1 2018-08-13 Ilya Leoshkevich <iii@linux.ibm.com>
3 * config/s390/s390.c (s390_decompose_constant_pool_ref):
5 (s390_decompose_address): Factor out constant pool ref
8 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
10 * config/nds32/nds32-predicates.c
11 (nds32_can_use_bclr_p): Change return type as bool.
12 (nds32_can_use_bset_p): Ditto.
13 (nds32_can_use_btgl_p): Ditto.
14 (nds32_can_use_bitci_p): Ditto.
15 * config/nds32/nds32-protos.h
16 (nds32_can_use_bclr_p): Change declaration.
17 (nds32_can_use_bset_p): Ditto.
18 (nds32_can_use_btgl_p): Ditto.
19 (nds32_can_use_bitci_p): Ditto.
21 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
23 * config/nds32/nds32.c (nds32_expand_prologue, nds32_expand_epilogue):
24 Support -msched-prolog-epilog option.
25 * config/nds32/nds32.opt (msched-prolog-epilog): New option.
27 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
29 * common/config/nds32/nds32-common.c
30 (nds32_option_optimization_table): Enalbe -malways-align.
32 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
34 * config.gcc (nds32*): Add nds32_isr.h and nds32_init.inc in
36 * common/config/nds32/nds32-common.c (nds32_handle_option): Handle
37 OPT_misr_secure_ case.
38 * config/nds32/nds32-isr.c: Implementation of backward compatibility.
39 * config/nds32/nds32-protos.h (nds32_isr_function_critical_p): New.
40 * config/nds32/nds32.c (nds32_attribute_table): Add critical and
42 * config/nds32/nds32.h (nds32_isr_nested_type): Add NDS32_CRITICAL.
43 (nds32_isr_info): New field security_level.
44 (TARGET_ISR_VECTOR_SIZE_4_BYTE): New macro.
45 * config/nds32/nds32.md (return_internal): Consider critical attribute.
46 * config/nds32/nds32.opt (misr-secure): New option.
47 * config/nds32/nds32_init.inc: New file.
48 * config/nds32/nds32_isr.h: New file.
50 2018-08-11 John David Anglin <danglin@gcc.gnu.org>
52 * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
53 Update comment for atomic instructions.
54 (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
55 atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
57 (atomic_loaddi): Revise fence expansion to only emit fence prior to
58 load for __ATOMIC_SEQ_CST model.
59 (atomic_loaddi_1): Remove float register target.
60 (atomic_storedi): Handle CONST_INT values.
61 (atomic_storedi_1): Remove float register source. Add special case
63 (memory_barrier): New expander and insn.
65 2018-08-11 Jakub Jelinek <jakub@redhat.com>
67 PR tree-optimization/86835
68 * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
69 new_stmt after def_gsi, make sure to insert new_square_stmt after
70 that stmt, not 2 stmts before it.
72 2018-08-10 Alexander Monakov <amonakov@ispras.ru>
75 * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
78 2018-08-10 Martin Liska <mliska@suse.cz>
81 * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
83 * builtins.c (expand_builtin_expect_with_probability):
85 (expand_builtin_expect_with_probability): New function.
86 (build_builtin_expect_predicate): Add new argumnet probability
87 for BUILT_IN_EXPECT_WITH_PROBABILITY.
88 (fold_builtin_expect):
91 * builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY):
92 * builtins.h (fold_builtin_expect): Set new argument.
93 * doc/extend.texi: Document __builtin_expect_with_probability.
94 * doc/invoke.texi: Likewise.
95 * gimple-fold.c (gimple_fold_call): Pass new argument.
96 * ipa-fnsummary.c (find_foldable_builtin_expect): Handle
97 also BUILT_IN_EXPECT_WITH_PROBABILITY.
98 * predict.c (get_predictor_value): New function.
99 (expr_expected_value): Add new argument probability. Assume
100 that predictor and probability are always non-null.
101 (expr_expected_value_1): Likewise. For __builtin_expect and
102 __builtin_expect_with_probability set probability. Handle
103 combination in binary expressions.
104 (tree_predict_by_opcode): Simplify code by simply calling
106 (pass_strip_predict_hints::execute): Add handling of
107 BUILT_IN_EXPECT_WITH_PROBABILITY.
108 * predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add
110 * tree.h (DECL_BUILT_IN_P): New function.
112 2018-08-10 Martin Liska <mliska@suse.cz>
114 PR tree-optimization/85799
115 * passes.def: Add argument for pass_strip_predict_hints.
116 * predict.c (class pass_strip_predict_hints): Add new argument
118 (strip_predictor_early): New function.
119 (pass_strip_predict_hints::execute): Call the function to
121 (strip_predict_hints): New function.
122 * predict.def: Fix comment.
124 2018-08-10 Thomas Preud'homme <thomas.preudhomme@linaro.org>
126 * Makefile.in: Clarify which tm.texi to copy over to assert the
127 right to grant a GFDL license for all.
129 2018-08-09 Jeff Law <law@redhat.com>
131 * config/m68k/m68k.c (m68k_adjust_decorated_operand): Remove
134 2018-08-09 Andreas Schwab <schwab@linux-m68k.org>
136 * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
139 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
141 * tree-vect-loop.c (vectorizable_reduction): Allow inner-loop
142 reductions for variable-length vectors.
144 2018-08-09 David Malcolm <dmalcolm@redhat.com>
147 * common.opt (fdiagnostics-show-line-numbers): New option.
148 * diagnostic-show-locus.c (class layout): Add fields
149 "m_show_line_numbers_p" and "m_linenum_width";
150 (num_digits): New function.
151 (test_num_digits): New function.
152 (layout::layout): Initialize new fields. Update m_x_offset
153 logic to handle any left margin.
154 (layout::print_source_line): Print line number when requested.
155 (layout::start_annotation_line): New member function.
156 (layout::print_annotation_line): Call it.
157 (layout::print_leading_fixits): Likewise.
158 (layout::print_trailing_fixits): Likewise. Update calls to
159 move_to_column for new parameter.
160 (layout::get_x_bound_for_row): Add "add_left_margin" param and use
161 it to potentially call start_annotation_line.
162 (layout::show_ruler): Call start_annotation_line.
163 (selftest::test_line_numbers_multiline_range): New selftest.
164 (selftest::diagnostic_show_locus_c_tests): Call test_num_digits
165 and selftest::test_line_numbers_multiline_range.
166 * diagnostic.c (diagnostic_initialize): Initialize
168 * diagnostic.h (struct diagnostic_context): Add field
169 "show_line_numbers_p".
170 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
171 -fno-diagnostics-show-line-numbers.
172 * dwarf2out.c (gen_producer_string): Add
173 OPT_fdiagnostics_show_line_numbers to the ignored options.
174 * lto-wrapper.c (merge_and_complain): Likewise to the "pick
175 one setting" options.
176 (append_compiler_options): Likewise to the dropped options.
177 (append_diag_options): Likewise to the passed-on options.
178 * opts.c (common_handle_option): Handle the new option.
179 * toplev.c (general_init): Set up global_dc->show_line_numbers_p.
181 2018-08-09 Kelvin Nilsen <kelvin@gcc.gnu.org>
183 * doc/extend.texi (PowerPC AltiVec Built-in Functions Available on
184 ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub. Add
185 third argument of type "const signed char" to descriptions of
186 __builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq,
187 __builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub,
188 __builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt,
189 __builtin_bcdsub_ov functions.
191 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
193 PR tree-optimization/86858
194 * tree-vect-loop.c (vect_is_simple_reduction): Restore
195 flow_bb_inside_loop_p calls.
197 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
199 PR tree-optimization/86871
200 * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
201 instead of gimple_assign_lhs.
203 2018-08-09 Richard Earnshaw <rearnsha@arm.com>
206 * config/aarch64/aarch64.md (add<mode>3_carryinC_zero): Add missing
207 register constraint to operand 0.
208 (add<mode>3_carryinC): Likewise.
209 (add<mode>3_carryinV_zero, add<mode>3_carryinV): Likewise.
211 2018-08-09 Martin Liska <mliska@suse.cz>
214 * common.opt: Remove extra line.
216 2018-08-09 Martin Liska <mliska@suse.cz>
218 * params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots
219 at the end of a line, make first letter capital and end up
220 a sentence with a dot.
221 (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise.
222 (PARAM_LOOP_BLOCK_TILE_SIZE): Likewise.
223 (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise.
224 (PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise.
225 (PARAM_MAX_ISL_OPERATIONS): Likewise.
226 (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise.
227 (PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise.
228 (PARAM_INDIR_CALL_TOPN_PROFILE): Likewise.
229 (PARAM_SLP_MAX_INSNS_IN_BB): Likewise.
230 (PARAM_IPA_CP_EVAL_THRESHOLD): Likewise.
231 (PARAM_IPA_CP_RECURSION_PENALTY): Likewise.
232 (PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
233 (PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise.
234 (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise.
235 (PARAM_TREE_REASSOC_WIDTH): Likewise.
236 (PARAM_HSA_GEN_DEBUG_STORES): Likewise.
237 (PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise.
238 (PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise.
240 2018-08-09 Andreas Krebbel <krebbel@linux.ibm.com>
243 * config/s390/s390.c (s390_option_override_internal): Reduce the
244 stack-clash-protection-probe-interval param if it would be too big
247 2018-08-08 Andreas Schwab <schwab@linux-m68k.org>
250 * config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define.
251 * config/m68k/m68k.c (handle_move_double): Don't call
252 m68k_final_prescan_insn.
253 (m68k_adjust_decorated_operand): Renamed from
254 m68k_final_prescan_insn, remove first and third operand and
256 (print_operand): Call it.
257 (print_operand_address): Call it.
259 2018-08-08 Nathan Sidwell <nathan@acm.org>
261 * diagnostic.c (diagnostic_report_current_module): Use
262 linemap_included_from & linemap_included_from_linemap.
264 2018-08-08 Hongbo Zhang <hongbo.zhang@linaro.org>
266 * config/aarch64/aarch64-cores.def: Add phecda core.
267 * config/aarch64/aarch64-tune.md: Regenerate.
268 * doc/invoke.texi: Add phecda core.
270 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
273 * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
275 * config/s390/s390.md ("movti"): Add more alternatives for
276 constant to GPR copies.
278 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
280 * config/s390/s390.c: Fix whitespace damage throughout the file.
281 * config/s390/s390.h: Likewise.
282 * config/s390/tpf.h: Likewise.
284 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
286 * config/s390/s390.c (s390_loadrelative_operand_p):
287 Remove TARGET_CPU_ZARCH usages.
288 (s390_rtx_costs): Likewise.
289 (s390_legitimate_constant_p): Likewise.
290 (s390_cannot_force_const_mem): Likewise.
291 (legitimate_reload_constant_p): Likewise.
292 (s390_preferred_reload_class): Likewise.
293 (legitimize_pic_address): Likewise.
294 (legitimize_tls_address): Likewise.
295 (s390_split_branches): Removed.
296 (s390_add_execute): Removed.
297 (s390_dump_pool): Remove TARGET_CPU_ZARCH usages.
298 (s390_mainpool_start): Likewise.
299 (s390_mainpool_finish): Likewise.
300 (s390_mainpool_cancel): Removed.
301 (s390_chunkify_start): Remove TARGET_CPU_ZARCH usages.
302 (s390_chunkify_cancel): Likewise.
303 (s390_return_addr_rtx): Likewise.
304 (s390_register_info): Remove split_branches_pending_p uages.
305 (s390_optimize_register_info): Likewise.
306 (s390_init_frame_layout): Remove TARGET_CPU_ZARCH and
307 split_branches_pending_p usages.
308 (s390_can_eliminate): Remove TARGET_CPU_ZARCH usages.
309 (s390_load_got): Likewise.
310 (s390_expand_split_stack_prologue): Likewise.
311 (output_asm_nops): Likewise.
312 (s390_function_profiler): Likewise.
313 (s390_emit_call): Likewise.
314 (s390_conditional_register_usage): Likewise.
315 (s390_optimize_prologue): Likewise.
316 (s390_reorg): Remove TARGET_CPU_ZARCH and
317 split_branches_pending_p usages.
318 (s390_option_override_internal): Remove TARGET_CPU_ZARCH
320 (s390_output_indirect_thunk_function): Likewise.
321 * config/s390/s390.h (TARGET_CPU_ZARCH): Removed.
322 (TARGET_CPU_ZARCH_P): Removed.
323 (struct machine_function): Remove split_branches_pending_p.
324 * config/s390/s390.md: Remove TARGET_CPU_ZARCH usages.
326 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
328 * common/config/s390/s390-common.c (processor_flags_table):
330 * config.gcc: Remove with_arch/with_tune support.
331 * config/s390/2064.md: Remove cpu attribute comparisons.
332 * config/s390/driver-native.c (s390_host_detect_local_cpu):
334 * config/s390/linux.h (ASM_SPEC):
335 Remove -march support.
336 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal):
337 Use a table to get an arch level.
338 * config/s390/s390-opts.h (enum processor_type):
341 (processor_table): Remove entries, add arch_level values.
342 (s390_issue_rate): Remove cases.
343 (s390_option_override): Adjust
344 s390_option_override_internal() call.
345 (s390_option_override_internal): Remove deprecation warning.
346 (s390_valid_target_attribute_tree): Adjust
347 s390_option_override_internal() call.
348 * config/s390/s390.h (struct s390_processor):
349 Share with s390-c.c, add arch_level field.
350 * config/s390/s390.md:
351 Remove occurrences in cpu attribute.
352 * config/s390/s390.opt: Remove -march/-mtune support.
353 * config/s390/tpf.h (ASM_SPEC): Remove -march support.
354 * doc/invoke.texi: Remove deprecation warning.
356 2018-08-08 Luis Machado <luis.machado@linaro.org>
358 * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static
360 (qdf24xx_tunings): Set vector cost structure to
363 * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
364 <register_sextend>: Set to 3.
366 2018-08-07 Richard Sandiford <richard.sandiford@arm.com>
369 * config/aarch64/iterators.md (FRECP, frecp_suffix): Delete.
370 * config/aarch64/aarch64-simd.md
371 (aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into...
372 (@aarch64_frecpe<mode>): ...here and the move FRECPX to...
373 (aarch64_frecpx<mode>): ...this new pattern.
374 * config/aarch64/aarch64-simd-builtins.def: Remove comment
375 about aarch64_frecp<FRECP:frecp_suffix><mode>.
377 2018-08-07 Martin Liska <mliska@suse.cz>
380 * predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC,
381 BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW.
382 * predict.def (PRED_MALLOC_NONNULL): New predictor.
383 * doc/extend.texi: Document that malloc attribute adds
386 2018-08-06 John David Anglin <danglin@gcc.gnu.org>
389 * config/pa/pa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
390 Define to speculation_safe_value_not_needed.
392 2018-08-06 Jeff Law <law@redhat.com>
394 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down
395 the vr_values instance to cprop_into_stmt.
396 (cprop_into_stmt): Pass vr_values instance down to cprop_operand.
397 (cprop_operand): Also query EVRP to determine if OP is a constant.
399 2018-08-06 Nathan Sidwell <nathan@acm.org>
401 * diagnostic.c (diagnostic_report_current_module): Reroll
402 included-at loop. Translate text.
404 2018-08-06 David Malcolm <dmalcolm@redhat.com>
406 * function-tests.c (selftest::test_expansion_to_rtl): Call
407 free_after_compilation.
409 2018-08-06 Alan Hayward <alan.hayward@arm.com>
411 * config/aarch64/aarch64.md: Add clobber highs to tls_desc.
413 2018-08-06 Andreas Krebbel <krebbel@linux.ibm.com>
415 * config/s390/s390.c (s390_loop_unroll_adjust): Prevent small
416 loops with memory block operations from getting unrolled.
418 2018-08-06 Ulrich Weigand <uweigand@de.ibm.com>
421 * config/spu/spu.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
422 Define to speculation_safe_value_not_needed.
424 2018-08-06 Jeff Law <law@redhat.com>
426 * reload1.c (forget_old_reloads_1): Adjust CLOBBER_HIGH
429 2018-08-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
432 * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
433 with all enabled __intN types.
435 * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
437 2018-08-06 Alan Hayward <alan.hayward@arm.com>
439 * alias.c (record_set): Check for clobber high.
440 * cfgexpand.c (expand_gimple_stmt): Likewise.
441 * combine-stack-adj.c (single_set_for_csa): Likewise.
442 * combine.c (find_single_use_1): Likewise.
443 (set_nonzero_bits_and_sign_copies): Likewise.
444 (get_combine_src_dest): Likewise.
445 (is_parallel_of_n_reg_sets): Likewise.
446 (try_combine): Likewise.
447 (record_dead_and_set_regs_1): Likewise.
448 (reg_dead_at_p_1): Likewise.
449 (reg_dead_at_p): Likewise.
450 * dce.c (deletable_insn_p): Likewise.
451 (mark_nonreg_stores_1): Likewise.
452 (mark_nonreg_stores_2): Likewise.
453 * df-scan.c (df_find_hard_reg_defs): Likewise.
454 (df_uses_record): Likewise.
455 (df_get_call_refs): Likewise.
456 * dwarf2out.c (mem_loc_descriptor): Likewise.
457 * haifa-sched.c (haifa_classify_rtx): Likewise.
458 * ira-build.c (create_insn_allocnos): Likewise.
459 * ira-costs.c (scan_one_insn): Likewise.
460 * ira.c (equiv_init_movable_p): Likewise.
461 (rtx_moveable_p): Likewise.
462 (interesting_dest_for_shprep): Likewise.
463 * jump.c (mark_jump_label_1): Likewise.
464 * postreload-gcse.c (record_opr_changes): Likewise.
465 * postreload.c (reload_cse_simplify): Likewise.
466 (struct reg_use): Add source expr.
467 (reload_combine): Check for clobber high.
468 (reload_combine_note_use): Likewise.
469 (reload_cse_move2add): Likewise.
470 (move2add_note_store): Likewise.
471 * print-rtl.c (print_pattern): Likewise.
472 * recog.c (decode_asm_operands): Likewise.
473 (store_data_bypass_p): Likewise.
474 (if_test_bypass_p): Likewise.
475 * regcprop.c (kill_clobbered_value): Likewise.
476 (kill_set_value): Likewise.
477 * reginfo.c (reg_scan_mark_refs): Likewise.
478 * reload1.c (maybe_fix_stack_asms): Likewise.
479 (eliminate_regs_1): Likewise.
480 (elimination_effects): Likewise.
481 (mark_not_eliminable): Likewise.
482 (scan_paradoxical_subregs): Likewise.
483 (forget_old_reloads_1): Likewise.
484 * reorg.c (find_end_label): Likewise.
485 (try_merge_delay_insns): Likewise.
486 (redundant_insn): Likewise.
487 (own_thread_p): Likewise.
488 (fill_simple_delay_slots): Likewise.
489 (fill_slots_from_thread): Likewise.
490 (dbr_schedule): Likewise.
491 * resource.c (update_live_status): Likewise.
492 (mark_referenced_resources): Likewise.
493 (mark_set_resources): Likewise.
494 * rtl.c (copy_rtx): Likewise.
495 * rtlanal.c (reg_referenced_p): Likewise.
496 (single_set_2): Likewise.
497 (noop_move_p): Likewise.
498 (note_stores): Likewise.
499 * sched-deps.c (sched_analyze_reg): Likewise.
500 (sched_analyze_insn): Likewise.
502 2018-08-06 Alan Hayward <alan.hayward@arm.com>
504 * cse.c (invalidate_reg): New function extracted from...
505 (invalidate): ...here.
506 (canonicalize_insn): Check for clobber high.
507 (invalidate_from_clobbers): invalidate clobber highs.
508 (invalidate_from_sets_and_clobbers): Likewise.
509 (count_reg_usage): Check for clobber high.
510 (insn_live_p): Likewise.
511 * cselib.c (cselib_expand_value_rtx_1):Likewise.
512 (cselib_invalidate_regno): Check for clobber in setter.
513 (cselib_invalidate_rtx): Pass through setter.
514 (cselib_invalidate_rtx_note_stores):
515 (cselib_process_insn): Check for clobber high.
516 * cselib.h (cselib_invalidate_rtx): Add operand.
518 2018-08-06 Alan Hayward <alan.hayward@arm.com>
520 * lra-eliminations.c (lra_eliminate_regs_1): Check for clobber high.
521 (mark_not_eliminable): Likewise.
522 * lra-int.h (struct lra_insn_reg): Add clobber high marker.
523 * lra-lives.c (process_bb_lives): Check for clobber high.
524 * lra.c (new_insn_reg): Remember clobber highs.
525 (collect_non_operand_hard_regs): Check for clobber high.
526 (lra_set_insn_recog_data): Likewise.
527 (add_regs_to_insn_regno_info): Likewise.
528 (lra_update_insn_regno_info): Likewise.
530 2018-08-06 Alan Hayward <alan.hayward@arm.com>
532 * rtl.h (reg_is_clobbered_by_clobber_high): Add declarations.
533 * rtlanal.c (reg_is_clobbered_by_clobber_high): Add function.
535 2018-08-06 Alan Hayward <alan.hayward@arm.com>
537 * emit-rtl.c (verify_rtx_sharing): Check for CLOBBER_HIGH.
538 (copy_insn_1): Likewise.
539 (gen_hard_reg_clobber_high): New gen function.
540 * genconfig.c (walk_insn_part): Check for CLOBBER_HIGH.
541 * genemit.c (gen_exp): Likewise.
542 (gen_emit_seq): Pass through info.
543 (gen_insn): Check for CLOBBER_HIGH.
544 (gen_expand): Pass through info.
545 (gen_split): Likewise.
546 (output_add_clobbers): Likewise.
547 * genrecog.c (validate_pattern): Check for CLOBBER_HIGH.
548 (remove_clobbers): Likewise.
549 * rtl.h (gen_hard_reg_clobber_high): New declaration.
551 2018-08-06 Alan Hayward <alan.hayward@arm.com>
553 * doc/rtl.texi (clobber_high): Add.
554 (parallel): Add in clobber high
555 * rtl.c (rtl_check_failed_code3): Add function.
556 * rtl.def (CLOBBER_HIGH): Add expression.
557 * rtl.h (RTL_CHECKC3): Add macro.
558 (rtl_check_failed_code3): Add declaration.
561 2018-08-05 H.J. Lu <hongjiu.lu@intel.com>
564 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
565 cfun->machine->max_used_stack_alignment if needed.
567 2018-08-04 Martin Sebor <msebor@redhat.com>
569 PR tree-optimization/86571
570 * gimple-ssa-sprintf.c (format_floating): Extend upper bound of
573 2018-08-03 Sandra Loosemore <sandra@codesourcery.com>
576 * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
579 2018-08-03 Jeff Law <law@redhat.com>
582 * config/mn10300/mn10300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
583 Define to speculation_safe_value_not_needed.
585 2018-08-03 David Malcolm <dmalcolm@redhat.com>
587 * doc/gcov.texi (-x): Remove duplicate "to".
588 * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
589 (-Wif-not-aligned): Remove duplicate "is".
590 (-flto): Remove duplicate "the".
591 (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
593 (MSP430 Options): Remove duplicate "and" from the description
594 of "-mgprel-sec=regexp".
595 (x86 Options): Remove duplicate copies of "vmldLog102" and
596 vmlsLog104 from description of "-mveclibabi=type".
598 2018-08-03 Richard Sandiford <richard.sandiford@arm.com>
600 * internal-fn.h (first_commutative_argument): Declare.
601 * internal-fn.c (first_commutative_argument): New function.
602 * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove extra
603 restrictions for pattern statements. Use first_commutative_argument
604 to look for commutative operands in calls to internal functions.
606 2018-08-03 Aldy Hernandez <aldyh@redhat.com>
608 * Makefile.in (wide-int-range.o): New.
609 * tree-vrp.c: Move all the wide_int_* functions to...
610 * wide-int-range.cc: ...here.
611 * tree-vrp.h: Move all the wide_int_* prototypes to...
612 * wide-int-range.h: ...here.
614 2018-08-03 Tom de Vries <tdevries@suse.de>
616 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): Return
618 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Remove define.
619 * except.c (output_function_exception_table): Do early exit if
620 targetm_common.except_unwind_info (&global_options) == UI_NONE.
622 2018-08-03 Martin Liska <mliska@suse.cz>
624 * predict.c (dump_prediction): Change to 2 digits
627 2018-08-03 Siddhesh Poyarekar <siddhesh@sourceware.org>
629 * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
631 (falkor_am_1_gtov_gtov): ... a new insn reservation.
633 2018-07-19 Ilya Leoshkevich <iii@linux.ibm.com>
635 * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Replace > with >=.
636 * df-problems.c (df_remove_dead_eq_notes): Replace > with >=.
637 * dwarf2out.c (mem_loc_descriptor): Replace > with >=.
638 * lra-constraints.c (spill_hard_reg_in_range): Replace <= with <.
639 * lra-remat.c (call_used_input_regno_present_p): Replace <= with <.
641 2018-08-02 David Malcolm <dmalcolm@redhat.com>
643 * diagnostic-show-locus.c (diagnostic_show_locus): Use
644 pp_take_prefix when saving the existing prefix.
645 * diagnostic.c (diagnostic_append_note): Likewise.
646 * langhooks.c (lhd_print_error_function): Likewise.
647 * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
648 param's type. Free the existing prefix.
649 (pp_take_prefix): New function.
650 (pretty_printer::pretty_printer): Drop the prefix parameter.
651 Rename the length parameter to match the comment.
652 (pretty_printer::~pretty_printer): Free the prefix.
653 * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
655 (struct pretty_printer): Drop the "const" from "prefix" field's
656 type and clarify memory management.
657 (pp_set_prefix): Drop the "const" from the 2nd param.
658 (pp_take_prefix): New decl.
660 2018-08-02 Aaron Sawdey <acsawdey@linux.ibm.com>
662 * config/rs6000/rs6000-string.c (select_block_compare_mode): Move test
663 for word_mode_ok here instead of passing as argument.
664 (expand_block_compare): Change select_block_compare_mode() call.
665 (expand_strncmp_gpr_sequence): New function.
666 (expand_strn_compare): Make use of expand_strncmp_gpr_sequence.
668 2018-08-02 Jeff Law <law@redhat.com>
671 * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
672 Define to speculation_safe_value_not_needed.
675 * config/h8300/h8300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
676 Define to speculation_safe_value_not_needed.
678 2018-08-02 Tom de Vries <tdevries@suse.de>
681 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): New
682 function. Return UI_TARGET unconditionally.
683 (TARGET_EXCEPT_UNWIND_INFO): Redefine to nvptx_except_unwind_info.
684 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Emit commented out '.byte'.
686 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
688 * genemit.c (print_overload_test): Fix typo.
690 2018-08-02 Richard Biener <rguenther@suse.de>
692 PR tree-optimization/86816
693 * tree-ssa-tail-merge.c (tail_merge_valueize): New function
694 which checks for value availability before querying it.
695 (gvn_uses_equal): Use it.
696 (same_succ_hash): Likewise.
697 (gimple_equal_p): Likewise.
699 2018-08-02 Nick Clifton <nickc@redhat.com>
702 * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
703 Define to speculation_safe_value_not_needed.
706 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
707 Define to speculation_safe_value_not_needed.
710 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
711 Define to speculation_safe_value_not_needed.
714 * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
715 Define to speculation_safe_value_not_needed.
718 * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
719 Define to speculation_safe_value_not_needed.
722 * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
723 Define to speculation_safe_value_not_needed.
726 * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
727 Define to speculation_safe_value_not_needed.
730 * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
731 Define to speculation_safe_value_not_needed.
734 * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
735 speculation_safe_value_not_needed.
738 * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define
739 to speculation_safe_value_not_needed.
741 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
743 * doc/md.texi: Expand the documentation of instruction names
744 to mention port-local uses. Document '@' in pattern names.
745 * read-md.h (overloaded_instance, overloaded_name): New structs.
747 (md_reader::handle_overloaded_name): New member function.
748 (md_reader::get_overloads): Likewise.
749 (md_reader::m_first_overload): New member variable.
750 (md_reader::m_next_overload_ptr): Likewise.
751 (md_reader::m_overloads_htab): Likewise.
752 * read-md.c (md_reader::md_reader): Initialize m_first_overload,
753 m_next_overload_ptr and m_overloads_htab.
754 * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields.
755 (get_mode_token, get_code_token, get_int_token): New functions.
756 (map_attr_string): Add an optional argument that passes back
757 the associated iterator.
758 (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p):
759 (md_reader::handle_overloaded_name, add_overload_instance): New
761 (apply_iterators): Handle '@' names. Report an error if '@'
762 is used without iterators.
763 (initialize_iterators): Initialize the new iterator_group fields.
764 * genopinit.c (handle_overloaded_code_for)
765 (handle_overloaded_gen): New functions.
766 (main): Use them to print declarations of maybe_code_for_* and
767 maybe_gen_* functions, and inline definitions of code_for_* and gen_*.
768 * genemit.c (print_overload_arguments, print_overload_test)
769 (handle_overloaded_code_for, handle_overloaded_gen): New functions.
770 (main): Use it to print definitions of maybe_code_for_* and
771 maybe_gen_* functions.
772 * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use
773 gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high}
774 instead of explicit mode checks.
775 (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine.
776 (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov.
777 (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive.
778 (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive.
779 (aarch64_expand_compare_and_swap): Likewise
780 gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse
781 (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas.
782 (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp.
783 (aarch64_constant_pool_reload_icode): Delete.
784 (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp
785 instead of aarch64_constant_pool_reload_icode. Use
786 code_for_aarch64_reload_mov instead of explicit mode checks.
787 (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete.
788 (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of
789 get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type.
790 (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete.
791 (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of
792 get_recpe_type and gen_aarch64_frecps instead of get_recps_type.
793 (aarch64_atomic_load_op_code): Delete.
794 (aarch64_emit_atomic_load_op): Likewise.
795 (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of
796 aarch64_atomic_load_op_code. Use gen_aarch64_atomic_load
797 instead of aarch64_emit_atomic_load_op.
798 * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>)
799 (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>)
800 (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high)
801 (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@'
802 character before the pattern name.
803 * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
804 (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>)
805 (aarch64_simd_combine<mode>, aarch64_frecpe<mode>)
806 (aarch64_frecps<mode>): Likewise.
807 * config/aarch64/atomics.md (atomic_compare_and_swap<mode>)
808 (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse)
809 (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>)
810 (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>)
811 (aarch64_atomic_load<atomic_ldop><mode>): Likewise.
813 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
815 * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
816 Allow HFmode constants if TARGET_FP_F16INST.
818 2018-08-02 Jackson Woodruff <jackson.woodruff@arm.com>
821 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp):
822 No longer check last store for clobber of address register.
824 2018-08-02 Martin Liska <mliska@suse.cz>
826 PR gcov-profile/86817
827 * gcov.c (process_all_functions): New function.
829 (process_file): Move functions processing to
830 process_all_functions.
832 2018-08-02 David Malcolm <dmalcolm@redhat.com>
834 * dumpfile.c (dump_user_location_t::dump_user_location_t): Add
835 "const" to the "gimple *" and "rtx_insn *" parameters.
836 * dumpfile.h (dump_user_location_t::dump_user_location_t):
838 (dump_location_t::dump_location_t): Likewise.
840 2018-08-01 Martin Sebor <msebor@redhat.com>
842 PR tree-optimization/86650
843 * gimple-pretty-print.c (percent_G_format): Accept a "gimple *"
844 rather than a "gcall *". Directly pass the data of interest
845 to percent_K_format, rather than building a temporary CALL_EXPR
847 * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust.
848 (gimple_fold_builtin_strncat): Adjust.
849 * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace
850 gcall* argument with gimple*.
851 * gimple-ssa-warn-restrict.c (check_call): Same.
852 (wrestrict_dom_walker::before_dom_children): Same.
853 (builtin_access::builtin_access): Same.
854 (check_bounds_or_overlap): Same
855 (maybe_diag_overlap): Same.
856 (maybe_diag_offset_bounds): Same.
857 * tree-diagnostic.c (default_tree_printer): Move usage of
858 EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format
860 * tree-pretty-print.c (percent_K_format): Add argument.
861 * tree-pretty-print.h: Add argument.
862 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust.
863 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust.
864 (maybe_diag_stxncpy_trunc): Same.
865 (handle_builtin_stxncpy): Same.
866 (handle_builtin_strcat): Same.
868 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
870 * match.pd: Optimise pointer range checks.
872 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
874 PR tree-optimization/86758
875 * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try
876 to remove pattern statements.
878 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
880 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the
881 result of dfs_enumerate_from when constructing stmt_vec_infos,
882 instead of additionally calling get_loop_body.
884 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
886 * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p
888 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise.
889 When creating an iv, assert that the step is not known to be zero.
890 (vect_setup_realignment): Update call accordingly.
891 * tree-vect-stmts.c (vectorizable_store): Likewise.
892 (vectorizable_load): Likewise. Handle VMAT_INVARIANT separately.
894 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
896 * tree-vectorizer.h (vect_stmt_to_vectorize): New function.
897 * tree-vect-loop.c (vect_update_vf_for_slp): Use it.
898 (vectorizable_reduction): Likewise.
899 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
900 (vect_detect_hybrid_slp_stmts): Likewise.
901 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
903 2018-08-01 Aldy Hernandez <aldyh@redhat.com>
905 * tree-vrp (zero_nonzero_bits_from_bounds): Rename to...
906 (wide_int_set_zero_nonzero_bits): ...this.
907 (zero_nonzero_bits_from_vr): Rename to...
908 (vrp_set_zero_nonzero_bits): ...this.
909 (extract_range_from_multiplicative_op_1): Abstract wide int
911 (wide_int_range_multiplicative_op): ...here.
912 (extract_range_from_binary_expr_1): Extract wide int binary
913 operations into their own functions.
914 (wide_int_range_lshift): New.
915 (wide_int_range_can_optimize_bit_op): New.
916 (wide_int_range_shift_undefined_p): New.
917 (wide_int_range_bit_xor): New.
918 (wide_int_range_bit_ior): New.
919 (wide_int_range_bit_and): New.
920 (wide_int_range_trunc_mod): New.
921 (extract_range_into_wide_ints): New.
922 (vrp_shift_undefined_p): New.
923 (extract_range_from_multiplicative_op): New.
924 (vrp_can_optimize_bit_op): New.
925 * tree-vrp.h (value_range::dump): New.
926 (wide_int_range_multiplicative_op): New.
927 (wide_int_range_lshift):New.
928 (wide_int_range_shift_undefined_p): New.
929 (wide_int_range_bit_xor): New.
930 (wide_int_range_bit_ior): New.
931 (wide_int_range_bit_and): New.
932 (wide_int_range_trunc_mod): New.
933 (zero_nonzero_bits_from_bounds): Rename to...
934 (wide_int_set_zero_nonzero_bits): ...this.
935 (zero_nonzero_bits_from_vr): Rename to...
936 (vrp_set_zero_nonzero_bits): ...this.
937 (range_easy_mask_min_max): Rename to...
938 (wide_int_range_can_optimize_bit_op): this.
939 * vr-values.c (simplify_bit_ops_using_ranges): Rename
940 zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits.
942 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
944 * tree-vectorizer.h (vect_orig_stmt): New function.
945 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it.
946 * tree-vect-loop.c (vect_model_reduction_cost): Likewise.
947 (vect_create_epilog_for_reduction): Likewise.
948 (vectorizable_live_operation): Likewise.
949 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
950 (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise.
951 * tree-vect-stmts.c (vectorizable_call): Likewise.
952 (vectorizable_simd_clone_call, vect_remove_stores): Likewise.
954 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
956 * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store
958 * tree-vect-stmts.c (vect_transform_stmt): Likewise.
959 * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly.
960 (vect_transform_loop): Likewise.
961 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
963 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
965 * tree-vectorizer.h (vect_schedule_slp): Return void.
966 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
967 (vect_schedule_slp): Likewise.
969 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
971 * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled
973 (vect_transform_loop): Update calls accordingly. Schedule SLP
974 instances before the main loop, if any exist.
976 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
978 PR tree-optimization/86749
979 * tree-vect-patterns.c (vect_determine_min_output_precision_1):
980 If the lhs is used in a COND_EXPR, check that it is being used
981 as the "then" or "else" value.
983 2018-08-01 Tom de Vries <tdevries@suse.de>
986 * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
987 speculation_safe_value_not_needed.
989 2018-08-01 Richard Biener <rguenther@suse.de>
991 * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses
994 2018-08-01 Martin Liska <mliska@suse.cz>
996 * value-prof.c (gimple_divmod_fixed_value_transform): Unify
997 format how successful transformation is dumped.
998 (gimple_mod_pow2_value_transform): Likewise.
999 (gimple_mod_subtract_transform): Likewise.
1000 (gimple_stringops_transform): Likewise.
1002 2018-08-01 Martin Liska <mliska@suse.cz>
1005 * value-prof.c (interesting_stringop_to_profile_p):
1006 Simplify the code and add BUILT_IN_MEMMOVE.
1007 (gimple_stringops_transform): Likewise.
1009 2018-08-01 Sam Tebbs <sam.tebbs@arm.com>
1011 * config/aarch64/aarch64-simd.md
1012 (*aarch64_get_lane_zero_extendsi<mode>): Rename to...
1013 (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and
1014 use GPI iterator instead of SI mode.
1016 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
1018 * config/rs6000/rs6000.md (speculation_barrier): Renamed from
1019 rs6000_speculation_barrier.
1020 * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for
1021 new barrier pattern name.
1023 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
1025 * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
1026 (speculation_barrier): New insn.
1028 2018-08-01 Richard Biener <rguenther@suse.de>
1031 * graphite.h: Include isl/id.h and isl/space.h to allow build
1034 2018-08-01 Jan Willem Jagersma <jwjagersma@gmail.com>
1037 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
1038 mode for COFF targets.
1039 * defaults.h (TARGET_COFF): Define.
1040 * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
1041 TARGET_COFF): Define.
1042 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
1043 * config/i386/djgpp.c (saved_debug_info_level): New static variable.
1044 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
1046 2018-07-31 Alexandre Oliva <oliva@adacore.com>
1048 * gimple-streamer-in.c (input_bb): Restore BB discriminator.
1049 * gimple-streamer-out.c (output_bb): Save it.
1050 * lto-streamer-in.c (input_struct_function_base): Restore
1051 instance discriminator if available. Create map on demand.
1052 * lto-streamer-out.c (output_struct_function_base): Save it if
1054 * final.c (decl_to_instance_map): Document LTO strategy.
1056 2018-07-31 Alexandre Oliva <oliva@adacore.com>
1057 Olivier Hainque <hainque@adacore.com>
1059 * debug.h (decl_to_instance_map_t): New type.
1060 (decl_to_instance_map): Declare.
1061 (maybe_create_decl_to_instance_map): New inline function.
1062 * final.c (bb_discriminator, last_bb_discriminator): New statics,
1063 to track basic block discriminators.
1064 (final_start_function_1): Initialize them.
1065 (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
1067 (decl_to_instance_map): New variable.
1068 (map_decl_to_instance, maybe_set_discriminator): New functions.
1069 (notice_source_line): Set discriminator.
1071 2018-07-31 Ian Lance Taylor <iant@golang.org>
1073 * targhooks.c (default_have_speculation_safe_value): Add
1076 2018-07-31 David Malcolm <dmalcolm@redhat.com>
1078 * dump-context.h: Include "pretty-print.h".
1079 (dump_context::refresh_dumps_are_enabled): New decl.
1080 (dump_context::emit_item): New decl.
1081 (class dump_context): Add fields "m_test_pp" and
1083 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
1084 (temp_dump_context::get_dumped_text): New decl.
1085 (class temp_dump_context): Add field "m_pp".
1086 * dumpfile.c (refresh_dumps_are_enabled): Convert to...
1087 (dump_context::refresh_dumps_are_enabled): ...and add a test for
1089 (set_dump_file): Update for above change.
1090 (set_alt_dump_file): Likewise.
1091 (dump_loc): New overload, taking a pretty_printer *.
1092 (dump_context::dump_loc): Call end_any_optinfo. Dump the location
1093 to any test pretty-printer.
1094 (make_item_for_dump_gimple_stmt): New function, adapted from
1095 optinfo::add_gimple_stmt.
1096 (dump_context::dump_gimple_stmt): Call it, and use the result,
1097 eliminating the direct usage of dump_file and alt_dump_file in
1098 favor of indirectly using them via emit_item.
1099 (make_item_for_dump_gimple_expr): New function, adapted from
1100 optinfo::add_gimple_expr.
1101 (dump_context::dump_gimple_expr): Call it, and use the result,
1102 eliminating the direct usage of dump_file and alt_dump_file in
1103 favor of indirectly using them via emit_item.
1104 (make_item_for_dump_generic_expr): New function, adapted from
1106 (dump_context::dump_generic_expr): Call it, and use the result,
1107 eliminating the direct usage of dump_file and alt_dump_file in
1108 favor of indirectly using them via emit_item.
1109 (make_item_for_dump_printf_va): New function, adapted from
1110 optinfo::add_printf_va.
1111 (make_item_for_dump_printf): New function.
1112 (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
1113 and use the result, eliminating the direct usage of dump_file and
1114 alt_dump_file in favor of indirectly using them via emit_item.
1115 (make_item_for_dump_dec): New function.
1116 (dump_context::dump_dec): Call it, and use the result,
1117 eliminating the direct usage of dump_file and alt_dump_file in
1118 favor of indirectly using them via emit_item.
1119 (make_item_for_dump_symtab_node): New function, adapted from
1120 optinfo::add_symtab_node.
1121 (dump_context::dump_symtab_node): Call it, and use the result,
1122 eliminating the direct usage of dump_file and alt_dump_file in
1123 favor of indirectly using them via emit_item.
1124 (dump_context::begin_scope): Reimplement, avoiding direct usage
1125 of dump_file and alt_dump_file in favor of indirectly using them
1127 (dump_context::emit_item): New member function.
1128 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
1129 Set up test pretty-printer on the underlying context. Call
1130 refresh_dumps_are_enabled.
1131 (temp_dump_context::~temp_dump_context): Call
1132 refresh_dumps_are_enabled.
1133 (temp_dump_context::get_dumped_text): New member function.
1134 (selftest::verify_dumped_text): New function.
1135 (ASSERT_DUMPED_TEXT_EQ): New macro.
1136 (selftest::test_capture_of_dump_calls): Run all tests twice, with
1137 and then without optinfo enabled. Add uses of
1138 ASSERT_DUMPED_TEXT_EQ to all tests. Add test of nested scopes.
1139 * dumpfile.h: Update comment for the dump_* API.
1140 * optinfo-emit-json.cc
1141 (selftest::test_building_json_from_dump_calls): Update for new
1142 param for temp_dump_context ctor.
1143 * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
1144 and "m_owned" field.
1145 (optinfo_item::~optinfo_item): Likewise.
1146 (optinfo::add_item): New member function.
1147 (optinfo::emit): Update comment.
1148 (optinfo::add_string): Delete.
1149 (optinfo::add_printf): Delete.
1150 (optinfo::add_printf_va): Delete.
1151 (optinfo::add_gimple_stmt): Delete.
1152 (optinfo::add_gimple_expr): Delete.
1153 (optinfo::add_tree): Delete.
1154 (optinfo::add_symtab_node): Delete.
1155 (optinfo::add_dec): Delete.
1156 * optinfo.h (class dump_context): New forward decl.
1157 (optinfo::add_item): New decl.
1158 (optinfo::add_string): Delete.
1159 (optinfo::add_printf): Delete.
1160 (optinfo::add_printf_va): Delete.
1161 (optinfo::add_gimple_stmt): Delete.
1162 (optinfo::add_gimple_expr): Delete.
1163 (optinfo::add_tree): Delete.
1164 (optinfo::add_symtab_node): Delete.
1165 (optinfo::add_dec): Delete.
1166 (optinfo::add_poly_int): Delete.
1167 (optinfo_item::optinfo_item): Remove "owned" param.
1168 (class optinfo_item): Remove field "m_owned".
1170 2018-07-31 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1173 * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
1174 MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
1175 requested variable alignment.
1176 (expand_one_ssa_partition): Likewise.
1177 (expand_one_var): Likewise.
1179 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1181 * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
1182 to speculation_safe_value_not_needed.
1184 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1186 * targhooks.h (speculation_safe_value_not_needed): New prototype.
1187 * targhooks.c (speculation_safe_value_not_needed): New function.
1188 * target.def (have_speculation_safe_value): Update documentation.
1189 * doc/tm.texi: Regenerated.
1191 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1193 * config/aarch64/iterators.md (ALLI_TI): New iterator.
1194 * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
1196 (despeculate_copy<ALLI:mode>_insn): New insn.
1197 (despeculate_copyti_insn): New insn.
1198 (despeculate_simple<ALLI:mode>): New insn
1199 (despeculate_simpleti): New insn.
1200 * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
1202 (TARGET_SPECULATION_SAFE_VALUE): Redefine to
1203 aarch64_speculation_safe_value.
1204 (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
1206 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1208 * config/aarch64/aarch64-speculation.cc: New file.
1209 * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
1210 before pass_reorder_blocks.
1211 * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
1213 * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
1214 X14 and X15 when tracking speculation.
1215 * config/aarch64/aarch64.md (register name constants): Add
1216 SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
1217 (unspec): Add UNSPEC_SPECULATION_TRACKER.
1218 (speculation_barrier): New insn attribute.
1219 (cmp<mode>): Allow SP in comparisons.
1220 (speculation_tracker): New insn.
1221 (speculation_barrier): Add speculation_barrier attribute.
1222 * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
1223 * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
1224 * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
1226 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1228 * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
1229 aarch64_track_speculation is true.
1230 (tb<optab><mode>1): Likewise.
1231 * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
1232 generate CB[N]Z when tracking speculation.
1233 (aarch64_split_compare_and_swap): Likewise.
1234 (aarch64_split_atomic_op): Likewise.
1236 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1238 * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
1240 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1242 * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
1243 (speculation_barrier): New insn.
1245 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1247 * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
1248 * config/arm/arm.md (speculation_barrier): New expand.
1249 (speculation_barrier_insn): New pattern.
1251 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1253 * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
1254 (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
1255 (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
1256 * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
1258 * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
1259 (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
1260 (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
1261 (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
1262 (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
1263 (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
1264 (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
1265 * builtins.c (expand_speculation_safe_value): New function.
1266 (expand_builtin): Call it.
1267 * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
1268 * doc/extend.texi: Document __builtin_speculation_safe_value.
1269 * doc/md.texi: Document "speculation_barrier" pattern.
1270 * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
1271 TARGET_HAVE_SPECULATION_SAFE_VALUE.
1272 * doc/tm.texi: Regenerated.
1273 * target.def (have_speculation_safe_value, speculation_safe_value): New
1275 * targhooks.c (default_have_speculation_safe_value): New function.
1276 (default_speculation_safe_value): New function.
1277 * targhooks.h (default_have_speculation_safe_value): Add prototype.
1278 (default_speculation_safe_value): Add prototype.
1280 2018-07-31 David Malcolm <dmalcolm@redhat.com>
1282 * dump-context.h (dump_context::dump_loc): New decl.
1283 * dumpfile.c (dump_context::dump_loc): New member function.
1284 (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
1285 and dump_gimple_stmt.
1286 (dump_context::dump_gimple_expr_loc): Likewise, using
1288 (dump_context::dump_generic_expr_loc): Likewise, using
1290 (dump_context::dump_printf_loc_va): Likewise, using
1292 (dump_context::begin_scope): Explicitly using the global function
1293 "dump_loc", rather than the member function.
1295 2018-07-31 Martin Sebor <msebor@redhat.com>
1297 PR tree-optimization/86741
1298 * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
1300 2018-07-31 Andreas Krebbel <krebbel@linux.ibm.com>
1302 * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
1303 depend on whether prefetch instructions will be emitted or not.
1304 Use TARGET_SETMEM_PFD for checking whether prefetch instructions
1305 will be emitted or not.
1306 * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
1307 (TARGET_SETMEM_PFD): New macros.
1309 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1311 * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
1312 (NULL_STMT_VEC_INFO): Delete.
1313 (stmt_vec_info::operator*): Likewise.
1314 (stmt_vec_info::operator gimple *): Likewise.
1315 * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
1316 of NULL_STMT_VEC_INFO.
1317 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
1318 (vect_reassociating_reduction_p): Likewise.
1319 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
1320 (vectorizable_store): Likewise.
1321 * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
1322 (vec_info::free_stmt_vec_infos): Likewise.
1324 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1326 * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
1327 * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
1329 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1331 * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
1332 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
1333 (vec_info::free_stmt_vec_info): New private member functions.
1334 (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
1335 (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
1336 * tree-parloops.c (gather_scalar_reductions): Remove calls to
1337 set_stmt_vec_info_vec and free_stmt_vec_infos.
1338 * tree-vect-loop.c (_loop_vec_info): Remove call to
1339 set_stmt_vec_info_vec.
1340 * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
1341 (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
1342 * tree-vectorizer.c (vec_info::new_stmt_vec_info)
1343 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
1344 (vec_info::free_stmt_vec_info): ...these new functions. Remove
1345 assignments in {vec_info::,}new_stmt_vec_info that are redundant
1346 with the clearing in the xcalloc.
1347 (stmt_vec_info_vec): Delete.
1348 (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
1349 (vectorize_loops): Likewise.
1350 (vec_info::~vec_info): Remove argument from call to
1351 free_stmt_vec_infos.
1352 (vec_info::add_stmt): Remove vinfo argument from call to
1355 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1357 * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
1358 rather than a gimple stmt.
1359 * tree-vect-stmts.c (free_stmt_vec_info): Likewise. Don't free
1360 information for pattern statements when passed the original
1361 statement; instead wait to be passed the pattern statement itself.
1362 Don't call set_vinfo_for_stmt here.
1363 (free_stmt_vec_infos): Update call to free_stmt_vec_info.
1364 * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
1365 stmt_vec_infos here.
1366 * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
1367 * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
1368 stmt_vec_infos entry.
1370 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1372 * tree-vectorizer.h (vec_info::replace_stmt): Declare.
1373 * tree-vectorizer.c (vec_info::replace_stmt): New function.
1374 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
1375 * tree-vect-stmts.c (vectorizable_call): Likewise.
1376 (vectorizable_simd_clone_call): Likewise.
1378 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1380 * tree-vectorizer.h (vec_info::remove_stmt): Declare.
1381 * tree-vectorizer.c (vec_info::remove_stmt): New function.
1382 * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
1383 * tree-vect-loop.c (vect_transform_loop): Likewise.
1384 * tree-vect-slp.c (vect_schedule_slp): Likewise.
1385 * tree-vect-stmts.c (vect_remove_stores): Likewise.
1387 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1389 * tree-vectorizer.h (vec_info::lookup_dr): New member function.
1390 (vect_dr_stmt): Delete.
1391 * tree-vectorizer.c (vec_info::lookup_dr): New function.
1392 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
1394 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
1395 (vect_analyze_data_ref_dependence, vect_record_base_alignments)
1396 (vect_verify_datarefs_alignment, vect_peeling_supportable)
1397 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
1398 (vect_analyze_data_refs): Likewise.
