14311 numbers of elements.
14313 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14314 Alan Hayward <alan.hayward@arm.com>
14315 David Sherwood <david.sherwood@arm.com>
14317 * match.pd: Cope with polynomial numbers of vector elements.
14319 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14320 Alan Hayward <alan.hayward@arm.com>
14321 David Sherwood <david.sherwood@arm.com>
14323 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
14324 in a POINTER_PLUS_EXPR.
14326 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14327 Alan Hayward <alan.hayward@arm.com>
14328 David Sherwood <david.sherwood@arm.com>
14330 * omp-simd-clone.c (simd_clone_subparts): New function.
14331 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
14332 (ipa_simd_modify_function_body): Likewise.
14334 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14335 Alan Hayward <alan.hayward@arm.com>
14336 David Sherwood <david.sherwood@arm.com>
14338 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
14339 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
14340 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
14341 (expand_vector_condition, vector_element): Likewise.
14342 (subparts_gt): New function.
14343 (get_compute_type): Use subparts_gt.
14344 (count_type_subparts): Delete.
14345 (expand_vector_operations_1): Use subparts_gt instead of
14346 count_type_subparts.
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-data-refs.c (vect_no_alias_p): Replace with...
14353 (vect_compile_time_alias): ...this new function. Do the calculation
14354 on poly_ints rather than trees.
14355 (vect_prune_runtime_alias_test_list): Update call accordingly.
14357 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14358 Alan Hayward <alan.hayward@arm.com>
14359 David Sherwood <david.sherwood@arm.com>
14361 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
14363 (vect_schedule_slp_instance): Likewise.
14365 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14366 Alan Hayward <alan.hayward@arm.com>
14367 David Sherwood <david.sherwood@arm.com>
14369 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
14370 constant and extern definitions for variable-length vectors.
14371 (vect_get_constant_vectors): Note that the number of units
14372 is known to be constant.
14374 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14375 Alan Hayward <alan.hayward@arm.com>
14376 David Sherwood <david.sherwood@arm.com>
14378 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
14379 of units as polynomial. Choose between WIDE and NARROW based
14382 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14383 Alan Hayward <alan.hayward@arm.com>
14384 David Sherwood <david.sherwood@arm.com>
14386 * tree-vect-stmts.c (simd_clone_subparts): New function.
14387 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
14389 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14390 Alan Hayward <alan.hayward@arm.com>
14391 David Sherwood <david.sherwood@arm.com>
14393 * tree-vect-stmts.c (vectorizable_call): Treat the number of
14394 vectors as polynomial. Use build_index_vector for
14395 IFN_GOMP_SIMD_LANE.
14397 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14398 Alan Hayward <alan.hayward@arm.com>
14399 David Sherwood <david.sherwood@arm.com>
14401 * tree-vect-stmts.c (get_load_store_type): Treat the number of
14402 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
14403 for variable-length vectors.
14404 (vectorizable_mask_load_store): Treat the number of units as
14405 polynomial, asserting that it is constant if the condition has
14406 already been enforced.
14407 (vectorizable_store, vectorizable_load): Likewise.
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-loop.c (vectorizable_live_operation): Treat the number
14414 of units as polynomial. Punt if we can't tell at compile time
14415 which vector contains the final result.
14417 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14418 Alan Hayward <alan.hayward@arm.com>
14419 David Sherwood <david.sherwood@arm.com>
14421 * tree-vect-loop.c (vectorizable_induction): Treat the number
14422 of units as polynomial. Punt on SLP inductions. Use an integer
14423 VEC_SERIES_EXPR for variable-length integer reductions. Use a
14424 cast of such a series for variable-length floating-point
14427 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14428 Alan Hayward <alan.hayward@arm.com>
14429 David Sherwood <david.sherwood@arm.com>
14431 * tree.h (build_index_vector): Declare.
14432 * tree.c (build_index_vector): New function.
14433 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
14434 of units as polynomial, forcibly converting it to a constant if
14435 vectorizable_reduction has already enforced the condition.
14436 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
14437 to create a {1,2,3,...} vector.
14438 (vectorizable_reduction): Treat the number of units as polynomial.
14439 Choose vectype_in based on the largest scalar element size rather
14440 than the smallest number of units. Enforce the restrictions
14443 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14444 Alan Hayward <alan.hayward@arm.com>
14445 David Sherwood <david.sherwood@arm.com>
14447 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
14448 number of units as polynomial.
14450 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14451 Alan Hayward <alan.hayward@arm.com>
14452 David Sherwood <david.sherwood@arm.com>
14454 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
14455 * target.def (autovectorize_vector_sizes): Return the vector sizes
14456 by pointer, using vector_sizes rather than a bitmask.
14457 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
14458 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
14459 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
14461 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
14462 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
14463 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
14464 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
14465 * omp-general.c (omp_max_vf): Likewise.
14466 * omp-low.c (omp_clause_aligned_alignment): Likewise.
14467 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
14468 * tree-vect-loop.c (vect_analyze_loop): Likewise.
14469 * tree-vect-slp.c (vect_slp_bb): Likewise.
14470 * doc/tm.texi: Regenerate.
14471 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
14473 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
14474 the vector size as a poly_uint64 rather than an unsigned int.
14475 (current_vector_size): Change from an unsigned int to a poly_uint64.
14476 (get_vectype_for_scalar_type): Update accordingly.
14477 * tree.h (build_truth_vector_type): Take the size and number of
14478 units as a poly_uint64 rather than an unsigned int.
14479 (build_vector_type): Add a temporary overload that takes
14480 the number of units as a poly_uint64 rather than an unsigned int.
14481 * tree.c (make_vector_type): Likewise.
14482 (build_truth_vector_type): Take the number of units as a poly_uint64
14483 rather than an unsigned int.
14485 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14486 Alan Hayward <alan.hayward@arm.com>
14487 David Sherwood <david.sherwood@arm.com>
14489 * target.def (get_mask_mode): Take the number of units and length
14490 as poly_uint64s rather than unsigned ints.
14491 * targhooks.h (default_get_mask_mode): Update accordingly.
14492 * targhooks.c (default_get_mask_mode): Likewise.
14493 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
14494 * doc/tm.texi: Regenerate.
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 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
14501 * omp-general.c (omp_max_vf): Likewise.
14502 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
14503 (expand_omp_simd): Handle polynomial safelen.
14504 * omp-low.c (omplow_simd_context): Add a default constructor.
14505 (omplow_simd_context::max_vf): Change from int to poly_uint64.
14506 (lower_rec_simd_input_clauses): Update accordingly.
14507 (lower_rec_input_clauses): Likewise.
14509 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14510 Alan Hayward <alan.hayward@arm.com>
14511 David Sherwood <david.sherwood@arm.com>
14513 * tree-vectorizer.h (vect_nunits_for_cost): New function.
14514 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
14515 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
14516 (vect_analyze_slp_cost): Likewise.
14517 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
14518 (vect_model_load_cost): Likewise.
14520 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14521 Alan Hayward <alan.hayward@arm.com>
14522 David Sherwood <david.sherwood@arm.com>
14524 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
14525 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
14526 from an unsigned int * to a poly_uint64_pod *.
14527 (calculate_unrolling_factor): New function.
14528 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
14530 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14531 Alan Hayward <alan.hayward@arm.com>
14532 David Sherwood <david.sherwood@arm.com>
14534 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
14535 from an unsigned int to a poly_uint64.
14536 (_loop_vec_info::slp_unrolling_factor): Likewise.
14537 (_loop_vec_info::vectorization_factor): Change from an int
14539 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
14540 (vect_get_num_vectors): New function.
14541 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
14542 (vect_get_num_copies): Use vect_get_num_vectors.
14543 (vect_analyze_data_ref_dependences): Change max_vf from an int *
14544 to an unsigned int *.
14545 (vect_analyze_data_refs): Change min_vf from an int * to a
14547 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
14548 than an unsigned HOST_WIDE_INT.
14549 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
14550 (vect_analyze_data_ref_dependence): Change max_vf from an int *
14551 to an unsigned int *.
14552 (vect_analyze_data_ref_dependences): Likewise.
14553 (vect_compute_data_ref_alignment): Handle polynomial vf.
14554 (vect_enhance_data_refs_alignment): Likewise.
14555 (vect_prune_runtime_alias_test_list): Likewise.
14556 (vect_shift_permute_load_chain): Likewise.
14557 (vect_supportable_dr_alignment): Likewise.
14558 (dependence_distance_ge_vf): Take the vectorization factor as a
14559 poly_uint64 rather than an unsigned HOST_WIDE_INT.
14560 (vect_analyze_data_refs): Change min_vf from an int * to a
14562 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
14563 vfm1 as a poly_uint64 rather than an int. Make the same change
14564 for the returned bound_scalar.
14565 (vect_gen_vector_loop_niters): Handle polynomial vf.