1399 (vect_slp_analyze_data_ref_dependence): Likewise. Take a vec_info
1401 (vect_find_same_alignment_drs): Likewise.
1402 (vect_slp_analyze_node_dependences): Update calls accordingly.
1403 (vect_analyze_data_refs_alignment): Likewise. Use vec_info::lookup_dr
1404 instead of DR_VECT_AUX.
1405 (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
1406 of a vector data references. Use vec_info::lookup_dr instead of
1408 (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
1409 (vect_enhance_data_refs_alignment): Likewise. Use vec_info::lookup_dr
1410 instead of DR_VECT_AUX.
1412 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1414 * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
1416 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
1418 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
1419 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
1420 (vect_gen_prolog_loop_niters): Likewise.
1422 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1424 * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
1425 (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
1426 (vect_known_alignment_in_bytes, vect_dr_behavior)
1427 (vect_get_scalar_dr_size): Take references as dr_vec_infos
1428 instead of data_references. Update calls to other routines for
1429 which the same change has been made.
1430 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
1431 dr_vec_infos instead of stmt_vec_infos.
1432 (vect_analyze_data_ref_dependence): Update call accordingly.
1433 (vect_slp_analyze_data_ref_dependence)
1434 (vect_record_base_alignments): Use DR_VECT_AUX.
1435 (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
1436 (vect_update_misalignment_for_peel, verify_data_ref_alignment)
1437 (vector_alignment_reachable_p, vect_get_data_access_cost)
1438 (vect_peeling_supportable, vect_analyze_group_access_1)
1439 (vect_analyze_group_access, vect_analyze_data_ref_access)
1440 (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
1441 (vect_compile_time_alias, vect_small_gap_p)
1442 (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
1443 (vect_supportable_dr_alignment): Take references as dr_vec_infos
1444 instead of data_references. Update calls to other routines for
1445 which the same change has been made.
1446 (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
1447 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
1448 (vect_slp_analyze_and_verify_node_alignment)
1449 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
1450 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
1451 (vect_setup_realignment): Use dr_vec_infos. Update calls after
1453 (_vect_peel_info::dr): Replace with...
1454 (_vect_peel_info::dr_info): ...this new field.
1455 (vect_peeling_hash_get_most_frequent)
1456 (vect_peeling_hash_choose_best_peeling): Update accordingly.
1457 (vect_peeling_hash_get_lowest_cost):
1458 (vect_enhance_data_refs_alignment): Likewise. Update calls to other
1459 routines for which the same change has been made.
1460 (vect_peeling_hash_insert): Likewise. Take a dr_vec_info instead of a
1462 * tree-vect-loop-manip.c (get_misalign_in_elems)
1463 (vect_gen_prolog_loop_niters): Use dr_vec_infos. Update calls after
1465 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
1466 * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
1467 (vect_truncate_gather_scatter_offset, compare_step_with_zero)
1468 (get_group_load_store_type, get_negative_load_store_type)
1469 (vect_get_data_ptr_increment, vectorizable_store)
1470 (vectorizable_load): Likewise.
1471 (ensure_base_align): Take a dr_vec_info instead of a data_reference.
1472 Update calls to other routines for which the same change has been made.
1474 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1476 * tree-vectorizer.h (vec_info::move_dr): New member function.
1477 (dataref_aux): Rename to...
1478 (dr_vec_info): ...this and add "dr" and "stmt" fields.
1479 (_stmt_vec_info::dr_aux): Update accordingly.
1480 (_stmt_vec_info::data_ref_info): Delete.
1481 (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
1482 (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
1483 (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
1484 (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
1486 (STMT_VINFO_DATA_REF): Likewise. Turn into an lvalue.
1487 (STMT_VINFO_DR_INFO): New macro.
1488 (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
1489 (set_dr_misalignment): Update after rename of dataref_aux.
1490 (vect_dr_stmt): Move earlier in file. Return dr_aux.stmt.
1491 * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
1492 initialization of STMT_VINFO_DATA_REF.
1493 * tree-vectorizer.c (vec_info::move_dr): New function.
1494 * tree-vect-patterns.c (vect_recog_bool_pattern)
1495 (vect_recog_mask_conversion_pattern)
1496 (vect_recog_gather_scatter_pattern): Use it.
1497 * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
1498 the "dr" and "stmt" fields of dr_vec_info instead of
1499 STMT_VINFO_DATA_REF.
1501 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1503 * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
1504 (is_pattern_stmt_p): Use it.
1505 * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
1506 on pattern statements.
1508 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1510 * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
1511 original stmt as a stmt_vec_info rather than a gimple stmt.
1512 (vect_pattern_recog_1): Take the statement directly as a
1513 stmt_vec_info, rather than via a gimple_stmt_iterator.
1514 Update call to vect_mark_pattern_stmts.
1515 (vect_pattern_recog): Update calls accordingly.
1517 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1519 * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
1520 (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
1521 a vect_def_type for the first argument.
1522 * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
1523 (vect_get_vec_def_for_stmt_copy): Likewise. Return the original
1524 operand if it isn't defined by a vectorized statement.
1525 (vect_build_gather_load_calls): Remove the mask_dt argument and
1526 update calls to vect_get_vec_def_for_stmt_copy.
1527 (vectorizable_bswap): Likewise the dt argument.
1528 (vectorizable_call): Update calls to vectorizable_bswap and
1529 vect_get_vec_def_for_stmt_copy.
1530 (vectorizable_simd_clone_call, vectorizable_assignment)
1531 (vectorizable_shift, vectorizable_operation, vectorizable_condition)
1532 (vectorizable_comparison): Update calls to
1533 vect_get_vec_def_for_stmt_copy.
1534 (vectorizable_store): Likewise. Remove now-unnecessary calls to
1536 (vect_get_loop_based_defs): Remove dt argument and update call
1537 to vect_get_vec_def_for_stmt_copy.
1538 (vectorizable_conversion): Update calls to vect_get_loop_based_defs
1539 and vect_get_vec_def_for_stmt_copy.
1540 (vectorizable_load): Update calls to vect_build_gather_load_calls
1541 and vect_get_vec_def_for_stmt_copy.
1542 * tree-vect-loop.c (vect_create_epilog_for_reduction)
1543 (vectorizable_reduction, vectorizable_live_operation): Update calls
1544 to vect_get_vec_def_for_stmt_copy.
1546 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1548 * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
1549 and gimple stmt arguments with a stmt_vec_info.
1550 (vect_record_base_alignments): Update calls accordingly.
1551 * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
1552 and gimple stmt arguments with a stmt_vec_info.
1553 (vect_build_slp_tree_1): Remove vinfo argument and update call
1554 to vect_record_max_nunits.
1555 (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
1556 and vect_record_max_nunits.
1558 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1560 * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
1561 file and take a stmt_vec_info instead of a gimple stmt.
1562 (supportable_widening_operation, vect_finish_replace_stmt)
1563 (vect_finish_stmt_generation, vect_get_store_rhs)
1564 (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
1565 (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
1566 (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
1567 (vect_get_smallest_scalar_type, vect_check_gather_scatter)
1568 (vect_create_data_ref_ptr, bump_vector_ptr)
1569 (vect_permute_store_chain, vect_setup_realignment)
1570 (vect_transform_grouped_load, vect_record_grouped_load_vectors)
1571 (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
1572 (vectorizable_reduction, vectorizable_induction)
1573 (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
1574 (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
1575 than gimple stmts as arguments.
1576 * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
1577 (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
1578 (can_group_stmts_p, vect_check_gather_scatter)
1579 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
1580 (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
1581 (vect_permute_load_chain, vect_shift_permute_load_chain)
1582 (vect_transform_grouped_load)
1583 (vect_record_grouped_load_vectors): Likewise.
1584 * tree-vect-loop.c (vect_fixup_reduc_chain)
1585 (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
1586 (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
1587 (vectorizable_reduction, vectorizable_induction)
1588 (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
1589 * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
1590 (vect_get_load_store_mask): Likewise.
1591 * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
1592 (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
1593 * tree-vect-stmts.c (vect_mark_relevant)
1594 (is_simple_and_all_uses_invariant)
1595 (exist_non_indexing_operands_for_use_p, process_use)
1596 (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
1597 (vect_get_vec_def_for_operand, vect_get_vec_defs)
1598 (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
1599 (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
1600 (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
1601 (get_negative_load_store_type, get_load_store_type)
1602 (vect_check_load_store_mask, vect_check_store_rhs)
1603 (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
1604 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
1605 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
1606 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
1607 (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
1608 (vectorizable_load, vectorizable_condition, vectorizable_comparison)
1609 (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
1610 (supportable_widening_operation): Likewise.
1612 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1614 * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
1615 a stmt_vec_info instead of a gcall.
1616 (vect_check_gather_scatter): Update call accordingly.
1617 * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
1619 (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
1620 (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
1621 * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
1622 instead of a gimple stmt.
1623 (vect_transform_loop): Update calls accordingly.
1624 * tree-vect-slp.c (vect_split_slp_store_group): Take and return
1625 stmt_vec_infos instead of gimple stmts.
1626 (vect_analyze_slp_instance): Update use accordingly.
1627 * tree-vect-stmts.c (read_vector_array, write_vector_array)
1628 (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
1629 (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
1630 (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
1631 (vect_gen_widened_results_half, vect_get_loop_based_defs)
1632 (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
1633 Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
1634 down to subroutines.
1636 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1638 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
1639 of the worklist from a vector of gimple stmts to a vector of
1641 * tree-vect-stmts.c (vect_mark_relevant, process_use)
1642 (vect_mark_stmts_to_be_vectorized): Likewise
1644 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1646 * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
1647 statement before passing it to vect_analyze_stmt.
1648 (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
1649 the chain of phi vector definitions. Track the exit phi via its
1651 (vectorizable_reduction): Set cond_stmt_vinfo directly from the
1652 STMT_VINFO_REDUC_DEF.
1653 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
1654 stmt_vec_infos to handle the statement chains.
1655 (vect_get_slp_defs): Record the first statement in the node
1656 using a stmt_vec_info.
1657 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
1658 statements here and pass their stmt_vec_info down to subroutines.
1659 (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
1660 down to vect_finish_stmt_generation.
1661 (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
1662 (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
1663 stmt_vec_infos to subroutines.
1664 (vect_remove_stores): Use stmt_vec_infos to handle the statement
1667 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1669 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
1670 (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
1671 (vect_permute_store_chain, vect_setup_realignment)
1672 (vect_permute_load_chain, vect_shift_permute_load_chain)
1673 (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
1674 stmts internally, and when passing values to other vectorizer routines.
1675 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
1676 * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
1677 (vect_analyze_loop_operations, get_initial_def_for_reduction)
1678 (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
1679 (vectorizable_reduction, vectorizable_induction)
1680 (vectorizable_live_operation, vect_transform_loop_stmt)
1681 (vect_transform_loop): Likewise.
1682 * tree-vect-patterns.c (vect_reassociating_reduction_p)
1683 (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
1684 (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
1685 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
1686 (vect_slp_analyze_node_operations_1): Likewise.
1687 * tree-vect-stmts.c (vect_mark_relevant, process_use)
1688 (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
1689 (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
1690 (vect_finish_stmt_generation_1, get_group_load_store_type)
1691 (get_load_store_type, vect_build_gather_load_calls)
1692 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
1693 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
1694 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
1695 (vectorizable_store, vectorizable_load, vectorizable_condition)
1696 (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
1697 (supportable_widening_operation): Likewise.
1698 (vect_get_vector_types_for_stmt): Likewise.
1699 * tree-vectorizer.h (vect_dr_behavior): Likewise.
1701 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1703 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
1704 (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
1705 (vect_permute_store_chain, vect_permute_load_chain)
1706 (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
1707 repeated stmt_vec_info lookups.
1708 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
1709 (vect_update_ivs_after_vectorizer): Likewise.
1710 * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
1711 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
1712 * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
1713 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
1714 (vect_bb_slp_scalar_cost): Likewise.
1715 * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
1717 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1719 * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
1720 gcall rather than the generic gimple stmt to gimple_call_internal_fn.
1721 (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
1722 to get gassigns and gcalls, rather than operating on generc gimple
1724 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
1725 (vect_mark_stmts_to_be_vectorized, vectorizable_store)
1726 (vectorizable_load, vect_analyze_stmt): Likewise.
1727 * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
1729 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1731 * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
1732 return stmt_vec_infos rather than gimple stmts. Do not accept
1734 (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
1736 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
1737 Update use of get_later_stmt.
1738 (vect_get_constant_vectors): Update call accordingly.
1739 (vect_schedule_slp_instance): Likewise
1740 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
1741 (vect_slp_analyze_instance_dependence): Likewise.
1742 (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
1744 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1746 * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
1747 (stmt_info_for_cost::stmt_info): ...this new field.
1748 (add_stmt_costs): Update accordingly.
1749 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
1750 (vect_get_known_peeling_cost): Likewise.
1751 (vect_estimate_min_profitable_iters): Likewise.
1752 * tree-vect-stmts.c (record_stmt_cost): Likewise.
1754 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1756 * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
1757 from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
1758 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
1760 * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
1762 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1764 * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
1765 a gimple stmt to a stmt_vec_info.
1766 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
1768 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1770 * tree-vectorizer.h (vec_info::grouped_stores): Change from
1771 an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
1772 (_loop_vec_info::reduction_chains): Likewise.
1773 * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
1775 * tree-vect-slp.c (vect_analyze_slp): Likewise.
1777 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1779 * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
1780 a gimple stmt to a stmt_vec_info.
1781 (_stmt_vec_info::next_element): Likewise.
1782 * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
1783 (vect_slp_analyze_and_verify_node_alignment)
1784 (vect_analyze_group_access_1, vect_analyze_group_access)
1785 (vect_small_gap_p, vect_prune_runtime_alias_test_list)
1786 (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
1787 (vect_supportable_dr_alignment): Update accordingly.
1788 * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
1789 (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
1790 (vect_is_simple_reduction, vectorizable_reduction): Likewise.
1791 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
1792 * tree-vect-slp.c (vect_build_slp_tree_1)
1793 (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
1794 (vect_split_slp_store_group, vect_analyze_slp_instance)
1795 (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
1796 * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
1797 (get_group_load_store_type, get_load_store_type)
1798 (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
1799 (vect_transform_stmt, vect_remove_stores): Likewise.
1801 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1803 * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
1805 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
1806 (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
1807 (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
1808 (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
1809 (vector_alignment_reachable_p, vect_get_data_access_cost)
1810 (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
1811 (vect_peeling_supportable, vect_enhance_data_refs_alignment)
1812 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
1813 (vect_analyze_group_access_1, vect_analyze_group_access)
1814 (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
1815 (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
1816 (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
1817 result of vect_dr_stmt and use the stmt_vec_info instead of
1818 the associated gimple stmt.
1819 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
1820 (vect_gen_prolog_loop_niters): Likewise.
1821 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
1823 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1825 * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
1826 to a vec<stmt_vec_info>.
1827 * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
1828 (vect_create_new_slp_node): Take a vec<gimple *> instead of a
1830 (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
1831 to a vec<stmt_vec_info>.
1832 (bst_traits::value_type, bst_traits::value_type): Likewise.
1833 (bst_traits::hash): Update accordingly.
1834 (vect_get_and_check_slp_defs): Change the stmts parameter from
1835 a vec<gimple *> to a vec<stmt_vec_info>.
1836 (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
1837 (vect_build_slp_tree): Likewise.
1838 (vect_build_slp_tree_2): Likewise. Update uses of
1839 SLP_TREE_SCALAR_STMTS.
1840 (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
1841 (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
1842 (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
1843 (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
1844 (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
1845 (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
1846 (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
1847 (vect_get_constant_vectors, vect_get_slp_defs)
1848 (vect_transform_slp_perm_load, vect_schedule_slp_instance)
1849 (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
1850 (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
1851 instead of gimple stmts.
1852 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
1853 the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
1854 (vect_slp_analyze_instance_dependence): Update uses of
1855 SLP_TREE_SCALAR_STMTS.
1856 (vect_slp_analyze_and_verify_node_alignment): Likewise.
1857 (vect_slp_analyze_and_verify_instance_alignment): Likewise.
1858 * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
1859 (get_initial_defs_for_reduction): Likewise.
1860 (vect_create_epilog_for_reduction): Likewise.
1861 (vectorize_fold_left_reduction): Likewise.
1862 * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
1863 (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
1864 (can_vectorize_live_stmts): Likewise.
1866 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1868 * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
1869 auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
1870 (vect_force_simple_reduction): Take and return stmt_vec_infos rather
1872 * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
1874 (gather_scalar_reductions): Update after above interface changes.
1875 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
1876 (vect_is_simple_reduction): Take and return stmt_vec_infos rather
1878 (vect_force_simple_reduction): Likewise.
1879 * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
1880 LOOP_VINFO_REDUCTIONS.
1881 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
1883 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1885 * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
1886 a gimple stmt to a stmt_vec_info.
1887 * tree-vect-loop.c (vect_active_double_reduction_p)
1888 (vect_force_simple_reduction, vectorizable_reduction): Update
1891 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1893 * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
1894 vec<gimple *> to a vec<stmt_vec_info>.
1895 * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
1896 the reduction_phis argument from a vec<gimple *> to a
1898 (vectorizable_reduction): Likewise the phis local variable that
1899 is passed to vect_create_epilog_for_reduction. Update for new type
1900 of SLP_TREE_VEC_STMTS.
1901 (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
1902 (vectorizable_live_operation): Likewise.
1903 * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
1904 (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
1906 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1908 * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
1909 a gimple stmt to a stmt_vec_info.
1910 (vectorizable_condition, vectorizable_live_operation)
1911 (vectorizable_reduction, vectorizable_induction): Pass back the
1912 vectorized statement as a stmt_vec_info.
1913 * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
1914 use of STMT_VINFO_VEC_STMT.
1915 * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
1916 accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
1917 as stmt_vec_infos rather than gimple stmts.
1918 (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
1920 (vectorizable_live_operation): Likewise.
1921 (vectorizable_reduction, vectorizable_induction): Likewise,
1922 updating use of STMT_VINFO_VEC_STMT.
1923 * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
1924 of STMT_VINFO_VEC_STMT.
1925 (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
1926 (vectorizable_simd_clone_call, vectorizable_conversion)
1927 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
1928 (vectorizable_store, vectorizable_load, vectorizable_condition)
1929 (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
1930 from a gimple stmt to a stmt_vec_info.
1931 (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT. Pass a
1932 pointer to a stmt_vec_info to the vectorizable_* routines.
1934 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1936 * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
1937 a gimple stmt to a stmt_vec_info.
1938 (is_pattern_stmt_p): Update accordingly.
1939 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
1940 (vect_record_grouped_load_vectors): Likewise.
1941 * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
1942 (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
1943 (vect_model_reduction_cost): Likewise.
1944 (vect_create_epilog_for_reduction): Likewise.
1945 (vectorizable_reduction, vectorizable_induction): Likewise.
1946 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
1947 Return the stmt_vec_info for the pattern statement.
1948 (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
1949 (vect_split_statement, vect_mark_pattern_stmts): Likewise.
1950 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
1951 (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
1952 * tree-vect-stmts.c (vect_mark_relevant): Likewise.
1953 (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
1954 (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
1955 (free_stmt_vec_info, vect_is_simple_use): Likewise.
1957 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1959 * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
1960 (vect_finish_stmt_generation): Likewise.
1961 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
1962 (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
1963 (vect_build_gather_load_calls): Use the return value of the above
1964 functions instead of a separate call to vinfo_for_stmt. Use narrow
1965 scopes for the input gimple stmt and wider scopes for the associated
1966 stmt_vec_info. Use vec_info::lookup_def when setting these
1967 stmt_vec_infos from an SSA_NAME definition.
1968 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
1969 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
1970 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
1971 (vectorizable_store, vectorizable_load, vectorizable_condition)
1972 (vectorizable_comparison): Likewise.
1973 * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
1974 (vectorizable_reduction): Likewise.
1976 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1978 * tree-vectorizer.h (vect_is_simple_use): Add an optional
1979 stmt_vec_info * parameter before the optional gimple **.
1980 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
1981 (process_use, vect_get_vec_def_for_operand_1): Update callers.
1982 (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
1983 * tree-vect-loop.c (vectorizable_reduction): Likewise.
1984 (vectorizable_live_operation): Likewise.
1985 * tree-vect-patterns.c (type_conversion_p): Likewise.
1986 (vect_look_through_possible_promotion): Likewise.
1987 (vect_recog_rotate_pattern): Likewise.
1988 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
1990 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1992 * tree-vectorizer.h (stmt_vec_info): Temporarily change from
1993 a typedef to a wrapper class.
1994 (NULL_STMT_VEC_INFO): New macro.
1995 (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
1996 (stmt_vec_info::operator*): New function.
1997 (stmt_vec_info::operator gimple *): Likewise.
1998 (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
1999 (add_stmt_costs): Likewise.
2000 * tree-vect-loop-manip.c (iv_phi_p): Likewise.
2001 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
2002 (vect_get_known_peeling_cost): Likewise.
2003 (vect_estimate_min_profitable_iters): Likewise.
2004 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
2005 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
2006 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
2007 (vectorizable_store, free_stmt_vec_infos): Likewise.
2008 (new_stmt_vec_info): Change return type of xcalloc to
2011 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2013 * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
2014 * tree-vectorizer.c (vec_info::lookup_single_use): New function.
2015 * tree-vect-loop.c (vectorizable_reduction): Use it instead of
2016 a single_imm_use-based sequence.
2017 * tree-vect-stmts.c (supportable_widening_operation): Likewise.
2019 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2021 * tree-vectorizer.h (vec_info::lookup_def): Declare.
2022 * tree-vectorizer.c (vec_info::lookup_def): New function.
2023 * tree-vect-patterns.c (vect_get_internal_def): Use it.
2024 (vect_widened_op_tree): Likewise.
2025 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
2026 * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
2027 (vectorizable_reduction): Likewise.
2028 (vect_valid_reduction_input_p): Take a stmt_vec_info instead
2030 (vect_is_slp_reduction): Update calls accordingly. Use
2031 vec_info::lookup_def.
2032 (vect_is_simple_reduction): Likewise
2033 * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
2035 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2037 * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
2038 * tree-vectorizer.c (vec_info::lookup_stmt): New function.
2039 * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
2041 (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
2042 (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
2043 (vect_update_vf_for_slp, vect_analyze_loop_operations)
2044 (vect_is_slp_reduction, vectorizable_induction)
2045 (vect_transform_loop_stmt, vect_transform_loop): Likewise.
2046 * tree-vect-patterns.c (vect_init_pattern_stmt):
2047 (vect_determine_min_output_precision_1, vect_determine_precisions)
2048 (vect_pattern_recog): Likewise.
2049 * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
2050 * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
2051 * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
2052 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
2053 (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
2054 (vect_detect_hybrid_slp): Likewise. Change the walk_stmt_info
2055 info field from a loop to a loop_vec_info.
2057 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2059 * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
2060 (vec_info::add_stmt): Declare.
2061 * tree-vectorizer.c (vec_info::add_stmt): New function.
2062 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
2063 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
2064 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
2065 (vectorizable_induction): Likewise.
2066 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
2067 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
2068 (vectorizable_simd_clone_call, vectorizable_store): Likewise.
2069 (vectorizable_load): Likewise.
2070 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
2071 (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
2072 (vect_recog_gather_scatter_pattern): Likewise.
2073 (append_pattern_def_seq): Likewise. Remove a check that is
2074 performed by add_stmt itself.
2076 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2078 * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
2079 which make_ssa_name was called with new_stmt before new_stmt
2082 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2084 * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
2086 (vect_is_slp_reduction): ...here...
2087 (vect_is_simple_reduction): ...and here. Remove repetition of tests
2088 that are already known to be false.
2090 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2092 * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
2093 * tree-vect-slp.c (vect_free_slp_tree): Likewise. Don't update
2094 STMT_VINFO_NUM_SLP_USES when it's true.
2095 (vect_free_slp_instance): Add a final_p parameter and pass it to
2097 (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
2098 (vect_analyze_slp_instance): Likewise.
2099 (vect_slp_analyze_operations): Likewise.
2100 (vect_slp_analyze_bb_1): Likewise.
2101 * tree-vectorizer.c (vec_info): Likewise.
2102 * tree-vect-loop.c (vect_transform_loop): Likewise.
2104 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2106 * tree-vect-loop.c (vectorizable_reduction): Assert that the
2107 function is not called for second and subsequent members of
2110 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2112 * tree-vect-loop.c (get_initial_def_for_reduction): Move special
2113 cases for nested loops from here to ...
2114 (vect_create_epilog_for_reduction): ...here. Only call
2115 vect_is_simple_use for inner-loop reductions.
2117 2018-07-31 Martin Liska <mliska@suse.cz>
2119 PR gcov-profile/85338
2120 PR gcov-profile/85350
2121 PR gcov-profile/85372
2122 * profile.c (struct location_triplet): New.
2123 (struct location_triplet_hash): Likewise.
2124 (output_location): Do not output a BB that
2125 is already recorded for a line.
2126 (branch_prob): Use streamed_locations.
2128 2018-07-31 Martin Liska <mliska@suse.cz>
2130 PR gcov-profile/85370
2131 * coverage.c (coverage_begin_function): Do not mark target
2132 clones as artificial functions.
2134 2018-07-31 Martin Liska <mliska@suse.cz>
2136 PR gcov-profile/83813
2137 PR gcov-profile/84758
2138 PR gcov-profile/85217
2139 PR gcov-profile/85332
2140 * profile.c (branch_prob): Do not record GOTO expressions
2141 for GIMPLE statements which locations are already streamed.
2143 2018-07-31 Olivier Hainque <hainque@adacore.com>
2145 * gcc.c (handle_spec_function): Accept a soft_matched_part
2146 argument, as do_spec_1. Pass it down to ...
2147 (eval_spec_function): Accept a soft_matched_part argument,
2148 and pass it down to ...
2149 (do_spec_2): Accept a soft_matched_part argument, and pass
2150 it down to do_spec_1.
2151 (do_spec_1): Pass soft_matched_part to handle_spec_function.
2152 (handle_braces): Update call to handle_spec_function.
2153 (driver::set_up_specs): Update calls to do_spec_2.
2154 (compare_debug_dump_opt_spec_function): Likewise.
2155 (compare_debug_self_opt_spec_function): Likewise.
2157 2018-07-31 Olivier Hainque <hainque@adacore.com>
2159 * common.opt (nolibc): New option.
2160 * doc/invoke.texi (Link Options): Document it.
2161 * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
2162 * config/alpha/linux.h: Likewise.
2163 * config/arc/elf.h: Likewise.
2164 * config/arm/uclinux-elf.h: Likewise.
2165 * config/arm/unknown-elf.h: Likewise.
2166 * config/avr/avrlibc.h: Likewise.
2167 * config/bfin/bfin.h: Likewise.
2168 * config/bfin/linux.h: Likewise.
2169 * config/bfin/uclinux.h: Likewise.
2170 * config/darwin.h: Likewise.
2171 * config/darwin10.h: Likewise.
2172 * config/darwin12.h: Likewise.
2173 * config/gnu-user.h: Likewise.
2174 * config/lm32/uclinux-elf.h: Likewise.
2175 * config/pa/pa-hpux11.h: Likewise.
2176 * config/pa/pa64-hpux.h: Likewise.
2177 * config/sparc/sparc.h: Likewise.
2179 2018-07-31 Olivier Hainque <hainque@adacore.com>
2181 * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
2182 undefined variables.
2184 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
2187 * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
2190 2018-07-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
2192 * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
2193 terminated string literal.
2195 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
2197 PR rtl-optimization/85160
2198 * combine.c (is_just_move): New function.
2199 (try_combine): Allow combining two instructions into two if neither of
2200 the original instructions was a move.
2202 2018-07-30 Alexander Monakov <amonakov@ispras.ru>
2205 * doc/extend.texi (Global Register Variables): Discourage use of type
2207 (Local Register Variables): Likewise.
2209 2018-07-30 Richard Sandiford <richard.sandiford@arm.com>
2211 PR tree-optimization/86506
2212 * hwint.h (ceil_log2): Resync with hwint.c implementation.
2214 2018-07-30 Ilya Leoshkevich <iii@linux.ibm.com>
2217 * lra-constraints.c (spill_hard_reg_in_range): When selecting the
2218 hard_regno, make sure no insn between `from` and `to` clobbers it.
2220 2018-07-30 Cesar Philippidis <cesar@codesourcery.com>
2221 Tom de Vries <tdevries@suse.de>
2223 * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
2224 (PTX_DEFAULT_RUNTIME_DIM): ... this.
2225 (nvptx_goacc_validate_dims): Set default worker and gang dims to
2226 PTX_DEFAULT_RUNTIME_DIM.
2227 (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
2229 2018-07-29 John David Anglin <danglin@gcc.gnu.org>
2231 * config/pa/pa.c (pa_output_addr_vec): Align address table.
2232 * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
2233 * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
2235 2018-07-27 Michael Meissner <meissner@linux.ibm.com>
2237 * config/rs6000/constraints.md (wG constraint): Delete, no longer
2239 * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
2240 predicate to reflect toc fusion has been deleted.
2241 (toc_fusion_mem_raw): Delete, no longer used.
2242 (toc_fusion_mem_wrapped): Likewise.
2243 * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
2245 * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
2246 Delete, no longer used.
2247 * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
2248 meant to be used for toc fusion.
2249 (rs6000_debug_print_mode): Delete toc fusion debugging.
2250 (rs6000_debug_reg_global): Likewise.
2251 (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
2252 fusion and secondary reload support that were never used.
2253 (rs6000_option_override_internal): Delete TOC fusion, that was only
2254 partially defined, and it did not work unless you also used the
2256 (rs6000_legitimate_address_p): Delete TOC fusion support.
2257 (rs6000_opt_masks): Likewise.
2258 (fusion_wrap_memory_address): Delete function, no longer used.
2259 (fusion_split_address); Delete TOC fusion support.
2260 * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
2261 longer used with toc fusion being deleted.
2262 (TARGET_TOC_FUSION_FP): Likewise.
2263 * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
2265 (toc fusion spliter): Delete TOC fusion support.
2266 (toc_fusionload_<mode>): Likewise.
2267 (toc_fusionload_di): Likewise.
2268 (fusion_gpr_load_<mode>): Delete generator function, this insn no
2269 longer needs to be named. Rename predicate to delete TOC fusion.
2270 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
2271 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
2272 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
2273 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
2274 (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
2276 2018-07-27 Kelvin Nilsen <kelvin@gcc.gnu.org>
2278 * doc/extend.texi (Basic PowerPC Built-in Functions Available on
2279 ISA 2.05): Replace __uint128_t with __uint128 and __int128_t with
2280 __int128 in built-in function prototypes.
2281 (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
2282 (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
2284 2018-07-27 Martin Sebor <msebor@redhat.com>
2286 PR tree-optimization/86696
2287 * tree-ssa-strlen.c (get_min_string_length): Handle all integer
2288 types, including enums.
2289 (handle_char_store): Be prepared for the above function to fail.
2291 2018-07-26 Qing Zhao <qing.zhao@oracle.com>
2293 * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
2294 when optimization level is lower than 2 or optimize for size.
2296 2018-07-26 Martin Sebor <msebor@redhat.com>
2298 PR tree-optimization/86043
2299 PR tree-optimization/86042
2300 * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
2301 (get_string_cst_length): Rename...
2302 (get_min_string_length): ...to this. Add argument.
2303 (handle_char_store): Extend to handle multi-character stores by
2305 * tree.c (initializer_zerop): Use new argument. Handle MEM_REF.
2306 * tree.h (initializer_zerop): Add argument.
2308 2018-07-26 Jakub Jelinek <jakub@redhat.com>
2311 * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
2312 declare target to variables if they have always,{to,from,tofrom} map
2315 2018-07-26 Martin Liska <mliska@suse.cz>
2318 * lto-wrapper.c: Add linker_output as prefix
2319 for ltrans_output_file.
2321 2018-07-26 Segher Boessenkool <segher@kernel.crashing.org>
2323 PR rtl-optimization/85805
2324 * combine.c (reg_nonzero_bits_for_combine): Only use the last set
2325 value for hard registers if that was written in the same mode.
2327 2018-07-26 Martin Liska <mliska@suse.cz>
2329 PR gcov-profile/86536
2330 * gcov.c (format_gcov): Use printf format %.*f directly
2331 and do not handle special values.
2333 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
2335 * common/config/arc/arc-common.c (arc_option_optimization_table):
2336 Update default optimizations for size.
2338 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
2340 * config/arc/arc.md (movsf_insn): Add short instruction selection.
2341 * config/arc/constraints.md (CfZ): New constraint.
2342 * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
2343 (subsf3_fpu): Likewise.
2344 (cmpsf_fpu): Likewise.
2345 (cmpsf_fpu_uneq): Likewise.
2347 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
2349 * config/arc/arc.c (compact_memory_operand_p): Check for uncached
2351 (arc_is_uncached_mem_p): uncached applies to both the variable and
2354 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
2356 * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
2359 2018-07-25 David Malcolm <dmalcolm@redhat.com>
2361 * optinfo-emit-json.cc (class optrecord_json_writer): Convert
2362 field "m_scopes" from vec to auto_vec.
2364 2018-07-25 Martin Liska <mliska@suse.cz>
2366 * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
2369 2018-07-25 Richard Biener <rguenther@suse.de>
2372 * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
2373 special wrt context_die late.
2374 (gen_subprogram_die): Re-use DIEs in local scope.
2376 2018-07-25 Richard Sandiford <richard.sandiford@arm.com>
2378 PR tree-optimization/86644
2379 * hwint.c (ceil_log2): Fix comment. Return 0 for 0.
2381 2018-07-25 Martin Liska <mliska@suse.cz>
2384 * dumpfile.c: And excluded values with TDF_ALL_VALUES.
2385 * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
2387 2018-07-25 Martin Liska <mliska@suse.cz>
2390 * params.def: Explain ASan abbreviation and provide
2391 a documentation link.
2393 2018-07-24 Martin Sebor <msebor@redhat.com>
2395 PR tree-optimization/86622
2396 PR tree-optimization/86532
2397 * builtins.h (string_length): Declare.
2398 * builtins.c (c_strlen): Correct handling of non-constant offsets.
2399 (check_access): Be prepared for non-constant length ranges.
2400 (string_length): Make extern.
2401 * expr.c (string_constant): Only handle the minor non-constant
2402 array index. Use string_constant to compute the length of
2403 a generic string constant.
2405 2018-07-24 Richard Sandiford <richard.sandiford@arm.com>
2407 PR tree-optimization/86618
2408 * tree-vect-stmts.c (vectorizable_call): Don't take the address
2409 of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
2411 2018-07-24 David Malcolm <dmalcolm@redhat.com>
2413 PR tree-optimization/86636
2414 * json.cc (json::object::set): Fix comment. Add assertions.
2415 (json::array::append): Move here from json.h. Add comment and an
2417 (json::string::string): Likewise.
2418 * json.h (json::array::append): Move to json.cc.
2419 (json::string::string): Likewise.
2420 * optinfo-emit-json.cc
2421 (optrecord_json_writer::impl_location_to_json): Assert that we
2422 aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
2423 wrapper around it. Expand the location once, rather than three
2425 (optrecord_json_writer::inlining_chain_to_json): Fix the check for
2426 UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
2428 (optrecord_json_writer::optinfo_to_json): Likewise, in four
2429 places. Fix some overlong lines.
2431 2018-07-24 Matthew Malcomson <matthew.malcomson@arm.com>
2433 * config/aarch64/aarch64-simd.md
2434 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
2435 (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
2436 (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
2437 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
2438 (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
2439 (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
2440 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
2441 (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
2442 (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
2444 2018-07-24 Jakub Jelinek <jakub@redhat.com>
2447 * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
2448 and size > HOST_BITS_PER_WIDE_INT. For size > HOST_BITS_PER_WIDE_INT
2449 and abs_d == d, do the power of two handling if profitable.
2451 2018-07-24 Richard Biener <rguenther@suse.de>
2453 * match.pd: Add BIT_FIELD_REF canonicalizations.
2455 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
2458 * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
2460 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
2462 * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
2463 terminated STRING_CST object.
2465 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
2467 hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
2469 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
2471 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
2472 * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
2473 * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
2474 the elements into a register.
2475 (rs6000_split_v4si_init_di_reg): Delete.
2476 (rs6000_split_v4si_init): Delete.
2477 * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
2478 (vsx_init_v4si): Rewrite as a define_expand.
2480 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
2482 * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
2483 zero_extend argument from memory): New.
2485 2018-07-22 Martin Sebor <msebor@redhat.com>
2488 * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
2489 diagnosing calls with unknown arguments unless -Walloca-larger-than
2490 is restricted to less than PTRDIFF_MAX bytes.
2492 2018-07-22 Gerald Pfeifer <gerald@pfeifer.com>
2494 * doc/gcov.texi (Invoking Gcov): Editorial changes.
2496 2018-07-20 David Malcolm <dmalcolm@redhat.com>
2498 * pretty-print.c (text_info::set_location): Remove redundant
2499 "line_table" parameter from call to rich_location::set_range.
2501 2018-07-20 Martin Sebor <msebor@redhat.com>
2504 * builtins.c (expand_builtin_alloca): Adjust.
2505 * calls.c (alloc_max_size): Simplify.
2506 * cgraphunit.c (cgraph_node::expand): Adjust.
2507 * common.opt (larger_than_size, warn_frame_larger_than): Remove
2509 (frame_larger_than_size): Same.
2510 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
2511 to take a HOST_WIDE_INT argument and accept a byte-size suffix.
2513 * doc/invoke.texi (GCC Command Options): Document option arguments.
2514 Explain byte-size arguments and suffixes.
2515 (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
2516 (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
2517 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
2518 * doc/options.texi (UInteger): Expand.
2519 (Host_Wide_Int, ByteSize): Document new properties.
2520 * final.c (final_start_function_1): Include sizes in an error message.
2521 * function.c (frame_offset_overflow): Same.
2522 * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
2523 (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
2524 Diagnose unbounded alloca calls only for limits of less than
2526 (alloca_call_type): Adjust. Diagnose possibly out-of-bounds alloca
2527 calls and VLA size only for limits of less than PTRDIFF_MAX. Same
2529 (pass_walloca::execute): Adjust. Diagnose alloca calls in loops
2530 only for limits of less than PTRDIFF_MAX.
2531 * langhooks-def.h (lhd_handle_option): Change function argument
2533 * langhooks.c (lhd_handle_option): Same.
2534 * langhooks.h (handle_option): Same.
2535 * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
2537 (var_type, var_type_struct): Same.
2538 (var_set): Handle ByteSize flag.
2539 * optc-gen.awk: Add comments to output to ease debugging. Make
2540 use of HOST_WIDE_INT where appropriate.
2541 * opts-gen-save.awk: Use %lx to format unsigned long.
2542 * opth-gen.awk: Change function argument to HOST_WIDE_INT.
2543 * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
2544 arguments. Parse bytes-size suffixes.
2545 (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
2546 (enum_value_to_arg): Same.
2547 (decode_cmdline_option): Handle cl_host_wide_int. Adjust.
2548 (handle_option): Adjust.
2549 (generate_option): Change function argument to HOST_WIDE_INT.
2550 (cmdline_handle_error): Adjust.
2551 (read_cmdline_option): Change function argument to HOST_WIDE_INT.
2552 (set_option): Change function argument to HOST_WIDE_INT.
2553 (option_enabled): Handle cl_host_wide_int.
2554 (get_option_state): Handle CLVC_SIZE.
2555 (control_warning_option): Same.
2556 * opts.c (common_handle_option): Change function argument to
2557 HOST_WIDE_INT. Remove handling of OPT_Walloca_larger_than_ and
2558 OPT_Wvla_larger_than_.
2559 * opts.h (enum cl_var_type): Add an enumerator.
2560 * stor-layout.c (layout_decl): Print a more meaningful warning.
2561 * toplev.c (output_stack_usage): Adjust.
2563 2018-07-20 Qing Zhao <qing.zhao@oracle.com>
2565 * builtins.c (expand_builtin_memcmp): Delete the last parameter for
2566 call to inline_expand_builtin_string_cmp.
2567 (expand_builtin_strcmp): Likewise.
2568 (expand_builtin_strncmp): Likewise.
2569 (inline_string_cmp): Delete the last parameter, change char_type_node
2570 to unsigned_char_type_node for strcmp/strncmp, add conversions to the
2572 (inline_expand_builtin_string_cmp): Delete the last parameter, give up
2573 the inlining expansion on target where the type of the call has same or
2574 narrower precision than unsigned char.
2576 2018-07-20 David Malcolm <dmalcolm@redhat.com>
2578 * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
2579 (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
2580 * common.opt (fsave-optimization-record): New option.
2581 * coretypes.h (struct kv_pair): Move here from dumpfile.c.
2582 * doc/invoke.texi (-fsave-optimization-record): New option.
2583 * dumpfile.c: Include "optinfo-emit-json.h".
2584 (struct kv_pair): Move to coretypes.h.
2585 (optgroup_options): Make non-static.
2586 (dump_context::end_scope): Call
2587 optimization_records_maybe_pop_dump_scope.
2588 * dumpfile.h (optgroup_options): New decl.
2589 * json.cc: New file.
2591 * optinfo-emit-json.cc: New file.
2592 * optinfo-emit-json.h: New file.
2593 * optinfo.cc: Include "optinfo-emit-json.h".
2594 (optinfo::emit): Call optimization_records_maybe_record_optinfo.
2595 (optinfo_enabled_p): Check optimization_records_enabled_p.
2596 (optinfo_wants_inlining_info_p): Likewise.
2597 * optinfo.h: Update comment.
2598 * profile-count.c (profile_quality_as_string): New function.
2599 * profile-count.h (profile_quality_as_string): New decl.
2600 (profile_count::quality): New accessor.
2601 * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
2602 and optinfo_emit_json_cc_tests.
2603 * selftest.h (selftest::json_cc_tests): New decl.
2604 (selftest::optinfo_emit_json_cc_tests): New decl.
2605 * toplev.c: Include "optinfo-emit-json.h".
2606 (compile_file): Call optimization_records_finish.
2607 (do_compile): Call optimization_records_start.
2608 * tree-ssa-live.c: Include optinfo.h.
2609 (remove_unused_scope_block_p): Retain inlining information if
2610 optinfo_wants_inlining_info_p returns true.
2612 2018-07-20 Richard Biener <rguenther@suse.de>
2615 * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
2616 to cover -flto-partition=none.
2618 2018-07-20 Martin Liska <mliska@suse.cz>
2620 * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
2621 (get_decl_source_range): Remove unused function.
2623 2018-07-20 Richard Biener <rguenther@suse.de>
2625 * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
2626 (struct vn_phi_s): Likewise.
2627 (struct vn_reference_s): Likewise.
2628 * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
2629 for searching the slot of an entry known to be in the hash itself.
2630 (vn_phi_hasher::equal): Likewise.
2631 (vn_reference_hasher::equal): Likewise.
2632 (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
2634 (optimistic_info, current_info): Remove, keeping only valid_info.
2635 (vn_reference_lookup_1): Remove fallback lookup.
2636 (vn_reference_lookup_2): Likewise.
2637 (vn_nary_op_lookup_1): Likewise.
2638 (vn_phi_lookup): Likewise.
2639 (vn_nary_build_or_lookup_1): Make sure to not chain the built
2641 (vn_reference_insert): Adjust, chain the inserted hash element
2642 at last_inserted_ref.
2643 (vn_reference_insert_pieces): Likewise.
2644 (visit_reference_op_call): Likewise.
2645 (vn_nary_op_insert_into): Chain the inserted hash element at
2647 (vn_nary_op_insert_pieces): Adjust.
2648 (vn_nary_op_insert): Likewise.
2649 (vn_nary_op_insert_stmt): Likewise.
2650 (vn_phi_insert): Adjust, chain the inserted hash element at
2652 (process_scc): Remove clearing and copying the optimistic
2653 table. Instead remove elements inserted during an optimistic
2654 iteration from the single table we maintain.
2655 (init_scc_vn): Adjust.
2656 (free_scc_vn): Likewise.
2657 (sccvn_dom_walker::record_cond): Likewise.
2658 (sccvn_dom_walker::after_dom_children): Likewise.
2660 2018-07-19 Martin Sebor <msebor@redhat.com>
2662 PR tree-optimization/84047
2663 PR tree-optimization/83776
2664 * tree-vrp.c (vrp_prop::check_mem_ref): New function.
2665 (check_array_bounds): Call it.
2667 2018-07-19 Martin Sebor <msebor@redhat.com>
2669 * align.h (align_flags): Use member initialization.
2671 2018-07-19 David Malcolm <dmalcolm@redhat.com>
2673 * Makefile.in (OBJS): Add optinfo.o.
2674 * coretypes.h (class symtab_node): New forward decl.
2675 (struct cgraph_node): New forward decl.
2676 (class varpool_node): New forward decl.
2677 * dump-context.h: New file.
2678 * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
2680 (refresh_dumps_are_enabled): Use optinfo_enabled_p.
2681 (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
2682 (set_alt_dump_file): Likewise.
2683 (dump_context::~dump_context): New dtor.
2684 (dump_gimple_stmt): Move implementation to...
2685 (dump_context::dump_gimple_stmt): ...this new member function.
2686 Add the stmt to any pending optinfo, creating one if need be.
2687 (dump_gimple_stmt_loc): Move implementation to...
2688 (dump_context::dump_gimple_stmt_loc): ...this new member function.
2689 Start a new optinfo and add the stmt to it.
2690 (dump_gimple_expr): Move implementation to...
2691 (dump_context::dump_gimple_expr): ...this new member function.
2692 Add the stmt to any pending optinfo, creating one if need be.
2693 (dump_gimple_expr_loc): Move implementation to...
2694 (dump_context::dump_gimple_expr_loc): ...this new member function.
2695 Start a new optinfo and add the stmt to it.
2696 (dump_generic_expr): Move implementation to...
2697 (dump_context::dump_generic_expr): ...this new member function.
2698 Add the tree to any pending optinfo, creating one if need be.
2699 (dump_generic_expr_loc): Move implementation to...
2700 (dump_context::dump_generic_expr_loc): ...this new member
2701 function. Add the tree to any pending optinfo, creating one if
2703 (dump_printf): Move implementation to...
2704 (dump_context::dump_printf_va): ...this new member function. Add
2705 the text to any pending optinfo, creating one if need be.
2706 (dump_printf_loc): Move implementation to...
2707 (dump_context::dump_printf_loc_va): ...this new member function.
2708 Start a new optinfo and add the stmt to it.
2709 (dump_dec): Move implementation to...
2710 (dump_context::dump_dec): ...this new member function. Add the
2711 value to any pending optinfo, creating one if need be.
2712 (dump_context::dump_symtab_node): New member function.
2713 (dump_context::get_scope_depth): New member function.
2714 (dump_context::begin_scope): New member function.
2715 (dump_context::end_scope): New member function.
2716 (dump_context::ensure_pending_optinfo): New member function.
2717 (dump_context::begin_next_optinfo): New member function.
2718 (dump_context::end_any_optinfo): New member function.
2719 (dump_context::s_current): New global.
2720 (dump_context::s_default): New global.
2721 (dump_scope_depth): Delete global.
2722 (dumpfile_ensure_any_optinfo_are_flushed): New function.
2723 (dump_symtab_node): New function.
2724 (get_dump_scope_depth): Reimplement in terms of dump_context.
2725 (dump_begin_scope): Likewise.
2726 (dump_end_scope): Likewise.
2727 (selftest::temp_dump_context::temp_dump_context): New ctor.
2728 (selftest::temp_dump_context::~temp_dump_context): New dtor.
2729 (selftest::verify_item): New function.
2730 (ASSERT_IS_TEXT): New macro.
2731 (ASSERT_IS_TREE): New macro.
2732 (ASSERT_IS_GIMPLE): New macro.
2733 (selftest::test_capture_of_dump_calls): New test.
2734 (selftest::dumpfile_c_tests): Call it.
2735 * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
2736 (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
2737 (dump_gimple_stmt, dump_dec): Gather these related decls and add a
2738 descriptive comment.
2739 (dump_function, print_combine_total_stats, enable_rtl_dump_file)
2740 (dump_node, dump_bb): Move these unrelated decls.
2741 (class dump_manager): Add leading comment.
2742 * optinfo.cc: New file.
2743 * optinfo.h: New file.
2745 2018-07-19 Michael Collison <michael.collison@arm.com>
2746 Richard Henderson <rth@redhat.com>
2748 * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
2749 (subti): Handle op1 zero.
2750 (subvti4, usub4ti4): New.
2751 (*sub<GPI>3_compare1_imm): New.
2752 (sub<GPI>3_carryinCV): New.
2753 (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
2754 (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
2756 2018-07-19 Michael Collison <michael.collison@arm.com>
2757 Richard Henderson <rth@redhat.com>
2759 * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
2760 (addti3): Create simpler code if low part is already known to be 0.
2761 (addvti4, uaddvti4): New.
2762 (*add<GPI>3_compareC_cconly_imm): New.
2763 (*add<GPI>3_compareC_cconly): New.
2764 (*add<GPI>3_compareC_imm): New.
2765 (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
2766 handle constants within this pattern..
2767 (*add<GPI>3_compareV_cconly_imm): New.
2768 (*add<GPI>3_compareV_cconly): New.
2769 (*add<GPI>3_compareV_imm): New.
2770 (add<GPI>3_compareV): New.
2771 (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
2772 (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
2773 (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
2774 ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
2775 with 'comparison' operator.
2776 (*add<GPI>3_compareV_cconly_imm): Ditto.
2777 (*add<GPI>3_compareV_cconly): Ditto.
2778 (*add<GPI>3_compareV_imm): Ditto.
2779 (add<GPI>3_compareV): Ditto.
2780 (add<mode>3_carryinC): Ditto.
2781 (*add<mode>3_carryinC_zero): Ditto.
2782 (*add<mode>3_carryinC): Ditto.
2783 (add<mode>3_carryinV): Ditto.
2784 (*add<mode>3_carryinV_zero): Ditto.
2785 (*add<mode>3_carryinV): Ditto.
2787 2018-07-19 Michael Collison <michael.collison@arm.com>
2788 Richard Henderson <rth@redhat.com>
2790 * config/aarch64/aarch64-modes.def (CC_V): New.
2791 * config/aarch64/aarch64-protos.h
2792 (aarch64_addti_scratch_regs): Declare
2793 (aarch64_subvti_scratch_regs): Declare.
2794 (aarch64_expand_subvti): Declare.
2795 (aarch64_gen_unlikely_cbranch): Declare
2796 * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
2797 for signed overflow using CC_Vmode.
2798 (aarch64_get_condition_code_1): Handle CC_Vmode.
2799 (aarch64_gen_unlikely_cbranch): New function.
2800 (aarch64_addti_scratch_regs): New function.
2801 (aarch64_subvti_scratch_regs): New function.
2802 (aarch64_expand_subvti): New function.
2804 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
2806 * config/aarch64/aarch64-option-extensions.def: New entry for profile
2808 * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
2809 * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
2812 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
2815 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
2816 address check not strict.
2818 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
2820 * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
2822 (store_pair_lanes<mode>): Likewise.
2823 * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
2824 enum value 'ADDR_QUERY_LDP_STP_N'.