14566 (vect_do_peeling): Likewise. Update call to
14567 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
14568 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
14570 * tree-vect-loop.c (vect_determine_vectorization_factor)
14571 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
14572 (vect_get_known_peeling_cost): Likewise.
14573 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
14574 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
14575 (vect_transform_loop): Likewise. Use the lowest possible VF when
14576 updating the upper bounds of the loop.
14577 (vect_min_worthwhile_factor): Make static. Return an unsigned int
14578 rather than an int.
14579 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
14580 polynomial unroll factors.
14581 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
14582 (vect_make_slp_decision): Likewise.
14583 (vect_supported_load_permutation_p): Likewise, and polynomial
14585 (vect_analyze_slp_cost): Handle polynomial vf.
14586 (vect_slp_analyze_node_operations): Likewise.
14587 (vect_slp_analyze_bb_1): Likewise.
14588 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
14589 than an unsigned HOST_WIDE_INT.
14590 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
14591 (vectorizable_load): Handle polynomial vf.
14592 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
14594 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
14596 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14597 Alan Hayward <alan.hayward@arm.com>
14598 David Sherwood <david.sherwood@arm.com>
14600 * match.pd: Handle bit operations involving three constants
14601 and try to fold one pair.
14603 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14605 * tree-vect-loop-manip.c: Include gimple-fold.h.
14606 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
14607 niters_maybe_zero parameters. Handle other cases besides a step of 1.
14608 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
14609 Add a path that uses a step of VF instead of 1, but disable it
14611 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
14612 and niters_no_overflow parameters. Update calls to
14613 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
14614 Create a new SSA name if the latter choses to use a ste other
14615 than zero, and return it via niters_vector_mult_vf_var.
14616 * tree-vect-loop.c (vect_transform_loop): Update calls to
14617 vect_do_peeling, vect_gen_vector_loop_niters and
14618 slpeel_make_loop_iterate_ntimes.
14619 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
14620 (vect_gen_vector_loop_niters): Update declarations after above changes.
14622 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
14624 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
14625 128-bit round to integer instructions.
14626 (ceil<mode>2): Likewise.
14627 (btrunc<mode>2): Likewise.
14628 (round<mode>2): Likewise.
14630 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14632 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
14633 unaligned VSX load/store on P8/P9.
14634 (expand_block_clear): Allow the use of unaligned VSX
14635 load/store on P8/P9.
14637 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14639 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
14641 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
14642 swap associated with both a load and a store.
14644 2018-01-02 Andrew Waterman <andrew@sifive.com>
14646 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
14647 * config/riscv/riscv.md (clear_cache): Use it.
14649 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
14651 * web.c: Remove out-of-date comment.
14653 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14655 * expr.c (fixup_args_size_notes): Check that any existing
14656 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
14657 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
14658 (emit_single_push_insn): ...here.
14660 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14662 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
14663 (const_vector_encoded_nelts): New function.
14664 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
14665 (const_vector_int_elt, const_vector_elt): Declare.
14666 * emit-rtl.c (const_vector_int_elt_1): New function.
14667 (const_vector_elt): Likewise.
14668 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
14669 of CONST_VECTOR_ELT.
14671 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14673 * expr.c: Include rtx-vector-builder.h.
14674 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
14675 directly on the tree encoding.
14676 (const_vector_from_tree): Likewise.
14677 * optabs.c: Include rtx-vector-builder.h.
14678 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
14679 sequence of "u" values.
14680 * vec-perm-indices.c: Include rtx-vector-builder.h.
14681 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
14682 directly on the vec_perm_indices encoding.
14684 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14686 * doc/rtl.texi (const_vector): Describe new encoding scheme.
14687 * Makefile.in (OBJS): Add rtx-vector-builder.o.
14688 * rtx-vector-builder.h: New file.
14689 * rtx-vector-builder.c: Likewise.
14690 * rtl.h (rtx_def::u2): Add a const_vector field.
14691 (CONST_VECTOR_NPATTERNS): New macro.
14692 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
14693 (CONST_VECTOR_DUPLICATE_P): Likewise.
14694 (CONST_VECTOR_STEPPED_P): Likewise.
14695 (CONST_VECTOR_ENCODED_ELT): Likewise.
14696 (const_vec_duplicate_p): Check for a duplicated vector encoding.
14697 (unwrap_const_vec_duplicate): Likewise.
14698 (const_vec_series_p): Check for a non-duplicated vector encoding.
14699 Say that the function only returns true for integer vectors.
14700 * emit-rtl.c: Include rtx-vector-builder.h.
14701 (gen_const_vec_duplicate_1): Delete.
14702 (gen_const_vector): Call gen_const_vec_duplicate instead of
14703 gen_const_vec_duplicate_1.
14704 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
14705 (gen_const_vec_duplicate): Use rtx_vector_builder.
14706 (gen_const_vec_series): Likewise.
14707 (gen_rtx_CONST_VECTOR): Likewise.
14708 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
14709 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
14710 Build a new vector rather than modifying a CONST_VECTOR in-place.
14711 (handle_special_swappables): Update call accordingly.
14712 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
14713 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
14714 Build a new vector rather than modifying a CONST_VECTOR in-place.
14715 (handle_special_swappables): Update call accordingly.
14717 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14719 * simplify-rtx.c (simplify_const_binary_operation): Use
14720 CONST_VECTOR_ELT instead of XVECEXP.
14722 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14724 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
14725 the selector elements to be different from the data elements
14726 if the selector is a VECTOR_CST.
14727 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
14728 ssizetype for the selector.
14730 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14732 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
14733 before testing each element individually.
14734 * tree-vect-generic.c (lower_vec_perm): Likewise.
14736 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14738 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
14739 * selftest-run-tests.c (selftest::run_tests): Call it.
14740 * vector-builder.h (vector_builder::operator ==): New function.
14741 (vector_builder::operator !=): Likewise.
14742 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
14743 (vec_perm_indices::all_from_input_p): New function.
14744 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
14745 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
14746 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
14747 instead of reading the VECTOR_CST directly. Detect whether both
14748 vector inputs are the same before constructing the vec_perm_indices,
14749 and update the number of inputs argument accordingly. Use the
14750 utility functions added above. Only construct sel2 if we need to.
14752 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14754 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
14755 the broadcast of the low byte.
14756 (expand_mult_highpart): Use an explicit encoding for the permutes.
14757 * optabs-query.c (can_mult_highpart_p): Likewise.
14758 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
14759 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14760 (vectorizable_bswap): Likewise.
14761 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
14762 explicit encoding for the power-of-2 permutes.
14763 (vect_permute_store_chain): Likewise.
14764 (vect_grouped_load_supported): Likewise.
14765 (vect_permute_load_chain): Likewise.
14767 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14769 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
14770 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
14771 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
14772 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
14773 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
14774 (vect_gen_perm_mask_any): Likewise.
14776 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14778 * int-vector-builder.h: New file.
14779 * vec-perm-indices.h: Include int-vector-builder.h.
14780 (vec_perm_indices): Redefine as an int_vector_builder.
14781 (auto_vec_perm_indices): Delete.
14782 (vec_perm_builder): Redefine as a stand-alone class.
14783 (vec_perm_indices::vec_perm_indices): New function.
14784 (vec_perm_indices::clamp): Likewise.
14785 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
14786 (vec_perm_indices::new_vector): New function.
14787 (vec_perm_indices::new_expanded_vector): Update for new
14788 vec_perm_indices class.
14789 (vec_perm_indices::rotate_inputs): New function.
14790 (vec_perm_indices::all_in_range_p): Operate directly on the
14791 encoded form, without computing elided elements.
14792 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
14793 encoding. Update for new vec_perm_indices class.
14794 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
14795 the given vec_perm_builder.
14796 (expand_vec_perm_var): Update vec_perm_builder constructor.
14797 (expand_mult_highpart): Use vec_perm_builder instead of
14798 auto_vec_perm_indices.
14799 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
14800 vec_perm_indices instead of auto_vec_perm_indices. Use a single
14801 or double series encoding as appropriate.
14802 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
14803 vec_perm_indices instead of auto_vec_perm_indices.
14804 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14805 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14806 (vect_permute_store_chain): Likewise.
14807 (vect_grouped_load_supported): Likewise.
14808 (vect_permute_load_chain): Likewise.
14809 (vect_shift_permute_load_chain): Likewise.
14810 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
14811 (vect_transform_slp_perm_load): Likewise.
14812 (vect_schedule_slp_instance): Likewise.
14813 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14814 (vectorizable_mask_load_store): Likewise.
14815 (vectorizable_bswap): Likewise.
14816 (vectorizable_store): Likewise.
14817 (vectorizable_load): Likewise.
14818 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
14819 vec_perm_indices instead of auto_vec_perm_indices. Use
14820 tree_to_vec_perm_builder to read the vector from a tree.