2825 * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
2826 (aarch64_print_address_internal): Add declaration.
2827 (aarch64_print_ldpstp_address): Remove.
2828 (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
2829 (aarch64_print_operand): Change printing of 'y'.
2830 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
2831 new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
2832 'true' rather than '1'.
2833 * config/aarch64/constraints.md (Uml): Likewise.
2834 (Uml): Rename to Umn.
2837 2018-07-19 Richard Biener <rguenther@suse.de>
2839 * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
2841 * tree-ssa-sccvn.c: Remove alloc-pool.h use.
2842 (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
2843 (vn_reference_hasher): Likewise.
2844 (struct vn_tables_s): Remove obstack and alloc-pool members.
2845 (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
2846 (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
2847 (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
2848 (vn_reference_insert_pieces): Likewise.
2849 (alloc_vn_nary_op_noinit): Adjust.
2850 (vn_nary_op_insert_stmt): Allocate phiargs in-place.
2851 (vn_phi_eq): Adjust.
2852 (shared_lookup_phiargs): Remove.
2853 (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
2854 (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
2855 (visit_reference_op_call): Likewise.
2856 (copy_nary, copy_phi, copy_reference): Remove.
2857 (process_scc): Rewind the obstack when iterating. Do not
2858 copy the elements to valid_info but just move them from one
2859 hashtable to the other.
2860 (allocate_vn_table): Adjust.
2861 (free_vn_table): Likewise.
2862 (init_scc_vn): Likewise.
2863 (free_scc_vn): Likewise.
2865 2018-07-19 H.J. Lu <hongjiu.lu@intel.com>
2868 * config/i386/i386.c (rest_of_insert_endbranch): Lookup
2869 indirect_return as function type attribute.
2870 (ix86_attribute_table): Change indirect_return to function
2872 * doc/extend.texi: Update indirect_return attribute.
2874 2018-07-19 Aldy Hernandez <aldyh@redhat.com>
2876 * wide-int.h (widest2_int): New.
2877 * gimple-fold.c (arith_overflowed_p): Use it.
2878 * tree.h (widest2_int_cst): New.
2879 * tree-vrp.c (wide_int_binop_overflow): Rename from
2880 vrp_int_const_binop.
2881 Rewrite to work on trees.
2882 (extract_range_from_multiplicative_op_1): Abstract code to...
2883 (wide_int_range_min_max): ...here.
2884 (wide_int_range_cross_product): ...and here.
2885 (extract_range_from_binary_expr_1): Abstract overflow code to...
2886 (wide_int_range_mult_wrapping): ...here.
2887 * tree-vrp.h (wide_int_range_cross_product): New.
2888 (wide_int_range_mult_wrapping): New.
2890 2018-07-19 Andrew Senkevich <andrew.senkevich@intel.com>
2891 Julia Koval <julia.koval@intel.com>
2893 * config/i386/x86-tune-costs.h (skylake_memcpy,
2894 skylake_memset): Replace rep_prefix with unrolling for size 512.
2896 2018-07-18 Kugan Vivekanandarajah <kuganv@linaro.org>
2899 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle comparision with EQ_EXPR
2902 2018-07-18 Kelvin Nilsen <kelvin@gcc.gnu.org>
2904 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
2905 this subsection to "PowerPC AltiVec/VSX Built-in Functions".
2906 (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
2907 previously known as "PowerPC AltiVec Built-in Functions". Move
2908 some material to new subsubsections "PowerPC AltiVec Built-in
2909 Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
2911 (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
2912 (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
2913 (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
2914 (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
2916 2018-07-18 Richard Biener <rguenther@suse.de>
2918 PR tree-optimization/86557
2919 * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
2922 2018-07-18 Ilya Leoshkevich <iii@linux.ibm.com>
2924 * config/s390/s390.c (s390_function_profiler): Generate CFI.
2926 2018-07-17 Jeff Law <law@redhat.com>
2928 * config/arm/arm.c (get_label_padding): Update for recent
2929 changes to label_to_alignment.
2931 PR tree-optimization/86010
2932 * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
2934 * config/mips/mips.c (vr4130_align_insns): Update for recent
2935 changes to label_to_alignment.
2937 * config/frv/frv.c (frv_label_align): Update for recent changes
2938 to label_to_alignment.
2940 * config/nios2/nios2.c (nios2_label_align): Update for recent
2941 changes which dropped ALIGN_LABELS_LOG.
2943 2018-07-17 Andreas Schwab <schwab@linux-m68k.org>
2945 * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
2946 (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
2948 2018-07-17 Claudiu Zissulescu <claziss@synopsys.com>
2950 * config/arc/arc.c (arc_label_align): Use align_labels instead of
2951 deprecated align_labels_log.
2953 2018-07-17 Richard Biener <rguenther@suse.de>
2956 * dwarf2out.c (init_sections_and_labels): Always generate
2957 a debug_line_str_section for early LTO debug.
2958 (dwarf2out_finish): Reset debug_line_str_hash output early.
2959 Bump counter for extra dwarf5 .debug_loc labels to not conflict
2961 (dwarf2out_early_finish): Output debug_line_str.
2963 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
2965 * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
2966 index register on z196 or later.
2968 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
2970 * config/s390/s390.c (s390_default_align): Set default function
2972 (s390_override_options_after_change): Call s390_default align.
2973 (s390_option_override_internal): Call s390_default_align.
2974 (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
2976 2018-07-17 Jakub Jelinek <jakub@redhat.com>
2979 * omp-low.c (create_task_copyfn): Copy over also fields corresponding
2980 to _looptemp_ clauses, other than the first two.
2982 2018-07-17 Martin Liska <mliska@suse.cz>
2984 * opts.c: Do not enable OPT_falign_* for -Os.
2986 2018-07-17 Martin Liska <mliska@suse.cz>
2988 * align.h (MAX_CODE_ALIGN): New.
2989 (MAX_CODE_ALIGN_VALUE): New.
2990 * common/config/i386/i386-common.c (ix86_handle_option):
2991 (MAX_CODE_ALIGN): Moved to align.h.
2992 * final.c (MAX_CODE_ALIGN): Likewise.
2993 * opts.c (parse_and_check_align_values):
2994 (MAX_CODE_ALIGN): Likewise.
2995 (MAX_CODE_ALIGN_VALUE): Likewise.
2997 2018-07-17 Martin Liska <mliska@suse.cz>
2999 * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
3000 in order to fulfil coding style.
3001 * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
3002 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
3003 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
3004 * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
3005 * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
3006 * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
3007 * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
3008 * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
3009 (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
3011 2018-07-17 Martin Liska <mliska@suse.cz>
3013 * align.h: New file.
3014 * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions directly.
3015 * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
3016 align_flags of label_to_alignment.
3017 * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into align_flags
3019 * config/m68k/m68k.c: Do not use removed align_labels_value and
3021 * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
3022 (LOOP_ALIGN): Likewise.
3023 (LABEL_ALIGN): Likewise.
3024 * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
3025 Remove not used macro.
3026 (rs6000_loop_align): Change return type to align_flags.
3027 (rs6000_loop_align_max_skip): Remove.
3028 * config/rs6000/rs6000-protos.h (rs6000_loop_align):
3029 Change return type to align_flags.
3030 * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
3031 Remove not used macro.
3032 (rs6000_loop_align): Change return type to align_flags.
3033 (rs6000_loop_align_max_skip): Remove.
3034 * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
3035 * config/rx/rx-protos.h (rx_align_for_label): Make it
3037 * config/rx/rx.c (rx_align_for_label): Change return type
3039 (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
3041 into align_flags class.
3042 (LABEL_ALIGN): Likewise.
3043 (LOOP_ALIGN): Likewise.
3044 * config/s390/s390.c (s390_label_align): Use align_flags
3046 (s390_asm_output_function_label): Likewise.
3047 * config/sh/sh.c (sh_override_options_after_change):
3048 Use align_flags class directly without macros.
3049 (find_barrier): Likewise.
3050 (barrier_align): Likewise.
3051 (sh_loop_align): Likewise.
3052 * config/spu/spu.c (spu_option_override):
3053 Use align_flags_tuple::get_value instead of removed macros.
3054 (spu_sched_init): Likewise.
3055 * config/spu/spu.h (GTY): Likewise.
3056 * config/visium/visium.c (visium_option_override):
3057 Set "8" as default secondary alignment.
3058 * config/visium/visium.h (SUBALIGN_LOG): Define to 3
3059 in order to guarantee secondary alignment of 8.
3060 * coretypes.h: Include align.h header file.
3061 * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
3062 TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
3063 and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
3064 * doc/tm.texi.in: Likewise.
3065 * final.c (struct label_alignment): Remove not used structure.
3066 (LABEL_ALIGN): Change type to align_flags.
3067 (LOOP_ALIGN): Likewise.
3068 (JUMP_ALIGN): Likewise.
3069 (default_loop_align_max_skip): Remove.
3070 (default_label_align_max_skip): Likewise.
3071 (default_jump_align_max_skip): Likewise.
3072 (default_label_align_after_barrier_max_skip):
3073 (LABEL_TO_ALIGNMENT): Change to access label_align vector.
3074 (LABEL_TO_MAX_SKIP): Remove.
3075 (label_to_alignment): Return align_flags type instead of integer.
3076 (label_to_max_skip): Remove.
3077 (align_fuzz): Use align_flags type.
3078 (compute_alignments): Use align_flags type and use align_flags::max
3079 to combine multiple alignments.
3080 (grow_label_align): Grow vec instead of C array.
3081 (update_alignments): Assign just LABEL_TO_ALIGNMENT.
3082 (shorten_branches): Use align_flags type and use align_flags::max
3083 to combine multiple alignments.
3084 (final_scan_insn_1): Remove usage of secondary alignment that comes
3085 from label alignment, but instead use proper secondary alignment
3086 which is computed in grow_label_align.
3087 * flags.h (struct align_flags_tuple): Move to align.h.
3088 (struct align_flags): Likewise.
3089 (state_align_loops): Rename to align_loops.
3090 (state_align_jumps): Rename to align_jumps.
3091 (state_align_labels): Rename to align_labels.
3092 (state_align_functions): Rename to align_functions.
3093 (align_loops_log): Remove.
3094 (align_jumps_log): Remove.
3095 (align_labels_log): Remove.
3096 (align_functions_log): Remove.
3097 (align_loops_max_skip): Remove.
3098 (align_jumps_max_skip): Remove.
3099 (align_labels_max_skip): Remove.
3100 (align_functions_max_skip): Remove.
3101 (align_loops_value): Remove.
3102 (align_jumps_value): Remove.
3103 (align_labels_value): Remove.
3104 (align_functions_value): Remove.
3105 * output.h (label_to_alignment): Change return type to align_flags.
3106 (label_to_max_skip): Remove.
3107 * target.def: Remove loop_align_max_skip, label_align_max_skip,
3108 jump_align_max_skip macros.
3109 * targhooks.h (default_loop_align_max_skip): Remove.
3110 (default_label_align_max_skip): Likewise.
3111 (default_jump_align_max_skip): Likewise.
3112 (default_label_align_after_barrier_max_skip): Remove.
3113 * toplev.c (read_log_maxskip): Use ::normalize function.
3114 (parse_N_M): Remove not used argument and also call ::normalize.
3115 (parse_alignment_opts): Do not pass unused arguments.
3116 * varasm.c (assemble_start_function): Use directly align_functions
3117 instead of removed macros.
3118 * system.h: Do not poison removed macros.
3120 2018-07-17 Jakub Jelinek <jakub@redhat.com>
3123 * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
3124 and cond temporaries don't have reference type if iterator has
3125 pointer type. For init use &for_pre_body instead of pre_p if
3126 for_pre_body is non-empty.
3128 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
3130 * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
3131 double-double modes to SFmode directly directly.
3132 (trunc<mode>sf2_fprs): Delete.
3134 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
3136 * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
3137 for conversions between IFmode and the decimal floating point modes.
3138 (init_float128_ieee): Use the correct names for conversions between
3139 KFmode and the decimal floating point modes.
3141 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
3143 * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
3144 for the conversions between TDmode and IFmode.
3145 (init_float128_ieee): Use more correct names for the conversions
3146 between TDmode and KFmode.
3148 2018-07-16 Jakub Jelinek <jakub@redhat.com>
3150 PR tree-optimization/86526
3151 * builtins.c (expand_builtin_memcmp): Formatting fixes.
3152 (inline_expand_builtin_string_cmp): Likewise.
3153 (inline_string_cmp): Likewise. Use c_readstr instead of
3154 builtin_memcpy_read_str. Add unit_mode temporary.
3156 2018-07-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
3159 * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
3160 * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
3162 2018-07-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
3164 * doc/extend.texi (PowerPC AltiVec Built-in Functions):
3165 Alphabetize prototypes of built-in functions, separating out
3166 built-in functions that are listed in this section but should be
3167 described elsewhere.
3169 2018-07-16 Uros Bizjak <ubizjak@gmail.com>
3172 * expmed.c (emit_store_flag): Do not emit setcc followed by a
3173 conditional move when trapping comparison was split to a
3174 non-trapping one (and vice versa).
3176 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
3178 * config/s390/s390.c (s390_function_profiler): Generate nops
3179 instead of profiler call sequences.
3180 * config/s390/s390.opt: Add the new option.
3182 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
3184 * config/s390/s390.c (s390_function_profiler): Generate
3185 __mcount_loc section.
3186 * config/s390/s390.opt: Add the new option.
3188 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
3190 * common.opt: Add the new warning.
3191 * config/s390/s390.c (s390_function_profiler): Emit "brasl
3192 %r0,__fentry__" when -mfentry is specified.
3193 (s390_option_override_internal): Disallow -mfentry for 31-bit
3195 * config/s390/s390.opt: Add the new option.
3197 2018-07-16 Richard Biener <rguenther@suse.de>
3200 * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
3201 for function-local FUNCTION_DECL and RESULT_DECL immediately.
3203 2018-07-16 Martin Liska <mliska@suse.cz>
3206 * ipa-pure-const.c (malloc_candidate_p): Revert ::get
3209 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
3211 * config/arc/arcHS.md: Update ARCHS scheduling rules.
3213 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
3215 * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
3217 * config/arc/arc-cpus.def (hs4x): New cpu.
3219 * config/arc/arc-tables.opt: Regenerate.
3220 * config/arc/arc.c (arc_sched_issue_rate): New function.
3221 (TARGET_SCHED_ISSUE_RATE): Define.
3222 (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
3223 * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
3225 (attr tune): Add ARCHS4x tune values.
3226 (attr tune_dspmpy): Define.
3227 (*tst): Correct instruction type.
3228 * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
3229 * config/arc/arcHS4x.md: New file.
3230 * config/arc/fpu.md: Update instruction type attributes.
3231 * config/arc/t-multilib: Regenerate.
3233 2018-07-16 Tom de Vries <tdevries@suse.de>
3236 * var-tracking.c (vt_initialize): Fix pre_dec handling.
3238 2018-07-16 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3240 * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
3243 2018-07-16 Eric Botcazou <ebotcazou@adacore.com>
3245 PR tree-optimization/86514
3246 * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
3247 conversion to a boolean type from a type with greater precision.
3249 2018-07-16 Tom de Vries <tdevries@suse.de>
3251 * var-tracking.c (vt_initialize): Print adjusted insn slim if
3252 dump_flags request TDF_SLIM.
3254 2018-07-16 Aldy Hernandez <aldyh@redhat.com>
3256 * fold-const.c (int_const_binop_1): Abstract...
3257 (wide_int_binop): ...wide int code here.
3258 (poly_int_binop): ...poly int code here.
3259 Abstract the rest of int_const_binop_1 into int_const_binop.
3260 * fold-const.h (wide_int_binop): New.
3261 * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
3262 Remove useless PLUS/MINUS_EXPR case.
3263 (zero_nonzero_bits_from_vr): Move wide int code...
3264 (zero_nonzero_bits_from_bounds): ...here.
3265 (extract_range_from_binary_expr_1): Move mask optimization code...
3266 (range_easy_mask_min_max): ...here.
3267 * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
3268 (range_easy_mask_min_max): New.
3270 2018-07-15 Jeff Law <law@redhat.com>
3273 * config/sh/sh.c (output_mi_thunk): Remove dead conditional
3276 2018-07-14 Jim Wilson <jimw@sifive.com>
3278 * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
3280 2018-07-14 Paul Koning <ni1d@arrl.net>
3282 * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
3284 2018-07-13 Jan Hubicka <hubicka@ucw.cz>
3286 * lto-streamer-out.c (copy_function_or_variable): Dump info about
3289 2018-07-13 Bill Schmidt <wschmidt@linux.ibm.com>
3290 Steve Munroe <munroesj52@gmail.com>
3292 * config/rs6000/emmintrin.h (_mm_and_si128): New function.
3293 (_mm_andnot_si128): Likewise.
3294 (_mm_or_si128): Likewise.
3295 (_mm_xor_si128): Likewise.
3297 2018-07-13 Qing Zhao <qing.zhao@oracle.com>
3300 * builtins.c (expand_builtin_memcmp): Inline the calls first
3301 when result_eq is false.
3302 (expand_builtin_strcmp): Inline the calls first.
3303 (expand_builtin_strncmp): Likewise.
3304 (inline_string_cmp): New routine. Expand a string compare
3305 call by using a sequence of char comparison.
3306 (inline_expand_builtin_string_cmp): New routine. Inline expansion
3307 a call to str(n)cmp/memcmp.
3308 * doc/invoke.texi (--param builtin-string-cmp-inline-length):
3310 * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
3312 2018-07-13 Richard Earnshaw <rearnsha@arm.com>
3314 * config/arm/driver-arm.c: Include arm-native.h.
3315 (host_detect_local_cpu): Use auto-generated data tables.
3316 (vendors, arm_cpu_table): Delete. Move part information to ...
3317 * config/arm/arm-cpus.in: ... here.
3318 * config/arm/parsecpu.awk (gen_native): New function.
3319 (vendor, part): New CPU fields.
3320 (END): Add support for building the native CPU detection tables.
3321 * config/arm/t-arm (arm-native.h): Add build rule.
3322 (driver-arm.o): Add dependency on arm-native.h.
3324 2018-07-13 Richard Biener <rguenther@suse.de>
3327 * match.pd (addr1 - addr2): Allow either of the operand to
3330 2018-07-13 Tom de Vries <tdevries@suse.de>
3332 * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
3335 2018-07-13 Jackson Woodruff <jackson.woodruff@arm.com>
3337 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
3338 arrays instead of numbered variables.
3340 2018-07-13 Eric Botcazou <ebotcazou@adacore.com>
3342 * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
3343 * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
3345 2018-07-13 Richard Biener <rguenther@suse.de>
3348 * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
3349 instead of get_context_die.
3351 2018-07-13 Kugan Vivekanandarajah <kuganv@linaro.org>
3352 Richard Biener <rguenther@suse.de>
3355 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
3356 that the loop latch destination where phi is defined.
3358 2018-07-12 Kito Cheng <kito.cheng@gmail.com>
3360 * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
3361 (riscv_expand_epilogue): Add assertion to check interrupt mode.
3362 (riscv_set_current_function): Extract getting interrupt type to new
3364 (riscv_get_interrupt_type): New function.
3365 (riscv_merge_decl_attributes): New function, checking interrupt type is
3367 (TARGET_MERGE_DECL_ATTRIBUTES): Define.
3369 2018-07-12 Paul Koning <ni1d@arrl.net>
3371 * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
3374 2018-07-12 Paul Koning <ni1d@arrl.net>
3376 * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
3377 zero reference, add doloop_end instead.
3378 * doc/md.texi (decrement_and_branch_until_zero): Remove.
3379 (Looping patterns): Remove decrement_and_branch_until_zero. Add
3380 detail for doloop_end.
3382 2018-07-12 Martin Sebor <msebor@redhat.com>
3385 * attribs.c (decl_attributes): Reject conflicting attributes before
3386 calling attribute handlers.
3388 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
3390 * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
3392 (gcc::dump_manager::get_dump_file_name): likewise.
3393 (dump_begin): Likewise.
3394 * dumpfile.h (dump_begin): Update prototype.
3395 (gcc::dump_manager::get_dump_file_name,
3396 gcc::dump_manager::get_dump_file_name): Update prototype.
3398 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
3400 * internal-fn.h (vectorizable_internal_fn_p): New function.
3401 * tree-vect-slp.c (compatible_calls_p): Likewise.
3402 (vect_build_slp_tree_1): Remove nops argument. Handle calls
3403 to internal functions.
3404 (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
3406 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
3408 * fold-const.h (inverse_conditions_p): Declare.
3409 * fold-const.c (inverse_conditions_p): New function.
3410 * match.pd: Use inverse_conditions_p. Add folds of view_converts
3411 that test the inverse condition of a conditional internal function.
3412 * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
3413 * internal-fn.c (internal_fn_mask_index): Handle conditional
3415 (vectorized_internal_fn_supported_p): New function.
3416 * tree-if-conv.c: Include internal-fn.h and fold-const.h.
3417 (any_pred_load_store): Replace with...
3418 (need_to_predicate): ...this new variable.
3419 (redundant_ssa_names): New variable.
3420 (ifcvt_can_use_mask_load_store): Move initial checks to...
3421 (ifcvt_can_predicate): ...this new function. Handle tree codes
3422 for which a conditional internal function exists.
3423 (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
3424 instead of ifcvt_can_use_mask_load_store. Update after variable
3426 (predicate_load_or_store): New function, split out from
3427 predicate_mem_writes.
3428 (check_redundant_cond_expr): New function.
3429 (value_available_p): Likewise.
3430 (predicate_rhs_code): Likewise.
3431 (predicate_mem_writes): Rename to...
3432 (predicate_statements): ...this. Use predicate_load_or_store
3433 and predicate_rhs_code.
3434 (combine_blocks, tree_if_conversion): Update after above name changes.
3435 (ifcvt_local_dce): Handle redundant_ssa_names.
3436 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
3437 general conditional functions.
3438 * tree-vect-stmts.c (vectorizable_call): Likewise.
3440 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
3441 Alan Hayward <alan.hayward@arm.com>
3442 David Sherwood <david.sherwood@arm.com>
3444 * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
3445 * internal-fn.c (can_interpret_as_conditional_op_p): New function.
3446 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
3447 plus and minus and convert them into IFN_COND_FMA-based sequences.
3448 (convert_mult_to_fma): Handle conditional plus and minus.
3450 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
3452 * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
3453 * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
3454 (cond_fnms_optab): New optabs.
3455 * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
3457 (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
3458 * internal-fn.h (get_conditional_internal_fn): Declare.
3459 (get_unconditional_internal_fn): Likewise.
3460 * internal-fn.c (cond_ternary_direct): New macro.
3461 (expand_cond_ternary_optab_fn): Likewise.
3462 (direct_cond_ternary_optab_supported_p): Likewise.
3463 (FOR_EACH_COND_FN_PAIR): Likewise.
3464 (get_conditional_internal_fn): New function.
3465 (get_unconditional_internal_fn): Likewise.
3466 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
3467 (gimple_match_op::gimple_match_op): Add a new overload for 5
3469 (gimple_match_op::set_op): Likewise.
3470 (gimple_resimplify5): Declare.
3471 * genmatch.c (decision_tree::gen): Generate simplifications for
3473 * gimple-match-head.c (gimple_simplify): Define an overload for
3474 5 operands. Handle calls with 5 arguments in the top-level overload.
3475 (convert_conditional_op): Handle conversions from unconditional
3476 internal functions to conditional ones.
3477 (gimple_resimplify5): New function.
3478 (build_call_internal): Pass a fifth operand.
3479 (maybe_push_res_to_seq): Likewise.
3480 (try_conditional_simplification): Try converting conditional
3481 internal functions to unconditional internal functions.
3482 Handle 3-operand unconditional forms.
3483 * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
3484 Define ternary equivalents of the current rules for binary conditional
3486 * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
3488 * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
3489 (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
3490 (optab): Handle them.
3491 (SVE_COND_FP_TERNARY): New int iterator.
3492 (sve_fmla_op, sve_fmad_op): New int attributes.
3493 * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
3494 (*cond_<optab><mode>_2, *cond_<optab><mode_4)
3495 (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
3497 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
3499 * target.def (preferred_else_value): New target hook.
3500 * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
3501 * doc/tm.texi: Regenerate.
3502 * targhooks.h (default_preferred_else_value): Declare.
3503 * targhooks.c (default_preferred_else_value): New function.
3504 * internal-fn.h (conditional_internal_fn_code): Declare.
3505 * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
3506 (get_conditional_internal_fn): Use it.
3507 (conditional_internal_fn_code): New function.
3508 * gimple-match.h (gimple_match_cond): New struct.
3509 (gimple_match_op): Add a cond member function.
3510 (gimple_match_op::gimple_match_op): Update all forms to take a
3512 * genmatch.c (expr::gen_transform): Use the same condition as res_op
3513 for the suboperation, but don't specify a particular else_value.
3514 * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
3515 (visit_nary_op, visit_reference_op_load): Pass
3516 gimple_match_cond::UNCOND to the gimple_match_op constructor.
3517 * gimple-match-head.c: Include tree-eh.h
3518 (convert_conditional_op): New function.
3519 (maybe_resimplify_conditional_op): Likewise.
3520 (gimple_resimplify1): Call maybe_resimplify_conditional_op.
3521 (gimple_resimplify2): Likewise.
3522 (gimple_resimplify3): Likewise.
3523 (gimple_resimplify4): Likewise.
3524 (maybe_push_res_to_seq): Return null for conditional operations.
3525 (try_conditional_simplification): New function.
3526 (gimple_simplify): Call it. Pass conditions to the gimple_match_op
3528 * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
3530 * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
3532 (TARGET_PREFERRED_ELSE_VALUE): Redefine.
3534 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
3536 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
3538 (hash_tree): Do not hash DECL_FCONTEXT
3539 * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
3540 Do not stream DECL_FCONTEXT.
3541 * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
3542 * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
3544 2018-07-12 Richard Biener <rguenther@suse.de>
3547 * dwarf2out.c (gen_block_die): Only output blocks when they have
3548 at least one !DECL_IGNORED_P variable.
3550 2018-07-12 Richard Biener <rguenther@suse.de>
3553 * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
3554 Remove -mieee-fp handling.
3556 2018-07-12 Richard Biener <rguenther@suse.de>
3558 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
3559 left-over from last patch.
3561 2018-07-12 Jakub Jelinek <jakub@redhat.com>
3563 PR tree-optimization/86492
3564 * gimple-ssa-store-merging.c
3565 (imm_store_chain_info::coalesce_immediate_stores): Call
3566 check_no_overlap even for the merge_overlapping case. Formatting fix.
3568 2018-07-12 Richard Biener <rguenther@suse.de>
3571 * fold-const.c (fold_binary_op_with_conditional_arg): Do not
3572 move possibly trapping operations into the conditional.
3574 2018-07-12 Richard Biener <rguenther@suse.de>
3576 * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
3577 (vn_lookup_simplify_result): Remove recursion limit applied
3579 (vn_nary_build_or_lookup_1): Adjust.
3580 (try_to_simplify): Likewise.
3581 * gimple-match-head.c (gimple_resimplify1): Instead apply one
3583 (gimple_resimplify2): Likewise.
3584 (gimple_resimplify3): Likewise.
3585 (gimple_resimplify4): Likewise.
3587 2018-07-11 Jakub Jelinek <jakub@redhat.com>
3589 * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
3590 Use __mmask64 type instead of __mmask8 for __M argument.
3591 * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
3592 _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
3594 (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
3595 __mmask16 for __M argument.
3596 (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
3597 _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
3598 to __mmask16 instead of __mmask8.
3599 * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
3600 _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
3601 _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
3602 _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
3603 instead of __mmask16 for __U argument.
3604 * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
3605 __mmask16 instead of __mmask8 for __U argument.
3606 (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
3608 (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
3610 (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
3612 (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
3614 (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
3616 (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
3618 (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
3620 (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
3622 (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
3624 (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
3626 (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
3627 _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
3628 return type as well as __M argument type and all casts from __mmask8
3630 (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
3631 _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
3632 return type as well as __M argument type and all casts from __mmask8
3634 (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
3635 _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
3636 return type as well as __M argument type and all casts from __mmask8
3638 (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
3639 _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
3640 return type as well as __M argument type and all casts from __mmask8
3642 * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
3643 _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
3646 2018-07-11 Grazvydas Ignotas <notasas@gmail.com>
3648 * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
3649 _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
3652 2018-07-11 Paul Koning <ni1d@arrl.net>
3654 * doc/md.texi (define_subst): Document how multiple occurrences of
3655 the same argument in the replacement pattern are handled.
3657 2018-07-11 Paul Koning <ni1d@arrl.net>
3659 * doc/extend.texi (Common Variable Attributes): Move "mode" into
3661 (Common Type Attributes): Add "mode" attribute.
3663 2018-07-11 Jan Hubicka <hubicka@ucw.cz>
3665 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
3666 stream DECL_ORIGINAL_TYPE.
3667 (DFS::DFS_write_tree_body): Drop hack handling local external decls.
3668 (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
3669 * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
3670 Do not walk original type.
3671 * tree-streamer-out.c (streamer_write_chain): Drop hack handling
3673 (write_ts_decl_non_common_tree_pointers): Do not stream
3675 * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
3676 (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
3678 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
3680 * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
3681 thread twice from the same starting edge.
3683 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
3685 * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
3686 * gimple.c (gimple_call_nonnull_result_p): ...here...
3687 (gimple_call_nonnull_arg): ...and here.
3688 * gimple.h (gimple_call_nonnull_result_p): New.
3689 (gimple_call_nonnull_arg): New.
3691 2018-07-11 Richard Earnshaw <rearnsha@arm.com>
3693 * config/arm/arm-cpus.in: Move information from fpu field of each
3694 cpu definition to the isa field.
3695 * config/arm/parsecpu.awk (fpu): Delete match rule.
3696 (gen_comm_data): Don't add bits from the CPU's FPU entry.
3698 2018-07-11 Richard Biener <rguenther@suse.de>
3701 * dwarf2out.c (init_sections_and_labels): Use
3702 output_asm_line_debug_info consistently.
3703 (dwarf2out_early_finish): Likewise.
3704 (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
3707 2018-07-11 Richard Biener <rguenther@suse.de>
3709 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
3710 Rework father_bb setting in a way to avoid propagating constants
3711 multiple times on a loop body.
3713 2018-07-10 Mark Wielaard <mark@klomp.org>
3716 * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
3718 2018-07-10 Richard Biener <rguenther@suse.de>
3720 * hash-map.h (hash_map::iterator::operator*): Return
3721 references to key and value.
3723 2018-07-10 Jakub Jelinek <jakub@redhat.com>
3726 * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
3727 to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
3728 BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
3729 (gimplify_omp_for): For composite loops, move outer
3730 OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
3731 OMP_FOR/OMP_SIMD if there are any non-trivial wrappers. For class
3732 iterators add any needed clauses. Allow OMP_FOR_ORIG_DECLS to contain
3733 TREE_LIST for both the original class iterator and the "last" helper
3734 var. Gimplify OMP_FOR_PRE_BODY before the outermost composite
3735 loop, remember has_decl_expr from outer composite loops for the
3736 innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
3738 2018-07-09 Martin Sebor <msebor@redhat.com>
3742 * builtins.c (c_strlen): Avoid out-of-bounds warnings when
3743 accessing implicitly initialized array elements.
3744 * expr.c (string_constant): Handle string initializers of
3745 character arrays within aggregates.
3746 * gimple-fold.c (fold_array_ctor_reference): Add argument.
3747 Store element offset. As a special case, handle zero size.
3748 (fold_nonarray_ctor_reference): Same.
3749 (fold_ctor_reference): Add argument. Store subobject offset.
3750 * gimple-fold.h (fold_ctor_reference): Add argument.
3752 2018-07-09 Paul Koning <ni1d@arrl.net>
3754 * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
3755 (pdp11_insn_cost): New function.
3756 (pdp11_md_asm_adjust): New function.
3757 (TARGET_INVALID_WITHIN_DOLOOP): Define.
3758 (pdp11_rtx_costs): Update to match machine better.
3759 (output_addr_const_pdp11): Correct format mismatch warnings.
3760 * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
3761 * config/pdp11/pdp11.md: General change to add base_cost and/or
3762 length attributes for use by new pdp11_insn_cost function.
3763 (MIN_BRANCH): Correct definition.
3765 (doloop_end): Use standard pattern name for looping pattern.
3766 (doloop_end_nocc): New.
3767 (movsf): Add another constraint alternative.
3768 (zero_extendqihi2): Add constraint alternatives for not in place
3770 (zero_extendhisi2): Remove.
3771 (shift patterns): Add CC handling variants.
3775 (define_peephole2): New peephole to recognize mov that sets CC for
3778 2018-07-09 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3780 * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
3781 SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
3784 2018-07-09 Jakub Jelinek <jakub@redhat.com>
3787 * real.c (real_nextafter): Return true if result is denormal.
3789 2018-07-09 Martin Liska <mliska@suse.cz>
3791 * common.opt: Add back wrongly removed attribute.
3793 2018-07-09 Richard Biener <rguenther@suse.de>
3796 * dwarf2out.c (gen_block_die): For an early generated DIE
3797 always output high/low PC attributes.
3799 2018-07-09 Tom de Vries <tdevries@suse.de>
3801 * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
3802 * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
3803 onto VAR_DECL with abstract origin.
3805 2018-07-07 Jim Wilson <jimw@sifive.com>
3807 * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
3809 2018-07-07 Tom de Vries <tdevries@suse.de>
3811 * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
3813 2018-07-07 Aldy Hernandez <aldyh@redhat.com>
3815 * tree-vrp.c (vrp_int_const_binop): Change overflow type to
3817 (combine_bound): Use wide-int overflow calculation instead of
3819 * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
3821 * fold-const.c (int_const_binop_2): Same.
3822 (extract_muldiv_1): Same.
3823 (fold_div_compare): Same.
3824 (fold_abs_const): Same.
3826 * poly-int.h (add): Same.
3830 * predict.c (predict_iv_comparison): Same.
3831 * profile-count.c (slow_safe_scale_64bit): Same.
3832 * simplify-rtx.c (simplify_const_binary_operation): Same.
3833 * tree-chrec.c (tree_fold_binomial): Same.
3834 * tree-data-ref.c (split_constant_offset_1): Same.
3835 * tree-if-conv.c (idx_within_array_bound): Same.
3836 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
3837 * tree-ssa-phiopt.c (minmax_replacement): Same.
3838 * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
3839 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
3840 * vr-values.c (vr_values::adjust_range_with_scev): Same.
3841 * wide-int.cc (wi::add_large): Same.
3842 (wi::mul_internal): Same.
3843 (wi::sub_large): Same.
3844 (wi::divmod_internal): Same.
3845 * wide-int.h: Change overflow type to overflow_type for neg, add,
3846 mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
3847 mod_trunc, mod_ceil, mod_round, add_large, sub_large,
3848 mul_internal, divmod_internal.
3849 (overflow_type): New enum.
3850 (accumulate_overflow): New.
3852 2018-07-06 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
3854 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
3855 (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
3857 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
3859 * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
3860 argument is checked for zero before entering loop, avoid checking again.
3862 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
3864 * gimplify.h (generic_expr_could_trap_p): Set as global function.
3865 * gimplify.h (generic_expr_could_trap_p): Likwise.
3866 * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
3868 2018-07-06 Jakub Jelinek <jakub@redhat.com>
3870 PR tree-optimization/86401
3871 * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
3872 ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
3873 (fold_bit_and_mask): ... here. New helper function for match.pd.
3874 * fold-const.h (fold_bit_and_mask): Declare.
3875 * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
3877 2018-07-06 Peter Bergner <bergner@linux.ibm.com>
3880 * target.def (translate_mode_attribute): New hook.
3881 * targhooks.h (default_translate_mode_attribute): Declare.
3882 * targhooks.c (default_translate_mode_attribute): New function.
3883 * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
3884 * doc/tm.texi: Regenerate.
3885 * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
3886 (rs6000_translate_mode_attribute): New function.
3888 2018-07-06 Paul Koning <ni1d@arrl.net>
3890 * doc/md.texi (define_split): Document DONE and FAIL.
3891 (define_peephole2): Ditto.
3893 2018-07-05 Jeff Law <law@redhat.com>
3895 PR tree-optimization/86010
3896 * tree-ssa-dse.c (compute_trims): More aggressively trim at
3897 both the head and tail of mem* and str* calls.
3899 2018-07-05 Jim Wilson <jimw@sifive.com>
3901 * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
3903 2018-07-05 Indu Bhagat <indu.bhagat@oracle.com>
3905 * config/aarch64/aarch64-simd.md: correct flags text for
3906 MIN_EXPR replacement.
3908 2018-07-05 James Clarke <jrtc27@jrtc27.com>
3910 * configure: Regenerated.
3912 2018-07-05 Carl Love <cel@us.ibm.com>
3914 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
3915 float argument to VSX_BUILTIN_DOUBLEH_V4SF.
3916 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
3917 VSX_BUILTIN_DOUBLEL_V4SF.
3919 2018-07-05 Martin Sebor <msebor@redhat.com>
3922 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
3923 than its domain to compute its the upper bound of a char array.
3925 2018-07-05 Nathan Sidwell <nathan@acm.org>
3927 Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
3928 * doc/cpp.texi: Update comment.
3929 * doc/tm.texi: Rebuilt.
3930 * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
3931 (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
3932 * doc/extend.texi (Backwards Compatibility): Clarify it is system
3933 headers affected by extern "C".
3934 * system.h: Poison NO_IMPLICIT_EXTERN_C.
3935 * config/alpha/alpha.h, config/arm/uclinux-elf.h,
3936 config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
3937 config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
3938 config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
3939 config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
3940 config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
3941 config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
3942 config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
3943 config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
3944 config/sparc/openbsd64.h, config/sparc/sp-elf.h,
3945 config/sparc/sp64-elf.h, config/spu/spu.h,
3946 config/stormy16/stormy16.h, config/v850/v850.h,
3947 config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
3948 define NO_IMPLICIT_EXTERN_C.
3949 * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
3951 2018-07-05 Tamar Christina <tamar.christina@arm.com>
3954 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
3955 instead of GET_MODE_SIZE when comparing Units.
3957 2018-07-05 Tamar Christina <tamar.christina@arm.com>
3960 * rtlanal.c (set_noop_p): Constrain on mode change,
3961 include hard-reg-set.h
3963 2018-07-05 Tamar Christina <tamar.christina@arm.com>
3965 * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
3967 2018-07-05 Jakub Jelinek <jakub@redhat.com>
3970 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
3973 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
3975 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
3977 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
3980 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
3982 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
3984 2018-07-04 Eric Botcazou <ebotcazou@adacore.com>
3987 * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
3989 2018-07-04 Aldy Hernandez <aldyh@redhat.com>
3991 * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
3992 neg_*_op* variables.
3994 2018-07-04 Martin Liska <mliska@suse.cz>
3996 * tree-switch-conversion.c: Define
3997 max_ratio_for_speed and max_ratio_for_size constants.
3999 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
4000 Martin Liska <mliska@suse.cz>
4005 * common.opt: Rename align options with 'str_' prefix.
4006 * common/config/i386/i386-common.c (set_malign_value): New
4008 (ix86_handle_option): Use it to set -falign-* options/
4009 * config/aarch64/aarch64-protos.h (struct tune_params): Change
4010 type from int to string.
4011 * config/aarch64/aarch64.c: Update default values from int
4013 * config/alpha/alpha.c (alpha_override_options_after_change):
4015 * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
4016 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
4017 max skip conditionally.
4018 * config/i386/freebsd.h (SUBALIGN_LOG): New.
4019 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
4020 max skip conditionally.
4021 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
4022 max skip conditionally.
4023 * config/i386/gnu-user.h (SUBALIGN_LOG): New.
4024 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
4025 max skip conditionally.
4026 * config/i386/i386.c (struct ptt): Change type from int to
4028 (ix86_default_align): Set default values.
4029 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
4030 max skip conditionally.
4031 * config/i386/iamcu.h (SUBALIGN_LOG): New.
4032 (ASM_OUTPUT_MAX_SKIP_ALIGN):
4033 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
4034 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
4035 max skip conditionally.
4036 * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
4037 (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
4038 * config/i386/x86-64.h (SUBALIGN_LOG): New.
4039 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
4040 max skip conditionally.
4041 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
4042 * config/ia64/ia64.c (ia64_option_override): Set default values
4043 for alignment options.
4044 * config/m68k/m68k.c: Handle new str_align_* options.
4045 * config/mips/mips.c (mips_set_compression_mode): Change
4047 (mips_option_override): Set default values for options.
4048 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
4050 * config/rs6000/rs6000.c (rs6000_option_override_internal):
4052 * config/rx/rx.c (rx_option_override): Likewise.
4053 * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
4054 (LABEL_ALIGN): Use align_labels_log.
4055 (LOOP_ALIGN): Use align_loops_align.
4056 * config/s390/s390.c (s390_asm_output_function_label): Use new
4058 * config/sh/sh.c (sh_override_options_after_change):
4059 Change type of constants.
4060 * config/spu/spu.c (spu_sched_init): Likewise.
4061 * config/sparc/sparc.c (sparc_option_override): Set default
4063 * config/visium/visium.c (visium_option_override): Likewise.
4064 * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
4065 emit p2align format with last argument if it's not needed.
4066 * doc/invoke.texi: Document extended format of -falign-*.
4067 * final.c: Use align_labels alignment.
4068 * flags.h (struct target_flag_state): Change type to use
4070 (struct align_flags_tuple): New.
4071 (struct align_flags): Likewise.
4072 (align_loops_log): Redefine macro to use new types.
4073 (align_loops_max_skip): Redefine macro to use new types.
4074 (align_jumps_log): Redefine macro to use new types.
4075 (align_jumps_max_skip): Redefine macro to use new types.
4076 (align_labels_log): Redefine macro to use new types.
4077 (align_labels_max_skip): Redefine macro to use new types.
4078 (align_functions_log): Redefine macro to use new types.
4079 (align_loops): Redefine macro to use new types.
4080 (align_jumps): Redefine macro to use new types.
4081 (align_labels): Redefine macro to use new types.
4082 (align_functions): Redefine macro to use new types.
4083 (align_functions_max_skip): Redefine macro to use new types.
4084 (align_loops_value): New macro.
4085 (align_jumps_value): New macro.
4086 (align_labels_value): New macro.
4087 (align_functions_value): New macro.
4088 * function.c (invoke_set_current_function_hook): Propagate
4089 alignment values from flags to global variables default in
4091 * ipa-icf.c (sem_function::equals_wpa): Use
4092 cl_optimization_option_eq instead of memcmp.
4093 * lto-streamer.h (cl_optimization_stream_out): Support streaming
4095 (cl_optimization_stream_in): Likewise.
4096 * optc-save-gen.awk: Support strings in cl_optimization.
4097 * opth-gen.awk: Likewise.
4098 * opts.c (finish_options): Remove error checking of invalid
4100 (MAX_CODE_ALIGN): Remove.
4101 (MAX_CODE_ALIGN_VALUE): Likewise.
4102 (parse_and_check_align_values): New function.
4103 (check_alignment_argument): Likewise.
4104 (common_handle_option): Use check_alignment_argument.
4105 * opts.h (parse_and_check_align_values): Declare.
4106 * toplev.c (init_alignments): Remove.
4107 (read_log_maxskip): New.
4108 (parse_N_M): Likewise.
4109 (parse_alignment_opts): Likewise.
4110 (backend_init_target): Remove usage of init_alignments.
4111 * toplev.h (parse_alignment_opts): Declare.
4112 * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
4114 * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
4115 * tree.c (cl_option_hasher::equal): New.
4116 * varasm.c: Use new global macros.
4118 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
4120 * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
4121 Use a simpler align directive also if MAXSKIP = ALIGN-1.
4122 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4123 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4124 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4125 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
4126 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
4127 is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
4128 define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
4129 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4130 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4131 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4132 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4134 2018-07-04 Martin Liska <mliska@suse.cz>
4135 Jonathan Wakely <jwakely@redhat.com>
4137 * coverage.c: Use correct type.
4138 * doc/invoke.texi: Language correction.
4140 2018-07-03 H.J. Lu <hongjiu.lu@intel.com>
4143 * config/i386/i386.c (rest_of_insert_endbranch): Also generate
4144 ENDBRANCH for non-tail call which may return via indirect branch.
4145 * doc/extend.texi: Document indirect_return attribute.
4147 2018-07-03 Martin Sebor <msebor@redhat.com>
4149 PR tree-optimization/86274
4150 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
4152 (format_floating): Correct handling of infinities and NaNs.
4154 2018-07-03 Martin Sebor <msebor@redhat.com>
4156 * print-tree.c (print_real_cst): New function.
4157 (print_node_brief): Call it.
4158 (print_node): Ditto.
4160 2018-07-03 Jeff Law <law@redhat.com>
4162 * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
4163 into a single pattern.
4165 * config/h8300/h8300.md (ors code_iterator): New.
4166 (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
4167 a single pattern and single splitter.
4168 (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
4169 (iorqi3_1, xorqi3_1): Likewise.
4170 (iorqi3, xorqi3 expanders): Similarly.
4172 * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
4173 (movmd_internal) into a single pattern using the P mode iterator.
4174 (movmd splitters): Similarly.
4175 (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
4176 (movsd splitters): Similarly.
4178 * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
4179 ADDB, ADDW and ADDL into a single ADD attribute which selects the
4180 right table based on the size of the operand.
4181 * config/h8300/h8300.md (length_table): Corresponding changes. All
4182 references to "addb", "addw" and "addl" changed to "add".
4183 (btst patterns): Merge two variants into a single pattern.
4184 (tstqi, tsthi): Likewise.
4185 (addhi3_incdec, addsi3_incdec): Likewise.
4186 (subhi3_h8300hs, subsi3_h8300hs): Likewise.
4187 (mulhi3, mulsi3): Likewise.
4188 (udivhi3, udivsi3): Likewise.
4189 (divhi3, divsi3): Likewise.
4190 (andorqi3, andorhi3, andorsi3): Likewise.
4192 2018-07-03 Uros Bizjak <ubizjak@gmail.com>
4195 * config/i386/sse.md (uavg<mode>3_ceil): New expander.
4196 (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
4198 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
4200 PR tree-optimization/85694
4201 * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
4202 (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
4204 * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
4205 (<u>avg<mode>3_ceil): New patterns.
4207 2018-07-03 David Malcolm <dmalcolm@redhat.com>
4209 * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
4210 scan-tree-dump directive.
4211 * gcc.dg/vect/slp-perm-2.c: Likewise.
4212 * gcc.dg/vect/slp-perm-3.c: Likewise.
4213 * gcc.dg/vect/slp-perm-5.c: Likewise.
4214 * gcc.dg/vect/slp-perm-6.c: Likewise.
4215 * gcc.dg/vect/slp-perm-7.c: Likewise.
4216 * gcc.dg/vect/slp-perm-8.c: Likewise.
4218 2018-07-03 Marek Polacek <polacek@redhat.com>
4221 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
4223 2018-07-03 Richard Biener <rguenther@suse.de>
4226 * tree-ssa-structalias.c (find_func_clobbers): Properly
4227 handle indirect calls.
4229 2018-07-03 Jeff Law <law@redhat.com>
4231 * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
4232 (shifts): New code iterator.
4233 (movqi, movhi, movsi, movsf expanders): Consolidate into a single
4234 expander. Fix HImode handling on H8/SX.
4235 (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
4236 (subqi3, subhi3, subsi3 expanders): Likewise.
4237 (andqi3, andhi3, andsi3 expanders): Likewise.
4238 (iorqi3, iorhi3, iorsi3 expanders): Likewise.
4239 (xorqi3, xorhi3, xorsi3 expanders): Likewise.
4240 (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
4241 (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
4242 (zero_extendqihi2, zero_extendqisi2): Likewise.
4243 (extendqihi2, extendqisi2): Likewise.
4244 (rotlqi3, rotlhi3, rotlsi3): Likewise.
4245 (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
4246 (rotlqi3_1, rotlhi3_1): Likewise.
4247 (logicalhi3_sn, logicalsi3_sn): Likewise.
4248 (logicalhi3, logicalsi3): Likewise.
4250 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
4252 * tree-vect-patterns.c (vect_recog_rotate_pattern)
4253 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
4254 (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
4255 (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
4256 type to append_pattern_def_seq instead of creating a stmt_vec_info
4258 (build_mask_conversion): Likewise. Remove vinfo argument.
4259 (vect_add_conversion_to_patterm): Likewise, renaming to...
4260 (vect_add_conversion_to_pattern): ...this.
4261 (vect_recog_mask_conversion_pattern): Update call to
4262 build_mask_conversion. Pass the vector type to
4263 append_pattern_def_seq here too.
4264 (vect_recog_gather_scatter_pattern): Update call to
4265 vect_add_conversion_to_pattern.
4267 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
4269 * tree-vect-patterns.c (new_pattern_def_seq): Delete.
4270 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
4271 (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
4272 (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
4273 STMT_VINFO_PATTERN_DEF_SEQ to null here.
4274 (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
4275 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
4276 append_pattern_def_seq instead of new_pattern_def_seq.
4277 (vect_recog_divmod_pattern): Do both of the above.
4278 (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
4281 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
4283 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
4284 (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
4285 (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
4286 (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
4287 (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
4288 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
4289 (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
4290 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
4291 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
4292 (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
4293 parameter with a single stmt_vec_info.
4294 (vect_recog_func_ptr): Likewise.
4295 (vect_recog_gather_scatter_pattern): Likewise, folding in...
4296 (vect_try_gather_scatter_pattern): ...this.
4297 (vect_pattern_recog_1): Remove stmts_to_replace and just pass
4298 the stmt_vec_info of the statement to be matched. Don't clear
4299 STMT_VINFO_RELATED_STMT.
4300 (vect_pattern_recog): Update call accordingly.
4302 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
4304 PR tree-optimization/85694
4305 * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
4306 (uavgM3_ceil): Document new optabs.
4307 * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
4308 * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
4310 * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
4311 (savg_ceil_optab): New optabs.
4312 * tree-vect-patterns.c (vect_recog_average_pattern): New function.
4313 (vect_vect_recog_func_ptrs): Add it.
4314 * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
4315 constant directly from the associated lhs.
4317 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
4319 * tree-vect-patterns.c (vect_split_statement): New function.
4320 (vect_convert_input): Use it to try to split an existing cast.
4322 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
4324 * poly-int.h (print_hex): New function.
4325 * dumpfile.h (dump_dec, dump_hex): Declare.
4326 * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
4327 * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
4328 min_input_precision, operation_precision and operation_sign.
4329 * tree-vect-patterns.c (vect_get_range_info): New function.
4330 (vect_same_loop_or_bb_p, vect_single_imm_use)
4331 (vect_operation_fits_smaller_type): Delete.
4332 (vect_look_through_possible_promotion): Add an optional
4333 single_use_p parameter.
4334 (vect_recog_over_widening_pattern): Rewrite to use new
4335 stmt_vec_info infomration. Handle one operation at a time.
4336 (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
4337 (vect_truncatable_operation_p, vect_set_operation_type)
4338 (vect_set_min_input_precision): New functions.
4339 (vect_determine_min_output_precision_1): Likewise.
4340 (vect_determine_min_output_precision): Likewise.
4341 (vect_determine_precisions_from_range): Likewise.
4342 (vect_determine_precisions_from_users): Likewise.
4343 (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
4344 (vect_vect_recog_func_ptrs): Put over_widening first.
4346 (vect_pattern_recog): Call vect_determine_precisions.
4348 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
4350 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
4351 statements that have been replaced by further pattern statements.
4352 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
4354 2018-07-03 Richard Biener <rguenther@suse.de>
4356 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
4357 always set *dt. Dump vectype in vectype overload.
4358 * dumpfile.h (dump_gimple_expr): New function.
4359 (dump_gimple_expr_loc): Likewise.
4360 * dumpfile.c (dump_gimple_expr): New function.
4361 (dump_gimple_expr_loc): Likewise.
4363 2018-07-02 Jeff Law <law@redhat.com>
4365 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
4366 the H8/300, H8/300H and H8/S variants into a single pattern.
4367 (movhi_h8300, movqi_h8300hs): Similarly.
4368 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
4369 (QHI mode iterator): New.
4371 * config/h8300/h8300.md: Remove trailing whitespace.
4373 2018-07-02 Jim Wilson <jimw@sifive.com>
4375 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
4376 instead of emit_insn for interrupt returns.
4377 * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
4378 (riscv_sret, riscv_uret): Likewise.
4380 2018-07-02 David Malcolm <dmalcolm@redhat.com>
4382 * pretty-print.c (selftest::test_pp_format): Move save and restore
4383 of quotes to class auto_fix_quotes, and add an instance.
4384 * selftest.c: Include "intl.h".
4385 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
4386 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
4387 * selftest.h (selftest::auto_fix_quotes): New class.
4389 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
4391 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
4392 (aarch64_sve_prepare_conditional_op): Remove.
4393 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
4394 Allow aarch64_simd_reg_or_zero as select operand; remove
4395 the aarch64_sve_prepare_conditional_op call.
4396 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
4397 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
4398 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
4399 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
4400 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
4401 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
4402 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
4403 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
4404 and a splitters to match all of the *_any patterns.
4405 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
4407 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
4408 (SVE_COND_FP_BINARY_REV): Remove.
4409 (sve_int_op_rev, sve_fp_op_rev): New.
4410 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
4411 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
4412 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
4413 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
4414 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
4415 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
4416 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
4417 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
4418 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
4420 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
4421 Remove match_dup 1 from the inner unspec.
4422 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
4424 * config/aarch64/aarch64.md (movprfx): New attr.
4425 (length): Default movprfx to 8.
4426 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
4427 (*madd<SVE_I>, *msub<SVE_I): Likewise.
4428 (*<su>mul<SVE_I>3_highpart): Likewise.
4429 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
4430 (*v<ASHIFT><SVE_I>3): Likewise.
4431 (*<su><MAXMIN><SVE_I>3): Likewise.
4432 (*<su><MAXMIN><SVE_F>3): Likewise.
4433 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
4434 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
4435 (*div<SVE_F>4): Likewise.
4437 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
4439 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
4442 2018-07-02 Richard Biener <rguenther@suse.de>
4444 PR tree-optimization/86363
4445 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
4446 memset argument refers to a non-variable address.
4448 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
4450 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
4451 {PLUS,MINUS}_EXPR code to...
4452 (adjust_symbolic_bound): ...here,
4453 (combine_bound): ...here,
4454 (set_value_range_with_overflow): ...and here.
4456 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
4458 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
4460 (extract_range_from_abs_expr): ...here.
4462 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
4464 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
4465 -fno-omit-frame-pointer when not optimizing.
4467 2018-07-02 Martin Liska <mliska@suse.cz>
4470 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
4471 (propagate_nothrow): Likewise.
4473 2018-07-02 Martin Liska <mliska@suse.cz>
4476 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
4478 2018-07-02 David Malcolm <dmalcolm@redhat.com>
4480 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
4481 function in r262149, changing "loc" param from source_location to
4482 const dump_location_t &.
4483 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
4484 declaration, as above.
4486 2018-07-01 Paul Koning <ni1d@arrl.net>
4488 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
4489 -munit-asm, -mgnu-asm, -mdec-asm.
4490 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
4491 (pdp11_output_labelref): New.
4492 (pdp11_output_def): New.
4493 (pdp11_output_addr_vec_elt): New.
4494 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
4495 %# and %@ format codes.
4496 (pdp11_option_override): New.
4497 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
4498 (pdp11_output_ident): New.
4499 (pdp11_asm_named_section): New.
4500 (pdp11_asm_init_sections): New.
4501 (pdp11_file_start): New.
4502 (pdp11_file_end): New.
4503 (output_ascii): Use .ascii/.asciz for -mdec-asm.
4504 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
4505 %o, like %c but octal.
4506 (pdp11_option_override): New.
4507 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
4509 (DATA_SECTION_ASM_OP): Ditto.
4510 (READONLY_DATA_SECTION_ASM_OP): New.
4511 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
4512 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
4513 (ASM_OUTPUT_LABELREF): Ditto.
4514 (ASM_OUTPUT_DEF): Ditto.
4515 (ASM_OUTPUT_EXTERNAL): New.
4516 (ASM_OUTPUT_SOURCE_FILENAME): New.
4517 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
4518 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
4519 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
4520 %# and %@ format codes.
4521 * config/pdp11/pdp11.opt (mgnu-asm): New.
4522 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
4523 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
4524 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
4526 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
4528 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
4529 dereferencing path[] beyond its length.
4531 (debug_all_paths): New.
4532 (rewire_first_differing_edge): New.
4533 (adjust_paths_after_duplication): New.
4534 (duplicate_thread_path): Call adjust_paths_after_duplication.
4536 (thread_through_all_blocks): Add new argument to
4537 duplicate_thread_path.
4539 2018-06-30 Jim Wilson <jimw@sifive.com>
4541 * config/riscv/predicates.md (p2m1_shift_operand): New.
4542 (high_mask_shift_operand): New.
4543 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
4544 pattern using p2m1_shift_operand.
4545 (lshsi3_zero_extend_3+2): New combiner pattern using
4546 high_mask_shift_operand.
4548 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
4550 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
4552 (vect_recog_rotate_pattern): ...here.
4553 (vect_convert_input): Try to insert casts of invariants in the
4555 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
4556 preheader to be empty.
4558 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
4560 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
4561 vector type. If given, install it in the new statement's
4563 (vect_element_precision): New function.
4564 (vect_unpromoted_value): New struct.
4565 (vect_unpromoted_value::vect_unpromoted_value): New function.
4566 (vect_unpromoted_value::set_op): Likewise.
4567 (vect_look_through_possible_promotion): Likewise.
4568 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
4569 (vect_widened_op_tree, vect_convert_input): Likewise.
4570 (vect_convert_inputs, vect_convert_output): Likewise.
4571 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
4572 to handle the optional cast of the multiplication result and
4573 vect_widened_op_tree to detect the widened multiplication itself.
4574 Do not require the input and output of promotion casts to have
4575 the same sign, but base the signedness of the operation on the
4576 input rather than the result. If the pattern includes two
4577 promotions, check that those promotions have the same sign.
4578 Do not restrict the MULT_EXPR handling to a double-width result;
4579 handle quadruple-width results and wider. Use vect_convert_inputs
4580 to convert the inputs to the common type.
4581 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
4582 to handle the optional cast of the ABS result. Also allow a sign
4583 change or a sign extension between the ABS and MINUS.
4584 Use vect_widened_op_tree to detect the widened subtraction and use
4585 vect_convert_inputs to convert the inputs to the common type.
4586 (vect_handle_widen_op_by_const): Delete.
4587 (vect_recog_widen_op_pattern): New function.
4588 (vect_recog_widen_mult_pattern): Use it.
4589 (vect_recog_widen_shift_pattern): Likewise.
4590 (vect_recog_widen_sum_pattern): Use
4591 vect_look_through_possible_promotion to handle the promoted
4594 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
4596 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
4597 the containing gimple_seq *.
4598 * gimple-iterator.h (gsi_for_stmt): Declare it.
4599 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
4600 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
4601 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
4602 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
4603 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
4605 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
4607 (vect_mark_pattern_stmts): ...here. Handle cases in which the
4608 statement being replaced is part of an existing pattern
4609 definition sequence, inserting the new pattern statements before
4611 (vect_pattern_recog_1): Don't return a bool. If the statement
4612 is already part of a pattern, instead apply pattern matching
4613 to the pattern definition statements. Don't clear the
4614 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
4615 (vect_pattern_recog): Don't break after the first match;
4616 continue processing the pattern definition statements instead.
4617 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
4619 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
4621 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
4622 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
4623 (vect_recog_widen_sum_pattern): Use it.
4625 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
4627 * tree-vect-loop.c (vectorizable_reduction): Assert that the
4628 phi is not a pattern statement and has not been replaced by
4629 a pattern statement.
4630 * tree-vect-patterns.c (type_conversion_p): Don't check
4631 STMT_VINFO_IN_PATTERN_P.
4632 (vect_recog_vector_vector_shift_pattern): Likewise.
4633 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
4634 the pattern statement rather than the original statement; check
4635 directly for a WIDEN_MULT_EXPR here.
4636 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
4637 vect_is_simple_use to return the pattern statement rather
4638 than the original statement; use is_pattern_stmt_p to check
4639 for such a pattern statement.
4640 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
4641 to return the pattern statement rather than the original statement;
4642 don't do the same transformation here.
4643 (vect_is_simple_use): If the defining statement has been replaced
4644 by a pattern statement, return the pattern statement instead.
4645 Remove the corresponding (local) transformation from the vectype
4648 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
4650 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
4651 end and default to null.
4652 * tree-vect-loop.c (vect_create_epilog_for_reduction)
4653 (vectorizable_reduction): Update calls accordingly, dropping the
4654 gimple ** argument if the passed-back statement isn't needed.
4655 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
4656 (vect_recog_rotate_pattern): Likewise.
4657 (vect_recog_mask_conversion_pattern): Likewise.
4658 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
4659 (vect_mask_constant_operand_p): Likewise.
4660 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
4661 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
4662 (get_group_load_store_type, get_load_store_type): Likewise.
4663 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
4664 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
4665 (vectorizable_conversion, vectorizable_assignment): Likewise.
4666 (vectorizable_shift, vectorizable_operation): Likewise.
4667 (vectorizable_store, vect_is_simple_cond): Likewise.
4668 (vectorizable_condition, vectorizable_comparison): Likewise.
4669 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
4670 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
4671 and move it to the end. Cope with null def_stmt_outs.
4673 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
4675 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
4677 2018-06-29 Jeff Law <law@redhat.com>
4679 * config/v850/v850.c (v850_legitimate_address_p): Handle large
4680 displacements for TARGET_V850E2V3 and newer.
4681 (TARGET_LRA_P): Remove. Defaults to LRA now.
4682 * config/v850/v850.md (sign23byte_load): Remove.
4683 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
4684 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
4686 2018-06-29 Martin Liska <mliska@suse.cz>
4689 * coverage.c (coverage_init): Mangle full path name.
4690 * doc/invoke.texi: Document the change.
4691 * gcov-io.c (mangle_path): New.
4692 * gcov-io.h (mangle_path): Likewise.
4693 * gcov.c (mangle_name): Use mangle_path for path mangling.
4695 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4697 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
4698 if starting source register is not even.
4700 2018-06-29 Martin Liska <mliska@suse.cz>
4702 PR tree-optimization/86263
4703 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
4704 Make edge redirection.
4706 2018-06-29 David Malcolm <dmalcolm@redhat.com>
4708 * dumpfile.c (dump_loc): Add indentation based on scope depth.
4709 (dump_scope_depth): New variable.
4710 (get_dump_scope_depth): New function.
4711 (dump_begin_scope): New function.
4712 (dump_end_scope): New function.
4713 * dumpfile.h (get_dump_scope_depth): New declaration.
4714 (dump_begin_scope): New declaration.
4715 (dump_end_scope): New declaration.
4716 (class auto_dump_scope): New class.
4717 (AUTO_DUMP_SCOPE): New macro.
4718 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
4721 2018-06-29 Richard Biener <rguenther@suse.de>
4723 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
4724 compute_all_dependences succeeds.
4725 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
4726 exceed --param loop-max-datarefs-for-datadeps.
4728 2018-06-29 Jakub Jelinek <jakub@redhat.com>
4730 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
4732 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
4735 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
4736 alternative 0 in preferred_for_speed attribute.
4738 2018-06-28 Paul Koning <ni1d@arrl.net>
4740 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
4741 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
4742 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
4743 * config/pdp11/pdp11.md: Correct "length" attribute calculation
4744 for shift insn patterns.
4746 2018-06-28 David Malcolm <dmalcolm@redhat.com>
4748 * cgraph.c (cgraph_node::get_body): Replace assignments to
4749 "dump_file" with calls to set_dump_file.
4750 * dumpfile.c (alt_dump_file): Make static, and group with...
4751 (alt_flags): ...this definition.
4752 (dumps_are_enabled): New variable.
4753 (refresh_dumps_are_enabled): New function.
4754 (set_dump_file): New function.
4755 (set_alt_dump_file): New function.
4756 (gcc::dump_manager::dump_start): Replace assignments to
4757 "dump_file" and "alt_dump_file" with calls to set_dump_file and
4759 (gcc::dump_manager::dump_finish): Likewise.
4760 * dumpfile.h (alt_dump_file): Delete decl.
4761 (dumps_are_enabled): New variable decl.
4762 (set_dump_file): New function decl.
4763 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
4765 * tree-nested.c (lower_nested_functions): Replace assignments to
4766 "dump_file" with calls to set_dump_file.
4768 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
4770 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
4771 goto_locus of each outgoing edge of each basic block.
4773 2018-06-28 Richard Biener <rguenther@suse.de>
4775 * dwarf2out.c (decl_scope_table): Remove.
4776 (push_decl_scope): Likewise.
4777 (pop_decl_scope): Likewise.
4778 (gen_type_die_for_member): Do not call push/pop_decl_scope.
4779 (gen_struct_or_union_type_die): Likewise.
4780 (gen_tagged_type_die): Likewise.
4781 (dwarf2out_init): Do not initialize decl_scope_table.
4782 (dwarf2out_c_finalize): Do not free it.
4784 2018-06-28 Richard Biener <rguenther@suse.de>
4786 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
4787 deciding whether to not re-use a DIE.
4789 2018-06-28 Richard Biener <rguenther@suse.de>
4791 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
4792 DW_AT_abstract_origin attribute.
4794 2018-06-28 Martin Liska <mliska@suse.cz>
4796 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
4797 Use newly introduced constants.
4798 * tree-switch-conversion.h (struct jump_table_cluster):
4799 Define max_ratio_for_size and max_ratio_for_speed.
4801 2018-06-28 Martin Liska <mliska@suse.cz>
4803 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
4804 Add new checking assert to catch invalid state.
4805 (jump_table_cluster::can_be_handled): Handle single case
4807 (jump_table_cluster::is_beneficial): Bail out for such case.
4808 (bit_test_cluster::find_bit_tests):
4809 Add new checking assert to catch invalid state.
4810 (bit_test_cluster::can_be_handled): Handle single case
4812 (bit_test_cluster::is_beneficial): Bail out for such case.
4813 (switch_decision_tree::analyze_switch_statement):
4816 2018-06-28 Martin Liska <mliska@suse.cz>
4818 * common.opt: Introduce -completion option.
4819 * gcc.c (driver_handle_option): Handle it.
4820 (driver::main): Print completions if completion
4822 * opt-suggestions.c (option_proposer::get_completions):
4824 (option_proposer::suggest_completion): Likewise.
4825 (option_proposer::find_param_completions): Likewise.
4826 (verify_autocompletions): Likewise.
4827 (test_completion_valid_options): Likewise.
4828 (test_completion_valid_params): Likewise.
4829 (in_completion_p): Likewise.
4830 (empty_completion_p): Likewise.
4831 (test_completion_partial_match): Likewise.
4832 (test_completion_garbage): Likewise.
4833 (opt_proposer_c_tests): Likewise.
4834 * opt-suggestions.h: Declare new functions.
4835 * opts.c (common_handle_option): Handle OPT__completion_.
4836 * selftest-run-tests.c (selftest::run_tests): Add
4837 opt_proposer_c_tests.
4838 * selftest.c (assert_str_startswith): New.
4839 * selftest.h (assert_str_startswith): Likewise.
4840 (opt_proposer_c_tests): New.
4841 (ASSERT_STR_STARTSWITH): Likewise.
4843 2018-06-28 Martin Liska <mliska@suse.cz>
4845 * Makefile.in: Add opt-suggestions.o.
4846 * gcc-main.c: Include opt-suggestions.h.
4847 * gcc.c (driver::driver): Likewise.
4848 (driver::~driver): Remove m_option_suggestions.
4849 (driver::build_option_suggestions): Moved to option_proposer.
4850 (driver::suggest_option): Likewise.
4851 (driver::handle_unrecognized_options): Use option_proposer.
4852 * gcc.h (class driver): Add new memver m_option_proposer.
4853 * opt-suggestions.c: New file.
4854 * opt-suggestions.h: New file.
4856 2018-06-28 Martin Liska <mliska@suse.cz>
4858 * vec.h (class auto_string_vec): New (moved from auto_argvec).
4859 (auto_string_vec::~auto_string_vec): Likewise.
4861 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
4863 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
4864 prevent_decl_creation_for_types fields up and add reset_location field.
4865 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
4866 statement if id->reset_location is true.
4867 (copy_edges_for_bb): Do not set goto_locus on the new edges if
4868 id->reset_location is true.
4869 (copy_phis_for_bb): Force input_location on the arguments if
4870 id->reset_location is true.
4871 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
4872 is set on the function to be inlined.
4874 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
4876 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
4878 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
4880 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
4881 registers for Pmode.
4882 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
4883 hard registers for the clobbered pseudo.
4885 2018-06-27 Paul Koning <ni1d@arrl.net>
4887 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
4888 mutually exclusive options.
4889 * config/pdp11/constraints.md (h): New constraint.
4890 (O): Update definition to match shift code generation.
4891 (D): New constraint.
4892 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
4894 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
4896 (output_jump): Change arguments.
4897 (pdp11_fixed_cc_regs): New function.
4898 (pdp11_cc_mode): Ditto.
4899 (pdp11_expand_shift): Ditto.
4900 (pdp11_assemble_shift): Ditto.
4901 (pdp11_small_shift): Ditto.
4902 (pdp11_branch_cost): Remove.
4903 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
4905 (pdp11_register_move_cost): Update for CC registers.
4906 (pdp11_rtx_costs): Add case for LSHIFTRT.
4907 (pdp11_output_jump): Add CCNZ mode conditional branches.
4908 (notice_update_cc_on_set): Remove.
4909 (pdp11_cc_mode): New function.
4910 (simple_memory_operand): Correct pre/post decrement case.
4911 (no_side_effect_operand): New function.
4912 (pdp11_regno_reg_class): Add CC_REGS class.
4913 (pdp11_fixed_cc_regs): New function.
4914 (pdp11_small_shift): New function.
4915 (pdp11_expand_shift): New function to expand shift insns.
4916 (pdp11_assemble_shift): New function to output shifts.
4917 (pdp11_branch_cost): Remove.
4918 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
4919 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
4920 (WCHAR_TYPE): Ditto.
4921 (PTRDIFF_TYPE): Ditto.
4922 (ADJUST_INSN_LENGTH): New macro.
4923 (FIXED_REGISTERS): Add CC registers.
4924 (CALL_USED_REGISTERS): Ditto.
4926 (REG_CLASS_NAMES): Ditto.
4927 (REG_CLASS_CONTENTS): Ditto.
4928 (SELECT_CC_MODE): Use new function.
4929 (TARGET_FLAGS_REGNUM): New macro.
4930 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
4931 (cc0_reg_rtx): Remove.
4932 (CC_STATUS_MDEP): Remove.
4933 (CC_STATUS_MDEFP_INIT): Remove.
4934 (CC_IN_FPU): Remove.
4935 (NOTICE_UPDATE_CC): Remove.
4936 (REGISTER_NAMES): Add CC registers.
4937 (BRANCH_COST): Change to constant 1.
4938 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
4940 * config/pdp11/pdp11.opt (mbcopy): Remove.
4941 (mbcopy-builtin): Remove.
4942 (mbranch-cheap): Remove.
4943 (mbranch-expensive): Remove.
4944 * config/pdp11/predicates.md (expand_shift_operand): Update to
4945 match shift code generation.
4946 (ccnz_operator): New predicate.
4947 * doc/invoke.texi (PDP-11 Options): Remove deleted options
4948 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
4949 Remove non-existent option -mabshi, -mno-abshi. Document mutually
4951 * doc/md.texi (PDP-11): Document new D and h constraints. Update
4952 description of O constraint.
4954 2018-06-27 Jeff Law <law@redhat.com>
4955 Austin Law <austinklaw@gmail.com>
4957 * config/v850/v850.md (addsi3_set_flags): New pattern.
4958 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
4959 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
4960 (zero_extendhisi2_v850_set_flags): Likewise.
4961 (zero_extendqisi2_v850_set_flags): Likewise.
4962 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
4963 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
4964 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
4966 * config/v850/v850-protos.h (notice_update_cc): Remove.
4967 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
4968 (v850_print_operand): Handle 'D' and "d".
4969 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
4970 Add handling of arithmetic/logical operations compared against zero.
4971 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
4972 Do not look at v850_compare_op, instead get mode from last argument.
4973 (v850_gen_compare): Remove
4974 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
4975 after reload for prologue insns.
4976 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
4978 (construct_save_jarl): Likewise.
4979 (TARGET_FLAGS_REGNUM): Define.
4980 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
4981 (NOTICE_UPDATE_CC): Remove.
4982 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
4983 than cc0. Conditionalize on reload_completed.
4984 (cmpsi_insn, setfcc_insn): Likewise.
4985 (tst1 splitter): Turn into define_and_split which sets the flags
4987 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
4988 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
4989 (cstoresf4, cstoredf4): Clobber the flags.
4990 (cmpsi, cmpsf, cmpdf): Remove expanders.
4991 (setf_insn): Remove pattern.
4992 (addsi3): Turn into define_and_split which clobbers the flags after
4993 reload and a suitable pattern (addsi3_clobber_flags) for use after
4995 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
4996 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
4997 (ashrsi3, ashrsi3_v850e2): Likewise.
4998 (bins): Clobber the flags.
4999 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
5000 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
5001 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
5002 (call_value_internal_short, call_value_internal_long): Likewise.
5003 (callt_save_interrupt, callt_return_interrupt): Likewise.
5004 (save_interrupt, return_interrupt): Likewise.
5005 (callt_save_all_interrupt, save_all_interrupt): Likewise.
5006 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
5007 (restore_all_interrupt, _restore_all_interrupt): Likewise.
5008 (All FP comparisons): Only allow after reload has completed.
5010 (divh, divhu): Tweak output template.
5011 (branch_z_normal, branch_z_invert): Remove
5012 (branch_nz_normal, branch_nz_invert): Likewise.
5013 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
5015 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
5016 * config/v850/v850.c (notice_update_cc): Remove.
5017 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
5018 (CC_NO_CARRY): Likewise.
5019 (NOTICE_UPDATE_CC): Define to nothing.
5020 * config/v850/v850.md: Remove block comment on cc0 handling
5021 Remove "cc" attribute from all patterns. Remove cc_status handling
5022 from all patterns. Minor formatting fixes.
5024 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5026 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
5027 (cortex-a76.cortex-a55): Likewise.
5028 * config/aarch64/aarch64-tune.md: Regenerate.
5029 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
5030 cortex-a76.cortex-a55.
5032 2018-06-27 Jeff Law <law@redhat.com>
5034 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
5035 (MULTILIB_DIRNAMES): Similarly.
5037 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
5039 * gimple.h (gimple_return_retbnd): Delete.
5040 (gimple_return_set_retbnd): Likewise.
5041 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
5042 gimple_return_set_retbnd.
5043 * gimple-pretty-print.c (dump_gimple_return): Remove call to
5044 gimple_return_retbnd and adjust.
5045 * tree-inline.h (struct copy_body_data): Remove retbnd field.
5046 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
5047 Explicitly return NULL in a couple more cases. Move assertion
5048 on debug statements and remove unreachable code.
5049 (reset_debug_binding): Do not test id->retbnd.
5050 (expand_call_inline): Do not set it.
5052 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
5054 * configure.ac: Add --disable-gcov option.
5055 * configure: Regenerate.
5056 * Makefile.in: Honour @enable_gcov@.
5057 * doc/install.texi: Document --disable-gcov.
5059 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5061 * config/arm/arm-cpus.in (cortex-a76): New entry.
5062 (cortex-a76.cortex-a55): Likewise.
5063 * config/arm/arm-tables.opt: Regenerate.
5064 * config/arm/arm-tune.md: Likewise.
5065 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
5066 * doc/invoke.texi (ARM Options): Document cortex-a76 and
5067 cortex-a76.cortex-a55.
5069 2018-06-27 Tamar Christina <tamar.christina@arm.com>
5072 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
5074 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
5076 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
5078 (EPILOGUE_USES): Likewise.
5080 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
5082 * tree-inline.c (remap_location): New function extracted from...
5083 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
5084 (copy_phis_for_bb): ...here. Call remap_location.
5085 (copy_cfg_body): Adjust call to copy_edges_for_bb.
5087 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
5089 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
5090 unaligned vsx for 16B memset.
5092 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
5095 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
5096 ieee128_float_type_node to long_double_type_node unless
5097 TARGET_LONG_DOUBLE_128 is set.
5099 2018-06-26 David Malcolm <dmalcolm@redhat.com>
5101 * cfgloop.c (get_loop_location): Convert return type from
5102 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
5103 by implicit construction from rtx_insn *, and using
5104 dump_user_location_t::from_function_decl for the fallback case.
5105 * cfgloop.h (get_loop_location): Convert return type from
5106 location_t to dump_user_location_t.
5107 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
5108 dump_printf_loc to pass in a dump_location_t rather than a
5109 location_t, via the gimple stmt.
5110 * coverage.c (get_coverage_counts): Update calls to
5111 dump_printf_loc to pass in dump_location_t rather than a
5113 * doc/optinfo.texi (Dump types): Convert example of
5114 dump_printf_loc from taking "locus" to taking "insn". Update
5115 description of the "_loc" calls to cover dump_location_t.
5116 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
5118 (dump_user_location_t::dump_user_location_t): New constructors,
5119 from gimple *stmt and rtx_insn *.
5120 (dump_user_location_t::from_function_decl): New function.
5121 (dump_loc): Make static.
5122 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
5123 const dump_location_t &.
5124 (dump_generic_expr_loc): Delete.
5125 (dump_printf_loc): Convert param "loc" from location_t to
5126 const dump_location_t &.
5127 (selftest::test_impl_location): New function.
5128 (selftest::dumpfile_c_tests): New function.
5129 * dumpfile.h: Include "profile-count.h".
5130 (class dump_user_location_t): New class.
5131 (struct dump_impl_location_t): New struct.
5132 (class dump_location_t): New class.
5133 (dump_printf_loc): Convert 2nd param from source_location to
5134 const dump_location_t &.
5135 (dump_generic_expr_loc): Delete.
5136 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
5137 const dump_location_t &.
5138 * gimple-fold.c (fold_gimple_assign): Update call to
5139 dump_printf_loc to pass in a dump_location_t rather than a
5140 location_t, via the gimple stmt.
5141 (gimple_fold_call): Likewise.
5142 * gimple-loop-interchange.cc
5143 (loop_cand::analyze_iloop_reduction_var): Update for change to
5144 check_reduction_path.
5145 (tree_loop_interchange::interchange): Update for change to
5147 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
5148 change in return-type of find_loop_location.
5149 (graphite_regenerate_ast_isl): Likewise.
5150 * graphite-optimize-isl.c (optimize_isl): Likewise.
5151 * graphite.c (graphite_transform_loops): Likewise.
5152 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
5153 pass in a dump_location_t rather than a location_t, via the
5155 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
5156 * ipa.c (walk_polymorphic_call_targets): Likewise.
5157 * loop-unroll.c (report_unroll): Convert "locus" param from
5158 location_t to dump_location_t.
5159 (decide_unrolling): Update for change to get_loop_location's
5161 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
5162 location_t to dump_user_location_t.
5163 (grid_find_single_omp_among_assignments_1): Updates calls to
5164 dump_printf_loc to pass in a dump_location_t rather than a
5165 location_t, via the gimple stmt.
5166 (grid_parallel_clauses_gridifiable): Convert "tloc" from
5167 location_t to dump_location_t. Updates calls to dump_printf_loc
5168 to pass in a dump_location_t rather than a location_t, via the
5170 (grid_inner_loop_gridifiable_p): Likewise.
5171 (grid_dist_follows_simple_pattern): Likewise.
5172 (grid_gfor_follows_tiling_pattern): Likewise.
5173 (grid_target_follows_gridifiable_pattern): Likewise.
5174 (grid_attempt_target_gridification): Convert initialization
5175 of local "grid" from memset to zero-initialization; FIXME: does
5176 this require C++11? Update call to dump_printf_loc to pass in a
5177 optinfo_location rather than a location_t, via the gimple stmt.
5178 * profile.c (read_profile_edge_counts): Updates call to
5179 dump_printf_loc to pass in a dump_location_t rather than a
5181 (compute_branch_probabilities): Likewise.
5182 * selftest-run-tests.c (selftest::run_tests): Call
5184 * selftest.h (dumpfile_c_tests): New decl.
5185 * tree-loop-distribution.c (pass_loop_distribution::execute):
5186 Update for change in return type of find_loop_location.
5187 * tree-parloops.c (parallelize_loops): Likewise.
5188 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
5189 "locus" from location_t to dump_user_location_t.
5190 (canonicalize_loop_induction_variables): Likewise.
5191 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
5192 for change in return type of find_loop_location.
5193 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
5194 to dump_printf_loc to pass in a dump_location_t rather than a
5195 location_t, via the stmt.
5196 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
5198 * tree-vect-loop-manip.c (find_loop_location): Convert return
5199 type from source_location to dump_user_location_t.
5200 (vect_do_peeling): Update for above change.
5201 (vect_loop_versioning): Update for change in type of
5203 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
5204 from location_t to dump_user_location_t.
5205 (vect_estimate_min_profitable_iters): Update for change in type
5207 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
5208 location_t to dump_location_t.
5209 (vect_slp_bb): Update for change in type of vect_location.
5210 * tree-vectorizer.c (vect_location): Convert from source_location
5211 to dump_user_location_t.
5212 (try_vectorize_loop_1): Update for change in vect_location's type.
5213 (vectorize_loops): Likewise.
5214 (increase_alignment): Likewise.
5215 * tree-vectorizer.h (vect_location): Convert from source_location
5216 to dump_user_location_t.
5217 (find_loop_location): Convert return type from source_location to
5218 dump_user_location_t.
5219 (check_reduction_path): Convert 1st param from location_t to
5220 dump_user_location_t.
5221 * value-prof.c (check_counter): Update call to dump_printf_loc to
5222 pass in a dump_user_location_t rather than a location_t; update
5223 call to error_at for change in type of "locus".
5224 (check_ic_target): Update call to dump_printf_loc to
5225 pass in a dump_user_location_t rather than a location_t, via the
5228 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
5230 * config/s390/s390.h (enum processor_flags): Do not use
5231 default tune parameter when -march was specified.
5233 2018-06-26 Jakub Jelinek <jakub@redhat.com>
5236 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
5237 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
5240 2018-06-26 Richard Biener <rguenther@suse.de>
5242 PR tree-optimization/86287
5244 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
5245 (vect_analyze_loop): Initialize n_stmts.
5247 2018-06-26 Richard Biener <rguenther@suse.de>
5250 * fold-const.c (fold_convertible_p): Pointer extension
5253 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
5256 * dwarf2out.c (loc_list_has_views): Adjust comments.
5257 (dw_loc_list): Split single cross-partition range with
5260 2018-06-25 Jeff Law <law@redhat.com>
5262 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
5263 on -mbig-switch by default.
5265 * config/v850/predicates.md (const_float_1_operand): Fix match_code
5267 (const_float_0_operand): Remove unused predicate.
5268 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
5269 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
5270 (recipsf2): New expander. Original pattern now called
5272 (recipdf2, recipdf2_insn): Similarly.
5273 (rsqrtsf2, rsqrtsf2_insn): Similarly
5274 (rsqrtdf2, rsqrtdf2_insn): Similarly
5276 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
5278 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
5279 Simplify logic for FreeBSD (twice).
5281 2018-06-25 Martin Sebor <msebor@redhat.com>
5283 PR tree-optimization/86204
5284 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
5285 a strnlen result if it's less than the length of the string.
5287 2018-06-25 Martin Sebor <msebor@redhat.com>
5289 PR tree-optimization/85700
5290 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
5291 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
5292 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
5294 2018-06-25 Martin Sebor <msebor@redhat.com>
5296 * doc/extend.texi (Zero-length arrays): Update and clarify.
5298 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
5300 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
5301 added IEEE/IBM long double multilib support on PowerPC little
5302 endian Linux systems.
5303 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
5304 (MULTILIB_DEFAULTS): Likewise.
5305 * config/rs6000/rs6000.c (rs6000_option_override_internal):
5307 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
5308 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
5309 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
5311 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
5314 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
5315 (REORDER_45): Likewise.
5317 2018-06-25 Jeff Law <law@redhat.com>
5319 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
5320 dividend to 32 bits. Adjust length.
5321 (udivmodhi4): Cleanup output template. Fix length.
5323 2018-06-25 Carl Love <cel@us.ibm.com>
5325 * config/rs6000/vsx.md: Change word selector to prefered location.
5327 2018-06-25 Richard Biener <rguenther@suse.de>
5329 PR tree-optimization/86304
5330 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
5331 epilogue-if-converted loops as well.
5333 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
5335 * lto-section-out.c (lto_begin_section): Do not print section
5336 name for noaddr and unnumbered dumps.
5338 2018-06-25 Richard Biener <rguenther@suse.de>
5340 * tree-vectorizer.h (struct vec_info_shared): New structure
5341 with parts split out from struct vec_info and loop_nest from
5342 struct _loop_vec_info.
5343 (struct vec_info): Adjust accordingly.
5344 (struct _loop_vec_info): Likewise.
5345 (LOOP_VINFO_LOOP_NEST): Adjust.
5346 (LOOP_VINFO_DATAREFS): Likewise.
5347 (LOOP_VINFO_DDRS): Likewise.
5348 (struct _bb_vec_info): Likewise.
5349 (BB_VINFO_DATAREFS): Likewise.
5350 (BB_VINFO_DDRS): Likewise.
5351 (struct _stmt_vec_info): Add dr_aux member.
5352 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
5353 (DR_MISALIGNMENT_UNINITIALIZED): New.
5354 (set_dr_misalignment): Adjust.
5355 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
5356 (vect_analyze_loop): Adjust prototype.
5357 (vect_analyze_loop_form): Likewise.
5358 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
5359 Compute dependences lazily.
5360 (vect_record_base_alignments): Use shared datarefs/ddrs.
5361 (vect_verify_datarefs_alignment): Likewise.
5362 (vect_analyze_data_refs_alignment): Likewise.
5363 (vect_analyze_data_ref_accesses): Likewise.
5364 (vect_analyze_data_refs): Likewise.
5365 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
5366 constructor parameter for shared part.
5367 (vect_analyze_loop_form): Pass in shared part and adjust.
5368 (vect_analyze_loop_2): Pass in storage for the number of
5369 stmts. Move loop nest finding to the caller. Compute
5371 (vect_analyze_loop): Pass in shared part.
5372 (vect_transform_loop): Verify shared datarefs are unchanged.
5373 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
5374 constructor parameter for shared part.
5375 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
5376 (vect_slp_bb): Verify shared datarefs are unchanged before
5378 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
5380 (new_stmt_vec_info): Initialize DR_AUX misalignment to
5381 DR_MISALIGNMENT_UNINITIALIZED.
5382 * tree-vectorizer.c (vec_info::vec_info): Add constructor
5383 parameter for shared part.
5384 (vec_info::~vec_info): Adjust.
5385 (vec_info_shared::vec_info_shared): New.
5386 (vec_info_shared::~vec_info_shared): Likewise.
5387 (vec_info_shared::save_datarefs): Likewise.
5388 (vec_info_shared::check_datarefs): Likewise.
5389 (try_vectorize_loop_1): Construct shared part live for analyses
5390 of a single loop for multiple vector sizes.
5391 * tree-parloops.c (gather_scalar_reductions): Adjust.
5393 2018-06-25 Richard Biener <rguenther@suse.de>
5395 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
5396 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
5397 (vect_analyze_data_refs): Remove similar code from here and
5398 simplify accordingly.
5400 2018-06-25 Richard Biener <rguenther@suse.de>
5402 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
5403 for reverse storage order accesses rather than asserting
5404 they cannot happen here.
5406 2018-06-25 Tom de Vries <tdevries@suse.de>
5409 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
5410 Use data16 instead of .byte for insn prefix.
5412 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
5415 * parser.c (make_char_string_pack): Pass this literal chars
5416 through cpp_interpret_string.
5417 (cp_parser_userdef_numeric_literal): Check the result of
5418 make_char_string_pack.
5420 2018-06-24 Maya Rashish <coypu@sdf.org>
5422 * ginclude/stddef.h: Simplify conditions around avoiding
5423 re-definition of __size_t.
5425 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
5427 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
5428 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
5430 2018-06-22 Maya Rashish <coypu@sdf.org>
5432 * doc/invoke.texi (mno-fancy-math-387): Update for changes
5433 made to OpenBSD and NetBSD through the years.
5435 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
5437 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
5438 behavior of vec_pack (vector double, vector double) to match
5439 behavior of vec_float2 (vector double, vector double).
5441 2018-06-22 Olivier Hainque <hainque@adacore.com>
5443 * gimplify.c (gimplify_function_tree): Prevent creation
5444 of a trampoline for the address of the current function
5445 passed to entry/exit instrumentation hooks.
5447 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
5450 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
5453 2018-06-22 Martin Liska <mliska@suse.cz>
5455 PR tree-optimization/86263
5456 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
5457 Bail out if is_enabled is false.
5458 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
5460 (jump_table_cluster::is_enabled): New function.
5462 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
5464 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
5465 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
5466 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
5467 (lto_input_ts_binfo_tree_pointers): Likewise.
5468 * tree-streamer-out.c (streamer_write_tree_bitfields,
5469 write_ts_binfo_tree_pointers): Likewise.
5470 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
5472 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
5474 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
5476 2018-06-22 Martin Liska <mliska@suse.cz>
5478 * symbol-summary.h (get): Make it pure and inline move
5479 functionality from ::get function.
5480 (get): Remove and inline into ::get and ::get_create.
5481 (get_create): Move code from ::get function.
5483 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5486 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
5487 -x assembler-with-cpp.
5489 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5491 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
5492 _FILE_OFFSET_BITS=64 for C++.
5494 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
5496 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
5497 conversion insn that shows up when pr85657-3.c is compiled using
5498 IEEE 128-bit long double.
5499 (neg<mode>2_internal): Use the correct mode to check whether the
5500 mode is IBM extended.
5501 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
5502 multiply and divide external functions from being created more
5505 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
5507 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
5509 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
5510 the edge can be forwarded.
5511 (cfg_layout_merge_blocks): Likewise.
5513 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
5515 * except.c (finish_eh_generation): Commit edge insertions only after
5516 the EH edges have been redirected from post-landing to landing pads.
5518 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
5520 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
5521 create_tmp_var_for to create the FRAME decl.
5522 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
5524 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
5526 * tree-inline.c (copy_edges_for_bb): Minor tweak.
5527 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
5528 debug statement when resetting its value.
5529 (expand_call_inline): Copy the locus of the call onto the assignment
5530 of the return value, if any. Use local variable in more cases.
5532 2018-06-21 Martin Liska <mliska@suse.cz>
5534 * ipa-pure-const.c (propagate_nothrow): Use
5535 funct_state_summaries->get.
5536 (dump_malloc_lattice): Likewise.
5537 (propagate_malloc): Likewise.
5539 2018-06-21 Richard Biener <rguenther@suse.de>
5541 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
5542 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
5543 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
5545 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
5546 BLOCK_ABSTRACT_ORIGIN unconditionally.
5548 2018-06-21 David Malcolm <dmalcolm@redhat.com>
5550 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
5552 * ipa-reference.c (ipa_reference_c_finalize): Delete
5553 ipa_ref_opt_sum_summaries and set it to NULL.
5555 2018-06-21 Tom de Vries <tdevries@suse.de>
5557 PR tree-optimization/85859
5558 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
5559 test with comment from bb_no_side_effects_p.
5561 2018-06-21 Richard Biener <rguenther@suse.de>
5563 PR tree-optimization/86232
5564 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
5565 max for constant niter.
5567 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
5569 * config/aarch64/aarch64-simd.md
5570 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
5572 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
5574 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
5575 Make opernads of the unspec commutative.
5577 2018-06-21 Richard Biener <rguenther@suse.de>
5579 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
5580 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
5581 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
5582 (vect_analyze_data_ref_dependence): Re-order checks to deal with
5584 (vect_record_base_alignments): Do not record base alignment
5585 for gathers or scatters.
5586 (vect_compute_data_ref_alignment): Drop return value that is always
5587 true. Bail out early for gathers or scatters.
5588 (vect_enhance_data_refs_alignment): Bail out early for gathers
5590 (vect_find_same_alignment_drs): Likewise.
5591 (vect_analyze_data_refs_alignment): Remove dead code.
5592 (vect_slp_analyze_and_verify_node_alignment): Likewise.
5593 (vect_analyze_data_refs): For possible gathers or scatters do
5594 not create an alternate DR, just check their possible validity
5595 and mark them. Adjust DECL_NONALIASED handling to not rely
5597 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
5598 update inits of gathers or scatters.
5599 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
5600 Also copy gather/scatter flag to pattern vinfo.
5602 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
5604 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
5605 behavior of vec_packsu (vector unsigned long long, vector unsigned
5606 long long) to match behavior of vec_packs with same signature.
5608 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
5609 Thomas Schwinge <thomas@codesourcery.com>
5610 Cesar Philippidis <cesar@codesourcery.com>
5612 * gimplify.c (gimplify_scan_omp_clauses): Add support for
5613 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
5614 (gimplify_adjust_omp_clauses): Likewise.
5615 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
5616 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
5617 (gimplify_omp_target_update): Update handling of acc update and
5619 * omp-low.c (install_var_field): Remove unused parameter
5620 base_pointers_restrict.
5621 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
5622 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
5624 (omp_target_base_pointers_restrict_p): Delete.
5625 (scan_omp_target): Update call to scan_sharing_clauses.
5626 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
5628 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
5629 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
5630 (convert_local_omp_clauses): Likewise.
5631 * tree-pretty-print.c (dump_omp_clause): Likewise.
5632 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
5634 (omp_clause_code_name): Likewise.
5636 2018-06-20 Jakub Jelinek <jakub@redhat.com>
5639 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
5642 PR tree-optimization/86231
5643 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
5644 anti-range don't overwrite *vr0min before using it to compute *vr0max.
5646 2018-06-20 Tom de Vries <tdevries@suse.de>
5648 PR tree-optimization/86097
5649 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
5650 iv type if signedness of iv type is not the same as that of *nit.
5652 2018-06-20 Jakub Jelinek <jakub@redhat.com>
5654 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
5655 EDGE_EH edges, verify they are all EDGE_EH.
5657 2018-06-20 Maya Rashish <coypu@sdf.org>
5659 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
5661 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5663 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
5664 * config/aarch64/aarch64.c (xgene1_tunings): Add
5665 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
5666 (aarch64_mode_valid_for_sched_fusion_p):
5667 Allow 16-byte modes.
5668 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
5669 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
5671 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
5673 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
5674 * config/aarch64/iterators.md (VQ2): New mode iterator.
5676 2018-06-20 Martin Liska <mliska@suse.cz>
5678 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
5679 Change default ratio from 10 to 8.
5681 2018-06-20 Martin Liska <mliska@suse.cz>
5683 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
5685 (bit_test_cluster::find_bit_tests): Likewise.
5686 (switch_decision_tree::analyze_switch_statement): Find clusters.
5687 * tree-switch-conversion.h (struct jump_table_cluster): Document
5690 2018-06-20 Martin Liska <mliska@suse.cz>
5692 * tree-switch-conversion.c (switch_conversion::collect):
5693 Record m_uniq property.
5694 (switch_conversion::expand): Bail out for special conditions.
5695 (group_cluster::~group_cluster): New.
5696 (group_cluster::group_cluster): Likewise.
5697 (group_cluster::dump): Likewise.
5698 (jump_table_cluster::emit): New.
5699 (switch_decision_tree::fix_phi_operands_for_edges): New.
5700 (struct case_node): Remove struct.
5701 (jump_table_cluster::can_be_handled): New.
5702 (case_values_threshold): Moved to header.
5703 (reset_out_edges_aux): Likewise.
5704 (jump_table_cluster::is_beneficial): New.
5705 (bit_test_cluster::can_be_handled): Likewise.
5706 (add_case_node): Remove.
5707 (bit_test_cluster::is_beneficial): New.
5708 (case_bit_test::cmp): New.
5709 (bit_test_cluster::emit): New.
5710 (expand_switch_as_decision_tree_p): Remove.
5711 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
5712 (fix_phi_operands_for_edge): Likewise.
5713 (switch_decision_tree::analyze_switch_statement): New.
5714 (compute_cases_per_edge): Move ...
5715 (switch_decision_tree::compute_cases_per_edge): ... here.
5716 (try_switch_expansion): Likewise.
5717 (switch_decision_tree::try_switch_expansion): Likewise.
5718 (record_phi_operand_mapping): Likewise.
5719 (switch_decision_tree::record_phi_operand_mapping): Likewise.
5720 (emit_case_decision_tree): Likewise.
5721 (switch_decision_tree::emit): Likewise.
5722 (balance_case_nodes): Likewise.
5723 (switch_decision_tree::balance_case_nodes): Likewise.
5724 (dump_case_nodes): Likewise.
5725 (switch_decision_tree::dump_case_nodes): Likewise.
5726 (emit_jump): Likewise.
5727 (switch_decision_tree::emit_jump): Likewise.
5728 (emit_cmp_and_jump_insns): Likewise.
5729 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
5730 (emit_case_nodes): Likewise.
5731 (switch_decision_tree::emit_case_nodes): Likewise.
5732 (conditional_probability): Remove.
5733 * tree-switch-conversion.h (enum cluster_type): New.
5735 (struct cluster): Likewise.
5736 (cluster::cluster): Likewise.
5737 (struct simple_cluster): Likewise.
5738 (simple_cluster::simple_cluster): Likewise.
5739 (struct group_cluster): Likewise.
5740 (struct jump_table_cluster): Likewise.
5741 (struct bit_test_cluster): Likewise.
5742 (struct min_cluster_item): Likewise.
5743 (struct case_tree_node): Likewise.
5744 (case_tree_node::case_tree_node): Likewise.
5745 (jump_table_cluster::case_values_threshold): Likewise.
5746 (struct case_bit_test): Likewise.
5747 (struct switch_decision_tree): Likewise.
5748 (struct switch_conversion): Likewise.
5749 (switch_decision_tree::reset_out_edges_aux): Likewise.
5751 2018-06-20 Martin Liska <mliska@suse.cz>
5753 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
5754 (hoist_edge_and_branch_if_true): Likewise.
5755 (expand_switch_using_bit_tests_p): Likewise.
5756 (struct case_bit_test): Likewise.
5757 (case_bit_test_cmp): Likewise.
5758 (emit_case_bit_tests): Likewise.
5759 (switch_conversion::switch_conversion): New class.
5760 (struct switch_conv_info): Remove old struct.
5761 (collect_switch_conv_info): More to ...
5762 (switch_conversion::collect): ... this.
5763 (check_range): Likewise.
5764 (switch_conversion::check_range): Likewise.
5765 (check_all_empty_except_final): Likewise.
5766 (switch_conversion::check_all_empty_except_final): Likewise.
5767 (check_final_bb): Likewise.
5768 (switch_conversion::check_final_bb): Likewise.
5769 (create_temp_arrays): Likewise.
5770 (switch_conversion::create_temp_arrays): Likewise.
5771 (free_temp_arrays): Likewise.
5772 (gather_default_values): Likewise.
5773 (switch_conversion::gather_default_values): Likewise.
5774 (build_constructors): Likewise.
5775 (switch_conversion::build_constructors): Likewise.
5776 (constructor_contains_same_values_p): Likewise.
5777 (switch_conversion::contains_same_values_p): Likewise.
5778 (array_value_type): Likewise.
5779 (switch_conversion::array_value_type): Likewise.
5780 (build_one_array): Likewise.
5781 (switch_conversion::build_one_array): Likewise.
5782 (build_arrays): Likewise.
5783 (switch_conversion::build_arrays): Likewise.
5784 (gen_def_assigns): Likewise.
5785 (switch_conversion::gen_def_assigns): Likewise.
5786 (prune_bbs): Likewise.
5787 (switch_conversion::prune_bbs): Likewise.
5788 (fix_phi_nodes): Likewise.
5789 (switch_conversion::fix_phi_nodes): Likewise.
5790 (gen_inbound_check): Likewise.
5791 (switch_conversion::gen_inbound_check): Likewise.
5792 (process_switch): Use the newly created class.
5793 (switch_conversion::expand): New.
5794 (switch_conversion::~switch_conversion): New.
5795 * tree-switch-conversion.h: New file.
5797 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
5799 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
5800 tree-vect-patterns.c.
5801 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
5802 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
5803 (vect_recog_sad_pattern): Likewise.
5804 (vect_recog_widen_sum_pattern): Likewise.
5805 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
5806 (vect_recog_widen_shift_pattern): Remove the type_in argument.
5807 (vect_recog_rotate_pattern): Likewise.
5808 (vect_recog_mult_pattern): Likewise.
5809 (vect_recog_vector_vector_shift_pattern): Likewise.
5810 (vect_recog_divmod_pattern): Likewise.
5811 (vect_recog_mixed_size_cond_pattern): Likewise.
5812 (vect_recog_bool_pattern): Likewise.
5813 (vect_recog_mask_conversion_pattern): Likewise.
5814 (vect_try_gather_scatter_pattern): Likewise.
5815 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
5816 (vect_recog_over_widening_pattern): Likewise.
5817 (vect_recog_gather_scatter_pattern): Likewise.
5818 (vect_recog_func_ptr): Move from tree-vectorizer.h
5819 (vect_vect_recog_func_ptrs): Move further down the file.
5820 (vect_recog_func): Likewise. Remove the third argument.
5821 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
5822 (vect_pattern_recog_1): Expect the pattern function to do any
5823 necessary target tests. Also expect it to provide a vector type.
5824 Remove the type_in handling.
5826 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
5828 * tree-vect-patterns.c (vect_pattern_detected): New function.
5829 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
5830 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
5831 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
5832 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
5833 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
5834 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
5835 (vect_recog_mask_conversion_pattern)
5836 (vect_try_gather_scatter_pattern): Likewise.
5838 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
5840 * tree-vect-patterns.c (vect_get_internal_def): New function.
5841 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
5842 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
5843 (search_type_for_mask_1): Use it.
5845 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
5847 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
5848 redundant WIDEN_SUM_EXPR handling.