14821 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
14822 vec_perm_builder instead of a vec_perm_indices.
14823 (have_whole_vector_shift): Use vec_perm_builder and
14824 vec_perm_indices instead of auto_vec_perm_indices. Leave the
14825 truncation to calc_vec_perm_mask_for_shift.
14826 (vect_create_epilog_for_reduction): Likewise.
14827 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
14828 from auto_vec_perm_indices to vec_perm_indices.
14829 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
14830 instead of changing individual elements.
14831 (aarch64_vectorize_vec_perm_const): Use new_vector to install
14832 the vector in d.perm.
14833 * config/arm/arm.c (expand_vec_perm_d::perm): Change
14834 from auto_vec_perm_indices to vec_perm_indices.
14835 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
14836 instead of changing individual elements.
14837 (arm_vectorize_vec_perm_const): Use new_vector to install
14838 the vector in d.perm.
14839 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
14840 Update vec_perm_builder constructor.
14841 (rs6000_expand_interleave): Likewise.
14842 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
14843 (rs6000_expand_interleave): Likewise.
14845 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14847 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
14848 to qimode could truncate the indices.
14849 * optabs.c (expand_vec_perm_var): Likewise.
14851 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14853 * Makefile.in (OBJS): Add vec-perm-indices.o.
14854 * vec-perm-indices.h: New file.
14855 * vec-perm-indices.c: Likewise.
14856 * target.h (vec_perm_indices): Replace with a forward class
14858 (auto_vec_perm_indices): Move to vec-perm-indices.h.
14859 * optabs.h: Include vec-perm-indices.h.
14860 (expand_vec_perm): Delete.
14861 (selector_fits_mode_p, expand_vec_perm_var): Declare.
14862 (expand_vec_perm_const): Declare.
14863 * target.def (vec_perm_const_ok): Replace with...
14864 (vec_perm_const): ...this new hook.
14865 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
14866 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
14867 * doc/tm.texi: Regenerate.
14868 * optabs.def (vec_perm_const): Delete.
14869 * doc/md.texi (vec_perm_const): Likewise.
14870 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
14871 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
14872 expand_vec_perm for constant permutation vectors. Assert that
14873 the mode of variable permutation vectors is the integer equivalent
14874 of the mode that is being permuted.
14875 * optabs-query.h (selector_fits_mode_p): Declare.
14876 * optabs-query.c: Include vec-perm-indices.h.
14877 (selector_fits_mode_p): New function.
14878 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
14879 is defined, instead of checking whether the vec_perm_const_optab
14880 exists. Use targetm.vectorize.vec_perm_const instead of
14881 targetm.vectorize.vec_perm_const_ok. Check whether the indices
14882 fit in the vector mode before using a variable permute.
14883 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
14884 vec_perm_indices instead of an rtx.
14885 (expand_vec_perm): Replace with...
14886 (expand_vec_perm_const): ...this new function. Take the selector
14887 as a vec_perm_indices rather than an rtx. Also take the mode of
14888 the selector. Update call to shift_amt_for_vec_perm_mask.
14889 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
14890 Use vec_perm_indices::new_expanded_vector to expand the original
14891 selector into bytes. Check whether the indices fit in the vector
14892 mode before using a variable permute.
14893 (expand_vec_perm_var): Make global.
14894 (expand_mult_highpart): Use expand_vec_perm_const.
14895 * fold-const.c: Includes vec-perm-indices.h.
14896 * tree-ssa-forwprop.c: Likewise.
14897 * tree-vect-data-refs.c: Likewise.
14898 * tree-vect-generic.c: Likewise.
14899 * tree-vect-loop.c: Likewise.
14900 * tree-vect-slp.c: Likewise.
14901 * tree-vect-stmts.c: Likewise.
14902 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
14904 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
14905 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
14906 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
14907 (aarch64_vectorize_vec_perm_const): ...this new function.
14908 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14909 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14910 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
14911 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
14912 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14913 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14914 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
14916 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
14917 check for NEON modes.
14918 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
14919 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
14920 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
14921 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
14923 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
14924 the old VEC_PERM_CONST conditions.
14925 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
14926 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
14927 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
14928 (ia64_vectorize_vec_perm_const_ok): Merge into...
14929 (ia64_vectorize_vec_perm_const): ...this new function.
14930 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
14931 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
14932 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
14933 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
14934 * config/mips/mips.c (mips_expand_vec_perm_const)
14935 (mips_vectorize_vec_perm_const_ok): Merge into...