5849 (vect_recog_sad_pattern): Likewise.
5851 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
5853 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
5854 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
5855 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
5856 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
5857 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
5859 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
5861 * tree-vect-stmts.c (vectorizable_call): Make sure that we
5862 use the stmt_vec_info of the original bb statement for the
5863 new zero assignment, even if the call is part of a pattern.
5865 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
5867 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
5868 that the sequence is attached to the original statement rather
5869 than the pattern statement.
5870 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
5871 PATTERN_DEF_SEQ from the original statement rather than
5872 the main pattern statement.
5873 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
5874 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
5875 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
5877 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
5879 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
5880 definition statements before the early exit for statements that aren't
5882 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
5884 (vect_transform_loop): ...here. Process pattern definition
5885 statements without first checking whether the main pattern
5886 statement is live or relevant.
5888 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
5890 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
5891 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
5893 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
5895 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
5896 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
5897 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
5898 (expand_block_compare): Change select_block_compare_mode call.
5899 (expand_strncmp_align_check): Use new functions, fix comment.
5900 (emit_final_str_compare_gpr): New function.
5901 (expand_strn_compare): Refactor and clean up code.
5902 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
5904 2018-06-19 Tony Reix <tony.reix@atos.com>
5905 Damien Bergamini <damien.bergamini@atos.com>
5906 David Edelsohn <dje.gcc@gmail.com>
5908 * collect2.c (static_obj): New variable.
5909 (static_libs): New variable.
5910 (is_in_list): Uncomment declaration.
5911 (main): Track AIX libraries linked statically.
5912 (is_in_list): Uncomment definition.
5913 (scan_prog_file): Don't add AIX shared libraries initializer
5914 to constructor list if linking statically.
5916 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
5918 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
5920 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
5922 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
5924 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
5927 2018-06-19 Martin Liska <mliska@suse.cz>
5929 * config/i386/i386.c (ix86_can_inline_p): Do not use
5930 ipa_fn_summaries::get_create.
5931 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
5933 (devirtualization_time_bonus): Likewise.
5934 (ipcp_propagate_stage): Likewise.
5935 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
5936 (edge_set_predicate): Likewise.
5937 (evaluate_conditions_for_known_args): Likewise.
5938 (evaluate_properties_for_edge): Likewise.
5939 (ipa_call_summary::reset): Tranform to ...
5940 (ipa_call_summary::~ipa_call_summary): ... this.
5941 (ipa_fn_summary::reset): Transform to ...
5942 (ipa_fn_summary::~ipa_fn_summary): ... this.
5943 (ipa_fn_summary_t::remove): Rename to ...
5944 (ipa_fn_summary_t::remove_callees): ... this.
5945 (ipa_fn_summary_t::duplicate): Use placement new
5946 instead of memory copy.
5947 (ipa_call_summary_t::duplicate): Likewise.
5948 (ipa_call_summary_t::remove): Remove.
5949 (dump_ipa_call_summary): Change get_create to get.
5950 (ipa_dump_fn_summary): Dump only when summary exists.
5951 (analyze_function_body): Use symbol_summary::get instead
5953 (compute_fn_summary): Likewise.
5954 (estimate_edge_devirt_benefit): Likewise.
5955 (estimate_edge_size_and_time): Likewise.
5956 (inline_update_callee_summaries): Likewise.
5957 (remap_edge_change_prob): Likewise.
5958 (remap_edge_summaries): Likewise.
5959 (ipa_merge_fn_summary_after_inlining): Likewise.
5960 (write_ipa_call_summary): Likewise.
5961 (ipa_fn_summary_write): Likewise.
5962 (ipa_free_fn_summary): Likewise.
5963 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
5964 (struct ipa_call_summary): Likewise.
5965 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
5967 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
5968 (estimate_size_after_inlining): Likewise.
5969 (estimate_growth): Likewise.
5970 (growth_likely_positive): Likewise.
5971 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
5972 (inline_call): Likewise.
5973 * ipa-inline.c (caller_growth_limits): Likewise.
5974 (can_inline_edge_p): Likewise.
5975 (can_inline_edge_by_limits_p): Likewise.
5976 (compute_uninlined_call_time): Likewise.
5977 (compute_inlined_call_time): Likewise.
5978 (want_inline_small_function_p): Likewise.
5979 (edge_badness): Likewise.
5980 (update_caller_keys): Likewise.
5981 (update_callee_keys): Likewise.
5982 (inline_small_functions): Likewise.
5983 (inline_to_all_callers_1): Likewise.
5984 (dump_overall_stats): Likewise.
5985 (early_inline_small_functions): Likewise.
5986 (early_inliner): Likewise.
5987 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
5988 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
5989 * ipa-pure-const.c (malloc_candidate_p): Likewise.
5990 * ipa-split.c (execute_split_functions): Likewise.
5991 * symbol-summary.h: Likewise.
5992 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
5994 2018-06-19 Richard Biener <rguenther@suse.de>
5996 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
5997 (vectorize_loops): ... here. Fix dbgcnt handling.
5998 (try_vectorize_loop): Wrap try_vectorize_loop_1.
6000 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
6003 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
6004 ieee128 argument takes up only one (vector) register, not two (floating
6007 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
6009 * gimplify.c (gimplify_init_constructor): Really never clear for an
6010 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
6012 2018-06-19 Richard Biener <rguenther@suse.de>
6014 PR tree-optimization/86179
6015 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
6016 after failed recognition.
6018 2018-06-18 Martin Sebor <msebor@redhat.com>
6021 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
6022 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
6023 Handle integer subtraction.
6024 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
6025 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
6027 2018-06-18 David Malcolm <dmalcolm@redhat.com>
6029 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
6030 param from rtx to rtx_insn *.
6031 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
6033 (frv_ifcvt_modify_insn): Likwise.
6034 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
6035 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
6036 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
6037 as_a <rtx_insn *> cast to local "unprotected_region" once
6038 it's been established that it's not NULL or pc_rtx.
6039 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
6040 param "sethi" from rtx to rtx_insn *.
6041 (nds32_group_float_insns): Likewise for param "insn".
6042 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
6044 (vax_output_int_subtract): Likewise.
6045 * config/vax/vax.c (vax_output_int_add): Likewise for param
6047 (vax_output_int_subtract): Likewise.
6048 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
6049 (emit_pattern_after): Likewise for param "after".
6050 (emit_insn_after): Likewise.
6051 (emit_jump_insn_after): Likewise.
6052 (emit_call_insn_after): Likewise.
6053 (emit_debug_insn_after): Likewise.
6054 (emit_pattern_before): Likewise for param "before".
6055 (emit_insn_before): Likewise.
6056 (emit_jump_insn_before): Likewise.
6057 * final.c (get_insn_template): Likewise for param "insn", removing
6059 * output.h (get_insn_template): Likewise for 2nd param.
6060 * rtl.h (emit_insn_before): Likewise.
6061 (emit_jump_insn_before): Likewise.
6062 (emit_debug_insn_before_noloc): Likewise.
6063 (emit_insn_after): Likewise.
6064 (emit_jump_insn_after): Likewise.
6065 (emit_call_insn_after): Likewise.
6066 (emit_debug_insn_after): Likewise.
6067 (set_insn_deleted): Likewise for param.
6069 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
6072 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
6073 floating point modes, so that IFmode is numerically greater than
6074 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
6075 to declare the ordering. This prevents IFmode from being
6076 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
6077 machine. Include rs6000-modes.h to share the fractional values
6078 between genmodes* and the rest of the compiler.
6082 * config/rs6000/rs6000-modes.h: New file.
6083 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
6084 meaning of rs6000_long_double_size so that 126..128 selects an
6085 appropriate 128-bit floating point type.
6086 (rs6000_option_override_internal): Likewise.
6087 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
6088 (TARGET_LONG_DOUBLE_128): Change the meaning of
6089 rs6000_long_double_size so that 126..128 selects an appropriate
6090 128-bit floating point type.
6091 (LONG_DOUBLE_TYPE_SIZE): Update comment.
6092 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
6093 source and destination to match the standard usage.
6094 (truncifkf2): Likewise.
6095 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
6096 ISA 2.07 to use an explicit clobber, instead of passing in a
6098 (copysign<mode>3_soft): Likewise.
6100 2018-06-18 David Malcolm <dmalcolm@redhat.com>
6102 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
6103 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
6104 (vect_slp_analyze_instance_dependence): Likewise.
6105 (vect_enhance_data_refs_alignment): Likewise.
6106 (vect_analyze_data_refs_alignment): Likewise.
6107 (vect_slp_analyze_and_verify_instance_alignment
6108 (vect_analyze_data_ref_accesses): Likewise.
6109 (vect_prune_runtime_alias_test_list): Likewise.
6110 (vect_analyze_data_refs): Likewise.
6111 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
6112 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
6113 (vect_analyze_scalar_cycles_1): Likewise.
6114 (vect_get_loop_niters): Likewise.
6115 (vect_analyze_loop_form_1): Likewise.
6116 (vect_update_vf_for_slp): Likewise.
6117 (vect_analyze_loop_operations): Likewise.
6118 (vect_analyze_loop): Likewise.
6119 (vectorizable_induction): Likewise.
6120 (vect_transform_loop): Likewise.
6121 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
6122 * tree-vect-slp.c (vect_analyze_slp): Likewise.
6123 (vect_make_slp_decision): Likewise.
6124 (vect_detect_hybrid_slp): Likewise.
6125 (vect_slp_analyze_operations): Likewise.
6126 (vect_slp_bb): Likewise.
6127 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
6128 (vectorizable_bswap): Likewise.
6129 (vectorizable_call): Likewise.
6130 (vectorizable_simd_clone_call): Likewise.
6131 (vectorizable_conversion): Likewise.
6132 (vectorizable_assignment): Likewise.
6133 (vectorizable_shift): Likewise.
6134 (vectorizable_operation): Likewise.
6135 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
6137 2018-06-18 Martin Sebor <msebor@redhat.com>
6139 PR tree-optimization/81384
6140 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
6141 * builtins.c (expand_builtin_strnlen): New function.
6142 (expand_builtin): Call it.
6143 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
6144 * builtins.def (BUILT_IN_STRNLEN): New.
6145 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
6146 Warn for bounds in excess of maximum object size.
6147 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
6148 single-value ranges. Handle strnlen.
6149 (handle_builtin_strlen): Handle strnlen.
6150 (strlen_check_and_optimize_stmt): Same.
6151 * doc/extend.texi (Other Builtins): Document strnlen.
6153 2018-06-18 Maya Rashish <coypu@sdf.org>
6155 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
6156 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
6157 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
6159 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
6161 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
6163 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6165 * tree.c (escaped_string::escape): Replace cast to char * by
6166 const_cast<char *> (unescaped).
6168 2018-06-18 Nick Clifton <nickc@redhat.com>
6171 * tree.c (escaped_string): New class. Converts an unescaped
6172 string into its escaped equivalent.
6173 (warn_deprecated_use): Use the new class to convert the
6174 deprecation message, if present.
6175 (test_escaped_strings): New self test.
6176 (test_c_tests): Add test_escaped_strings.
6177 * doc/extend.texi (deprecated): Add a note that the
6178 deprecation message is affected by the -fmessage-length
6179 option, and that control characters will be escaped.
6180 (#pragma GCC error): Document this pragma.
6181 (#pragma GCC warning): Likewise.
6182 * doc/invoke.texi (-fmessage-length): Document this option's
6183 effect on the #warning and #error preprocessor directives and
6184 the deprecated attribute.
6186 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
6188 * tree.c (decl_value_expr_lookup): Revert latest change.
6189 (decl_value_expr_insert): Likewise.
6191 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
6193 * gimplify.c (nonlocal_vlas): Delete.
6194 (nonlocal_vla_vars): Likewise.
6195 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
6197 (gimplify_body): Do not create and destroy nonlocal_vlas.
6198 * tree-nested.c: Include diagnostic.h.
6199 (use_pointer_in_frame): Tweak.
6200 (lookup_field_for_decl): Add assertion and declare the transformation.
6201 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
6202 internal error when the reference is in a wrong context. Do not
6203 create a debug decl by default.
6204 (note_nonlocal_block_vlas): Delete.
6205 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
6206 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
6207 create a debug decl by default.
6208 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
6209 call is in a wrong context.
6210 (fixup_vla_decls): New function.
6211 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
6212 debug variables were created.
6213 * tree.c (decl_value_expr_lookup): Add checking assertion.
6214 (decl_value_expr_insert): Likewise.
6216 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
6219 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
6220 * tree-scalar-evolution.c (interpret_expr): Likewise.
6221 (expression_expensive_p): Likewise.
6222 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
6223 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
6224 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
6225 (ssa_defined_by_minus_one_stmt_p): New.
6227 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
6230 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
6231 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
6232 * dojump.c (do_jump): Likewise.
6233 * expr.c (expand_expr_real_2): Check operand type's sign.
6234 * fold-const.c (const_unop): Handle ABSU_EXPR.
6235 (fold_abs_const): Likewise.
6236 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
6237 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
6238 (strip_sign_op_1): Likesise.
6239 * match.pd: Add new pattern to generate ABSU_EXPR.
6240 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
6241 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
6242 * tree-eh.c (operation_could_trap_helper_p): Likewise.
6243 * tree-inline.c (estimate_operator_cost): Likewise.
6244 * tree-pretty-print.c (dump_generic_node): Likewise.
6245 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
6246 * tree.def (ABSU_EXPR): New.
6248 2018-06-16 Jakub Jelinek <jakub@redhat.com>
6251 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
6252 documented as preserved for backward compatibility only.
6253 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
6255 PR rtl-optimization/86108
6256 * bb-reorder.c (create_forwarder_block): Renamed to ...
6257 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
6258 jump from new landing pad to the second part.
6259 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
6262 2018-06-15 Jakub Jelinek <jakub@redhat.com>
6265 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
6266 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
6267 Only call store_expr for halves if the mode is the same.
6270 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
6271 Fix up comment formatting.
6273 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
6275 * typed-splay-tree.h (typed_splay_tree::remove): New function.
6276 (typed_splay_tree::closure,
6277 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
6278 (typed_splay_tree::typed_splay_tree,
6279 typed_splay_tree::operator =): Declared private.
6280 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
6281 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
6282 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
6283 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
6284 typed_splay_tree::splay_tree_splay,
6285 typed_splay_tree::splay_tree_foreach_helper,
6286 typed_splay_tree::splay_tree_insert,
6287 typed_splay_tree::splay_tree_remove,
6288 typed_splay_tree::splay_tree_lookup,
6289 typed_splay_tree::splay_tree_predecessor,
6290 typed_splay_tree::splay_tree_successor,
6291 typed_splay_tree::splay_tree_min,
6292 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
6293 (typed_splay_tree::root, typed_splay_tree::comp,
6294 typed_splay_tree::delete_key,
6295 typed_splay_tree::delete_value): New data members.
6296 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
6297 typed_splay_tree::remove.
6299 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
6301 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
6302 -mginv and -mno-ginv to the assembler.
6303 * config/mips/mips.opt (-mcrc): New option.
6305 * doc/invoke.text (-mcrc): Document.
6308 2018-06-15 Nick Clifton <nickc@redhat.com>
6311 * tree.c (escaped_string): New class. Converts an unescaped
6312 string into its escaped equivalent.
6313 (warn_deprecated_use): Use the new class to convert the
6314 deprecation message, if present.
6315 (test_escaped_strings): New self test.
6316 (test_c_tests): Add test_escaped_strings.
6317 * doc/extend.texi (deprecated): Add a note that the
6318 deprecation message is affected by the -fmessage-length
6319 option, and that control characters will be escaped.
6320 (#pragma GCC error): Document this pragma.
6321 (#pragma GCC warning): Likewise.
6322 * doc/invoke.texi (-fmessage-length): Document this option's
6323 effect on the #warning and #error preprocessor directives and
6324 the deprecated attribute.
6326 2018-06-15 Richard Biener <rguenther@suse.de>
6328 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
6329 here, also noting vector size used.
6330 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
6331 size used in MSG_OPTIMIZED_LOCATIONS dump.
6332 (pass_slp_vectorize::execute): Adjust.
6334 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
6337 * config/arc/arc.c (arc_return_address_register): Fix
6340 2018-06-15 Richard Biener <rguenther@suse.de>
6343 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
6344 leave useless conversion stripping to force_gimple_operand_gsi.
6345 (gimplify_build2): Likewise.
6346 (gimplify_build1): Likewise.
6348 2018-06-15 Richard Biener <rguenther@suse.de>
6351 * tree-cfg.c (move_stmt_op): unshare invariant addresses
6352 before adjusting their block.
6354 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
6356 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
6357 multilibs for *-*-rtems*.
6358 * config/riscv/t-rtems: New file.
6360 2018-06-14 Jakub Jelinek <jakub@redhat.com>
6363 * match.pd ((A +- CST1) +- CST2): Punt if last resort
6364 unsigned_type_for returns NULL.
6367 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
6368 subregs of multi-word pseudos unless the float mode has word size.
6370 2018-06-14 Richard Biener <rguenther@suse.de>
6373 * tree-vect-generic.c (build_word_mode_vector_type): Remove
6374 duplicate and harmful type_hash_canon.
6375 * tree.c (type_hash_canon): Assert we didn't find ourselves.
6377 2018-06-14 Richard Biener <rguenther@suse.de>
6380 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
6383 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
6385 * config/rtems.h (STDINT_LONG32): Define.
6387 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
6388 Prachi Godbole <prachi.godbole@imgtec.com>
6390 * config/mips/mips-cpus.def: Define P6600.
6391 * config/mips/mips-tables.opt: Regenerate.
6392 * config/mips/mips.c (mips_ucbranch_type): New enum.
6393 (mips_rtx_cost_data): Add support for P6600.
6394 (mips_issue_rate): Likewise.
6395 (mips_multipass_dfa_lookahead): Likewise.
6396 (mips_avoid_hazard): Likewise.
6397 (mips_reorg_process_insns): Likewise.
6398 (mips_classify_branch_p6600): New function.
6399 * config/mips/mips.h (TUNE_P6600): New define.
6400 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
6401 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
6402 * config/mips/mips.md: Include p6600.md.
6403 (processor): Add p6600.
6404 * config/mips/p6600.md: New file.
6405 * doc/invoke.texi: Add p6600 to supported architectures.
6407 2018-06-13 Martin Sebor <msebor@redhat.com>
6409 PR tree-optimization/86114
6410 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
6412 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
6414 2018-06-13 Richard Biener <rguenther@suse.de>
6416 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
6417 Properly set vector type of the intermediate stmt.
6418 * tree-vect-stmts.c (vectorizable_operation): The destination
6419 var always has vectype_out type.
6421 2018-06-13 Jeff Law <law@redhat.com>
6423 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
6424 integer 0 for argument to print_rtl_with_bb.
6425 (rl78_reorg): Likewise.
6427 2018-06-13 David Malcolm <dmalcolm@redhat.com>
6429 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
6430 from rtx to rtx_insn *.
6431 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
6433 (add_sched_insns_for_speculation): Likewise for local "target",
6434 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
6435 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
6436 from rtx_insn ** to rtx_code_label **.
6437 (reorg_emit_nops): Likewise.
6438 (c6x_reorg): Likewise for local "call_labels".
6439 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
6441 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
6442 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
6443 the loops over LABEL_REFS.
6444 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
6446 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
6447 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
6448 (split_branches): Strengthen local "olabel" from rtx to
6449 rtx_insn *, adding a safe_as_a cast.
6450 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
6452 (add_insn_after): Likewise for first two params.
6453 (add_insn_before): Likewise.
6454 (remove_insn): Likewise for param.
6455 (emit_pattern_before_noloc): Likewise for second and third params.
6456 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
6457 (emit_call_insn_before_noloc): Likewise.
6458 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
6460 (emit_barrier_before): Likewise.
6461 (emit_label_before): Strengthen "label" param from "rtx" to
6462 "rtx_code_label *". Strengthen "before" param from "rtx" to
6464 (emit_insn_after_1): Strengthen "after" param from "rtx" to
6466 (emit_pattern_after_noloc): Likewise.
6467 (emit_insn_after_noloc): Likewise.
6468 (emit_jump_insn_after_noloc): Likewise.
6469 (emit_call_insn_after_noloc): Likewise.
6470 (emit_debug_insn_after_noloc): Likewise.
6471 (emit_barrier_after): Likewise.
6472 (emit_label_after): Likewise for both params.
6473 (emit_pattern_after_setloc): Likewise for "after" param. Convert
6474 "loc" param from "int" to "location_t".
6475 (emit_insn_after_setloc): Likewise.
6476 (emit_jump_insn_after_setloc): Likewise.
6477 (emit_call_insn_after_setloc): Likewise.
6478 (emit_debug_insn_after_setloc): Likewise.
6479 (emit_pattern_before_setloc): Likewise for "before" param. Convert
6480 "loc" param from "int" to "location_t".
6481 (emit_pattern_before): Convert NULL_RTX to NULL.
6482 (emit_insn_before_setloc): Convert "loc" param from "int" to
6484 (emit_jump_insn_before_setloc): Likewise.
6485 (emit_call_insn_before_setloc): Likewise.
6486 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
6487 rtx_insn *. Convert "loc" param from "int" to "location_t".
6488 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
6489 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
6490 Convert 3rd param from "int" to "location_t".
6491 (emit_barrier_before, emit_barrier_after, next_real_insn):
6492 Strengthen param from rtx to rtx_insn *.
6493 (emit_label_before): Strengthen 1st param from "rtx" to
6494 "rtx_code_label *". Strengthen 2nd param from "rtx" to
6496 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
6497 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
6498 Strengthen 2nd param from "rtx" to "rtx_insn *".
6499 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
6500 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
6501 Likewise. Convert 3rd param from "int" to "location_t".
6502 (emit_label_after): Strengthen 1st param from "rtx" to
6504 (next_real_insn, remove_insn): Strengthen param from "rtx" to
6506 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
6507 from "rtx" to "rtx_insn *".
6509 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
6511 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
6512 bodies streamed in with -Q.
6513 * dumpfile.c (dump_files): Add lto-stream-out dump file.
6514 * dumpfile.h (tree_dump_index): Add lto_stream_out.
6515 * gimple-streamer-out.c: Include gimple-pretty-print.h
6516 (output_bb): Dump stmts streamed.
6517 * lto-section-out.c: Include print-tree.h
6518 (lto_begin_section): Dump sections created.
6519 (lto_output_decl_index): Dump decl encoded.
6520 * lto-streamer-out.c: Include print-tree.h
6521 (create_output_block): Dump output block created.
6522 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
6523 (output_function): Dump function output.
6524 (output_constructor): Dump constructor streamed.
6525 (write_global_stream): Output indexes encoded.
6526 (produce_asm_for_decls): Dump streams encoded.
6527 * lto-streamer.c (streamer_dump_file): New global var.
6528 * lto-streamer.h (streamer_dump_file): Declare.
6529 * passes.c (ipa_write_summaries): Initialize streamer dump.
6530 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
6533 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
6536 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
6537 offsets for register save directives. Emit a second batch of save
6538 directives, if need be, when the function accesses prior frames.
6540 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
6542 * config/arc/fpu.md (fmasf4): Force operand to register.
6543 (fnmasf4): Likewise.
6545 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
6547 * config/arc/arc-protos.h (arc_pad_return): Remove.
6548 * config/arc/arc.c (machine_function): Remove force_short_suffix
6550 (arc_print_operand): Adjust printing of '&'.
6551 (arc_verify_short): Remove conditional printing of short suffix.
6552 (arc_final_prescan_insn): Remove reference to size_reason.
6553 (pad_return): New function.
6554 (arc_reorg): Call pad_return.
6555 (arc_pad_return): Remove.
6556 (arc_init_machine_status): Remove reference to force_short_suffix.
6557 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
6558 (attr length): When attribute iscompact is true force to 2
6559 regardless; in the case of maybe check if we want to force the
6560 instruction to have 4 bytes length.
6561 (nopv): Change it to generate 4 byte long nop as well.
6562 (blockage): New pattern.
6563 (simple_return): Remove call to arc_pad_return.
6564 (p_return_i): Likewise.
6566 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
6568 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
6570 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
6572 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
6575 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
6577 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
6578 for ARC700 and ARCv2.
6580 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
6583 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
6584 operands[2] instead of operands[1].
6587 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
6589 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
6590 case, check whether the outer register overlaps an unallocatable
6591 register, not just whether it fits the required class.
6593 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
6595 * poly-int.h (can_div_trunc_p): Add new overload in which all values
6597 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
6598 (memrefs_conflict_p): Likewise.
6599 (init_alias_analysis): Likewise.
6600 * cfgexpand.c (expand_debug_expr): Likewise.
6601 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
6602 * cse.c (fold_rtx): Likewise.
6603 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
6604 * expr.c (emit_block_move_hints): Likewise.
6605 (clear_storage_hints, push_block, emit_push_insn): Likewise.
6606 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
6607 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
6608 (emit_group_store): Likewise.
6609 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
6610 to read the PRE/POST_MODIFY increment.
6611 * calls.c (store_one_arg): Use strip_offset.
6612 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
6614 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
6616 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
6617 (simplify_binary_operation_1): Extend CONST_INT handling to
6619 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
6620 than a HOST_WIDE_INT.
6621 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
6623 (adjust_mems, add_stores): Update accodingly.
6624 (vt_canonicalize_addr): Track polynomial offsets.
6625 (emit_note_insn_var_location): Likewise.
6626 (vt_add_function_parameter): Likewise.
6627 (vt_initialize): Likewise.
6629 2018-06-12 Jeff Law <law@redhat.com>
6631 * config.gcc (alpha*-*-freebsd*): Remove.
6632 * config/alpha/freebsd.h: Remove.
6634 2018-06-12 David Malcolm <dmalcolm@redhat.com>
6637 * spellcheck-tree.c (levenshtein_distance): Rename to...
6638 (get_edit_distance): ...this, and update for underlying renaming.
6639 * spellcheck-tree.h (levenshtein_distance): Rename to...
6640 (get_edit_distance): ...this.
6641 * spellcheck.c (levenshtein_distance): Rename to...
6642 (get_edit_distance): ...this. Convert from Levenshtein distance
6643 to Damerau-Levenshtein distance by supporting transpositions of
6644 adjacent characters. Rename "v1" to "v_next" and "v0" to
6646 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
6647 (selftest::test_edit_distance_unit_test_oneway): ...this, and
6648 update for underlying renaming.
6649 (selftest::levenshtein_distance_unit_test): Rename to...
6650 (selftest::test_get_edit_distance_unit): ...this, and update for
6651 underlying renaming.
6652 (selftest::test_find_closest_string): Add example from PR 69968
6653 where transposition helps
6654 (selftest::test_metric_conditions): Update for renaming.
6655 (selftest::test_metric_conditions): Likewise.
6656 (selftest::spellcheck_c_tests): Likewise.
6657 * spellcheck.h (levenshtein_distance): Rename both overloads to...
6658 (get_edit_distance): ...this.
6659 (best_match::consider): Update for renaming.
6661 2018-06-12 Martin Sebor <msebor@redhat.com>
6663 PR tree-optimization/85259
6664 * builtins.c (compute_objsize): Handle constant offsets.
6665 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
6666 true iff a warning has been issued.
6667 * gimple.h (gimple_nonartificial_location): New function.
6668 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
6669 gimple_nonartificial_location and handle -Wno-system-headers.
6670 (handle_builtin_stxncpy): Same.
6672 2018-06-12 Martin Sebor <msebor@redhat.com>
6675 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
6677 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
6679 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
6680 BUILTIN_VEC_XST entries for pointer to double and long long.
6682 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
6685 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
6687 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
6690 2018-06-12 Martin Liska <mliska@suse.cz>
6692 * doc/options.texi: Document IntegerRange.
6694 2018-06-12 Martin Liska <mliska@suse.cz>
6696 * config/i386/i386.opt: Make MPX-related options as Deprecated.
6697 * opt-functions.awk: Handle Deprecated flag.
6698 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
6700 (read_cmdline_option): Report warning for a deprecated option.
6701 * opts.h (struct cl_option): Add new field cl_deprecated.
6702 (CL_ERR_DEPRECATED): New.
6704 2018-06-12 Martin Liska <mliska@suse.cz>
6706 * doc/options.texi: Document Deprecated option flag.
6708 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
6710 * config/arc/arc-arch.h (arc_extras): New enum.
6711 (arc_cpu_t):Add field extra.
6712 (arc_cpu_types): Consider the extras.
6713 * config/arc/arc-cpus.def: Add extras info.
6714 * config/arc/arc-opts.h (processor_type): Consider extra field.
6715 * config/arc/arc.c (arc_override_options): Handle extra field.
6717 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
6719 * config/arc/arc-arch.h: Update ARC_OPTX macro.
6720 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
6722 * config/arc/arc.c (arc_init): Update pic warning.
6723 (irq_range): Update irq range parsing warnings.
6724 (arc_override_options): Update various warning messages.
6725 (arc_handle_aux_attribute): Likewise.
6727 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
6729 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
6731 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
6733 * doc/sourcebuild.texi: Document usage of line number 0 in verify
6734 compiler messages directives.
6736 2018-06-12 Matthew Fortune <mfortune@gmail.com>
6738 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
6739 * config/mips/mips-tables.opt: Regenerate.
6740 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
6742 * doc/invoke.texi: Document -march=i6500.
6744 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
6746 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
6747 (i6400_gpmul): Add cpu_unit.
6748 (i6400_gpdiv): Likewise.
6749 (i6400_msa_add_d): Update reservations.
6750 (i6400_msa_int_add) Likewise.
6751 (i6400_msa_short_logic3) Likewise.
6752 (i6400_msa_short_logic2) Likewise.
6753 (i6400_msa_short_logic) Likewise.
6754 (i6400_msa_move) Likewise.
6755 (i6400_msa_cmp) Likewise.
6756 (i6400_msa_short_float2) Likewise.
6757 (i6400_msa_div_d) Likewise.
6758 (i6400_msa_long_logic1) Likewise.
6759 (i6400_msa_long_logic2) Likewise.
6760 (i6400_msa_mult) Likewise.
6761 (i6400_msa_long_float2) Likewise.
6762 (i6400_msa_long_float4) Likewise.
6763 (i6400_msa_long_float5) Likewise.
6764 (i6400_msa_long_float8) Likewise.
6765 (i6400_fpu_fadd): Include frint type.
6766 (i6400_fpu_store): New define_insn_reservation.
6767 (i6400_fpu_load): Likewise.
6768 (i6400_fpu_move): Likewise.
6769 (i6400_fpu_fcmp): Likewise.
6770 (i6400_fpu_fmadd): Likewise.
6771 (i6400_int_mult): Include imul3nc type and update reservation.
6772 (i6400_int_div): Include idiv3 type and update reservation.
6773 (i6400_int_load): Update to check type not move_type.
6774 (i6400_int_store): Likewise.
6775 (i6400_int_prefetch): Set zero latency.
6777 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
6779 * gcc.c: Document new %@{...} sequence.
6780 (LINK_COMMAND_SPEC): Use it for the -L switches.
6781 (cpp_unique_options): Use it for the -I switches.
6782 (at_file_argbuf): New global variable.
6783 (in_at_file): Likewise.
6784 (alloc_args): Create at_file_argbuf.
6785 (clear_args): Truncate at_file_argbuf.
6786 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
6787 (open_at_file): New function.
6788 (close_at_file): Likewise.
6789 (create_at_file): Delete.
6790 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
6793 (validate_switches_from_spec): Deal with %@{...} sequence.
6794 (validate_switches): Likewise.
6795 (driver::finalize): Call clear_args.
6797 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
6799 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
6801 2018-06-11 Martin Sebor <msebor@redhat.com>
6803 * doc/invoke.texi (-Wall): List -Wc++17-compat.
6804 (Wno-class-memaccess): Add @opindex.
6805 (Wno-templates, Wno-multiple-inheritance): Same.
6806 (Wno-virtual-inheritance, Wno-namespaces): Same.
6807 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
6808 (Wno-format-overflow, Wno-format-truncation): Same.
6809 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
6810 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
6811 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
6812 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
6813 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
6814 (Wno-misspelled-isr): Same.
6816 2018-06-11 Martin Sebor <msebor@redhat.com>
6818 * PR tree-optimization/86083
6819 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
6821 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
6823 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
6825 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
6828 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
6829 on the correct operand.
6830 (*movdi_internal64): Ditto.
6832 2018-06-11 Martin Liska <mliska@suse.cz>
6834 PR tree-optimization/86089
6835 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
6837 2018-06-11 Julia Koval <julia.koval@intel.com>
6839 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
6840 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
6841 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
6843 2018-06-11 Olivier Hainque <hainque@adacore.com>
6845 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
6846 for Ada with strict dwarf2.
6848 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
6851 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
6854 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
6856 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
6858 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
6860 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
6863 2018-06-08 Martin Liska <mliska@suse.cz>
6865 * tree-cfg.h (debug_function): Fix argument type to match
6868 2018-06-08 Martin Liska <mliska@suse.cz>
6870 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
6871 Remove usage of MPX-related (and removed) fields.
6872 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
6874 2018-06-08 David Malcolm <dmalcolm@redhat.com>
6876 * cfg.c (debug): Use TDF_NONE rather than 0.
6877 * cfghooks.c (debug): Likewise.
6878 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
6879 (struct dump_option_value_info): Convert to...
6880 (struct kv_pair): ...this template type.
6881 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
6883 (optinfo_verbosity_options): Likewise.
6884 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
6886 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
6887 than int for "optgroup_flags" param.
6888 (dump_generic_expr_loc): Use dump_flags_t rather than int for
6890 (dump_dec): Likewise.
6891 (dump_finish): Use TDF_NONE rather than 0.
6892 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
6893 rather than int for "optgroup_flags" param. Use TDF_NONE rather
6894 than 0. Update for change to option_ptr.
6895 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
6896 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
6897 0. Update for changes to optinfo_verbosity_options and
6899 (opt_info_switch_p): Convert optgroup_flags from int to
6901 (dump_basic_block): Use dump_flags_t rather than int
6902 for "dump_kind" param.
6903 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
6904 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
6905 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
6906 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
6907 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
6908 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
6909 TDF_NONE): Convert from macros to...
6910 (enum dump_flag): ...this new enum.
6911 (dump_flags_t): Update to use enum.
6912 (operator|, operator&, operator~, operator|=, operator&=):
6913 Implement for dump_flags_t.
6914 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
6915 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
6916 Convert from macros to...
6917 (enum optgroup_flag): ...this new enum.
6918 (optgroup_flags_t): New typedef.
6919 (operator|, operator|=): Implement for optgroup_flags_t.
6920 (struct dump_file_info): Convert field "alt_flags" to
6921 dump_flags_t. Convert field "optgroup_flags" to
6923 (dump_basic_block): Use dump_flags_t rather than int for param.
6924 (dump_generic_expr_loc): Likewise.
6925 (dump_dec): Likewise.
6926 (dump_register): Convert param "optgroup_flags" to
6928 (opt_info_enable_passes): Likewise.
6929 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
6931 * gimple-pretty-print.c (debug): Likewise.
6932 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
6933 (merged_store_group::apply_stores): Likewise.
6934 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
6935 * gimple.c (verify_gimple_pp): Likewise.
6936 * graphite-poly.c (print_pbb_body): Likewise.
6937 * passes.c (pass_manager::register_one_dump_file): Convert
6938 local "optgroup_flags" to optgroup_flags_t.
6939 * print-tree.c (print_node): Use TDF_NONE rather than 0.
6941 (debug_body): Likewise.
6942 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
6943 to optgroup_flags_t.
6944 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
6946 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
6947 (convert_mult_to_fma): Likewise.
6948 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
6949 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
6950 * tree-vect-data-refs.c (dump_lower_bound): Convert param
6951 "dump_kind" to dump_flags_t.
6953 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
6955 * config/rs6000/rs6000.c (min, max): Delete.
6957 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
6959 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
6960 -mabi=spe and -mabi=no-spe.
6962 2018-06-08 Martin Liska <mliska@suse.cz>
6964 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
6965 where we expect an existing summary.
6967 2018-06-08 Martin Liska <mliska@suse.cz>
6969 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
6970 * ipa-inline.h (estimate_edge_growth): Likewise.
6972 2018-06-08 Martin Liska <mliska@suse.cz>
6974 * cgraph.c (function_version_hasher::hash): Use
6975 cgraph_node::get_uid ().
6976 (function_version_hasher::equal):
6977 * cgraph.h (cgraph_node::get_uid): New method.
6978 * ipa-inline.c (update_caller_keys): Use
6979 cgraph_node::get_uid ().
6980 (update_callee_keys): Likewise.
6981 * ipa-utils.c (searchc): Likewise.
6982 (ipa_reduced_postorder): Likewise.
6983 * lto-cgraph.c (input_node): Likewise.
6984 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
6985 * symbol-summary.h (symtab_insertion): Likewise.
6986 (symtab_removal): Likewise.
6987 (symtab_duplication): Likewise.
6988 * tree-pretty-print.c (dump_function_header): Likewise.
6989 * tree-sra.c (convert_callers_for_node): Likewise.
6991 2018-06-08 Martin Liska <mliska@suse.cz>
6993 * cgraph.c (symbol_table::create_edge): Always assign a new
6995 (symbol_table::free_edge): Do not recycle numbers.
6996 * cgraph.h (cgraph_edge::get): New method.
6997 * symbol-summary.h (symtab_removal): Use it.
6998 (symtab_duplication): Likewise.
6999 (call_summary::hashable_uid): Remove.
7001 2018-06-08 Martin Liska <mliska@suse.cz>
7003 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
7004 (initialize_growth_caches): Remove.
7005 (free_growth_caches): Likewise.
7006 (do_estimate_edge_time): Use edge_growth_cache.
7007 (do_estimate_edge_size): Likewise.
7008 (do_estimate_edge_hints): Likewise.
7009 * ipa-inline.c (reset_edge_caches): Likewise.
7010 (recursive_inlining): Likewise.
7011 (inline_small_functions): Likewise.
7012 * ipa-inline.h (initialize_growth_caches): Remove.
7013 (estimate_edge_size): Likewise.
7014 (estimate_edge_time): Likewise.
7015 (estimate_edge_hints): Likewise.
7016 (reset_edge_growth_cache): Likewise.
7017 * symbol-summary.h (call_summary::remove): New method.
7019 2018-06-08 Martin Liska <mliska@suse.cz>
7021 * ipa-cp.c (class edge_clone_summary): New summary.
7022 (grow_edge_clone_vectors): Remove.
7023 (ipcp_edge_duplication_hook): Remove.
7024 (class edge_clone_summary_t): New call_summary class.
7025 (ipcp_edge_removal_hook): Remove.
7026 (edge_clone_summary_t::duplicate): New function.
7027 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
7028 (create_specialized_node): Likewise.
7029 (ipcp_driver): Initialize edge_clone_summaries and do not
7032 2018-06-08 Martin Liska <mliska@suse.cz>
7034 * symbol-summary.h (get): New function.
7035 (call_summary::m_initialize_when_cloning): New class member.
7037 2018-06-08 Martin Liska <mliska@suse.cz>
7039 * cgraph.c (cgraph_node::remove): Do not recycle uid.
7040 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
7041 (symbol_table::allocate_cgraph_symbol): Do not set uid.
7042 * passes.c (uid_hash_t): Record removed_nodes by their uids.
7043 (remove_cgraph_node_from_order): Use the removed_nodes set.
7044 (do_per_function_toporder): Likwise.
7045 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
7046 instead of summary_uid.
7047 (symtab_removal): Likewise.
7048 (symtab_duplication): Likewise.
7050 2018-06-08 Martin Liska <mliska@suse.cz>
7052 * ipa-cp.c (ipcp_store_bits_results): Use
7053 ipcp_transformation_sum.
7054 (ipcp_store_vr_results): Likewise.
7055 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
7057 (ipcp_transformation_initialize): ... this.
7058 (ipa_set_node_agg_value_chain):
7059 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
7060 (write_ipcp_transformation_info): Likewise.
7061 (read_ipcp_transformation_info): Likewise.
7062 (ipcp_update_bits): Likewise.
7063 (ipcp_update_vr): Likewise.
7064 (ipcp_transform_function): Likewise.
7065 * ipa-prop.h: Rename ipcp_transformation_summary to
7066 ipcp_transformation.
7067 (class ipcp_transformation_t): New function summary.
7068 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
7069 (ipa_get_agg_replacements_for_node): Likewise.
7071 2018-06-08 Martin Liska <mliska@suse.cz>
7073 * ipa-pure-const.c (struct funct_state_d): Do it class instead
7075 (class funct_state_summary_t): New function_summary class.
7076 (has_function_state): Remove.
7077 (get_function_state): Likewise.
7078 (set_function_state): Likewise.
7079 (add_new_function): Likewise.
7080 (funct_state_summary_t::insert): New function.
7081 (duplicate_node_data): Remove.
7082 (remove_node_data): Remove.
7083 (funct_state_summary_t::duplicate): New function.
7084 (register_hooks): Create new funct_state_summaries.
7085 (pure_const_generate_summary): Use it.
7086 (pure_const_write_summary): Likewise.
7087 (pure_const_read_summary): Likewise.
7088 (propagate_pure_const): Likewise.
7089 (propagate_nothrow): Likewise.
7090 (dump_malloc_lattice): Likewise.
7091 (propagate_malloc): Likewise.
7092 (execute): Do not register hooks, just remove summary
7094 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
7097 2018-06-08 Martin Liska <mliska@suse.cz>
7099 * ipa-reference.c (remove_node_data): Remove.
7100 (duplicate_node_data): Likewise.
7101 (class ipa_ref_var_info_summary_t): New class.
7102 (class ipa_ref_opt_summary_t): Likewise.
7103 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
7104 (get_reference_optimization_summary): Use
7105 ipa_ref_opt_sum_summaries.
7106 (set_reference_vars_info): Remove.
7107 (set_reference_optimization_summary): Likewise.
7108 (ipa_init): Create summaries.
7109 (init_function_info): Use function summary.
7110 (ipa_ref_opt_summary_t::duplicate): New function.
7111 (ipa_ref_opt_summary_t::remove): New function.
7112 (get_read_write_all_from_node): Fix GNU coding style.
7113 (propagate): Use function summary.
7114 (write_node_summary_p): Fix GNU coding style.
7115 (stream_out_bitmap): Likewise.
7116 (ipa_reference_read_optimization_summary): Use function summary.
7117 (ipa_reference_c_finalize): Do not release hooks.
7119 2018-06-08 Martin Liska <mliska@suse.cz>
7121 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
7122 (analyze_function_body): Extract multiple calls of get_create.
7123 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
7124 * ipa-inline.c (recursive_inlining): Use ::get method.
7125 * ipa-inline.h (estimate_edge_growth): Likewise.
7127 2018-06-08 Martin Liska <mliska@suse.cz>
7129 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
7131 (hsa_function_summary::hsa_function_summary): Use the new enum
7133 (hsa_gpu_implementation_p): Use hsa_summaries::get.
7134 * hsa-gen.c (hsa_get_host_function): Likewise.
7135 (get_brig_function_name): Likewise.
7136 * ipa-hsa.c (process_hsa_functions): Likewise.
7137 (ipa_hsa_write_summary): Likewise.
7138 * symbol-summary.h (symtab_duplication): Use ::get function/
7139 (get): New function.
7141 2018-06-08 Martin Liska <mliska@suse.cz>
7143 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
7145 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
7146 (hsa_register_kernel): Likewise.
7147 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
7148 * hsa-gen.c (hsa_get_host_function): Likewise.
7149 (get_brig_function_name): Likewise.
7150 (generate_hsa): Likewise.
7151 (pass_gen_hsail::execute): Likewise.
7152 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
7153 (devirtualization_time_bonus): Likewise.
7154 (ipcp_propagate_stage): Likewise.
7155 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
7156 (edge_set_predicate): Likewise.
7157 (evaluate_conditions_for_known_args): Likewise.
7158 (evaluate_properties_for_edge): Likewise.
7159 (ipa_fn_summary::reset): Likewise.
7160 (ipa_fn_summary_t::duplicate): Likewise.
7161 (dump_ipa_call_summary): Likewise.
7162 (ipa_dump_fn_summary): Likewise.
7163 (analyze_function_body): Likewise.
7164 (compute_fn_summary): Likewise.
7165 (estimate_edge_devirt_benefit): Likewise.
7166 (estimate_edge_size_and_time): Likewise.
7167 (estimate_calls_size_and_time): Likewise.
7168 (estimate_node_size_and_time): Likewise.
7169 (inline_update_callee_summaries): Likewise.
7170 (remap_edge_change_prob): Likewise.
7171 (remap_edge_summaries): Likewise.
7172 (ipa_merge_fn_summary_after_inlining): Likewise.
7173 (ipa_update_overall_fn_summary): Likewise.
7174 (read_ipa_call_summary): Likewise.
7175 (inline_read_section): Likewise.
7176 (write_ipa_call_summary): Likewise.
7177 (ipa_fn_summary_write): Likewise.
7178 (ipa_free_fn_summary): Likewise.
7179 * ipa-hsa.c (process_hsa_functions): Likewise.
7180 (ipa_hsa_write_summary): Likewise.
7181 (ipa_hsa_read_section): Likewise.
7182 * ipa-icf.c (sem_function::merge): Likewise.
7183 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
7184 (do_estimate_edge_time): Likewise.
7185 (estimate_size_after_inlining): Likewise.
7186 (estimate_growth): Likewise.
7187 (growth_likely_positive): Likewise.
7188 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
7189 (inline_call): Likewise.
7190 * ipa-inline.c (caller_growth_limits): Likewise.
7191 (can_inline_edge_p): Likewise.
7192 (can_inline_edge_by_limits_p): Likewise.
7193 (compute_uninlined_call_time): Likewise.
7194 (compute_inlined_call_time): Likewise.
7195 (want_inline_small_function_p): Likewise.
7196 (edge_badness): Likewise.
7197 (update_caller_keys): Likewise.
7198 (update_callee_keys): Likewise.
7199 (recursive_inlining): Likewise.
7200 (inline_small_functions): Likewise.
7201 (inline_to_all_callers_1): Likewise.
7202 (dump_overall_stats): Likewise.
7203 (early_inline_small_functions): Likewise.
7204 (early_inliner): Likewise.
7205 * ipa-inline.h (estimate_edge_growth): Likewise.
7206 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
7207 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
7208 * ipa-prop.h (IPA_NODE_REF): Likewise.
7209 (IPA_EDGE_REF): Likewise.
7210 * ipa-pure-const.c (malloc_candidate_p): Likewise.
7211 (propagate_malloc): Likewise.
7212 * ipa-split.c (execute_split_functions): Likewise.
7213 * symbol-summary.h: Rename get to get_create.
7215 (get_create): Likewise.
7216 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
7218 2018-06-08 Martin Liska <mliska@suse.cz>
7220 * symbol-summary.h (release): Move definition out of class
7222 (symtab_removal): Likewise.
7223 (symtab_duplication): Likewise.
7225 2018-06-08 Martin Liska <mliska@suse.cz>
7227 * symbol-summary.h (function_summary): Move constructor
7228 implementation out of class declaration.
7229 (release): Likewise.
7230 (symtab_insertion): Likewise.
7231 (symtab_removal): Likewise.
7232 (symtab_duplication): Likewise.
7235 2018-06-08 Martin Liska <mliska@suse.cz>
7237 * Makefile.in: Remove support for MPX (macros, related functions,
7238 fields in cgraph_node, ...).
7239 * builtin-types.def (BT_BND): Likewise.
7240 (BT_FN_BND_CONST_PTR): Likewise.
7241 (BT_FN_CONST_PTR_BND): Likewise.
7242 (BT_FN_VOID_PTR_BND): Likewise.
7243 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
7244 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
7245 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
7246 (expand_builtin_mempcpy_with_bounds): Likewise.
7247 (expand_builtin_memset_with_bounds): Likewise.
7248 (expand_builtin_memset_args): Likewise.
7249 (std_expand_builtin_va_start): Likewise.
7250 (expand_builtin): Likewise.
7251 (expand_builtin_with_bounds): Likewise.
7252 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
7253 (DEF_LIB_BUILTIN_CHKP): Likewise.
7254 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
7255 (DEF_CHKP_BUILTIN): Likewise.
7256 (BUILT_IN_MEMCPY): Likewise.
7257 (BUILT_IN_MEMMOVE): Likewise.
7258 (BUILT_IN_MEMPCPY): Likewise.
7259 (BUILT_IN_MEMSET): Likewise.
7260 (BUILT_IN_STPCPY): Likewise.
7261 (BUILT_IN_STRCAT): Likewise.
7262 (BUILT_IN_STRCHR): Likewise.
7263 (BUILT_IN_STRCPY): Likewise.
7264 (BUILT_IN_STRLEN): Likewise.
7265 (BUILT_IN_MEMCPY_CHK): Likewise.
7266 (BUILT_IN_MEMMOVE_CHK): Likewise.
7267 (BUILT_IN_MEMPCPY_CHK): Likewise.
7268 (BUILT_IN_MEMSET_CHK): Likewise.
7269 (BUILT_IN_STPCPY_CHK): Likewise.
7270 (BUILT_IN_STRCAT_CHK): Likewise.
7271 (BUILT_IN_STRCPY_CHK): Likewise.
7272 * calls.c (store_bounds): Likewise.
7273 (emit_call_1): Likewise.
7274 (special_function_p): Likewise.
7275 (maybe_warn_nonstring_arg): Likewise.
7276 (initialize_argument_information): Likewise.
7277 (finalize_must_preallocate): Likewise.
7278 (compute_argument_addresses): Likewise.
7279 (expand_call): Likewise.
7280 * cfgexpand.c (expand_call_stmt): Likewise.
7281 (expand_return): Likewise.
7282 (expand_gimple_stmt_1): Likewise.
7283 (pass_expand::execute): Likewise.
7284 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
7285 (cgraph_node::remove): Likewise.
7286 (cgraph_node::dump): Likewise.
7287 (cgraph_node::verify_node): Likewise.
7288 * cgraph.h (chkp_function_instrumented_p): Likewise.
7289 (symtab_node::get_alias_target): Likewise.
7290 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
7291 (cgraph_local_p): Likewise.
7292 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
7293 (cgraph_edge::rebuild_references): Likewise.
7294 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
7295 (walk_polymorphic_call_targets): Likewise.
7296 (cgraph_node::expand_thunk): Likewise.
7297 (symbol_table::output_weakrefs): Likewise.
7298 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
7299 (ix86_handle_option): Likewise.
7300 * config/i386/constraints.md: Likewise.
7301 * config/i386/i386-builtin-types.def (BND): Likewise.
7305 * config/i386/i386-builtin.def (BDESC_END): Likewise.
7306 (BDESC_FIRST): Likewise.
7308 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
7309 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
7310 * config/i386/i386.c (enum reg_class): Likewise.
7311 (ix86_target_string): Likewise.
7312 (ix86_option_override_internal): Likewise.
7313 (ix86_conditional_register_usage): Likewise.
7314 (ix86_valid_target_attribute_inner_p): Likewise.
7315 (ix86_set_indirect_branch_type): Likewise.
7316 (ix86_set_current_function): Likewise.
7317 (ix86_function_arg_regno_p): Likewise.
7318 (init_cumulative_args): Likewise.
7319 (ix86_function_arg_advance): Likewise.
7320 (ix86_function_arg): Likewise.
7321 (ix86_pass_by_reference): Likewise.
7322 (ix86_function_value_regno_p): Likewise.
7323 (ix86_function_value_1): Likewise.
7324 (ix86_function_value_bounds): Likewise.
7325 (ix86_return_in_memory): Likewise.
7326 (ix86_setup_incoming_vararg_bounds): Likewise.
7327 (ix86_va_start): Likewise.
7328 (indirect_thunk_need_prefix): Likewise.
7329 (print_reg): Likewise.
7330 (ix86_print_operand): Likewise.
7331 (ix86_expand_call): Likewise.
7332 (ix86_output_function_return): Likewise.
7333 (reg_encoded_number): Likewise.
7334 (BDESC_VERIFYS): Likewise.
7335 (ix86_init_mpx_builtins): Likewise.
7336 (ix86_init_builtins): Likewise.
7337 (ix86_emit_cmove): Likewise.
7338 (ix86_emit_move_max): Likewise.
7339 (ix86_expand_builtin): Likewise.
7340 (ix86_builtin_mpx_function): Likewise.
7341 (ix86_get_arg_address_for_bt): Likewise.
7342 (ix86_load_bounds): Likewise.
7343 (ix86_store_bounds): Likewise.
7344 (ix86_load_returned_bounds): Likewise.
7345 (ix86_store_returned_bounds): Likewise.
7346 (ix86_class_likely_spilled_p): Likewise.
7347 (ix86_hard_regno_mode_ok): Likewise.
7348 (x86_order_regs_for_local_alloc): Likewise.
7349 (ix86_mitigate_rop): Likewise.
7350 (ix86_bnd_prefixed_insn_p): Likewise.
7351 (ix86_mpx_bound_mode): Likewise.
7352 (ix86_make_bounds_constant): Likewise.
7353 (ix86_initialize_bounds): Likewise.
7354 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
7355 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
7356 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
7357 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
7358 (TARGET_CHKP_BOUND_MODE): Likewise.
7359 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
7360 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
7361 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
7362 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
7363 * config/i386/i386.h (TARGET_MPX): Likewise.
7364 (TARGET_MPX_P): Likewise.
7365 (VALID_BND_REG_MODE): Likewise.
7366 (FIRST_BND_REG): Likewise.
7367 (LAST_BND_REG): Likewise.
7368 (enum reg_class): Likewise.
7369 (BND_REG_P): Likewise.
7370 (BND_REGNO_P): Likewise.
7371 (BNDmode): Likewise.
7372 (ADJUST_INSN_LENGTH): Likewise.
7373 * config/i386/i386.md: Likewise.
7374 * config/i386/i386.opt: Likewise.
7375 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
7376 (defined): Likewise.
7377 (LINK_MPX): Likewise.
7378 (MPX_SPEC): Likewise.
7379 (LIBMPX_SPEC): Likewise.
7380 (LIBMPXWRAPPERS_SPEC): Likewise.
7381 (CHKP_SPEC): Likewise.
7382 * config/i386/predicates.md: Likewise.
7383 * dbxout.c (dbxout_type): Likewise.
7384 * doc/extend.texi: Likewise.
7385 * doc/invoke.texi: Likewise.
7386 * doc/md.texi: Likewise.
7387 * doc/tm.texi: Likewise.
7388 * doc/tm.texi.in: Likewise.
7389 * dwarf2out.c (is_base_type): Likewise.
7390 (gen_formal_types_die): Likewise.
7391 (gen_subprogram_die): Likewise.
7392 (gen_type_die_with_usage): Likewise.
7393 (gen_decl_die): Likewise.
7394 (dwarf2out_late_global_decl): Likewise.
7395 * expr.c (expand_assignment): Likewise.
7396 (emit_storent_insn): Likewise.
7397 (store_expr_with_bounds): Likewise.
7398 (store_expr): Likewise.
7399 (expand_expr_real_1): Likewise.
7400 * expr.h (store_expr_with_bounds): Likewise.
7401 * function.c (use_register_for_decl): Likewise.
7402 (struct bounds_parm_data): Likewise.
7403 (assign_parms_augmented_arg_list): Likewise.
7404 (assign_parm_find_entry_rtl): Likewise.
7405 (assign_parm_is_stack_parm): Likewise.
7406 (assign_parm_load_bounds): Likewise.
7407 (assign_bounds): Likewise.
7408 (assign_parms): Likewise.
7409 (expand_function_start): Likewise.
7410 * gcc.c (CHKP_SPEC): Likewise.
7411 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
7412 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
7413 (wrestrict_dom_walker::check_call): Likewise.
7414 * gimple.c (gimple_build_call_from_tree): Likewise.
7415 * gimple.h (enum gf_mask): Likewise.
7416 (gimple_call_with_bounds_p): Likewise.
7417 (gimple_call_set_with_bounds): Likewise.
7418 * gimplify.c (gimplify_init_constructor): Likewise.
7419 * ipa-cp.c (initialize_node_lattices): Likewise.
7420 (propagate_constants_across_call): Likewise.
7421 (find_more_scalar_values_for_callers_subset): Likewise.
7422 * ipa-hsa.c (process_hsa_functions): Likewise.
7423 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
7424 * ipa-icf.c (sem_function::merge): Likewise.
7425 * ipa-inline.c (early_inliner): Likewise.
7426 * ipa-pure-const.c (warn_function_noreturn): Likewise.
7427 (warn_function_cold): Likewise.
7428 (propagate_pure_const): Likewise.
7429 * ipa-ref.h (enum GTY): Likewise.
7430 * ipa-split.c (find_retbnd): Likewise.
7431 (consider_split): Likewise.
7432 (split_function): Likewise.
7433 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
7434 * ipa.c (walk_polymorphic_call_targets): Likewise.
7435 (symbol_table::remove_unreachable_nodes): Likewise.
7436 (process_references): Likewise.
7437 (cgraph_build_static_cdtor_1): Likewise.
7438 * lto-cgraph.c (lto_output_node): Likewise.
7439 (output_refs): Likewise.
7440 (compute_ltrans_boundary): Likewise.
7441 (input_overwrite_node): Likewise.
7442 (input_node): Likewise.
7443 (input_cgraph_1): Likewise.
7444 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
7445 * passes.c (pass_manager::execute_early_local_passes): Likewise.
7446 (class pass_chkp_instrumentation_passes): Likewise.
7447 (make_pass_chkp_instrumentation_passes): Likewise.
7448 * passes.def: Likewise.
7449 * rtl.h (struct GTY): Likewise.
7450 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
7451 * stor-layout.c (layout_type): Likewise.
7452 * symtab.c: Likewise.
7453 * target.def: Likewise.
7454 * targhooks.c (default_chkp_bound_type): Likewise.
7455 (default_chkp_bound_mode): Likewise.
7456 (default_builtin_chkp_function): Likewise.
7457 (default_chkp_function_value_bounds): Likewise.
7458 (default_chkp_make_bounds_constant): Likewise.
7459 (default_chkp_initialize_bounds): Likewise.
7460 * targhooks.h (default_chkp_bound_type): Likewise.
7461 (default_chkp_bound_mode): Likewise.
7462 (default_builtin_chkp_function): Likewise.
7463 (default_chkp_function_value_bounds): Likewise.
7464 (default_chkp_make_bounds_constant): Likewise.
7465 (default_chkp_initialize_bounds): Likewise.
7466 * toplev.c (compile_file): Likewise.
7467 (process_options): Likewise.
7468 * tree-core.h (DEF_BUILTIN): Likewise.
7469 (DEF_BUILTIN_CHKP): Likewise.
7470 * tree-inline.c (declare_return_variable): Likewise.
7471 (remap_gimple_stmt): Likewise.
7472 (copy_bb): Likewise.
7473 (initialize_inlined_parameters): Likewise.
7474 (expand_call_inline): Likewise.
7475 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
7476 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
7477 (make_pass_ipa_chkp_produce_thunks): Likewise.
7478 (make_pass_chkp): Likewise.
7479 (make_pass_chkp_opt): Likewise.
7480 (make_pass_chkp_instrumentation_passes): Likewise.
7481 * tree-pretty-print.c (dump_generic_node): Likewise.
7482 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
7483 * tree-ssa-dce.c (propagate_necessity): Likewise.
7484 (eliminate_unnecessary_stmts): Likewise.
7485 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
7486 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
7487 * tree-ssa-sccvn.h: Likewise.
7488 * tree-ssa-strlen.c (get_string_length): Likewise.
7489 (valid_builtin_call): Likewise.
7490 (adjust_last_stmt): Likewise.
7491 (handle_builtin_strchr): Likewise.
7492 (handle_builtin_strcpy): Likewise.
7493 (handle_builtin_stxncpy): Likewise.
7494 (handle_builtin_memcpy): Likewise.
7495 (handle_builtin_strcat): Likewise.
7496 (strlen_check_and_optimize_stmt): Likewise.
7497 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
7498 * tree-streamer-in.c: Likewise.
7499 * tree-streamer.c (record_common_node): Likewise.
7500 * tree.c (tree_code_size): Likewise.
7501 (wide_int_to_tree_1): Likewise.
7502 (type_contains_placeholder_1): Likewise.
7503 (build_common_tree_nodes): Likewise.
7504 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
7505 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
7506 (POINTER_BOUNDS_P): Likewise.
7507 (BOUNDED_TYPE_P): Likewise.
7508 (BOUNDED_P): Likewise.
7509 (CALL_WITH_BOUNDS_P): Likewise.
7510 (pointer_bounds_type_node): Likewise.
7511 * value-prof.c (gimple_ic): Likewise.
7512 * var-tracking.c (vt_add_function_parameters): Likewise.
7513 * varasm.c (make_decl_rtl): Likewise.
7514 (assemble_start_function): Likewise.
7515 (output_constant): Likewise.
7516 (maybe_assemble_visibility): Likewise.
7517 * varpool.c (ctor_for_folding): Likewise.
7518 * chkp-builtins.def: Remove.
7519 * ipa-chkp.c: Remove.
7520 * ipa-chkp.h: Remove.
7521 * rtl-chkp.c: Remove.
7522 * rtl-chkp.h: Remove.
7523 * tree-chkp-opt.c: Remove.
7524 * tree-chkp.c: Remove.
7525 * tree-chkp.h: Remove.
7527 2018-06-07 Carl Love <cel@us.ibm.com>
7529 * config/rs6000/vsx.md (vextract_fp_from_shorth,
7530 vextract_fp_from_shortl): Add BE support.
7532 2018-06-07 Paul Koning <ni1d@arrl.net>
7534 * compare-elim.c (try_merge_compare): Don't merge compare if
7535 address contains a side effect.
7536 (try_eliminate_compare): Likewise.
7538 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
7540 * config.gcc: Support "tremont".
7541 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
7542 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
7544 * config/i386/i386.c (m_TREMONT): Define.
7545 (processor_target_table): Add "tremont".
7546 (PTA_TREMONT): Define.
7547 (ix86_lea_outperforms): Add TARGET_TREMONT.
7548 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
7549 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
7550 and M_INTEL_GOLDMONT_PLUS.
7551 (fold_builtin_cpu): Add "tremont".
7552 (ix86_add_stmt_cost): Add TARGET_TREMONT.
7553 (ix86_option_override_internal): Add "tremont".
7554 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
7555 (processor_type): Add PROCESSOR_TREMONT.
7556 * config/i386/x86-tune.def: Add m_TREMONT.
7557 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
7559 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
7561 * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
7562 symbol defined for msp430i* devices to be lower case.
7564 2018-06-07 Richard Biener <rguenther@suse.de>
7566 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
7567 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
7568 Properly wrap signed arithmetic if overflow wraps.
7570 2018-06-07 Jakub Jelinek <jakub@redhat.com>
7572 PR tree-optimization/69615
7573 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
7574 of a cast from a same precision integral SSA_NAME in a bb dominated
7575 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
7576 cast to utype if rhs2 has already a compatible type.
7578 2018-06-07 Richard Biener <rguenther@suse.de>
7580 PR tree-optimization/85935
7581 * graphite-scop-detection.c (find_params_in_bb): Analyze
7582 condition operands with respect to the correct loop. Assert
7583 the analysis doesn't fail.
7585 2018-06-04 Carl Love <cel@us.ibm.com>
7587 * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
7588 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
7589 as it is slightly cheaper.
7590 (first_match_or_eos_index_<mode>):
7591 Calculate index using natural element order.
7592 (first_match_index_<mode>):
7593 Calculate index using natural element order.
7594 (first_match_or_eos_index_<mode>):
7595 Calculate index using natural order.
7596 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
7597 for BE and LE modes.
7598 * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
7599 P9V_BUILTIN_VCLZLSBB_V16QI.
7600 * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
7603 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
7605 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
7606 indentation and line wrap for many prototypes. Add missing
7607 @smallexample directives around block of prototypes for vec_xl and
7610 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
7612 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
7613 track if we pass or return IEEE 128-bit floating point.
7614 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
7615 C++ mangling that is compatible with GCC 8.1.
7616 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
7617 (init_cumulative_args): Note if we pass or return IEEE 128-bit
7618 floating point types.
7619 (rs6000_function_arg_advance_1): Likewise.
7620 (rs6000_mangle_type): Optionally generate mangled names that match
7621 what GCC 8.1 generated for IEEE 128-bit floating point types.
7622 (rs6000_globalize_decl_name): If we have an external function that
7623 passes or returns IEEE 128-bit types, generate a weak reference
7624 from the mangled name used in GCC 8.1 to the current mangled
7626 (rs6000_init_builtins): Make __ibm128 use the long double type if
7627 long double is IBM extended double. Make __float128 use the long
7628 double type if long double is IEEE 128-bit.
7631 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
7632 macro for __ibm128 built-in functions.
7633 (PACK_IF): Add __ibm128 pack/unpack functions.
7634 (UNPACK_IF): Likewise.
7635 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
7636 enable long double built-in functions if long double is IEEE
7637 128-bit floating point.
7638 (rs6000_invalid_builtin): Update long double built-in function
7640 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
7641 functions, adjust the built-in function to use the long double
7642 built-in function if __ibm128 and long double are the same type.
7643 * doc/extend.texi (PowerPC builtins): Update documention for
7644 __builtin_{,un}pack_longdouble. Add documentation for
7645 __builtin_{,un}pack_ibm128.
7647 2018-06-06 Jim Wilson <jimw@sifive.com>
7649 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
7650 (struct machine_function): New field interrupt_mode.
7651 (riscv_handle_type_attribute): New function. Add forward declaration.
7652 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
7653 (riscv_expand_epilogue): Check interrupt_mode field.
7654 (riscv_set_current_function): Check interrupt attribute args and
7655 set interrupt_mode field.
7656 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
7657 (riscv_sret, riscv_uret): New.
7658 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
7659 new arguments to interrupt attribute.
7661 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
7664 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
7665 Don't handle -mcpu=power8 if -mpower9-vector is also used.
7667 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
7669 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
7670 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
7671 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
7672 several redundant entries.
7674 2018-06-06 David Malcolm <dmalcolm@redhat.com>
7676 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
7677 type from "rtx" to "rtx_insn *".
7678 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
7679 for local "call_insn", removing cast.
7680 (ix86_expand_call): Likewise, introducing a "call_insn" local.
7682 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
7684 PR tree-optimization/86066
7685 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
7686 for BIT_INSERT_EXPR stores.
7688 2018-06-06 Richard Biener <rguenther@suse.de>
7690 PR tree-optimization/86062
7691 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
7692 component refs ontop
7693 of to be offsetted base.
7695 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
7697 * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
7698 to be static and remove check on interrupt attribute name.
7700 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
7702 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
7703 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
7705 2018-06-05 Steve Ellcey <sellcey@cavium.com>
7708 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
7710 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
7711 Remove second argument, change how error is called.
7712 (aarch64_layout_arg): Remove second argument from
7713 aarch64_err_no_fpadvsimd call.
7714 (aarch64_init_cumulative_args): Ditto.
7715 (aarch64_gimplify_va_arg_expr): Ditto.
7716 * config/aarch64/aarch64.md (mov<mode>): Ditto.
7718 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
7720 * config/i386/i386.md (simple_return_indirect_internal): New expander.
7721 (*simple_return_indirect_internal<mode>): Rename from
7722 simple_return_indirect_internal. Use W mode iterator.
7723 (rstorssp): New expander.
7724 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
7725 (clrssbsy): New expander.
7726 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
7728 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
7730 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
7732 (cmse_check_pointed_object): Likewise.
7734 2018-06-05 Martin Liska <mliska@suse.cz>
7736 PR gcov-profile/47618
7737 * doc/invoke.texi: Document how -fprofile-dir format
7740 2018-06-05 Richard Biener <rguenther@suse.de>
7742 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
7743 removal pretend DOM info isn't available so we do not update
7744 it and only remove edges, not dominated blocks. Actually free
7745 DOM info in case we removed something. Remove unreachable blocks.
7746 (mfb_keep_latches): Work with either DOM info or marked backedges.
7747 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
7748 first. Mark backedges if DOM info isn't available.
7749 (Re-)compute DOM info after cleanup_control_flow_pre.
7751 2018-06-05 Richard Biener <rguenther@suse.de>
7753 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
7754 (locus_discrim_hasher::hash): Adjust.
7755 (locus_discrim_hasher::equal): Likewise.
7756 (next_discriminator_for_locus): Work on line directly.
7757 (same_line_p): Pass in expanded locus1 as well.
7758 (assign_discriminators): Avoid redundant location expansions.
7760 2018-06-05 Richard Biener <rguenther@suse.de>
7762 PR tree-optimization/86046
7763 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
7764 if required after clearing TREE_ADDRESSABLE.
7766 2018-06-05 Richard Biener <rguenther@suse.de>
7768 PR tree-optimization/86047
7769 * tree-ssa-loop.c (for_each_index): Glob handling of all
7770 decls and constants and really handle all of them.
7772 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7775 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
7776 qualifier_void_pointer and qualifier_const_void_pointer.
7777 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
7778 (arm_init_builtins): Handle the above.
7779 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
7780 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
7781 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
7784 2018-06-05 Martin Liska <mliska@suse.cz>
7786 * auto-profile.c (read_autofdo_file): Do not use
7787 gcov_ctr_summary struct.
7788 (afdo_callsite_hot_enough_for_early_inline): Likewise.
7789 * coverage.c (struct counts_entry): Likewise.
7790 (read_counts_file): Read just single summary entry.
7791 (get_coverage_counts): Use gcov_summary struct.
7792 * coverage.h (get_coverage_counts): Likewise.
7793 * gcov-dump.c (dump_working_sets): Likewise.
7794 (tag_summary): Dump just single summary.
7795 * gcov-io.c (gcov_write_summary): Write just histogram
7797 (gcov_read_summary): Read just single summary.
7798 (compute_working_sets): Use gcov_summary struct.
7799 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
7800 of GCOV_COUNTERS_SUMMABLE.
7801 (GCOV_COUNTERS_SUMMABLE): Remove.
7802 (GCOV_FIRST_VALUE_COUNTER): Replace with
7803 GCOV_COUNTER_V_INTERVAL.
7804 (struct gcov_ctr_summary): Remove.
7805 (struct gcov_summary): Directly use fields of former
7807 (compute_working_sets): Use gcov_summary struct.
7808 * gcov.c (read_count_file): Do not use ctrs fields.
7809 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
7811 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
7813 * profile.c: Likewise.
7814 * profile.h: Likewise.
7816 2018-06-05 Martin Liska <mliska@suse.cz>
7818 PR gcov-profile/84846
7819 * gcov.c (output_lines): Print working directory only
7820 in intermediate format.
7822 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
7824 * config/s390/s390-builtin-types.def: Add void function type.
7825 * config/s390/s390-builtins.def: Use the function type for the
7828 2018-06-04 Jim Wilson <jimw@sifive.com>
7830 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
7832 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
7833 and maybe_eh_return. Change regno to unsigned int. Use new args to
7834 handle EH_RETURN_DATA_REGNO registers properly.
7835 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
7836 (riscv_expand_epilogue): Update comment. Change argument name and
7837 type. Update code to use new name and type. Pass new args to
7838 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
7840 * config/riscv/riscv.md (NORMAL_RETURN): New.
7841 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
7842 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
7843 (eh_return): Call gen_eh_return_internal and emit barrier.
7844 (eh_return_internal): Call riscv_expand_epilogue.
7846 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
7848 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
7849 bit_insertion field and declare can_be_merged_into method.
7850 (merged_store_group::can_be_merged_into): New method.
7851 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
7852 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
7853 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
7855 2018-06-04 Richard Biener <rguenther@suse.de>
7857 PR tree-optimization/85955
7858 * builtins.c (fold_builtin_sincos): Convert pointers to
7859 destination to appropriate type before dereferencing.
7861 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
7863 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
7865 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
7867 * expr.c (expand_expr_real_1): Force the operand into memory if
7868 its TYPE_MODE is BLKmode and if there is no integer mode for
7869 the number of bits being extracted.
7871 2018-06-04 Jakub Jelinek <jakub@redhat.com>
7875 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
7876 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
7878 2018-06-04 Richard Biener <rguenther@suse.de>
7880 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
7881 (cleanup_tree_cfg_noloop): ... single caller. Do
7882 start_recording_case_labels later.
7884 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
7886 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
7887 to _IMMINTRIN_H_INCLUDED.
7888 * config/i386/pconfigintrin.h: Ditto.
7889 * config/i386/waitpkgintrin.h: Ditto.
7890 * config/i386/immintrin.h: Add includes for sgxintrin.h,
7891 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
7892 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
7893 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
7894 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
7895 waitpkgintrin.h and cldemoteintrin.h.
7897 2018-06-04 Richard Biener <rguenther@suse.de>
7899 PR tree-optimization/86038
7900 * tracer.c (find_best_successor): Check probability for
7901 being initialized, bail out if not.
7903 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
7906 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
7907 of bits to ignore when comparing architectures.
7909 2018-06-04 Jakub Jelinek <jakub@redhat.com>
7911 PR tree-optimization/69615
7912 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
7913 maximum or minimum of the type, try to merge it also as if
7914 range1 is + [-, x - 1] or + [x + 1, -].
7917 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
7919 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
7921 PR tree-optimization/86034
7922 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
7923 the unsigned bitfield type in a bit insertion sequence if it does not
7924 have a larger precision than the bitfield size.
7925 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
7927 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
7929 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
7931 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
7933 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
7934 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
7935 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
7936 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
7938 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
7940 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
7941 Disable -fdelete-null-pointer-checks for ELF toolchain.
7943 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
7944 Kito Cheng <kito.cheng@gmail.com>
7946 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
7947 (nds32le-*-*, nds32be-*-*): Integrate checking process.
7948 (nds32*-*-*): Add glibc and uclibc conditions.
7949 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
7950 (TARGET_EXCEPT_UNWIND_INFO): Define.
7951 * config/nds32/elf.h: New file.
7952 * config/nds32/linux.h: New file.
7953 * config/nds32/nds32-elf.opt: New file.
7954 * config/nds32/nds32-linux.opt: New file.
7955 * config/nds32/nds32-fp-as-gp.c
7956 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
7957 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
7959 (nds32_asm_file_end): Ditto.
7960 (nds32_print_operand): Ditto.
7961 (nds32_insert_attributes): Ditto.
7962 (nds32_init_libfuncs): New function.
7963 (TARGET_HAVE_TLS): Define.
7964 (TARGET_INIT_LIBFUNCS): Define.
7965 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
7967 (TARGET_ELF): Apply different mcmodel setting.
7968 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
7969 been migrated into elf.h and linux.h files.
7970 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
7971 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
7972 (mcmodel): The content has been migrated into nds32-elf.opt and
7973 nds32-linux.opt files.
7974 * config/nds32/t-elf: New file.
7975 * config/nds32/t-linux: New file.
7977 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
7978 Shiva Chen <shiva0217@gmail.com>
7980 * config/nds32/constants.md (unspec_volatile_element): Add
7981 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
7982 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
7984 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
7985 (make_pass_nds32_fp_as_gp): Declare.
7986 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
7988 (nds32_asm_function_end_prologue): Remove unused asm output.
7989 (nds32_asm_function_begin_epilogue): Remove unused asm output.
7990 (nds32_asm_file_start): Output necessary fp_as_gp information.
7991 (nds32_option_override): Adjust register usage.
7992 (nds32_expand_prologue): Consider fp_as_gp situation.
7993 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
7994 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
7997 (simple_return): Ditto.
7998 (omit_fp_begin): Output special directive for fp_as_gp.
7999 (omit_fp_end): Output special directive for fp_as_gp.
8000 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
8001 mforbid-fp-as-gp): New options.
8003 2018-06-01 Mark Wielaard <mark@klomp.org>
8005 * dwarf2out.c (dwarf2out_finish): Remove generation of
8006 DW_AT_loclists_base.
8008 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
8010 * gimple-ssa-store-merging.c: Include gimple-fold.h.
8011 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
8012 (struct merged_store_group): Add bit_insertion field.
8013 (dump_char_array): Use standard hexadecimal format.
8014 (merged_store_group::merged_store_group): Set bit_insertion to false.
8015 (merged_store_group::apply_stores): Use optimal buffer size. Deal
8016 with BIT_INSERT_EXPR stores. Move up code updating the mask and
8017 also print the mask in the dump file.
8018 (pass_store_merging::gate): Minor tweak.
8019 (imm_store_chain_info::coalesce_immediate): Fix wrong association
8020 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
8021 stores with INTEGER_CST stores.
8022 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
8023 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
8024 and use it throughout. Generate bit insertion sequences if need be.
8025 (pass_store_merging::process_store): Remove redundant condition.
8026 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
8028 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
8030 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
8031 the 128-bit floating point types. Fix function comment.
8033 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8035 * config/aarch64/aarch64-simd.md
8036 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
8038 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
8041 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
8043 PR tree-optimization/85989
8044 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
8046 (backprop::intersect_uses): Check it when deciding whether this
8047 is a backedge reference.
8048 (backprop::process_block): Add each phi to m_visited_phis
8049 after visiting it, then clear it at the end.
8051 2018-06-01 Richard Biener <rguenther@suse.de>
8053 * tree-vectorizer.h (vect_dr_stmt): New function.
8054 (vect_get_load_cost): Adjust.
8055 (vect_get_store_cost): Likewise.
8056 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
8057 Use vect_dr_stmt instead of DR_SMTT.
8058 (vect_record_base_alignments): Likewise.
8059 (vect_calculate_target_alignment): Likewise.
8060 (vect_compute_data_ref_alignment): Likewise and make static.
8061 (vect_update_misalignment_for_peel): Likewise.
8062 (vect_verify_datarefs_alignment): Likewise.
8063 (vector_alignment_reachable_p): Likewise.
8064 (vect_get_data_access_cost): Likewise. Pass down
8065 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
8066 (vect_get_peeling_costs_all_drs): Likewise.
8067 (vect_peeling_hash_get_lowest_cost): Likewise.
8068 (vect_enhance_data_refs_alignment): Likewise.
8069 (vect_find_same_alignment_drs): Likewise.
8070 (vect_analyze_data_refs_alignment): Likewise.
8071 (vect_analyze_group_access_1): Likewise.
8072 (vect_analyze_group_access): Likewise.
8073 (vect_analyze_data_ref_access): Likewise.
8074 (vect_analyze_data_ref_accesses): Likewise.
8075 (vect_vfa_segment_size): Likewise.
8076 (vect_small_gap_p): Likewise.
8077 (vectorizable_with_step_bound_p): Likewise.
8078 (vect_prune_runtime_alias_test_list): Likewise.
8079 (vect_analyze_data_refs): Likewise.
8080 (vect_supportable_dr_alignment): Likewise.
8081 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
8082 (vect_gen_prolog_loop_niters): Likewise.
8083 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
8084 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
8086 (vect_recog_mask_conversion_pattern): Likewise.
8087 (vect_try_gather_scatter_pattern): Likewise.
8088 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
8089 to vect_get_store_cost.
8090 (vect_get_store_cost): Get stmt_info instead of DR.
8091 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
8092 (vect_get_load_cost): Get stmt_info instead of DR.
8094 2018-06-01 Richard Biener <rguenther@suse.de>
8097 * gimple-fold.c (var_decl_component_p): Also allow offsetted
8098 vars wrapped in MEM_REFs.
8100 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
8102 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
8103 Fix subreg tests so that we only return a choice between
8104 GENERAL_REGS and FP_REGS if the original classes included both.
8106 2018-06-01 Richard Biener <rguenther@suse.de>
8109 * tree-ssa-structalias.c (get_function_part_constraint):
8110 Handle NULL fi->decl.
8111 (find_func_aliases_for_call): Properly handle indirect
8112 fi from direct call.
8113 (find_func_clobbers): Likewise.
8114 (ipa_pta_execute): Likewise.
8115 (create_variable_info_for): For functions that are ifunc_resolver
8116 resolve to a varinfo that contains the result of the resolver call.
8117 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
8120 2018-05-31 Michael Collison <michael.collison@arm.com>
8122 * config/aarch64/aarch64.md:
8123 (*fix_to_zero_extenddfdi2): New pattern.
8124 * gcc.target/aarch64/fix_extend1.c: New testcase.
8126 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
8130 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
8131 and BUILT_IN_STRNCMP_EQ.
8132 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
8133 BUILT_IN_STRNCMP_EQ.
8134 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
8135 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
8136 (gimple_fold_builtin): Likewise.
8137 * tree-ssa-strlen.c (compute_string_length): New function.
8138 (determine_min_obsize): New function.
8139 (handle_builtin_string_cmp): New function to handle calls to
8140 string compare functions.
8141 (strlen_optimize_stmt): Add handling to builtin string compare
8143 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
8144 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
8145 * tree.c (build_common_builtin_nodes): Add new defines of
8146 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
8148 2018-05-31 Jakub Jelinek <jakub@redhat.com>
8151 * bb-reorder.c (pass_partition_blocks::gate): Return false for
8152 functions with naked attribute.
8154 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
8156 * config/i386/sse.md (avx_vec_concat<mode>):
8157 Substitute concat_tg_mode mode attribute with xtg_mode.
8158 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
8159 (concat_tg_mode): Remove mode attribute.
8161 2018-05-31 Martin Sebor <msebor@redhat.com>
8164 * calls.c (alloc_max_size): Correct a logic error/typo.
8165 Treat excessive arguments as infinite. Warn for invalid arguments.
8166 * doc/invoke.texi (-Walloc-size-larger-than): Update.
8168 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
8171 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
8172 and movx for Haswell.
8174 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
8175 Cesar Philippidis <cesar@codesourcery.com>
8178 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
8179 when emitting error on private/firstprivate reductions.
8180 * omp-low.c (lower_omp_target): Avoid reference-type processing
8181 on pointers for firstprivate clause.
8183 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
8185 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
8188 * config/aarch64/aarch64-simd.md
8189 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
8190 (aarch64_ld1_x3_<mode>): Likewise
8191 (aarch64_st1x2<VALLDIF:mode>): Likewise
8192 (aarch64_st1_x2_<mode>): Likewise
8193 (aarch64_st1x3<VALLDIF:mode>): Likewise
8194 (aarch64_st1_x3_<mode>): Likewise
8195 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
8196 (vld1_s8_x3): Likewise.
8197 (vld1_u16_x3): Likewise.
8198 (vld1_s16_x3): Likewise.
8199 (vld1_u32_x3): Likewise.
8200 (vld1_s32_x3): Likewise.
8201 (vld1_u64_x3): Likewise.
8202 (vld1_s64_x3): Likewise.
8203 (vld1_f16_x3): Likewise.
8204 (vld1_f32_x3): Likewise.
8205 (vld1_f64_x3): Likewise.
8206 (vld1_p8_x3): Likewise.
8207 (vld1_p16_x3): Likewise.
8208 (vld1_p64_x3): Likewise.
8209 (vld1q_u8_x3): Likewise.
8210 (vld1q_s8_x3): Likewise.
8211 (vld1q_u16_x3): Likewise.
8212 (vld1q_s16_x3): Likewise.
8213 (vld1q_u32_x3): Likewise.
8214 (vld1q_s32_x3): Likewise.
8215 (vld1q_u64_x3): Likewise.
8216 (vld1q_s64_x3): Likewise.
8217 (vld1q_f16_x3): Likewise.
8218 (vld1q_f32_x3): Likewise.
8219 (vld1q_f64_x3): Likewise.
8220 (vld1q_p8_x3): Likewise.
8221 (vld1q_p16_x3): Likewise.
8222 (vld1q_p64_x3): Likewise.
8223 (vst1_s64_x2): Likewise.
8224 (vst1_u64_x2): Likewise.
8225 (vst1_f64_x2): Likewise.
8226 (vst1_s8_x2): Likewise.
8227 (vst1_p8_x2): Likewise.
8228 (vst1_s16_x2): Likewise.
8229 (vst1_p16_x2): Likewise.
8230 (vst1_s32_x2): Likewise.
8231 (vst1_u8_x2): Likewise.
8232 (vst1_u16_x2): Likewise.
8233 (vst1_u32_x2): Likewise.
8234 (vst1_f16_x2): Likewise.
8235 (vst1_f32_x2): Likewise.
8236 (vst1_p64_x2): Likewise.
8237 (vst1q_s8_x2): Likewise.
8238 (vst1q_p8_x2): Likewise.
8239 (vst1q_s16_x2): Likewise.
8240 (vst1q_p16_x2): Likewise.
8241 (vst1q_s32_x2): Likewise.
8242 (vst1q_s64_x2): Likewise.
8243 (vst1q_u8_x2): Likewise.
8244 (vst1q_u16_x2): Likewise.
8245 (vst1q_u32_x2): Likewise.
8246 (vst1q_u64_x2): Likewise.
8247 (vst1q_f16_x2): Likewise.
8248 (vst1q_f32_x2): Likewise.
8249 (vst1q_f64_x2): Likewise.
8250 (vst1q_p64_x2): Likewise.
8251 (vst1_s64_x3): Likewise.
8252 (vst1_u64_x3): Likewise.
8253 (vst1_f64_x3): Likewise.
8254 (vst1_s8_x3): Likewise.
8255 (vst1_p8_x3): Likewise.
8256 (vst1_s16_x3): Likewise.
8257 (vst1_p16_x3): Likewise.
8258 (vst1_s32_x3): Likewise.
8259 (vst1_u8_x3): Likewise.
8260 (vst1_u16_x3): Likewise.
8261 (vst1_u32_x3): Likewise.
8262 (vst1_f16_x3): Likewise.
8263 (vst1_f32_x3): Likewise.
8264 (vst1_p64_x3): Likewise.
8265 (vst1q_s8_x3): Likewise.
8266 (vst1q_p8_x3): Likewise.
8267 (vst1q_s16_x3): Likewise.
8268 (vst1q_p16_x3): Likewise.
8269 (vst1q_s32_x3): Likewise.
8270 (vst1q_s64_x3): Likewise.
8271 (vst1q_u8_x3): Likewise.
8272 (vst1q_u16_x3): Likewise.
8273 (vst1q_u32_x3): Likewise.
8274 (vst1q_u64_x3): Likewise.
8275 (vst1q_f16_x3): Likewise.
8276 (vst1q_f32_x3): Likewise.
8277 (vst1q_f64_x3): Likewise.
8278 (vst1q_p64_x3): Likewise.
8280 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
8282 * config/msp430/msp430.c (msp430_output_labelref): Prepend
8283 user_label_prefix to name.
8285 * tree-core.h: Update comment about the format of NAME string
8286 passed to handler in attribute_spec.
8288 * config/msp430/msp430.md: Remove erroneous subreg expression from
8289 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
8290 zero_extend{q,h}isi2.
8292 2018-05-30 Borislav Petkov <bp@suse.de>
8294 * doc/extend.texi: Document some architecture specific
8295 constraints and sort entries.
8297 2018-05-30 Martin Sebor <msebor@redhat.com>
8300 * builtins.c (expand_builtin_stpcpy_1): New function.
8301 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
8302 only if the former succeeds.
8304 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
8306 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
8309 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
8311 * doc/invoke.texi (-flinker-output): Document
8313 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
8315 * passes.c (ipa_write_summaries): Only modify statements if body
8317 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
8318 incrementally linking.
8319 (ipa_passes): Likewise.
8320 * lto-cgraph.c (lto_output_node): When incrementally linking do not
8321 pass down resolution info.
8322 * common.opt (flag_incremental_link): Update info.
8323 * gcc.c (plugin specs): Turn flinker-output=* to
8324 -plugin-opt=-linker-output-known
8325 * toplev.c (compile_file): Also cut compilation when doing incremental
8327 * flag-types. (enum lto_partition_model): Add
8328 LTO_LINKER_OUTPUT_NOLTOREL.
8329 (invoke.texi): Add -flinker-output docs.
8330 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
8331 link same way as WPA; do not stream in dead initializers.
8333 * dwarf2out.c (dwarf2out_die_ref_for_decl,
8334 darf2out_register_external_decl): Support incremental link.
8336 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
8338 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
8340 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
8342 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
8343 it down to simple_object_copy_lto_debug_sections.
8344 (run_gcc): Determine incremental LTO link time and configure
8345 lto1 into non-wpa mode, disable renaming of debug sections.
8347 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
8349 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
8350 descriptions of various incorrectly documented functions.
8352 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
8355 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
8356 address check not strict.
8358 2018-05-30 Richard Biener <rguenther@suse.de>
8360 PR tree-optimization/85964
8361 * tracer.c (better_p): Drop initialized count check, we only
8362 call the function with initialized counts now.
8363 (find_best_successor): Do find a best edge if one
8364 has uninitialized count.
8365 (find_best_predecessor): Likewise. Do BB frequency check only
8366 if count is initialized.
8368 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
8370 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
8371 (aarch64_ldrstr_offset_compare): New.
8372 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
8373 load/store orderings.
8374 (aarch64_gen_adjusted_ldpstp): Likewise.
8376 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
8378 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
8379 Check for subset of GENERAL_REGS and FP_REGS.
8380 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
8383 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
8385 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
8386 and wi::to_poly_offset. Add the current offset and then check
8387 whether the sum fits, rather than using an unchecked addition of
8388 a checked term. Check for a shwi rather than a uhwi.
8389 * expr.c (get_bit_range): Use tree_to_poly_uint64.
8390 (store_constructor): Use poly_int_tree_p.
8391 (expand_expr_real_1): Likewise.
8392 * function.c (assign_temp): Likewise.
8393 * fold-const.c (const_binop): Use poly_int_tree_p and
8395 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
8397 * ipa-icf-gimple.c (func_checker::compare_operand): Use
8398 to_poly_offset for MEM offsets.
8399 * ipa-icf.c (sem_variable::equals): Likewise.
8400 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
8401 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
8402 wi::to_poly_offset for BIT_FIELD_REF offsets.
8403 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
8405 * var-tracking.c (emit_note_insn_var_location): Use
8406 tree_to_poly_uint64.
8408 2018-05-29 Jim Wilson <jimw@sifive.com>
8410 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
8412 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
8415 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
8416 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
8418 (sse4_1_round<mode>2): Use nonimmediate_operand
8419 for operand 1 predicate.
8421 2018-05-29 Martin Sebor <msebor@redhat.com>
8422 Richard Biener <rguenther@suse.de>
8425 * calls.c (get_size_range): Call determine_value_range instead
8426 of get_value_range..
8427 * tree-vrp.h (determine_value_range): Declared new function.
8428 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
8430 2018-05-29 Richard Biener <rguenther@suse.de>
8432 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
8433 sure to use non-pattern stmts for get_earlier_stmt arguments.
8434 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
8435 called on pattern stmts.
8436 (get_later_stmt): Likewise.
8438 2018-05-29 Martin Liska <mliska@suse.cz>
8440 PR gcov-profile/85759
8441 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
8444 2018-05-29 Jakub Jelinek <jakub@redhat.com>
8446 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
8448 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
8452 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
8453 VEC_PACK_FLOAT_EXPR): New tree codes.
8454 * tree-pretty-print.c (op_code_prio): Handle
8455 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
8456 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
8457 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
8458 * tree-inline.c (estimate_operator_cost): Likewise.
8459 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
8460 * fold-const.c (const_binop): Likewise.
8461 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
8462 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
8463 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
8464 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
8465 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
8466 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
8467 * expr.c (expand_expr_real_2): Likewise.
8468 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
8469 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
8470 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
8472 * optabs.c (expand_widen_pattern_expr): For
8473 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
8474 sign from result type rather than operand's type.
8475 (expand_binop_directly): For vec_packu_float_optab and
8476 vec_packs_float_optab allow result type to be different from operand's
8478 * optabs-tree.c (optab_for_tree_code): Handle
8479 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
8480 VEC_PACK_FLOAT_EXPR. Formatting fixes.
8481 * tree-vect-generic.c (expand_vector_operations_1): Handle
8482 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
8483 VEC_PACK_FLOAT_EXPR.
8484 * tree-vect-stmts.c (supportable_widening_operation): Handle
8486 (supportable_narrowing_operation): Handle FLOAT_EXPR.
8487 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
8488 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
8489 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
8490 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
8492 (vec_pack<floatprefix>_float_<mode>): New expander.
8493 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
8495 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
8496 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
8497 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
8498 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
8499 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
8501 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
8502 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
8503 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
8504 VEC_PACK_FLOAT_EXPR): Document.
8506 2018-05-29 Richard Biener <rguenther@suse.de>
8508 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
8510 (stmt_vec_info_vec): Make pointer.
8511 (init_stmt_vec_info_vec): Remove.
8512 (free_stmt_vec_info_vec): Likewise.
8513 (set_stmt_vec_info_vec): New function.
8514 (free_stmt_vec_infos): Likewise.
8515 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
8516 (set_vinfo_for_stmt): Likewise.
8517 (get_earlier_stmt): Likewise.
8518 (get_later_stmt): Likewise.
8519 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
8520 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
8521 (vec_info::~vec_info): Free stmt_vec_infos.
8522 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
8523 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
8524 (pass_slp_vectorize::execute): Likewise.
8525 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
8526 (free_stmt_vec_info_vec): Likewise.
8527 (set_stmt_vec_info_vec): New function.
8528 (free_stmt_vec_infos): Likewise.
8529 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
8530 the global stmt_vec_info_vec.
8531 * tree-parloops.c (gather_scalar_reductions): Use
8532 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
8535 2018-05-29 Richard Biener <rguenther@suse.de>
8537 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
8539 2018-05-29 Martin Liska <mliska@suse.cz>
8540 David Malcolm <dmalcolm@redhat.com>
8542 * vec.c (test_reverse): New.
8543 (vec_c_tests): Add new test.
8544 * vec.h (vl_ptr>::reverse): New function.
8546 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
8548 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
8550 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
8553 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
8555 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
8557 2018-05-28 Richard Biener <rguenther@suse.de>
8559 PR tree-optimization/85933
8560 * tree-vect-data-refs.c (vect_record_base_alignments): Only
8561 look at stmts marked as vectorizable.
8563 2018-05-28 Richard Biener <rguenther@suse.de>
8565 PR tree-optimization/85934
8566 * tree-vect-generic.c (expand_vector_operations_1): Hoist
8567 vector boolean check before scalar optimization.
8569 2018-05-28 Jakub Jelinek <jakub@redhat.com>
8571 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
8574 2018-05-28 Mark Wielaard <mark@klomp.org>
8576 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
8577 if it is an expression containing a minus sign.
8579 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
8581 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
8583 2018-05-27 Paul Koning <ni1d@arrl.net>
8585 * config/pdp11/pdp11.md (truncsihi2): Remove.
8587 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
8588 Chung-Ju Wu <jasonwucj@gmail.com>
8590 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
8592 (unaligned_store_dw): Ditto.
8593 * config/nds32/nds32-memory-manipulation.c
8594 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
8595 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
8596 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
8597 (emit_setmem_word_loop): Rename to ...
8598 (emit_setmem_doubleword_loop): ... this.
8599 (nds32_gen_dup_4_byte_to_word_value): New function.
8600 (nds32_gen_dup_8_byte_to_double_word_value): New function.
8601 (nds32_expand_setmem_loop): Refine implementation.
8602 (nds32_expand_setmem_loop_v3m): Ditto.
8603 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
8606 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
8608 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
8610 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
8612 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
8613 (nds32_init_machine_status): Initialize machine->attr_naked_p and
8614 machine->attr_no_prologue_p.
8615 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
8616 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
8617 (nds32_expand_epilogue): Consider attr_naked_p.
8618 (nds32_expand_epilogue_v3pop): Likewise.
8619 (nds32_can_use_return_insn): Likewise.
8620 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
8621 attr_no_prologue_p fields.
8622 * config/nds32/nds32.opt (mret-in-naked-func): New option.
8624 2018-05-27 Jakub Jelinek <jakub@redhat.com>
8627 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
8629 * config/i386/sse.md
8630 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
8632 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
8634 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
8636 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
8638 (*<floatsuffix>floatv2div2sf2): Rename to ...
8639 (*float<floatunssuffix>v2div2sf2): ... this.
8640 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
8641 (float<floatunssuffix>v2div2sf2_mask): ... this.
8642 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
8643 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
8644 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
8646 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
8648 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
8650 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
8652 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
8654 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
8656 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
8657 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
8658 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
8659 gen_ufix_truncv8dfv8si2.
8660 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
8661 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
8662 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
8663 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
8664 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
8665 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
8666 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
8667 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
8669 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
8673 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
8675 2018-05-25 Jim Wilson <jimw@sifive.com>
8677 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
8678 * config/riscv/riscv.c (struct machine_function): Add
8679 interrupt_handler_p and attribute_checked_p fields.
8680 (riscv_attribute_table): Add interrupt.
8681 (riscv_interrupt_type_p): New.
8682 (riscv_save_reg_p): Save extra regs for interrupt handler.
8683 (riscv_use_save_libcall): Return false for interrupt handler.
8684 (riscv_first_stack_step): Add forward declaration.
8685 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
8686 for interrupt handler with large frame. Use it for saved reg list.
8687 (riscv_expand_prologue): Move flag_stack_usage_info support to
8688 eliminate duplication.
8689 (riscv_expand_epilogue): Generate mret for interrupt handler.
8690 (riscv_epilogue_uses): New.
8691 (riscv_can_use_return_insn): Return false for interrupt handler.
8692 (riscv_function_ok_for_sibcall): Likewise.
8693 (riscv_set_current_function): Add interrupt handler support.
8694 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
8695 * config/riscv/riscv.md (UNSPECV_MRET): New.
8697 (riscv_frflags, riscv_fsflags): Use tab after opcode.
8699 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
8701 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
8703 PR tree-optimization/85712
8704 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
8705 this candidate has already been replaced in-situ by a copy.
8707 2018-05-25 Jason Merrill <jason@redhat.com>
8709 PR c++/80485 - inline function non-zero address.
8710 * symtab.c (nonzero_address): Check DECL_COMDAT.
8712 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
8715 * config/alpha/alpha.md (ashlsi3): New insn pattern.
8716 (*ashlsi_se): Rename from *ashldi_se. Define as sign
8717 extension of SImode operation. Use const123_operand predicate.
8718 (*saddsi_1): Remove.
8719 (*saddl_se_1): Ditto.
8721 (*ssubl_se_1): Ditto.
8722 * config/alpha/predicates.md (const123_operand): New predicate.
8723 * config/alpha/constraints.md (P): Use IN_RANGE.
8725 2018-05-25 Richard Biener <rguenther@suse.de>
8727 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
8729 (ref_maybe_used_by_stmt_p): Likewise.
8730 (stmt_may_clobber_ref_p): Likewise.
8731 (stmt_may_clobber_ref_p_1): Likewise.
8732 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
8734 (ref_maybe_used_by_stmt_p): Likewise.
8735 (stmt_may_clobber_ref_p): Likewise.
8736 (stmt_may_clobber_ref_p_1): Likewise.
8737 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
8738 the alias oracle to disambiguate DRs with stmts DR analysis
8740 (vect_analyze_data_refs): Do not give up on not analyzable
8741 DRs for BB vectorization. Remove code truncating the dataref
8744 2018-05-25 Jakub Jelinek <jakub@redhat.com>
8747 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
8748 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
8749 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
8751 2018-05-25 Richard Biener <rguenther@suse.de>
8753 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
8754 function, combining stmt data ref gathering and fatal analysis
8756 (vect_analyze_data_refs): Remove now redudnant code and simplify.
8757 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
8758 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
8759 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
8760 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
8762 2018-05-25 Bin Cheng <bin.cheng@arm.com>
8764 PR tree-optimization/85720
8765 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
8766 SCC if all partitions are builtins.
8767 (version_loop_by_alias_check): New parameter. Generate cancelable
8768 runtime alias check if all partitions are builtins.
8769 (distribute_loop): Update call to above function.
8771 2018-05-25 Bin Cheng <bin.cheng@arm.com>
8773 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
8774 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
8775 (parm_default_def_partition_arg): Ditto.
8776 (set_parm_default_def_partition): Ditto.
8777 (get_parm_default_def_partitions): Ditto and make it static.
8778 (get_undefined_value_partitions): Ditto and make it static.
8779 (remove_ssa_form): Refactor call to init_var_map here.
8780 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
8781 computation for loop region.
8782 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
8783 (register_default_def): Delete.
8784 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
8785 (parm_default_def_partition_arg): Ditto.
8786 (set_parm_default_def_partition): Ditto.
8787 (get_parm_default_def_partitions): Ditto and make it static.
8788 (get_undefined_value_partitions): Ditto and make it static.
8789 (coalesce_with_default, coalesce_with_default): Update comment.
8790 (create_coalesce_list_for_region): New func factored out from
8791 create_outofssa_var_map.
8792 (populate_coalesce_list_for_outofssa): New func factored out from
8793 create_outofssa_var_map and coalesce_ssa_name.
8794 (create_outofssa_var_map): Delete.
8795 (coalesce_ssa_name): Refactor to support live range computation.
8796 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
8797 (get_parm_default_def_partitions): Delete.
8798 (get_undefined_value_partitions): Ditto.
8799 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
8800 computation for loop region.
8801 (new_tree_live_info, loe_visit_block): Ditto.
8802 (live_worklist, set_var_live_on_entry): Ditto.
8803 (calculate_live_on_exit, verify_live_on_entry): Ditto.
8804 * tree-ssa-live.h (struct _var_map): New fields.
8805 (init_var_map): Change decl.
8806 (region_contains_p): New.
8808 2018-05-25 Bin Cheng <bin.cheng@arm.com>
8810 * tree-ssa-live.h (live_merge_and_clear): Delete.
8812 2018-05-25 Richard Biener <rguenther@suse.de>
8815 * tree-dump.c (dequeue_and_dump): Remove access to removed
8816 operand 2 of a SWITCH_EXPR.
8818 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
8820 * doc/sourcebuild.texi (vect_double_cond_arith): Include
8821 multiplication and division.
8822 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
8823 (cond_udiv@var{m}, cond_umod@var{m}): Document.
8824 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
8825 (cond_udiv_optab, cond_umod_optab): New optabs.
8826 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
8827 (IFN_COND_RDIV): New internal functions.
8828 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
8829 TRUNC_MOD_EXPR and RDIV_EXPR.
8830 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
8831 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
8833 (SVE_INT_BINARY): Include mult.
8834 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
8835 (optab, sve_int_op): Handle mult.