14936 (mips_vectorize_vec_perm_const): ...this new function.
14937 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
14938 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
14939 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
14940 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
14941 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
14942 (rs6000_expand_vec_perm_const): Delete.
14943 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
14945 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14946 (altivec_expand_vec_perm_const_le): Take each operand individually.
14947 Operate on constant selectors rather than rtxes.
14948 (altivec_expand_vec_perm_const): Likewise. Update call to
14949 altivec_expand_vec_perm_const_le.
14950 (rs6000_expand_vec_perm_const): Delete.
14951 (rs6000_vectorize_vec_perm_const_ok): Delete.
14952 (rs6000_vectorize_vec_perm_const): New function.
14953 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
14954 an element count and rtx array.
14955 (rs6000_expand_extract_even): Update call accordingly.
14956 (rs6000_expand_interleave): Likewise.
14957 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
14958 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
14959 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
14960 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
14961 (rs6000_expand_vec_perm_const): Delete.
14962 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14963 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14964 (altivec_expand_vec_perm_const_le): Take each operand individually.
14965 Operate on constant selectors rather than rtxes.
14966 (altivec_expand_vec_perm_const): Likewise. Update call to
14967 altivec_expand_vec_perm_const_le.
14968 (rs6000_expand_vec_perm_const): Delete.
14969 (rs6000_vectorize_vec_perm_const_ok): Delete.
14970 (rs6000_vectorize_vec_perm_const): New function. Remove stray
14971 reference to the SPE evmerge intructions.
14972 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
14973 an element count and rtx array.
14974 (rs6000_expand_extract_even): Update call accordingly.
14975 (rs6000_expand_interleave): Likewise.
14976 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
14977 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
14979 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14981 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14983 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
14984 vector mode and that that mode matches the mode of the data
14986 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
14987 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
14988 directly using expand_vec_perm_1 when forcing selectors into
14990 (expand_vec_perm_var): New function, split out from expand_vec_perm.
14992 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14994 * optabs-query.h (can_vec_perm_p): Delete.
14995 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
14996 * optabs-query.c (can_vec_perm_p): Split into...
14997 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
14998 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
14999 particular selector is valid.
15000 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15001 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
15002 (vect_grouped_load_supported): Likewise.
15003 (vect_shift_permute_load_chain): Likewise.
15004 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
15005 (vect_transform_slp_perm_load): Likewise.
15006 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
15007 (vectorizable_bswap): Likewise.
15008 (vect_gen_perm_mask_checked): Likewise.
15009 * fold-const.c (fold_ternary_loc): Likewise. Don't take
15010 implementations of variable permutation vectors into account
15011 when deciding which selector to use.
15012 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
15013 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
15014 with a false third argument.
15015 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
15016 to test whether the constant selector is valid and can_vec_perm_var_p
15017 to test whether a variable selector is valid.
15019 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15021 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
15022 * optabs-query.c (can_vec_perm_p): Likewise.
15023 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
15024 instead of vec_perm_indices.
15025 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
15026 (vect_gen_perm_mask_checked): Likewise,
15027 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
15028 (vect_gen_perm_mask_checked): Likewise,
15030 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15032 * optabs-query.h (qimode_for_vec_perm): Declare.
15033 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
15034 (qimode_for_vec_perm): ...this new function.
15035 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
15037 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15039 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
15040 does not have a conditional at the top.
15042 2018-01-02 Richard Biener <rguenther@suse.de>
15044 * ipa-inline.c (big_speedup_p): Fix expression.
15046 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
15049 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
15052 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
15056 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
15057 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
15058 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
15059 cond_taken_branch_cost 3->4.
15061 2018-01-01 Jakub Jelinek <jakub@redhat.com>
15063 PR tree-optimization/83581
15064 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
15065 TODO_cleanup_cfg if any changes have been made.
15067 PR middle-end/83608
15068 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
15069 convert_modes if target mode has the right side, but different mode
15072 PR middle-end/83609
15073 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
15074 last argument when extracting from CONCAT. If either from_real or
15075 from_imag is NULL, use expansion through memory. If result is not
15076 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
15077 the parts directly to inner mode, if even that fails, use expansion
15080 PR middle-end/83623
15081 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
15082 check for bswap in mode rather than HImode and use that in expand_unop
15085 Copyright (C) 2018 Free Software Foundation, Inc.
15087 Copying and distribution of this file, with or without modification,
15088 are permitted in any medium without royalty provided the copyright
15089 notice and this notice are preserved.