8836 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
8838 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
8839 for SVE_INT_BINARY_SD.
8841 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
8843 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
8844 (optab, sve_int_op): Handle div and udiv.
8845 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
8846 for SVE_INT_BINARY_SD.
8847 (*<optab><mode>3): New insn for the same.
8849 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
8851 * tree-vect-patterns.c: Include predict.h.
8852 (vect_recog_divmod_pattern): Restrict check for division support
8853 to when optimizing for size.
8855 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
8857 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
8858 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
8859 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
8860 (gimple_match_op::set_op): Likewise.
8861 (gimple_resimplify4): Declare.
8862 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
8863 (expr::gen_transform): Likewise.
8864 (decision_tree::gen): Generate a simplification routine for 4 operands.
8865 * gimple-match-head.c (gimple_simplify): Add an overload for
8866 4 operands. In the top-level function, handle up to 4 call
8867 arguments and call gimple_resimplify4.
8868 (gimple_resimplify4): New function.
8869 (build_call_internal): Pass a fourth operand.
8870 (maybe_push_to_seq): Likewise.
8871 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
8872 Fold VEC_COND_EXPRs of an operation and a default value into
8873 an IFN_COND_* function if possible.
8874 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
8876 (SVE_COND_FP_BINARY): Include them.
8877 (optab, sve_fp_op): Handle them.
8878 (SVE_INT_BINARY_REV): New code iterator.
8879 (SVE_COND_FP_BINARY_REV): New int iterator.
8880 (commutative): New int attribute.
8881 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
8883 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
8885 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
8886 (*cond_<optab><mode>): New patterns for reversed operands.
8888 2018-05-25 Richard Biener <rguenther@suse.de>
8890 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
8891 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
8892 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
8893 (STMT_VINFO_GROUPED_ACCESS): Adjust.
8894 * tree-vect-data-refs.c (everywhere): Adjust users.
8895 * tree-vect-loop.c (everywhere): Likewise.
8896 * tree-vect-slp.c (everywhere): Likewise.
8897 * tree-vect-stmts.c (everywhere): Likewise.
8898 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
8900 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8902 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
8904 (gcc_cv_as_section_exclude): ... this.
8905 Try Solaris as #exclude syntax.
8906 * configure: Regenerate.
8907 * config.in: Regenerate.
8908 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
8910 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
8911 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
8913 * varasm.c (default_elf_asm_named_section): Don't check if
8914 HAVE_GAS_SECTION_EXCLUDE is defined.
8916 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
8918 * doc/md.texi: Update the documentation of the cond_* optabs
8919 to mention the new final operand. Fix GET_MODE_NUNITS call.
8920 Describe the scalar case too.
8921 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
8922 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
8924 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
8925 (get_conditional_internal_fn): Update comment.
8926 * tree-vect-loop.c (vectorizable_reduction): Pass the original
8927 accumulator value as a final argument to conditional functions.
8928 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
8929 a define_expand and add an "else" operand. Assert for now that
8930 the else operand is equal to operand 2. Use SVE_INT_BINARY and
8931 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
8932 (*cond_<optab><mode>): New patterns.
8933 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
8934 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
8935 (UNSPEC_COND_EOR): Delete.
8936 (optab): Remove associated mappings.
8937 (SVE_INT_BINARY): New code iterator.
8938 (sve_int_op): Remove int attribute and add "minus" to the code
8940 (SVE_COND_INT_OP): Delete.
8941 (SVE_COND_FP_OP): Rename to...
8942 (SVE_COND_FP_BINARY): ...this.
8944 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
8946 * optabs.c (can_reuse_operands_p): New function.
8947 (maybe_legitimize_operands): Try to reuse the results for
8950 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
8952 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
8953 Add {q} suffix to insn mnemonic.
8955 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
8957 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
8958 (msp430_warn_func_return): New.
8960 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
8962 * fold-const.c (tree_nonzero_bits): New function.
8963 * fold-const.h (tree_nonzero_bits): Likewise.
8964 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
8965 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
8967 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
8971 * varasm.c (assemble_alias): Check ifunc_resolver only on
8974 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
8977 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
8978 when memory input operand is handled.
8980 2018-05-24 Luis Machado <luis.machado@linaro.org>
8982 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
8984 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
8986 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
8988 * match.pd: Delay FMA folds until after vectorization.
8990 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
8993 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
8994 address check not strict.
8996 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
8998 * gimple-match.h (gimple_match_op): New class.
8999 (mprts_hook): Replace parameters with a gimple_match_op *.
9000 (maybe_build_generic_op): Likewise.
9001 (gimple_simplified_result_is_gimple_val): Replace parameters with
9002 a const gimple_match_op *.
9003 (gimple_simplify): Replace code_helper * and tree * parameters with
9004 a gimple_match_op * parameter.
9005 (gimple_resimplify1): Replace code_helper *, tree and tree *
9006 parameters with a gimple_match_op * parameter.
9007 (gimple_resimplify2): Likewise.
9008 (gimple_resimplify3): Likewise.
9009 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
9010 parameters with a gimple_match_op * parameter.
9011 * gimple-match-head.c (gimple_simplify): Change prototypes of
9012 auto-generated functions to take a gimple_match_op * instead of
9013 separate code_helper * and tree * parameters. Make the same
9014 change in the top-level overload and update calls to the
9015 gimple_resimplify routines. Update calls to the auto-generated
9016 functions and to maybe_push_res_to_seq in the publicly-facing
9017 operation-specific gimple_simplify overloads.
9018 (gimple_match_op::MAX_NUM_OPS): Define.
9019 (gimple_resimplify1): Replace rcode and ops with a single res_op
9020 parameter. Update call to gimple_simplify.
9021 (gimple_resimplify2): Likewise.
9022 (gimple_resimplify3): Likewise.
9023 (mprts_hook): Replace parameters with a gimple_match_op *.
9024 (maybe_build_generic_op): Likewise.
9025 (build_call_internal): Replace type, nargs and ops with
9026 a gimple_match_op *.
9027 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
9028 with a single gimple_match_op *. Update calls to mprts_hook,
9029 build_call_internal and gimple_simplified_result_is_gimple_val.
9030 Factor out code that is common to the tree_code and combined_fn cases.
9031 * genmatch.c (expr::gen_transform): Replace tem_code and
9032 tem_ops with a gimple_match_op called tem_op. Update calls
9033 to the gimple_resimplify functions and maybe_push_res_to_seq.
9034 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
9035 res_ops. Update call to the gimple_resimplify functions.
9036 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
9037 (decision_tree::gen): Make the functions take a gimple_match_op *
9038 called res_op instead of separate res_code and res_ops parameters.
9039 Update call accordingly.
9040 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
9041 and ops with a single res_op parameter. Update calls to
9042 maybe_build_generic_op and maybe_push_res_to_seq.
9043 (fold_stmt_1): Update calls to gimple_simplify and
9044 replace_stmt_with_simplification.
9045 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
9046 and gimple_simplified_result_is_gimple_val.
9047 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
9049 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
9050 with a gimple_match_op *.
9051 (vn_nary_build_or_lookup): Likewise. Update call to
9052 vn_nary_build_or_lookup_1.
9053 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
9054 gimple_match_op *. Update calls to the gimple_resimplify routines
9055 and to gimple_simplified_result_is_gimple_val.
9056 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
9057 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
9058 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
9059 (visit_nary_op): Likewise.
9060 (visit_reference_op_load): Likewise.
9062 2018-05-23 Luis Machado <luis.machado@linaro.org>
9064 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
9065 modifier for printing the step amount.
9067 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
9070 * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
9073 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
9075 * doc/sourcebuild.texi (Endianness): New subsubsection.
9077 2018-05-23 Luis Machado <luis.machado@linaro.org>
9079 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
9080 <prefetch_dynamic_strides>: New const bool field.
9081 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
9082 prefetch_dynamic_strides.
9083 (exynosm1_prefetch_tune): Likewise.
9084 (thunderxt88_prefetch_tune): Likewise.
9085 (thunderx_prefetch_tune): Likewise.
9086 (thunderx2t99_prefetch_tune): Likewise.
9087 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
9089 (aarch64_override_options_internal): Update to set
9090 PARAM_PREFETCH_DYNAMIC_STRIDES.
9091 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
9092 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
9093 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
9094 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
9095 prefetch-dynamic-strides setting.
9097 2018-05-23 Luis Machado <luis.machado@linaro.org>
9099 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
9100 <minimum_stride>: New const int field.
9101 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
9102 minimum_stride field defaulting to -1.
9103 (exynosm1_prefetch_tune): Likewise.
9104 (thunderxt88_prefetch_tune): Likewise.
9105 (thunderx_prefetch_tune): Likewise.
9106 (thunderx2t99_prefetch_tune): Likewise.
9107 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
9108 <default_opt_level>: Set to 3.
9109 (aarch64_override_options_internal): Update to set
9110 PARAM_PREFETCH_MINIMUM_STRIDE.
9111 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
9112 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
9113 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
9114 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
9115 stride is constant and is below the minimum stride threshold.
9117 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9119 * config/arm/arm-cpus.in (mode26): Delete.
9120 (armv4): Delete mode26 reference.
9121 * config/arm/arm.c (arm_configure_build_target): Delete use of
9124 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
9126 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
9128 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
9129 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
9131 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
9132 (floatunsdidf2): Ditto.
9134 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
9136 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
9137 (fixuns_trunc<mode>si2_avx512f): Ditto.
9138 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
9139 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
9140 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
9142 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
9144 PR rtl-optimization/79985
9145 * df-scan.c (df_insn_refs_collect): Remove special case for
9146 global registers and asm statements.
9148 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
9150 * extend.texi (Global Register Variables): Rewrite the bullet list.
9151 Note that the register is available for allocation. Note that access
9152 via inline asm must use constraints. Add note about async-signal
9153 handlers. Remove paragraph about automagic register selection.
9155 2018-05-23 Richard Biener <rguenther@suse.de>
9157 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
9158 of fixed offset from memset VN.
9160 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
9162 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
9164 (alloc_cand_and_find_basis): Initialize first_interp field.
9165 (slsr_process_mul): Modify first_interp field.
9166 (slsr_process_add): Likewise.
9167 (slsr_process_cast): Modify first_interp field for each new
9169 (slsr_process_copy): Likewise.
9170 (dump_candidate): Dump first_interp field.
9171 (replace_mult_candidate): Process all interpretations, not just
9173 (replace_rhs_if_not_dup): Likewise.
9174 (replace_one_candidate): Likewise.
9176 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
9178 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
9180 (aarch64_needs_frame_chain): New function.
9181 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
9183 2018-05-23 Sudakshina Das <sudi.das@arm.com>
9186 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
9187 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
9188 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
9189 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
9190 as true for strict-align.
9191 (aarch64_can_inline_p): Perform checks even when callee has no
9192 attributes to check for strict alignment.
9193 * doc/extend.texi (AArch64 Function Attributes): Document
9195 * doc/invoke.texi: (AArch64 Options): Likewise.
9197 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
9199 PR tree-optimization/85853
9200 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
9201 the handling of the root of the node to...
9202 (vect_slp_analyze_node_operations_1): ...this new function,
9203 and run the whole thing with the child nodes' def types
9204 set according to their SLP node's def type.
9206 2018-05-23 Richard Biener <rguenther@suse.de>
9209 * tree-data-ref.c (create_runtime_alias_checks): Defer
9210 and ignore overflow warnings.
9212 2018-05-23 Yury Gribov <tetra2005@gmail.com>
9214 PR tree-optimization/85822
9215 * tree-vrp.c (is_masked_range_test): Fix handling of negative
9218 2018-05-23 Richard Biener <rguenther@suse.de>
9220 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
9221 memset constants via native_interpret_expr.
9223 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
9226 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
9228 (cgraph_node::create_alias): Likewise.
9229 (cgraph_node::get_availability): Check ifunc_resolver instead
9230 of looking up ifunc attribute.
9231 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
9232 * varasm.c (do_assemble_alias): Likewise.
9233 (assemble_alias): Likewise.
9234 (default_binds_local_p_3): Likewise.
9235 * cgraph.h (cgraph_node): Add ifunc_resolver.
9236 (cgraph_node::only_called_directly_or_aliased_p): Return false
9238 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
9240 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
9241 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
9242 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
9243 instead of looking up ifunc attribute.
9245 2018-05-22 Luis Machado <luis.machado@linaro.org>
9247 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
9249 2018-05-22 Martin Sebor <msebor@redhat.com>
9252 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
9253 only when expasion succeeds.
9254 (expand_builtin_strcmp): Same.
9255 (expand_builtin_strncmp): Same.
9257 2018-05-22 Martin Sebor <msebor@redhat.com>
9259 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
9261 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
9262 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9264 * config/aarch64/aarch64-ldpstp.md: Replace uses of
9265 aarch64_mem_pair_operand with memory_operand and delete operand swapping
9267 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
9268 Add check for legitimate_address.
9269 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
9270 (aarch64_swap_ldrstr_operands): New.
9271 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
9274 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
9275 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9277 * config/aarch64/aarch64.md: New patterns to generate stp
9279 (store_pair_sw, store_pair_dw): New patterns to generate stp for
9280 single words and double words.
9281 (load_pair_sw, load_pair_dw): Likewise.
9282 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
9284 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
9286 * config/aarch64/aarch64-ldpstp.md: Modify peephole
9287 for different mode ldpstp and add peephole for merged zero stores.
9289 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
9291 (aarch64_gen_store_pair): Rename calls to match new patterns.
9292 (aarch64_gen_load_pair): Rename calls to match new patterns.
9293 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
9294 (load_pair<DREG:mode><DREG2:mode>): ... This.
9295 (store_pair<mode>): Rename to...
9296 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
9297 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
9299 (V_INT_EQUIV): Handle SImode.
9300 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
9303 2018-05-22 Martin Sebor <msebor@redhat.com>
9306 * calls.c (maybe_warn_nonstring_arg): Use string length to set
9307 or ajust the presumed bound on an operation to avoid unnecessary
9310 2018-05-22 Martin Sebor <msebor@redhat.com>
9312 PR tree-optimization/85826
9313 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
9314 assuming that a DECL necesarily has a constant size.
9316 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
9319 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
9321 2018-05-22 Richard Biener <rguenther@suse.de>
9323 PR tree-optimization/85834
9324 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
9325 non-constant and non-zero memset arguments.
9327 2018-05-22 Martin Liska <mliska@suse.cz>
9330 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
9332 2018-05-22 Richard Biener <rguenther@suse.de>
9334 PR tree-optimization/85863
9335 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
9336 comparisons when vectype is specified.
9337 (vectorizable_condition): Do not specify vectype for
9338 vect_is_simple_cond when SLP vectorizing.
9340 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
9343 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
9344 define __ibm128 as long double.
9345 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
9346 as a distinct type when IEEE 128-bit support is enabled.
9347 (init_float128_ieee): Fix up conversions between IFmode and IEEE
9348 128-bit types to use the correct functions.
9349 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
9350 convert between 128-bit floating point types that have different
9351 modes but the same representation, instead of using gen_lowpart to
9353 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
9355 (IFKF_reg): New attributes to give the register constraints for
9357 (extend<mode>tf2_internal): New insns to mark an explicit
9358 conversion between 128-bit floating point types that have a
9359 different mode but share the same representation.
9361 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
9363 PR tree-optimization/85814
9364 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
9365 a null return from get_strinfo when unsharing the next
9366 strinfo in the chain.
9368 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
9371 * varasm.c (weak_finish): Clean up weak_decls.
9373 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9375 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
9376 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
9377 UNSPEC_UADALP values.
9378 * config/aarch64/iterators.md (ABAL): New int iterator.
9381 (sur): Add mappings for the above.
9382 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
9384 (aarch64_<sur>abal<mode>_4): Likewise.
9385 (aarch64_<sur>adalp<mode>_3): Likewise.
9386 (<sur>sadv16qi): New define_expand.
9388 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
9390 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
9391 (*movdf_internal): Ditto.
9392 (*rcpsf2_sse): Ditto.
9393 (*rsqrtsf2_sse): Ditto.
9394 (*sqrt<mode>2_sse): Ditto.
9396 2018-05-21 Tamar Christina <tamar.christina@arm.com>
9398 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
9400 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
9401 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
9402 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
9403 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
9405 * config/aarch64/arm_neon.h: Likewise.
9406 * config/aarch64/iterators.md (VQ_I): New.
9408 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
9410 * config.gcc: Add arc/t-multilib-linux to tmake_file for
9412 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
9415 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
9417 * config/nds32/constraints.md (S): New constraint.
9418 * config/nds32/nds32.md (call_internal): Use constraint S.
9419 (call_value_internal): Likewise.
9420 (sibcall_internal): Likewise.
9421 (sibcall_value_internal): Likewise.
9423 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
9424 Chung-Ju Wu <jasonwucj@gmail.com>
9426 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
9429 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
9430 Chung-Ju Wu <jasonwucj@gmail.com>
9432 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
9433 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
9434 (nds32_rtx_costs_impl): Simplify.
9435 (nds32_address_cost_impl): Simplify.
9436 (nds32_init_rtx_costs): New function.
9437 (nds32_rtx_costs_speed_prefer): Likewise.
9438 (nds32_rtx_costs_size_prefer): Likewise.
9439 (nds32_address_cost_speed_prefer): Likewise.
9440 (nds32_address_cost_speed_fwprop): Likewise.
9441 (nds32_address_cost_size_prefer): Likewise.
9442 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
9443 * config/nds32/nds32.c (nds32_option_override): Use
9444 nds32_init_rtx_costs function.
9446 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
9448 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
9449 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
9450 (TARGET_PIPELINE_N8): Likewise.
9451 (TARGET_PIPELINE_N10): Likewise.
9452 (TARGET_PIPELINE_N13): Likewise.
9453 (TARGET_PIPELINE_GRAYWOLF): Likewise.
9455 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
9457 * config/nds32/nds32-fpu.md: Update copyright year.
9459 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
9461 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
9463 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
9465 * config/nds32/nds32.c
9466 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
9467 * config/nds32/nds32.opt (minline-asm-r15): New option.
9469 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
9471 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
9473 * config/nds32/nds32.md (abssi2): New pattern.
9475 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
9477 * config/i386/i386.md (rex64namesuffix): New mode attribute.
9478 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
9479 Merge insn pattern from sse_cvtsi2ss<round_name> and
9480 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
9481 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
9482 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
9483 using SWI48 mode iterator.
9484 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
9485 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
9486 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
9487 pattern from sse_cvttss2si<round_saeonly_name>
9488 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
9489 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
9490 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
9491 using SWI48 mode iterator.
9492 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
9493 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
9494 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
9495 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
9496 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
9497 using SWI48 mode iterator.
9498 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
9499 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
9500 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
9501 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
9502 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
9503 SWI48 mode iterator.
9504 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
9505 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
9506 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
9507 pattern from sse_cvttsd2si<round_saeonly_name>
9508 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
9510 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
9512 * config/nds32/nds32-md-auxiliary.c
9513 (nds32_valid_smw_lwm_base_p): Refine.
9514 (nds32_output_smw_single_word): Refine.
9515 (nds32_output_smw_double_word): New.
9516 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
9518 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
9520 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
9521 (nds32_output_stack_pop): Refine.
9522 (nds32_expand_unaligned_load): Refine.
9523 (nds32_expand_unaligned_store): Refine.
9525 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
9526 Chung-Ju Wu <jasonwucj@gmail.com>
9528 * config/nds32/constants.md: Add TP_REGNUM constant.
9529 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
9530 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
9532 * config/nds32/nds32-doubleword.md: Consider flag_pic.
9533 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
9534 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
9535 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
9536 and PIC code generation.
9537 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
9539 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
9541 * config/nds32/nds32.md: Support TLS and PIC.
9542 * config/nds32/nds32.c: Support TLS and PIC.
9543 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
9544 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
9547 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
9549 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
9550 mode with E_ prefix.
9552 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
9553 Chung-Ju Wu <jasonwucj@gmail.com>
9555 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
9556 * config/nds32/nds32-md-auxiliary.c
9557 (symbolic_reference_mentioned_p): New.
9558 (nds32_legitimize_ict_address): New.
9559 (nds32_expand_ict_move): New.
9560 (nds32_indirect_call_referenced_p): New.
9561 (nds32_symbol_binds_local_p): Delete.
9562 (nds32_long_call_p): Modify.
9563 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
9564 * config/nds32/nds32-protos.h
9565 (symbolic_reference_mentioned_p): Declare.
9566 (nds32_legitimize_ict_address): Declare.
9567 (nds32_expand_ict_move): Declare.
9568 (nds32_indirect_call_referenced_p): Declare.
9569 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
9570 (nds32_relax_group): Use nds32_ict_const_p as condition.
9571 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
9572 (nds32_asm_file_start): Output ict_model directive in asm code.
9573 (nds32_legitimate_address_p): Consider indirect call.
9574 (nds32_print_operand): Consider indirect call.
9575 (nds32_print_operand_address): Consider indirect call.
9576 (nds32_insert_attributes): Handle "indirect_call" attribute.
9577 (TARGET_LEGITIMATE_ADDRESS_P): Define.
9578 (TARGET_LEGITIMATE_CONSTANT_P): Define.
9579 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
9580 (TARGET_DELEGITIMIZE_ADDRESS): Define.
9581 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
9582 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
9583 (TARGET_ICT_MODEL_SMALL): Define.
9584 (TARGET_ICT_MODEL_LARGE): Define.
9585 * config/nds32/nds32.md (movsi): Consider ict model.
9586 (call, call_value): Consider ict model.
9587 (sibcall, sibcall_value): Consider ict model.
9588 * config/nds32/nds32.opt (mict-model): New option.
9589 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
9592 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
9593 Monk Chiang <sh.chiang04@gmail.com>
9594 Jim Wilson <jimw@sifive.com>
9596 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
9597 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
9598 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
9599 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
9600 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
9601 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
9602 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
9603 compute save_libcall_adjustment properly.
9604 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
9605 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
9606 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
9607 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
9608 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
9609 (ABI_SPEC): Handle mabi=ilp32e.
9610 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
9611 (RVE): Add RVE mask.
9612 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
9613 <-march>: Add rv32e as an example.
9615 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
9618 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
9619 (intra_create_variable_infos): Handle C++ constructors.
9621 2018-05-18 Martin Liska <mliska@suse.cz>
9623 * passes.def: Remove a redundant pass.
9625 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
9628 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
9630 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9632 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
9634 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
9636 (armv2, armv2a, armv3, armv3m): Delete architectures.
9637 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
9638 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
9639 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
9641 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
9642 (*mulsidi3adddi): Likewise.
9643 (mulsidi3): Likewise.
9644 (*mulsidi3_nov6): Likewise.
9645 (umulsidi3): Likewise.
9646 (umulsidi3_nov6): Likewise.
9647 (umaddsidi4): Likewise.
9648 (*umulsidi3adddi): Likewise.
9649 (smulsi3_highpart): Likewise.
9650 (*smulsi3_highpart_nov6): Likewise.
9651 (umulsi3_highpart): Likewise.
9652 (*umulsi3_highpart_nov6): Likewise.
9653 * config/arm/arm.h (arm_arch3m): Delete.
9654 * config/arm/arm.c (arm_arch3m): Delete.
9655 (arm_option_override_internal): Update armv3-related comment.
9656 (arm_configure_build_target): Delete use of isa_bit_mode32.
9657 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
9658 (arm_rtx_costs_internal): Delete check of arm_arch3m.
9659 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
9661 (mulusa3): Likewise.
9662 * config/arm/arm-protos.h (arm_arch3m): Delete.
9663 * config/arm/arm-tables.opt: Regenerate.
9664 * config/arm/arm-tune.md: Likewise.
9665 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
9666 deleted architectures.
9668 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9670 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
9671 (armv5t, armv5te): New features.
9672 (ARMv5, ARMv5e): Delete fgroups.
9673 (ARMv5t, ARMv5te): Adjust for above changes.
9675 (armv5, armv5e): Delete arches.
9676 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
9678 (*call_reg_arm): Likewise.
9679 (*call_value_reg_armv5): Likewise.
9680 (*call_value_reg_arm): Likewise.
9681 (*call_symbol): Likewise.
9682 (*call_value_symbol): Likewise.
9683 (*sibcall_insn): Likewise.
9684 (*sibcall_value_insn): Likewise.
9686 (prefetch): Likewise.
9687 (define_split and define_peephole2 dependent on arm_arch5):
9689 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
9691 (TARGET_ARM_QBIT): Likewise.
9692 (TARGET_DSP_MULTIPLY): Likewise.
9693 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
9694 (arm_arch5, arm_arch5e): Delete.
9695 (arm_arch5t, arm_arch5te): Declare.
9696 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
9697 (arm_arch5t): Declare.
9698 (arm_option_reconfigure_globals): Update for the above.
9699 (arm_options_perform_arch_sanity_checks): Update comment, replace
9700 use of arm_arch5 with arm_arch5t.
9701 (use_return_insn): Likewise.
9702 (arm_emit_call_insn): Likewise.
9703 (output_return_instruction): Likewise.
9704 (arm_final_prescan_insn): Likewise.
9705 (arm_coproc_builtin_available): Likewise.
9706 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
9707 arm_arch5e with arm_arch5t and arm_arch5te.
9708 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
9709 (arm_arch5t, arm_arch5te): Declare.
9710 * config/arm/arm-tables.opt: Regenerate.
9711 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
9712 * config/arm/t-multilib: Likewise.
9713 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
9714 instead of arm_arch5.
9715 (*call_reg_thumb1): Likewise.
9716 (*call_value_reg_thumb1_v5): Likewise.
9717 (*call_value_reg_thumb1): Likewise.
9718 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
9720 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
9722 2018-05-18 Martin Liska <mliska@suse.cz>
9724 PR gcov-profile/84846
9725 * doc/gcov.texi: Document -t option of gcov tool.
9727 2018-05-18 Martin Liska <mliska@suse.cz>
9729 PR gcov-profile/84846
9730 * gcov.c (print_usage): Add new -t option.
9731 (process_args): Handle the option.
9732 (generate_results): Use stdout as output when requested by
9735 2018-05-18 Martin Liska <mliska@suse.cz>
9737 PR gcov-profile/84846
9738 * coverage.c (coverage_init): Write PWD to .gcno file.
9739 * doc/gcov.texi: Document how working directory is printed.
9740 * gcov-dump.c (dump_gcov_file): Print PWD.
9741 * gcov.c (output_intermediate_file): Likewise.
9742 (read_graph_file): Read PWD string.
9743 (output_lines): Print PWD.
9745 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9748 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
9749 for retval and return false if all args to phi are zero.
9751 2018-05-18 Richard Biener <rguenther@suse.de>
9753 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
9755 (evrp_dom_walker::before_dom_children): Call it.
9757 2018-05-18 Richard Biener <rguenther@suse.de>
9759 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
9760 results when processing array refs with variable index.
9762 2018-05-18 Toon Moene <toon@moene.org>
9764 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
9765 directly after that of -floop-interchange. Indicate that both
9766 options are enabled by default when specifying -O3.
9768 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9770 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
9771 iterator. Delete separate integer-mode vec_set<mode> expander.
9772 (aarch64_simd_vec_setv2di): Delete.
9773 (vec_setv2di): Delete.
9774 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
9775 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
9776 the "w, r" alternative.
9778 2018-05-18 Martin Liska <mliska@suse.cz>
9780 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
9781 * tree-pass.h (make_pass_lower_switch_O0): New function.
9782 * tree-switch-conversion.c (node_has_low_bound): Remove.
9783 (node_has_high_bound): Likewise.
9784 (node_is_bounded): Likewise.
9785 (class pass_lower_switch): Make it a template type and create
9787 (pass_lower_switch::execute): Add template argument.
9788 (make_pass_lower_switch): New function.
9789 (make_pass_lower_switch_O0): New function.
9790 (do_jump_if_equal): Remove.
9791 (emit_case_nodes): Simplify to just handle all 3 cases and leave
9792 all the hard work to tree optimization passes.
9794 2018-05-18 Martin Liska <mliska@suse.cz>
9796 * dbgcnt.c (limit_low): Renamed from limit.
9797 (limit_high): New variable.
9798 (dbg_cnt_is_enabled): Check for upper limit.
9799 (dbg_cnt): Adjust dumping.
9800 (dbg_cnt_set_limit_by_index): Add new argument for high
9802 (dbg_cnt_set_limit_by_name): Likewise.
9803 (dbg_cnt_process_single_pair): Parse new format.
9804 (dbg_cnt_process_opt): Use strtok.
9805 (dbg_cnt_list_all_counters): Remove 'value' and add
9807 * doc/invoke.texi: Document changes.
9809 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
9811 * doc/sourcebuild.texi (scalar_all_fma): Document.
9812 * tree.def (FMA_EXPR): Delete.
9813 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
9814 * internal-fn.c (ternary_direct): New macro.
9815 (expand_ternary_optab_fn): Likewise.
9816 (direct_ternary_optab_supported_p): Likewise.
9817 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
9818 * builtins.c (fold_builtin_fma): Delete.
9819 (fold_builtin_3): Don't call it.
9820 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
9821 * expr.c (expand_expr_real_2): Likewise.
9822 * fold-const.c (operand_equal_p): Likewise.
9823 (fold_ternary_loc): Likewise.
9824 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
9825 * gimple.c (DEFTREECODE): Likewise.
9826 * gimplify.c (gimplify_expr): Likewise.
9827 * optabs-tree.c (optab_for_tree_code): Likewise.
9828 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
9829 * tree-eh.c (operation_could_trap_p): Likewise.
9830 (stmt_could_throw_1_p): Likewise.
9831 * tree-inline.c (estimate_operator_cost): Likewise.
9832 * tree-pretty-print.c (dump_generic_node): Likewise.
9833 (op_code_prio): Likewise.
9834 * tree-ssa-loop-im.c (stmt_cost): Likewise.
9835 * tree-ssa-operands.c (get_expr_operands): Likewise.
9836 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
9837 * fold-const-call.h (fold_fma): Delete.
9838 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
9839 CFN_FNMA and CFN_FNMS.
9841 * genmatch.c (combined_fn): New enum.
9842 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
9843 (commutative_op): New function.
9844 (commutate): Use it. Handle more than 2 operands.
9845 (dt_operand::gen_gimple_expr): Use commutative_op.
9846 (parser::parse_expr): Allow :c to be used with non-binary
9847 operators if the commutative operand is known.
9848 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
9849 CFN_FMS, CFN_FNMA and CFN_FNMS.
9850 (backprop::process_assign_use): Remove FMA_EXPR handling.
9851 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
9852 (gen_hsa_fma): New function.
9853 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
9854 IFN_FNMA and IFN_FNMS.
9855 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
9856 * gimple-fold.h (follow_all_ssa_edges): Declare.
9857 * gimple-fold.c (follow_all_ssa_edges): New function.
9858 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
9859 gimple_build interface and use follow_all_ssa_edges to fold the result.
9860 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
9861 instead of checking for optabs directly.
9862 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
9863 rather than FMA_EXPRs.
9864 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
9865 call to IFN_FMA instead of an FMA_EXPR.
9867 2018-05-17 Jim Wilson <jimw@sifive.com>
9869 * expr.c (do_tablejump): When converting index to Pmode, if we have a
9870 sign extended promoted subreg, and the range does not have the sign bit
9871 set, then do a sign extend.
9873 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
9874 test, check for sign extended subreg and/or constant operands, and
9875 do a sign extend in that case.
9877 2018-05-17 Steve Ellcey <sellcey@cavium.com>
9879 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
9880 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
9882 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
9883 Change logics_shift_reg to logics_shift_imm.
9884 (thunderx2t99_fp_loadpair_basic): Delete.
9885 (thunderx2t99_fp_storepair_basic): Delete.
9886 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
9887 (thunderx2t99_asimd_polynomial): Delete.
9888 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
9889 and neon_fp_mul_d_scalar_q.
9890 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
9891 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
9892 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
9893 (thunderx2t99_asimd_lut): Add missing tbl types.
9894 (thunderx2t99_asimd_ext): Delete.
9895 (thunderx2t99_asimd_load1_1_mult): Delete.
9896 (thunderx2t99_asimd_load1_2_mult): Delete.
9897 (thunderx2t99_asimd_load1_ldp): New.
9898 (thunderx2t99_asimd_load1): New.
9899 (thunderx2t99_asimd_load2): Add missing *load2* types.
9900 (thunderx2t99_asimd_load3): New.
9901 (thunderx2t99_asimd_load4): New.
9902 (thunderx2t99_asimd_store1_1_mult): Delete.
9903 (thunderx2t99_asimd_store1_2_mult): Delete.
9904 (thunderx2t99_asimd_store2_mult): Delete.
9905 (thunderx2t99_asimd_store2_onelane): Delete.
9906 (thunderx2t99_asimd_store_stp): New.
9907 (thunderx2t99_asimd_store1): New.
9908 (thunderx2t99_asimd_store2): New.
9909 (thunderx2t99_asimd_store3): New.
9910 (thunderx2t99_asimd_store4): New.
9912 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
9914 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
9915 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
9917 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
9918 Segher Boessenkool <segher@kernel.crashing.org>
9921 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
9924 2018-05-17 Richard Biener <rguenther@suse.de>
9926 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
9927 for pruning loop and prune defs feeding only already visited PHIs.
9929 2018-05-17 Richard Biener <rguenther@suse.de>
9931 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
9933 2018-05-17 Bin Cheng <bin.cheng@arm.com>
9934 Richard Biener <rguenther@suse.de>
9936 PR tree-optimization/85793
9937 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
9938 for VMAT_ELEMENTWISE.
9940 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
9942 * internal-fn.h (lookup_internal_fn): Declare
9943 * internal-fn.c (lookup_internal_fn): New function.
9944 * gimple.c (gimple_build_call_from_tree): Handle calls to
9946 * gimple-pretty-print.c (dump_gimple_call): Print "." before
9947 internal function names.
9948 * tree-pretty-print.c (dump_generic_node): Likewise.
9949 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
9951 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
9953 * gimple-fold.h (gimple_build): Make the function forms take
9954 combined_fn rather than built_in_function.
9955 (gimple_simplify): Likewise.
9956 * gimple-match-head.c (gimple_simplify): Likewise.
9957 * gimple-fold.c (gimple_build): Likewise.
9958 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
9959 rather than gimple_build_call_internal.
9960 (get_initial_defs_for_reduction): Likewise.
9961 (vect_create_epilog_for_reduction): Likewise.
9962 (vectorizable_live_operation): Likewise.
9964 2018-05-17 Martin Liska <mliska@suse.cz>
9966 * gimple-ssa-sprintf.c (format_directive): Do not use
9967 space in between 'G_' and '('.
9969 2018-05-17 Jakub Jelinek <jakub@redhat.com>
9972 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
9973 even if the mask is not all ones.
9976 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
9978 (ix86_gimple_fold_builtin): Likewise.
9981 * config/i386/i386.c: Include tree-vector-builder.h.
9982 (ix86_vector_shift_count): New function.
9983 (ix86_fold_builtin): Fold shift builtins by scalar count.
9984 (ix86_gimple_fold_builtin): Likewise.
9986 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
9987 _mm512_setzero): New intrinsics.
9989 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
9990 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9992 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
9993 code generation for cases where splatting a value is not useful.
9994 * simplify-rtx.c (simplify_ternary_operation): Simplify
9995 vec_merge across a vec_duplicate and a paradoxical subreg forming
9996 a vector mode to a vec_concat.
9998 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
10000 * config.gcc: Support "goldmont-plus".
10001 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
10003 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
10004 PROCESSOR_GOLDMONT_PLUS.
10005 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
10006 (processor_target_table): Add "goldmont-plus".
10007 (PTA_GOLDMONT_PLUS): Define.
10008 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
10009 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
10010 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
10011 (fold_builtin_cpu): Add "goldmont-plus".
10012 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
10013 (ix86_option_override_internal): Add "goldmont-plus".
10014 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
10015 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
10016 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
10017 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
10019 2018-05-17 Richard Biener <rguenther@suse.de>
10021 PR tree-optimization/85757
10022 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
10023 remove defs that only feed that PHI from further processing.
10025 2018-05-16 Jim Wilson <jimw@sifive.com>
10027 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
10029 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
10031 2018-05-16 Mark Wielaard <mark@klomp.org>
10033 * dwarf2out.c (count_index_strings): New function.
10034 (output_indirect_strings): Call count_index_strings and generate
10035 header for dwarf_version >= 5.
10037 2018-05-16 Mark Wielaard <mark@klomp.org>
10039 * dwarf2out.c (dwarf_FORM): New function.
10040 (set_indirect_string): Use dwarf_FORM.
10041 (reset_indirect_string): Likewise.
10042 (size_of_die): Likewise.
10043 (value_format): Likewise.
10044 (output_die): Likewise.
10045 (add_skeleton_AT_string): Likewise.
10046 (output_macinfo_op): Likewise.
10047 (index_string): Likewise.
10048 (output_index_string_offset): Likewise.
10049 (output_index_string): Likewise.
10050 (count_index_strings): Likewise.
10052 2018-05-16 Carl Love <cel@us.ibm.com>
10054 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
10055 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
10057 2018-05-16 Martin Jambor <mjambor@suse.cz>
10059 * ipa-prop.c (ipa_free_all_edge_args): Remove.
10060 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
10062 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
10064 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
10065 (fnma<mode>4): Likewise.
10066 (fms<mode>4): Likewise.
10067 (fnms<mode>4): Likewise.
10068 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
10069 (aarch64_fnma<mode>4): Likewise.
10070 (aarch64_fms<mode>4): Likewise.
10071 (aarch64_fnms<mode>4): Likewise.
10072 (aarch64_fnmadd<mode>4): Likewise.
10074 2018-05-16 Jason Merrill <jason@redhat.com>
10076 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
10078 2018-05-16 Richard Biener <rguenther@suse.de>
10080 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
10081 (dump_stmt_cost): Declare.
10082 (add_stmt_cost): Dump cost we add.
10083 (add_stmt_costs): New function.
10084 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
10085 No longer exported.
10086 (vect_analyze_stmt): Adjust prototype.
10087 (vectorizable_condition): Likewise.
10088 (vectorizable_live_operation): Likewise.
10089 (vectorizable_reduction): Likewise.
10090 (vectorizable_induction): Likewise.
10091 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
10092 cost vector to pass to vectorizable_ and record afterwards.
10093 (vect_model_reduction_cost): Take cost vector argument and adjust.
10094 (vect_model_induction_cost): Likewise.
10095 (vectorizable_reduction): Likewise.
10096 (vectorizable_induction): Likewise.
10097 (vectorizable_live_operation): Likewise.
10098 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
10099 SLP_TREE_NUMBER_OF_VEC_STMTS.
10100 (vect_analyze_slp_cost_1): Remove.
10101 (vect_analyze_slp_cost): Likewise.
10102 (vect_slp_analyze_node_operations): Take visited args and
10103 a target cost vector. Avoid processing already visited stmt sets.
10104 (vect_slp_analyze_operations): Use a local cost vector to gather
10105 costs and register those of non-discarded instances.
10106 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
10107 (vect_schedule_slp_instance): Remove copying of
10108 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
10110 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
10111 adding cost. Record cost entry location.
10112 (vect_prologue_cost_for_slp_op): Function to compute cost of
10113 a constant or invariant generated for SLP vect in the prologue,
10114 split out from vect_analyze_slp_cost_1.
10115 (vect_model_simple_cost): Make static. Adjust for SLP costing.
10116 (vect_model_promotion_demotion_cost): Likewise.
10117 (vect_model_store_cost): Likewise, make static.
10118 (vect_model_load_cost): Likewise.
10119 (vectorizable_bswap): Add cost vector arg and adjust.
10120 (vectorizable_call): Likewise.
10121 (vectorizable_simd_clone_call): Likewise.
10122 (vectorizable_conversion): Likewise.
10123 (vectorizable_assignment): Likewise.
10124 (vectorizable_shift): Likewise.
10125 (vectorizable_operation): Likewise.
10126 (vectorizable_store): Likewise.
10127 (vectorizable_load): Likewise.
10128 (vectorizable_condition): Likewise.
10129 (vectorizable_comparison): Likewise.
10130 (can_vectorize_live_stmts): Likewise.
10131 (vect_analyze_stmt): Likewise.
10132 (vect_transform_stmt): Adjust calls to vectorizable_*.
10133 * tree-vectorizer.c: Include gimple-pretty-print.h.
10134 (dump_stmt_cost): New function.
10136 2018-05-16 Richard Biener <rguenther@suse.de>
10138 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
10139 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
10140 * tree-ssa-dse.c: Include tree-ssa-loop.h.
10141 (check_name): New callback.
10142 (dse_classify_store): Track cycles via a visited bitmap of PHI
10143 defs and simplify handling of in-loop and across loop dead stores
10144 and properly fail for loop-variant refs. Handle byte-tracking with
10145 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
10148 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
10150 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
10151 (vect_get_mask_type_for_stmt): Likewise.
10152 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
10154 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
10155 to determine the statement's vector type and the vector type that
10156 should be used for calculating nunits. Deal with cases in which
10157 the type has to be deferred.
10158 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
10159 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
10160 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
10161 (vect_determine_vf_for_stmt): New functions, split out from...
10162 (vect_determine_vectorization_factor): ...here.
10163 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
10164 (vect_get_mask_type_for_stmt): New functions, split out from
10165 vect_determine_vectorization_factor.
10167 2018-05-16 Richard Biener <rguenther@suse.de>
10169 * tree-cfg.c (verify_gimple_assign_ternary): Properly
10170 verify the [VEC_]COND_EXPR embedded comparison.
10172 2018-05-15 Martin Sebor <msebor@redhat.com>
10174 PR tree-optimization/85753
10175 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
10176 RECORD_TYPE in addition to ARRAY_TYPE.
10178 2018-05-15 Martin Sebor <msebor@redhat.com>
10180 PR middle-end/85643
10181 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
10183 2018-05-15 Richard Biener <rguenther@suse.de>
10185 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
10186 add by_clobber_p one. Change algorithm to collect all defs
10187 representing uses we need to walk and try reducing them to
10188 a single one before failing.
10189 (dse_dom_walker::dse_optimize_stmt): Adjust.
10191 2018-05-13 Mark Wielaard <mark@klomp.org>
10193 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
10194 (size_of_loc_descr): Likewise.
10195 (output_loc_operands): Likewise.
10196 (output_loc_operands_raw): Likewise.
10197 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
10198 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
10199 (hash_loc_operands): Likewise.
10200 (compare_loc_operands): Likewise.
10202 2018-05-14 Mark Wielaard <mark@klomp.org>
10204 * dwarf2out.c (count_index_addrs): New function.
10205 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
10207 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10209 PR tree-optimization/83648
10210 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
10211 return value as malloc candidate.
10213 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10216 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
10217 param as true in call to suggest_attribute.
10219 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
10221 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
10222 -mreadonly-in-sdata.
10224 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10226 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
10228 (aarch64_crypto_aesd_fused): Likewise.
10230 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
10232 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
10233 (movsi_aarch64): Likewise.
10234 (load_pairsi): Likewise.
10235 (load_pairdi): Likewise.
10236 (store_pairsi): Likewise.
10237 (store_pairdi): Likewise.
10238 (load_pairsf): Likewise.
10239 (load_pairdf): Likewise.
10240 (store_pairsf): Likewise.
10241 (store_pairdf): Likewise.
10242 (zero_extend): Likewise.
10243 (trunc): Swap alternatives.
10244 (fcvt_target): Add '?' to prefer w over r.
10246 2018-05-14 Jakub Jelinek <jakub@redhat.com>
10249 * config/i386/i386.md: Disallow non-commutative arithmetics in
10250 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
10251 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
10252 in the peephole2 before it.
10254 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
10256 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
10257 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
10258 (ix86_handle_option): Handle -mcldemote.
10259 * config.gcc: New header.
10260 * config/i386/cldemoteintrin.h: New file.
10261 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
10262 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
10264 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
10265 OPTION_MASK_ISA_CLDEMOTE.
10266 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
10267 (ix86_valid_target_attribute_inner_p): Ditto.
10268 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
10269 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
10270 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
10271 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
10272 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
10274 * config/i386/i386.opt: Add -mcldemote.
10275 * config/i386/x86intrin.h: New header.
10276 * doc/invoke.texi: Add -mcldemote.
10278 2018-05-14 Richard Biener <rguenther@suse.de>
10280 * doc/match-and-simplify.texi: Adjust :s documentation.
10282 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
10284 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
10285 intended memcpy size.
10286 (REORDER_45): Likewise.
10288 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
10290 * sort.cc: New file.
10291 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
10292 * vec.c (qsort_chk): Use gcc_qsort.
10293 * Makefile.in (OBJS-libcommon): Add sort.o.
10294 (build/sort.o): New target. Use it...
10295 (BUILD_RTL): ... here, and...
10296 (build/gencfn-macros): ... here, and...
10297 (build/genmatch): ... here.
10299 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
10300 Chung-Ju Wu <jasonwucj@gmail.com>
10302 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
10303 * config/nds32/nds32-graywolf.md: New file.
10304 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
10305 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
10307 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
10308 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
10309 * config/nds32/nds32.md (pipeline_model): Add graywolf.
10310 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
10311 * config/nds32/pipelines.md: Include n15 settings.
10313 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
10314 Chung-Ju Wu <jasonwucj@gmail.com>
10316 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
10317 * config/nds32/nds32-n13.md: New file.
10318 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
10319 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
10321 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
10322 * config/nds32/nds32.md (pipeline_model): Add n13.
10323 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
10324 * config/nds32/pipelines.md: Include n13 settings.
10326 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
10327 Chung-Ju Wu <jasonwucj@gmail.com>
10329 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
10330 * config/nds32/nds32-n10.md: New file.
10331 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
10332 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
10334 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
10335 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
10336 * config/nds32/nds32.md (pipeline_model): Add n10.
10337 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
10338 * config/nds32/pipelines.md: Include n10 settings.
10340 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
10341 Kito Cheng <kito.cheng@gmail.com>
10342 Chung-Ju Wu <jasonwucj@gmail.com>
10344 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
10345 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
10346 Add enum values for DSP extension instructions.
10347 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
10349 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
10350 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
10351 New code iterators.
10352 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
10353 * config/nds32/nds32-dspext.md: New file for DSP implementation.
10354 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
10355 * config/nds32/nds32-intrinsic.md: Likewise.
10356 * config/nds32/nds32_intrinsic.h: Likewise.
10357 * config/nds32/nds32-md-auxiliary.c: Likewise.
10358 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
10359 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
10360 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
10361 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
10362 * config/nds32/nds32-protos.h: New declarations for DSP extension.
10363 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
10364 TYPE_DMAC in switch statement.
10365 * config/nds32/nds32.c: New checking and implementation for DSP
10366 extension instructions.
10367 * config/nds32/nds32.h: Likewise.
10368 * config/nds32/nds32.md: Likewise.
10369 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
10370 * config/nds32/predicates.md: Implement new predicates for DSP
10373 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
10375 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
10376 Reformat alternatives and attributes so it is easier to identify
10377 which constraints/attributes go with which instruction.
10378 (mov<mode>_hardfloat32, FMOVE64): Likewise.
10379 (mov<mode>_softfloat32, FMOVE64): Likewise.
10380 (mov<mode>_hardfloat64, FMOVE64): Likewise.
10381 (mov<mode>_softfloat64, FMOVE64): Likewise.
10383 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
10385 * doc/extend.texi (PowerPC Built-in Functions): Rename this
10387 (Basic PowerPC Built-in Functions): The new name of the
10388 subsection previously known as "PowerPC Built-in Functions".
10389 (Basic PowerPC Built-in Functions Available on all Configurations):
10391 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
10392 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
10393 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
10394 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
10396 2018-05-11 Martin Jambor <mjambor@suse.cz>
10399 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
10402 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
10405 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
10407 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
10409 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
10410 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
10411 (ix86_handle_option): Handle -mwaitpkg.
10412 * config.gcc: New header.
10413 * config/i386/cpuid.h (bit_WAITPKG): New bit.
10414 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
10415 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
10417 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
10418 OPTION_MASK_ISA_WAITPKG.
10419 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
10420 (ix86_option_override_internal): Add PTA_WAITPKG.
10421 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
10422 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
10423 IX86_BUILTIN_TPAUSE.
10424 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
10425 __builtin_ia32_umwait and __builtin_ia32_tpause.
10426 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
10427 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
10428 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
10429 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
10430 UNSPECV_TPAUSE): New.
10431 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
10432 * config/i386/i386.opt: Add -mwaitpkg.
10433 * config/i386/waitpkgintrin.h: New file.
10434 * config/i386/x86intrin.h: New header.
10435 * doc/invoke.texi: Add -mwaitpkg.
10437 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
10440 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
10442 (cortex-m0): Use armv6s-m isa.
10443 (cortex-m0plus): Likewise.
10444 (cortex-m1): Likewise.
10445 (cortex-m0.small-multiply): Likewise.
10446 (cortex-m0plus.small-multiply): Likewise.
10447 (cortex-m1.small-multiply): Likewise.
10449 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
10450 Jakub Jelinek <jakub@redhat.com>
10452 PR tree-optimization/85692
10453 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
10454 source permute as well.
10456 2018-05-11 Martin Liska <mliska@suse.cz>
10459 * doc/extend.texi: Document LLVM style format for no_sanitize
10462 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
10464 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
10465 mode_supports_vsx_dform_quad to mode_supports_dq_form.
10466 (mode_supports_vsx_dform_quad): Likewise.
10467 (mode_supports_vmx_dform): Move these functions to be next to the
10468 other mode_supports functions.
10469 (mode_supports_dq_form): Likewise.
10470 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
10471 mode_supports_dq_form.
10472 (reg_offset_addressing_ok_p): Likewise.
10473 (offsettable_ok_by_alignment): Likewise.
10474 (rs6000_legitimate_offset_address_p): Likewise.
10475 (legitimate_lo_sum_address_p): Likewise.
10476 (rs6000_legitimize_address): Likewise.
10477 (rs6000_legitimize_reload_address): Likewise.
10478 (rs6000_secondary_reload_inner): Likewise.
10479 (rs6000_preferred_reload_class): Likewise.
10480 (rs6000_output_move_128bit): Likewise.
10482 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
10484 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
10485 Generate SImode target register for null target.
10486 <case IX86_BUILTIN_XGETBV>: Ditto.
10487 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
10488 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
10490 2018-05-10 Carl Love <cel@us.ibm.com>
10492 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
10493 dcbtt and dcbtstt if operands[2] is 0.
10495 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
10498 * config/i386/sse.md (usadv64qi): New expander.
10500 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
10502 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
10503 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
10504 -maltivec=be support.
10505 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
10506 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
10507 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
10508 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
10509 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
10510 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
10511 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
10512 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
10513 altivec_vsumsws): Adjust.
10514 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
10515 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
10516 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
10518 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
10519 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
10520 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
10521 (altivec_lve<VI_char>x): Delete expand.
10522 (*altivec_lve<VI_char>x_internal): Rename to...
10523 (altivec_lve<VI_char>x): ... this.
10524 (altivec_lvxl_<mode>): Delete expand.
10525 (*altivec_lvxl_<mode>_internal): Rename to ...
10526 (altivec_lvxl_<mode>): ... this.
10527 (altivec_stvxl_<mode>): Delete expand.
10528 (*altivec_stvxl_<mode>_internal): Rename to ...
10529 (altivec_stvxl_<mode>): ... this.
10530 (altivec_stve<VI_char>x): Delete expand.
10531 (*altivec_stve<VI_char>x_internal): Rename to ...
10532 (altivec_stve<VI_char>x): ... this.
10533 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
10534 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
10535 reduc_plus_scal_<mode>): Adjust.
10536 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
10538 (rs6000_cpu_cpp_builtins): Adjust.
10539 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
10540 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
10541 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
10542 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
10543 -maltivec=be support.
10544 (rs6000_split_vec_extract_var): Adjust.
10545 (rs6000_split_v4si_init): Adjust.
10546 (swap_selector_for_mode): Delete.
10547 (altivec_expand_lvx_be, altivec_expand_stvx_be,
10548 altivec_expand_stvex_be): Delete.
10549 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
10550 -maltivec=be support.
10551 (rs6000_gimple_fold_builtin): Ditto.
10552 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
10554 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
10555 (TARGET_DIRECT_MOVE_64BIT): Adjust.
10556 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
10557 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
10558 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
10559 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
10560 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
10561 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
10562 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
10563 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
10564 anonymous split): Adjust.
10565 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
10566 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
10568 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
10570 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
10571 when --with-gxx-include-dir is also specified.
10572 * configure: Regenerate.
10574 2018-05-09 Jim Wilson <jimw@sifive.com>
10577 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
10578 * config/riscv/t-withmultilib: New.
10579 * config/riscv/withmultilib.h: New.
10580 * doc/install.texi: Document RISC-V --with-multilib-list support.
10582 2018-05-09 Richard Biener <rguenther@suse.de>
10584 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
10586 (vect_bb_vectorization_profitable_p): Adjust. Compute
10587 actual scalar cost using the cost vector and the add_stmt_cost
10590 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
10592 PR rtl-optimization/85645
10593 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
10594 in the REG_CFA_REGISTER note for LR, don't leave it empty.
10596 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
10598 PR rtl-optimization/85645
10599 * shrink-wrap.c (spread_components): Return a boolean saying if
10600 anything was changed.
10601 (try_shrink_wrapping_separate): Iterate spread_components until
10602 nothing changes anymore.
10604 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
10606 PR rtl-optimization/85645
10607 * regrename.c (build_def_use): Also kill the chains that include the
10608 destination of a REG_CFA_REGISTER note.
10610 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
10612 PR rtl-optimization/85645
10613 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
10614 insn that has a REG_CFA_REGISTER note.
10616 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
10618 * cfgexpand.c (expand_clobber): New function.
10619 (expand_gimple_stmt_1): Use it.
10620 * tree-vect-stmts.c (vect_clobber_variable): New function,
10622 (vectorizable_simd_clone_call): ...here.
10623 (vectorizable_store): Emit a clobber either side of an
10624 IFN_STORE_LANES sequence.
10625 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
10627 2018-05-09 Tom de Vries <tom@codesourcery.com>
10630 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
10631 (define_insn "trap_if_false"): Add exit after trap.
10633 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
10635 PR rtl-optimization/85638
10636 * bb-reorder.c: Include common/common-target.h.
10637 (create_forwarder_block): New function extracted from...
10638 (fix_up_crossing_landing_pad): ...here. Rename into...
10639 (dw2_fix_up_crossing_landing_pad): ...this.
10640 (sjlj_fix_up_crossing_landing_pad): New function.
10641 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
10642 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
10643 from both partitions and exit the loop after one iteration.
10645 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
10648 * doc/extend.texi (PowerPC Built-in Functions): Rename this
10650 (Basic PowerPC Built-in Functions): The new name of the
10651 subsection previously known as "PowerPC Built-in Functions".
10652 (Basic PowerPC Built-in Functions Available on all Configurations):
10654 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
10656 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
10658 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
10660 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
10663 2018-05-08 Jim Wilson <jimw@sifive.com>
10665 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
10666 (LD_EMUL_SUFFIX): New.
10667 (LINK_SPEC): Use it.
10669 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
10671 * doc/extend.texi (PowerPC Built-in Functions): Rename this
10673 (Basic PowerPC Built-in Functions): The new name of the
10674 subsection previously known as "PowerPC Built-in Functions".
10675 (Basic PowerPC Built-in Functions Available on all Configurations):
10677 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
10679 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
10681 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
10683 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
10686 2018-05-08 Jakub Jelinek <jakub@redhat.com>
10689 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
10690 after cmpelim optimization.
10692 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
10694 * config.gcc: Support "goldmont".
10695 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
10696 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
10697 PROCESSOR_GOLDMONT.
10698 * config/i386/i386.c (m_GOLDMONT): Define.
10699 (processor_target_table): Add "goldmont".
10700 (PTA_GOLDMONT): Define.
10701 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
10702 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
10703 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
10704 (fold_builtin_cpu): Add "goldmont".
10705 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
10706 (ix86_option_override_internal): Add "goldmont".
10707 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
10708 (processor_type): Add PROCESSOR_GOLDMONT.
10709 * config/i386/i386.md: Add CPU "glm".
10710 * config/i386/glm.md: New file.
10711 * config/i386/x86-tune.def: Add m_GOLDMONT.
10712 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
10714 2018-05-08 Jakub Jelinek <jakub@redhat.com>
10717 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
10719 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
10720 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
10721 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
10724 * config/i386/i386.c (ix86_fold_builtin): Handle
10725 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
10728 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
10729 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
10731 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
10734 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
10735 (check_arch): Likewise.
10736 (check_fpu): Return the result rather than printing it.
10737 (end arch): Fix operator precedence.
10738 (end cpu): Likewise.
10739 (END): Print the result from check_fpu.
10741 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
10742 Alan Hayward <alan.hayward@arm.com>
10743 David Sherwood <david.sherwood@arm.com>
10745 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
10746 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
10747 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
10748 (*fcmuo<mode>_and): New patterns.
10750 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
10752 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
10753 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
10754 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
10755 (cmp_op, sve_imm_con): New code attributes.
10756 (SVE_COND_INT_CMP, imm_con): Delete.
10757 (cmp_op): Remove above unspecs from int attribute.
10758 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
10760 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
10761 comparison-specific unspecs.
10762 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
10763 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
10764 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
10765 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
10766 (*vec_fcm<cmp_op><mode>): Rename to...
10767 (*fcm<cmp_op><mode>): ...this and adjust likewise.
10768 (*vec_fcmuo<mode>): Rename to...
10769 (*fcmuo<mode>): ...this and adjust likewise.
10770 (*pred_fcm<cmp_op><mode>): New pattern.
10771 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
10772 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
10774 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
10776 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
10777 (aarch64_emit_sve_predicated_cond): New function.
10778 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
10779 (aarch64_emit_unspec_cond_or): Replace with...
10780 (aarch64_emit_sve_or_conds): ...this new function. Use
10781 aarch64_emit_sve_ptrue_op for the individual comparisons and
10782 aarch64_emit_binop to OR them together.
10783 (aarch64_emit_inverted_unspec_cond): Replace with...
10784 (aarch64_emit_sve_inverted_cond): ...this new function. Use
10785 aarch64_emit_sve_ptrue_op for the comparison and
10786 aarch64_emit_unop to invert the result.
10787 (aarch64_expand_sve_vec_cmp_float): Update after the above
10788 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
10790 2018-05-07 Nathan Sidwell <nathan@acm.org>
10792 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
10793 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
10794 (Backwards Compatibility): Likewise.
10796 2018-05-07 Luis Machado <luis.machado@linaro.org>
10800 2018-05-07 Luis Machado <luis.machado@linaro.org>
10802 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
10803 <prefetch_dynamic_strides>: New const bool field.
10804 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
10805 prefetch_dynamic_strides.
10806 (exynosm1_prefetch_tune): Likewise.
10807 (thunderxt88_prefetch_tune): Likewise.
10808 (thunderx_prefetch_tune): Likewise.
10809 (thunderx2t99_prefetch_tune): Likewise.
10810 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
10812 (aarch64_override_options_internal): Update to set
10813 PARAM_PREFETCH_DYNAMIC_STRIDES.
10814 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
10815 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
10816 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
10817 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
10818 prefetch-dynamic-strides setting.
10820 2018-05-07 Luis Machado <luis.machado@linaro.org>
10822 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
10823 <minimum_stride>: New const int field.
10824 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
10825 minimum_stride field.
10826 (exynosm1_prefetch_tune): Likewise.
10827 (thunderxt88_prefetch_tune): Likewise.
10828 (thunderx_prefetch_tune): Likewise.
10829 (thunderx2t99_prefetch_tune): Likewise.
10830 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
10831 (aarch64_override_options_internal): Update to set
10832 PARAM_PREFETCH_MINIMUM_STRIDE.
10833 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
10834 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
10835 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
10836 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
10837 stride is constant and is below the minimum stride threshold.
10839 2018-05-07 Luis Machado <luis.machado@linaro.org>
10841 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
10844 2018-05-07 Luis Machado <luis.machado@linaro.org>
10846 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
10847 <prefetch_dynamic_strides>: New const bool field.
10848 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
10849 prefetch_dynamic_strides.
10850 (exynosm1_prefetch_tune): Likewise.
10851 (thunderxt88_prefetch_tune): Likewise.
10852 (thunderx_prefetch_tune): Likewise.
10853 (thunderx2t99_prefetch_tune): Likewise.
10854 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
10856 (aarch64_override_options_internal): Update to set
10857 PARAM_PREFETCH_DYNAMIC_STRIDES.
10858 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
10859 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
10860 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
10861 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
10862 prefetch-dynamic-strides setting.
10864 2018-05-07 Luis Machado <luis.machado@linaro.org>
10866 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
10867 <minimum_stride>: New const int field.
10868 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
10869 minimum_stride field.
10870 (exynosm1_prefetch_tune): Likewise.
10871 (thunderxt88_prefetch_tune): Likewise.
10872 (thunderx_prefetch_tune): Likewise.
10873 (thunderx2t99_prefetch_tune): Likewise.
10874 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
10875 (aarch64_override_options_internal): Update to set
10876 PARAM_PREFETCH_MINIMUM_STRIDE.
10877 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
10878 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
10879 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
10880 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
10881 stride is constant and is below the minimum stride threshold.
10883 2018-05-06 Jakub Jelinek <jakub@redhat.com>
10886 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
10887 the type is addressable. Don't force op into register if it has
10890 2018-05-05 Roland McGrath <mcgrathr@google.com>
10893 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
10894 any section for which we don't know a specific type it should have,
10895 regardless of name. Previously this was done only for the exact
10896 names ".init_array", ".fini_array", and ".preinit_array".
10897 (default_elf_asm_named_section): Add comment about
10898 relationship with default_section_type_flags and SECTION_NOTYPE.
10899 (get_section): Don't consider it a type conflict if one side has
10900 SECTION_NOTYPE and the other doesn't, as long as neither has the
10901 SECTION_BSS et al used in the default_section_type_flags logic.
10903 2018-05-05 Tom de Vries <tom@codesourcery.com>
10906 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
10907 (workaround_barsyncs): New function.
10908 (nvptx_reorg): Use workaround_barsyncs.
10909 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
10910 (define_expand "nvptx_membar_cta"): New define_expand.
10911 (define_insn "*nvptx_membar_cta"): New insn.
10913 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
10915 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
10916 To improve optimization opportunities.
10917 * builtin-types.def: The new needed builtin types for the above.
10919 2018-05-04 Richard Biener <rguenther@suse.de>
10921 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
10922 * gimple-ssa-store-merging.c
10923 (imm_store_chain_info::output_merged_store): Remove redundant create,
10924 release split_store vector contents on failure.
10925 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
10926 scalar stmt vector on cache hit.
10928 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
10930 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
10932 * config.gcc (powerpc-xilinx-eabi*): Remove.
10933 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
10935 (fusion_addis_mem_combo_load): Ditto.
10936 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
10938 (rs6000_cpu_cpp_builtins): Ditto.
10939 * config/rs6000/rs6000-linux.c
10940 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
10941 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
10942 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
10944 (rs6000_setup_reg_addr_masks): Ditto.
10945 (rs6000_init_hard_regno_mode_ok): Ditto.
10946 (rs6000_option_override_internal): Ditto.
10947 (legitimate_lo_sum_address_p): Ditto.
10948 (rs6000_legitimize_address): Ditto.
10949 (rs6000_legitimize_reload_address): Ditto.
10950 (rs6000_legitimate_address_p): Ditto.
10951 (abi_v4_pass_in_fpr): Ditto.
10952 (setup_incoming_varargs): Ditto.
10953 (rs6000_gimplify_va_arg): Ditto.
10954 (rs6000_split_multireg_move): Ditto.
10955 (rs6000_savres_strategy): Ditto.
10956 (rs6000_emit_prologue_components): Ditto.
10957 (rs6000_emit_epilogue_components): Ditto.
10958 (rs6000_emit_prologue): Ditto.
10959 (rs6000_emit_epilogue): Ditto.
10960 (rs6000_elf_file_end): Ditto.
10961 (rs6000_function_value): Ditto.
10962 (rs6000_libcall_value): Ditto.
10963 * config/rs6000/rs6000.h: Ditto.
10964 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
10965 (TARGET_MINMAX): ... this. New.
10966 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
10967 * config/rs6000/rs6000.md: Remove Xilinx FP support.
10968 (*movsi_internal1_single): Delete.
10969 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
10970 mfpu=, mxilinx-fpu): Delete.
10971 * config/rs6000/singlefp.h: Delete.
10972 * config/rs6000/sysv4.h: Remove Xilinx FP support.
10973 * config/rs6000/t-rs6000: Ditto.
10974 * config/rs6000/t-xilinx: Delete.
10975 * config/rs6000/titan.md: Adjust for fp_type removal.
10976 * config/rs6000/vsx.md: Remove Xilinx FP support.
10977 (VStype_simple): Delete.
10978 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
10979 * config/rs6000/xfpu.h: Delete.
10980 * config/rs6000/xfpu.md: Delete.
10981 * config/rs6000/xilinx.h: Delete.
10982 * config/rs6000/xilinx.opt: Delete.
10983 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
10984 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
10986 2018-05-04 Tom de Vries <tom@codesourcery.com>
10989 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
10992 2018-05-04 Richard Biener <rguenther@suse.de>
10994 PR middle-end/85627
10995 * tree-complex.c (update_complex_assignment): We are always in SSA form.
10996 (expand_complex_div_wide): Likewise.
10997 (expand_complex_operations_1): Likewise.
10998 (expand_complex_libcall): Preserve EH info of the original stmt.
10999 (tree_lower_complex): Handle removed blocks.
11000 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
11001 on complex multiplication and division libcall builtins.
11003 2018-05-04 Richard Biener <rguenther@suse.de>
11005 PR middle-end/85574
11006 * fold-const.c (negate_expr_p): Restrict negation of operand
11007 zero of a division to when we know that can happen without
11009 (fold_negate_expr_1): Likewise.
11011 2018-05-04 Jakub Jelinek <jakub@redhat.com>
11014 * real.h (real_nextafter): Declare.
11015 * real.c (real_nextafter): New function.
11016 * fold-const-call.c (fold_const_nextafter): New function.
11017 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
11018 CASE_CFN_NEXTTOWARD.
11019 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
11020 even when arg1_mode is different from arg0_mode.
11022 2018-05-03 Nathan Sidwell <nathan@acm.org>
11024 * doc/extend.texi (Deprecated Features): Remove
11025 -ffriend-injection.
11026 (Backwards Compatibility): Likewise.
11027 * doc/invoke.texi (C++ Language Options): Likewise.
11028 (C++ Dialect Options): Likewise.
11030 2018-05-03 Jakub Jelinek <jakub@redhat.com>
11033 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
11034 _mm512_mask_mullox_epi64): New intrinsics.
11036 2018-05-03 Tom de Vries <tom@codesourcery.com>
11039 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
11040 dump files): Add offload-tree.
11042 2018-05-03 Richard Biener <rguenther@suse.de>
11044 PR tree-optimization/85615
11045 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
11046 to loops not nested in BBs loop father to avoid creating multi-entry
11049 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11051 PR tree-optimization/70291
11052 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
11053 arguments. Change return type to tree. Emit libcall as a new
11054 statement rather than replacing existing one when inplace_p is true.
11055 (expand_complex_multiplication_components): New function.
11056 (expand_complex_multiplication): Expand floating-point complex
11057 multiplication using the above.
11058 (expand_complex_division): Rename inner_type parameter to type.
11059 Update expand_complex_libcall call-site.
11060 (expand_complex_operations_1): Update expand_complex_multiplication
11061 and expand_complex_division call-sites.
11063 2018-05-02 Jakub Jelinek <jakub@redhat.com>
11066 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
11067 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
11068 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
11069 the highest significant bit of the shift count mask is clear. In
11070 check whether and[sq]i3 is needed verify that all significant bits
11071 of the shift count other than the highest are set.
11073 2018-05-02 Tom de Vries <tom@codesourcery.com>
11076 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
11077 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
11078 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
11079 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
11080 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
11081 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
11082 __builtin_goacc_parlevel_size.
11084 2018-05-02 Richard Biener <rguenther@suse.de>
11086 PR tree-optimization/85597
11087 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
11088 do not use split vect_get_vec_defs call but call vect_get_slp_defs
11091 2018-05-02 Tom de Vries <tom@codesourcery.com>
11094 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
11095 dump files): Add ltrans-tree.
11097 2018-05-02 Tom de Vries <tom@codesourcery.com>
11100 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
11101 dump files): Add wpa-ipa.
11103 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
11105 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
11106 powerpc*-*-linux*paired* target.
11107 * config/rs6000/750cl.h: Delete.
11108 * config/rs6000/paired.h: Delete.
11109 * config/rs6000/paired.md: Delete.
11110 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
11112 * config/rs6000/rs6000-builtin.def: Remove paired float support.
11113 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
11114 comment. Remove paired float support.
11115 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
11116 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
11118 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
11119 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
11121 * config/rs6000/rs6000.c: Remove paired float support.
11122 (paired_expand_vector_init, paired_expand_vector_move,
11123 paired_emit_vector_compare, paired_emit_vector_cond_expr,
11124 (paired_expand_lv_builtin, paired_expand_stv_builtin,
11125 paired_expand_builtin, paired_expand_predicate_builtin,
11126 paired_init_builtins): Delete.
11127 * config/rs6000/rs6000.h: Remove paired float support.
11128 * config/rs6000/rs6000.md: Remove paired float support.
11129 (move_from_CR_ov_bit): Delete.
11130 * config/rs6000/rs6000.opt (mpaired): Delete.
11131 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
11132 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
11134 2018-05-02 Richard Biener <rguenther@suse.de>
11136 PR middle-end/85567
11137 * gimplify.c (gimplify_save_expr): When in SSA form allow
11138 SAVE_EXPRs to compute to SSA vars.
11140 2018-05-02 Jakub Jelinek <jakub@redhat.com>
11143 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
11144 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
11145 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
11146 clobber operands[2], instead use a new pseudo. Formatting fixes.
11148 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
11150 PR tree-optimization/85586
11151 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
11152 exit early for statements in the same group if the accesses are
11155 2018-05-02 Tom de Vries <tom@codesourcery.com>
11158 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
11161 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
11163 PR tree-optimization/85143
11164 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
11166 2018-05-01 Tom de Vries <tom@codesourcery.com>
11169 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
11170 not found" error message.
11172 2018-05-01 Tom de Vries <tom@codesourcery.com>
11175 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
11176 * vec.c (test_ordered_remove_if): New function.
11177 (vec_c_tests): Call test_ordered_remove_if.
11178 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
11179 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
11180 * tree-vect-patterns.c (vect_pattern_recog_1): Use
11181 VEC_ORDERED_REMOVE_IF.
11183 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11185 PR tree-optimization/82665
11186 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
11187 pointer subtraction where arguments come from a memchr call.
11189 2018-05-01 Jakub Jelinek <jakub@redhat.com>
11191 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
11192 --push-state --as-needed and --pop-state instead of --as-needed and
11193 --no-as-needed if ld supports it.
11194 * configure: Regenerated.
11197 * doc/install.texi2html: Replace _002d with - and _002a with * in
11198 generated html files using sed.
11200 2018-04-30 David Malcolm <dmalcolm@redhat.com>
11203 * gcc-rich-location.c (blank_line_before_p): New function.
11204 (use_new_line): New function.
11205 (gcc_rich_location::add_fixit_insert_formatted): New function.
11206 * gcc-rich-location.h
11207 (gcc_rich_location::add_fixit_insert_formatted): New function.
11209 2018-04-30 David Malcolm <dmalcolm@redhat.com>
11211 * selftest.c (assert_streq): Rename "expected" and "actual" to
11212 "val1" and "val2". Extend NULL-handling to cover both inputs
11213 symmetrically, while still requiring both to be non-NULL for a pass.
11214 * selftest.h (assert_streq): Rename "expected" and "actual" to
11216 (ASSERT_EQ): Likewise.
11217 (ASSERT_EQ_AT): Likewise.
11218 (ASSERT_KNOWN_EQ): Likewise.
11219 (ASSERT_KNOWN_EQ_AT): Likewise.
11220 (ASSERT_NE): Likewise.
11221 (ASSERT_MAYBE_NE): Likewise.
11222 (ASSERT_MAYBE_NE_AT): Likewise.
11223 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
11224 the assertion to pass.
11225 (ASSERT_STREQ_AT): Likewise.
11227 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
11229 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
11230 interaction with -pie.
11232 2018-04-30 David Malcolm <dmalcolm@redhat.com>
11234 * selftest.h: Fix alphabetization of per-source-file selftest
11237 2018-04-30 Jason Merrill <jason@redhat.com>
11239 PR c++/61982 - dead stores to destroyed objects.
11240 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
11243 2018-04-30 Jason Merrill <jason@redhat.com>
11245 * tree.c (build_clobber): New.
11246 * tree.h: Declare it.
11247 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
11249 2018-04-30 David Malcolm <dmalcolm@redhat.com>
11251 * diagnostic-show-locus.c (layout::layout): Update for
11252 location_get_source_line returning a char_span.
11253 (struct char_span): Move to input.h.
11254 (struct correction): Update for fields in char_span becoming
11256 (struct source_line): Update for location_get_source_line
11257 returning a char_span.
11258 (layout::print_line): Likewise.
11259 * edit-context.c (edited_file::print_content): Likewise.
11260 (edited_file::print_diff_hunk): Likewise.
11261 (edited_file::print_run_of_changed_lines): Likewise.
11262 (edited_file::get_num_lines): Likewise.
11263 (edited_line::edited_line): Likewise.
11264 * final.c (asm_show_source): Likewise.
11265 * input.c (location_get_source_line): Convert return type
11266 from const char * to char_span, losing the final "line_len"
11268 (dump_location_info): Update for the above.
11269 (get_substring_ranges_for_loc): Likewise. Use a char_span
11270 when handling the literal within the line.
11271 (test_reading_source_line): Update for location_get_source_line
11272 returning a char_span.
11273 * input.h (class char_span): Move here from
11274 diagnostic-show-locus.c, converting from a struct to a class.
11275 Make data members private.
11276 (char_span::operator bool): New.
11277 (char_span::length): New.
11278 (char_span::get_buffer): New.
11279 (char_span::operator[]): New.
11280 (char_span::subspan): Make const.
11281 (char_span::xstrdup): New.
11282 (location_get_source_line): Convert return type from const char *
11283 to char_span, losing the final "line_size" param.
11285 2018-04-30 Jan Hubicka <jh@suse.cz>
11287 * lto-wrapper.c (ltrans_priorities): New static var.
11288 (cmp_priority): New.
11289 (run_gcc): Read priorities and if doing parallel build order
11290 the Makefile by them.
11292 2018-04-30 David Malcolm <dmalcolm@redhat.com>
11294 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
11296 2018-04-30 Richard Biener <rguenther@suse.de>
11298 * tree-cfg.c (verify_address): Remove base argument, add
11299 flag whether to check TREE_ADDRESSABLE and do that.
11300 (verify_expr): Remove.
11301 (verify_types_in_gimple_reference): Add pieces from verify_expr.
11302 (verify_gimple_assign_single): Likewise.
11303 (verify_gimple_switch): Likewise.
11304 (verify_expr_location_1): Dereference tp once. Add (disabled)
11305 piece from verify_expr.
11306 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
11308 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
11310 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
11312 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
11314 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
11315 (small_data_pattern): Likewise.
11316 (arc_rewrite_small_data): Likewise.
11317 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
11318 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
11319 (get_symbol_alignment): New function.
11320 (legitimate_small_data_address_p): Likewise.
11321 (legitimate_scaled_address): Update, call
11322 legitimate_small_data_address_p.
11323 (output_sdata): New static variable.
11324 (arc_print_operand): Update how we handle small data operands.
11325 (arc_print_operand_address): Likewise.
11326 (arc_legitimate_address_p): Update, use
11327 legitimate_small_data_address_p.
11328 (arc_rewrite_small_data_p): Remove.
11329 (arc_rewrite_small_data_1): Likewise.
11330 (arc_rewrite_small_data): Likewise.
11331 (small_data_pattern): Likewise.
11332 (compact_sda_memory_operand): Update to use
11333 legitimate_small_data_address_p and get_symbol_alignment.
11334 (prepare_move_operands): Don't rewite sdata pattern.
11335 (prepare_extend_operands): Remove.
11336 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
11338 (zero_extendqisi2): Likewise.
11339 (zero_extendhisi2): Likewise.
11340 (extendqihi2): Likewise.
11341 (extendqisi2): Likewise.
11342 (extendhisi2): Likewise.
11343 (addsi3): Likewise.
11344 (subsi3): Likewise.
11345 (andsi3): Likewise.
11346 * config/arc/constraints.md (Usd): Change it to memory constraint.
11348 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
11350 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
11351 as source of std instructions.
11352 * config/arc/arc.md (movsi_insn): Update pattern predicate to
11353 allow 6-bit constants as source for store instructions.
11354 (movdi_insn): Update instruction pattern to allow 6-bit constants
11355 as source for store instructions.
11357 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
11359 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
11361 2018-04-30 Nathan Sidwell <nathan@acm.org>
11362 Sandra Loosemore <sandra@codesourcery.com>
11364 * dumpfile.c (dump_open): Allow '-' for stdout.
11365 * doc/invoke.texi (Developer Options): Document dump filename
11366 determination early. Document stdin/stdout selection.
11368 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
11370 Microblaze Target: PIC data text relative
11372 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
11373 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
11375 * config/microblaze/microblaze.h (microblaze_constant_address_p):
11376 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
11377 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
11378 New addressing mode for data-text relative position indepenedent code.
11379 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
11380 'ADDRESS_SYMBOLIC_TXT_REL'.
11381 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
11382 (microblaze_legitimate_pic_operand): Exclude function calls from
11383 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
11384 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
11386 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
11387 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
11388 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
11389 for 'address + offset'.
11390 (microblaze_expand_prologue): Add new function prologue call for
11392 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
11393 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
11394 table in case of TARGET_PIC_DATA_TEXT_REL.
11395 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
11396 * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
11397 Add new macros 'UNSPEC_TEXT',
11398 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
11399 + exclude function calls from 'UNSPEC_PLT' in case of data text
11401 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
11402 new target hook for generating address diff vector tables in case of
11404 * doc/tm.texi : Regenerate.
11405 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
11406 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
11407 of addr diff vector generation.
11408 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
11409 target hook definition.
11410 * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
11411 Add default function for generate_pic_addr_diff_vec -> flag_pic.
11412 * doc/invoke.texi (Add new pic option): Add new microblaze pic
11413 option for data text relative.
11415 2018-04-30 Richard Biener <rguenther@suse.de>
11417 * tree-chrec.h (evolution_function_is_constant_p): Remove
11419 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
11421 2018-04-30 Richard Biener <rguenther@suse.de>
11424 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
11426 2018-04-30 Richard Biener <rguenther@suse.de>
11428 PR tree-optimization/28364
11429 PR tree-optimization/85275
11430 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
11431 copying first exit test.
11433 2018-04-28 Mark Wielaard <mark@klomp.org>
11435 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
11436 dwarf_version >= 5.
11437 (dwarf_AT): Handle DW_AT_addr_base.
11438 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
11440 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
11443 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
11444 (*ashl<dwi>3_doubleword_mask_1): Ditto.
11445 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
11446 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
11448 2018-04-28 Richard Biener <rguenther@suse.de>
11450 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
11451 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
11452 to reflect use. Only add interesting stmts.
11454 2018-04-27 Martin Jambor <mjambor@suse.cz>
11457 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
11458 the jump function allows for passing through aggregate values.
11460 2018-04-27 David Malcolm <dmalcolm@redhat.com>
11462 * input.h (in_system_header_at): Convert from macro to inline
11464 (from_macro_expansion_at): Likewise.
11465 (from_macro_definition_at): Likewise.
11467 2018-04-27 Jeff Law <law@redhat.com>
11469 * config.gcc: Mark tile* targets as deprecated/obsolete.
11471 2018-04-27 Richard Biener <rguenther@suse.de>
11473 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
11476 2018-04-27 Richard Biener <rguenther@suse.de>
11478 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
11480 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
11482 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
11483 with Yd constraint. Set "preferred_for_speed" attribute from
11484 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
11485 with Yd constraint.
11486 (*movdi_internal): Ditto.
11487 (movti_interunit splitters): Remove
11488 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
11489 (movdi_interunit splitters): Ditto.
11490 * config/i386/constraints.md (Ye): Remove.
11491 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
11493 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11496 * config/aarch64/constraints.md (Usg): Limit to 31.
11497 (Usj): Limit to 63.
11499 2018-04-27 Jakub Jelinek <jakub@redhat.com>
11501 PR tree-optimization/85529
11502 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
11503 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
11504 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
11505 zero extension or masking of the MSB bit.
11506 (optimize_range_tests): Add FIRST_BB argument, pass it through
11507 to optimize_range_tests_var_bound.
11508 (maybe_optimize_range_tests, reassociate_bb): Adjust
11509 optimize_range_tests callers.
11511 2018-04-26 Richard Biener <rguenther@suse.de>
11512 Jakub Jelinek <jakub@redhat.com>
11514 * cgraph.h (symbol_table): Just declare debug method here.
11515 * symtab.c (symbol_table::debug): Define.
11517 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
11519 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
11521 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
11523 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
11524 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
11525 (*movdi_internal): Substitute Yi and Yj constraint with x
11526 and Ym and Yn constraint with y constraint. Update "isa"
11527 attribute and set "preferred_for_speed" attribute from
11528 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
11529 (*movsi_internal): Ditto.
11530 (*movdf_internal): Ditto.
11531 (*movsf_internal): Ditto.
11532 (*zero_extendsidi2): Ditto.
11533 * config/i386/sse.md (vec_set<mode>_0): Ditto.
11534 (sse2_loadld): Ditto.
11535 (*vec_extract<ssevecmodelower>_0): Ditto.
11536 (*vec_extractv4si_0_zext_sse4): Ditto.
11537 (vec_concatv2di): Ditto.
11538 (*vec_dup<mode>): Ditto.
11539 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
11540 * config/i386/constraints.md (Yi): Remove.
11545 2018-04-26 Nathan Sidwell <nathan@acm.org>
11547 * dumpfile.c (dump_open): New.
11548 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
11549 (dump_finish): Detect stdio/stderr by value not name.
11551 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
11553 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
11555 2018-04-26 Tom de Vries <tom@codesourcery.com>
11558 * config/nvptx/nvptx.c (verify_neutering_jumps)
11559 (verify_neutering_labels): New function
11560 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
11562 2018-04-26 Tom de Vries <tom@codesourcery.com>
11565 * config/nvptx/nvptx.c (needs_neutering_p): New function.
11566 (nvptx_single): Use needs_neutering_p to skip over insns that do not
11569 2018-04-26 Richard Biener <rguenther@suse.de>
11570 Tom de Vries <tom@codesourcery.com>
11573 * lto-streamer-out.c (output_function): Fixup loops if required to match
11574 discovery done in the reader.
11576 2018-04-26 Richard Biener <rguenther@suse.de>
11578 PR tree-optimization/85116
11579 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
11580 have a loop exit from the single latch predecessor. Remove
11581 case of header with just condition.
11582 (ch_base::copy_headers): Exclude infinite loops from any
11584 (pass_ch::execute): Record exits.
11586 2018-04-26 Richard Biener <rguenther@suse.de>
11588 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
11589 prologue cost vector and pass it to vect_get_load_cost.
11590 (vect_get_peeling_costs_all_drs): Likewise.
11591 (vect_peeling_hash_get_lowest_cost): Likewise.
11592 (vect_enhance_data_refs_alignment): Likewise.
11594 2018-04-26 Richard Biener <rguenther@suse.de>
11596 PR middle-end/85450
11597 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
11598 checking of integer<->pointer conversions.
11599 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
11600 sign-/zero-extending pointer types.
11601 (expand_omp_for_static_chunk): Likewise.
11603 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
11604 Jean Lee <xiaoyur347@gmail.com>
11606 * config/mips/mips.c (mips_asan_shadow_offset): New function.
11607 (TARGET_ASAN_SHADOW_OFFSET): Define.
11608 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
11609 true for -fsanitize=address.
11611 2018-04-25 Mark Wielaard <mark@klomp.org>
11613 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
11616 2018-04-25 Jakub Jelinek <jakub@redhat.com>
11618 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
11619 than "alu", remove explicit "memory" and "imm_disp" attributes.
11620 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
11622 PR middle-end/85414
11623 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
11624 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
11625 gen_lowpart_no_emit.
11627 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
11630 * config/i386/i386.c (ix86_expand_builtin): Change memory
11631 operand to XI, extend p0 to Pmode.
11632 * config/i386/i386.md: Change unspec volatile and operand
11633 1 mode to XI, change operand 0 mode to P.
11635 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
11637 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
11638 GET_MODE_MASK before any checking.
11639 (nds32_can_use_bset_p): Likewise.
11640 (nds32_can_use_btgl_p): Likewise.
11642 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
11644 * config/nds32/nds32-doubleword.md: New define_split pattern for
11645 illegal register number.
11647 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
11649 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
11651 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
11653 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
11655 2018-04-25 Richard Biener <rguenther@suse.de>
11657 * lto-streamer.h (LTO_major_version): Bump to 8.
11659 2018-04-25 Jakub Jelinek <jakub@redhat.com>
11661 * BASE-VER: Set to 9.0.0.
11663 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
11665 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
11666 in __abskf2 and __powikf2.
11668 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11671 * config/aarch64/constraints.md (Usg, Usj): New constraints.
11672 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
11673 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
11674 Use the above on operand 2. Reindent.
11675 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
11677 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
11680 * common/config/i386/i386-common.c (ix86_handle_option): Don't
11682 * config/i386/i386.opt (mcet): Removed.
11683 * doc/install.texi: Remove -mcet documentation.
11684 * doc/invoke.texi: Likewise.
11686 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
11689 * doc/install.texi: Remove -mcet from bootstrap-cet.
11691 2018-04-24 Jakub Jelinek <jakub@redhat.com>
11694 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
11695 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
11699 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
11700 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
11701 containing a CONST_VECTOR.
11703 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
11705 * doc/install.texi: Update newlib dependency for nvptx.
11707 2018-04-24 Jakub Jelinek <jakub@redhat.com>
11710 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
11711 instead of INTVAL when shifting x left.
11713 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
11715 PR tree-optimization/85478
11716 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
11717 vect_grouped_store_supported for single element vectors.
11719 2018-04-24 Richard Biener <rguenther@suse.de>
11722 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
11723 load cost increase to the case of non-constant step.
11725 2018-04-24 Jakub Jelinek <jakub@redhat.com>
11728 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
11729 destination if any_malformed_asm.
11731 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
11733 PR middle-end/85496
11734 * expr.c (store_field): In the bitfield case, if the value comes from
11735 a function call and is returned in registers by means of a PARALLEL,
11736 do not change the mode of the temporary unless BLKmode and VOIDmode.
11738 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
11740 PR rtl-optimization/85423
11741 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
11742 dependencies to debug insns when the previous insn is non-debug.
11744 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
11746 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
11747 enums into a single definition.
11748 (fls): Fix predicates and printing.
11751 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
11753 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
11754 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
11755 and short u6 immediate.
11756 (check_if_valid_sleep_operand): Remove.
11757 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
11759 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
11761 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
11762 flag_always_save_lp condition.
11763 * config/nds32/nds32.opt (malways-save-lp): New option.
11765 2018-04-22 Shiva Chen <shiva0217@gmail.com>
11767 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
11768 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
11769 * config/nds32/nds32.h
11770 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
11771 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
11773 2018-04-22 Shiva Chen <shiva0217@gmail.com>
11775 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
11776 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
11778 2018-04-22 Shiva Chen <shiva0217@gmail.com>
11779 Chung-Ju Wu <jasonwucj@gmail.com>
11781 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
11783 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
11784 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
11786 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
11788 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
11790 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
11792 * config/nds32/nds32-protos.h (nds32_data_alignment,
11793 nds32_local_alignment): Declare.
11794 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
11795 nds32_local_alignment): New functions.
11796 (TARGET_CONSTANT_ALIGNMENT): Define.
11797 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
11799 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
11801 * config/nds32/nds32.c
11802 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
11803 (TARGET_MODES_TIEABLE_P): Likewise.
11805 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
11807 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
11808 level Ofast and Og.
11810 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
11811 Chung-Ju Wu <jasonwucj@gmail.com>
11813 * config/nds32/constants.md (unspec_volatile_element): Add enum values
11814 for unaligned access.
11815 * config/nds32/nds32-intrinsic.c: Implementation of expanding
11817 * config/nds32/nds32-intrinsic.md: Likewise.
11818 * config/nds32/nds32_intrinsic.h: Likewise.
11819 * config/nds32/nds32.h (nds32_builtins): Likewise.
11820 * config/nds32/nds32.opt (munaligned-access): New option.
11821 * config/nds32/nds32.c (nds32_asm_file_start): Display
11822 flag_unaligned_access status.
11824 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
11826 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
11827 -mno-relax is present.
11828 * config/riscv/linux.h (LINK_SPEC): Ditto.
11830 2018-04-20 Martin Sebor <msebor@redhat.com>
11833 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
11835 (gimple_fold_builtin_stxcpy_chk): Same.
11836 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
11838 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
11841 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
11842 __powikf2 when long double is IEEE 128-bit.
11844 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
11846 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
11847 step to make sure stack always aligned.
11849 2018-04-20 Carl Love <cel@us.ibm.com>
11852 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
11853 size check for arg0.
11855 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
11856 Tom de Vries <tom@codesourcery.com>
11859 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
11860 Emit insns for calls too.
11861 (nvptx_find_par): Always look for worker-level predecessor insn.
11862 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
11864 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
11865 (nvptx_process_pars): Propagate frames for calls.
11867 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
11870 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
11872 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
11873 (ix86_handle_option): Don't handle OPT_mibt.
11874 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
11876 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
11878 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
11879 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
11880 (ix86_target_macros): Define __CET__ with flag_cf_protection
11881 for -fcf-protection.
11882 * config/i386/i386.c (isa2_opts): Remove -mibt.
11883 * config/i386/i386.h (TARGET_IBT): Removed.
11884 (TARGET_IBT_P): Likewise.
11885 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
11886 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
11887 * config/i386/i386.opt (mcet): Update help message.
11888 (mshstk): Likewise.
11890 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
11891 -mcet as an alias for -mshstk.
11893 2018-04-20 Richard Biener <rguenther@suse.de>
11895 PR middle-end/85475
11896 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
11897 complexity by forcing a single use of the multiply operand.
11899 2018-04-20 Martin Jambor <mjambor@suse.cz>
11902 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
11903 recursion dependency to only apply to non-clones.
11905 2018-04-20 Martin Jambor <mjambor@suse.cz>
11908 * ipa-cp.c (create_specialized_node): Check that clones of
11909 self-recursive edges exist during IPA-CP.
11911 2018-04-19 Toon Moene <toon@moene.org>
11913 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
11916 2018-04-19 Jakub Jelinek <jakub@redhat.com>
11918 PR tree-optimization/85467
11919 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
11920 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
11921 VECTOR_CST element to type.
11923 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
11926 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
11927 * config/i386/i386.md (builtin_setjmp_setup): Removed.
11928 (builtin_longjmp): Likewise.
11929 (save_stack_nonlocal): New pattern.
11930 (restore_stack_nonlocal): Likewise.
11932 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
11935 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
11936 Replace ASM_OUTPUT_LABEL with fprintf.
11938 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
11941 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
11942 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
11943 * config/i386/i386-c.c (ix86_target_macros_internal): Also
11944 define __IBT__ and __SHSTK__ for -fcf-protection.
11945 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
11947 (ix86_trampoline_init): Likewise.
11948 (x86_output_mi_thunk): Likewise.
11949 (ix86_notrack_prefixed_insn_p): Likewise.
11950 (ix86_option_override_internal): Don't disallow -fcf-protection.
11951 * config/i386/i386.md (rdssp<mode>): Also enable for
11953 (incssp<mode>): Likewise.
11954 (nop_endbr): Likewise.
11955 * config/i386/i386.opt (mcet): Change help message to built-in
11958 (mshstk): Likewise.
11959 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
11960 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
11961 enable CET built-in functions.
11963 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
11965 * common/config/i386/i386-common.c
11966 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
11967 OPTION_MASK_ISA_MOVDIRI_UNSET,
11968 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
11969 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
11970 * config.gcc (movdirintrin.h): New header.
11971 * config/i386/cpuid.h (bit_MOVDIRI,
11972 bit_MOVDIR64B): New bits.
11973 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
11975 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
11976 (VOID, PVOID, PCVOID)): New function types.
11977 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
11978 __builtin_ia32_directstoreu_u64,
11979 __builtin_ia32_movdir64b): New builtins.
11980 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
11981 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
11983 (ix86_valid_target_attribute_inner_p): Ditto.
11984 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
11985 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
11986 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
11987 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
11988 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
11989 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
11990 (movdiri<mode>, movdir64b_<mode>): New.
11991 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
11992 * config/i386/immintrin.h: Include movdirintrin.h.
11993 * config/i386/movdirintrin.h: New file.
11994 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
11996 2018-04-19 Richard Biener <rguenther@suse.de>
11998 PR middle-end/85455
11999 * cfg.c (clear_bb_flags): When loop state says we have
12000 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
12002 2018-04-19 Richard Biener <rguenther@suse.de>
12004 PR tree-optimization/84737
12005 * tree-vect-data-refs.c (vect_copy_ref_info): New function
12006 copying restrict info.
12007 (vect_setup_realignment): Use it.
12008 * tree-vectorizer.h (vect_copy_ref_info): Declare.
12009 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
12010 the first DR to all generated stores.
12011 (vectorizable_load): Likewise for loads.
12013 2018-04-19 Jakub Jelinek <jakub@redhat.com>
12015 PR tree-optimization/85446
12016 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
12017 the integral and pointer types to have the same precision.
12019 * doc/install.texi: Document --disable-cet being the default and
12022 2018-04-18 Martin Liska <mliska@suse.cz>
12024 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
12027 2018-04-18 Martin Liska <mliska@suse.cz>
12030 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
12033 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
12034 arguments if they are comparable.
12036 2018-04-18 Martin Liska <mliska@suse.cz>
12039 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
12042 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
12045 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
12048 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
12049 ENDBR after calling __morestack.
12051 2018-04-18 David Malcolm <dmalcolm@redhat.com>
12054 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
12055 by using gcc_base_ver to generate a gcc_driver_version, and use
12056 it when generating GCC_DRIVER_NAME.
12057 * configure: Regenerate.
12059 2018-04-18 Jakub Jelinek <jakub@redhat.com>
12062 * config.gcc: Obsolete powerpc*-*-*spe*.
12064 2018-04-17 Jakub Jelinek <jakub@redhat.com>
12067 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
12068 (stabstr_D): Change type of unum from unsigned int to
12069 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
12072 2018-04-17 Jim Wilson <jimw@sifive.com>
12075 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
12076 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
12077 Set arg_pointer_offset after using pretend_args_size.
12079 2018-04-17 Jakub Jelinek <jakub@redhat.com>
12081 PR rtl-optimization/85431
12082 * dse.c (record_store): Ignore zero width stores.
12085 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
12086 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
12087 __builtin_stack_restore rather than after it.
12088 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
12089 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
12090 argument instead of virtual_dynamic_stack_rtx.
12092 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
12094 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
12096 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
12097 Add note to error message to explain internal mapping of overloaded
12098 built-in function name to non-overloaded built-in function name.
12099 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
12102 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
12105 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
12106 where the inputs overlap with the output.
12108 2018-04-17 Jakub Jelinek <jakub@redhat.com>
12110 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
12111 (=v, v) alternative and explicit "memory" attribute.
12112 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
12113 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
12115 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
12116 "sselog1" type instead of "sselog".
12117 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
12118 "sselog". Remove explicit "memory" attribute.
12119 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
12120 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
12122 (vec_extract_hi_v32hi): Merge all alternatives into one, use
12123 "sselog1" type instead of "sselog". Remove explicit "memory"
12125 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
12126 use "sselog1" type instead of "sselog". Remove explicit "memory"
12128 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
12129 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
12131 (vec_extract_hi_v64qi): Merge all alternatives into one, use
12132 "sselog1" type instead of "sselog". Remove explicit "memory"
12134 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
12135 use "sselog1" type instead of "sselog". Remove explicit "memory"
12139 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
12141 PR middle-end/85414
12142 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
12145 2018-04-17 Martin Jambor <mjambor@suse.cz>
12148 * ipa-cp.c (create_specialized_node): Call
12149 expand_all_artificial_thunks if necessary.
12151 2018-04-17 Martin Liska <mliska@suse.cz>
12154 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
12155 in message, remote space in between '_G' and '('.
12157 2018-04-17 Jakub Jelinek <jakub@redhat.com>
12160 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
12161 avx512f_vmcmp<mode>3<round_saeonly_name>,
12162 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
12163 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
12164 avx512f_rndscale<mode><round_saeonly_name>,
12165 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
12166 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
12167 Use %<iptr>2 instead of %2 for -masm=intel.
12168 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
12169 avx512f_vcvttss2usi<round_saeonly_name>,
12170 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
12172 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
12173 avx512f_vcvttsd2usi<round_saeonly_name>,
12174 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
12175 Use %q1 instead of %1 for -masm=intel.
12176 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
12177 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
12178 of %3 for -masm=intel.
12179 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
12181 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
12183 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
12185 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
12186 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
12188 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
12190 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
12191 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
12192 %g1 and one with %0 and %1.
12193 (avx512er_vmrcp28<mode><round_saeonly_name>,
12194 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
12195 %1 for -masm=intel.
12196 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
12197 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
12198 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
12199 of %0 and %{%4%} for -masm=intel.
12200 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
12201 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
12202 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
12203 order of %0 and %{%5%}%{z%} for -masm=intel.
12205 2018-04-17 Jan Hubicka <jh@suse.cz>
12208 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
12210 2018-04-17 Martin Liska <mliska@suse.cz>
12213 * multiple_target.c (create_dispatcher_calls): Set apostrophes
12214 for target_clone error message. Make default implementation
12215 clone to be a local declaration.
12216 (separate_attrs): Add new argument and check for an empty
12218 (expand_target_clones): Handle it.
12219 (ipa_target_clone): Make redirection just for target_clones
12222 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
12223 Tom de Vries <tom@codesourcery.com>
12225 PR middle-end/84955
12226 * omp-expand.c (expand_oacc_for): Add dummy false branch for
12227 tiled basic blocks without omp continue statements.
12229 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
12232 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
12233 vec_extract expression as having side effects to make sure it gets
12236 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
12239 * config/i386/i386.c (get_builtin_code_for_version): Check
12242 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
12245 * config.gcc: Support "skylake".
12246 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
12248 * config/i386/i386.c (m_SKYLAKE): Define.
12249 (processor_target_table): Add "skylake".
12250 (ix86_option_override_internal): Add "skylake".
12251 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
12252 PROCESSOR_CANNONLAKE.
12253 (get_builtin_code_for_version): Fix priority for
12254 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
12255 PROCESSOR_SKYLAKE-AVX512.
12256 * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
12257 (processor_type): Add PROCESSOR_SKYLAKE.
12259 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
12260 Jason Merrill <jason@redhat.com>
12263 * convert.c (convert_to_integer_1): Use direct recursion for
12264 enumeral types and types with a precision less than the number
12265 of bits in their mode.
12267 2018-04-16 Julia Koval <julia.koval@intel.com>
12270 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
12271 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
12273 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
12276 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
12277 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
12278 and -mno-direct-move.
12280 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
12283 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
12284 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
12285 Ensure negative shifts result in {0}.
12287 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
12289 PR rtl-optimization/79916
12290 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
12291 regs (if any) to define how to gnerate SD moves when LRA is in
12294 2018-04-13 Jakub Jelinek <jakub@redhat.com>
12296 PR rtl-optimization/85393
12297 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
12298 * except.c (expand_dw2_landing_pad_for_region): Make static.
12299 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
12300 a label and unconditional jump to old_bb, rather than
12301 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
12304 PR rtl-optimization/85376
12305 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
12306 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
12307 instead of a specific value.
12309 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
12310 Bin Cheng <bin.cheng@arm.com>
12312 PR tree-optimization/82965
12313 PR tree-optimization/83991
12314 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
12315 by_profile_only parameter.
12316 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
12317 information if the loop was predicted to iterate too many times.
12318 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
12320 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
12323 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
12326 2018-04-13 Martin Liska <mliska@suse.cz>
12327 Jakub Jelinek <jakub@redhat.com>
12329 PR middle-end/81657
12330 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
12331 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
12332 * builtins.c (expand_builtin_memory_copy_args): Use
12333 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
12334 handle dest_addr == pc_rtx.
12336 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
12339 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
12340 asked to not generate direct moves.
12341 (fix_trunc<mode>si2_stfiwx): Similar.
12342 (fix_trunc<mode>si2_internal): Similar.
12344 2018-04-12 Jakub Jelinek <jakub@redhat.com>
12347 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
12348 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
12349 lookup if dest in some wider mode is known to be const0_rtx and
12350 if so, record permanent equivalence for it to be ZERO_EXTEND of
12351 the narrower mode destination.
12353 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
12355 * lto-streamer-out.c (output_function): Revert 259346.
12356 * omp-expand.c (expand_oacc_for): Likewise.
12358 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
12360 PR rtl-optimization/85354
12361 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
12362 * sel-sched.c (sel_global_init): ... here.
12364 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
12367 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
12368 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
12369 mode for PE-COFF targets.
12370 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
12371 (i386_pe_asm_lto_end): Likewise.
12372 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
12373 (TARGET_ASM_LTO_END): Likewise.
12374 * config/i386/winnt.c (saved_debug_info_level): New static variable.
12375 (i386_pe_asm_lto_start): New function.
12376 (i386_pe_asm_lto_end): Likewise.
12378 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
12379 Richard Biener <rguenther@suse.de>
12381 PR middle-end/84955
12382 * lto-streamer-out.c (output_function): Fix CFG loop state before
12384 * omp-expand.c (expand_oacc_for): Handle calls to internal
12385 functions like regular functions.
12387 2018-04-12 Richard Biener <rguenther@suse.de>
12390 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
12391 for the early LTO debug to properly generate references to it
12392 during DIE emission. Do not re-use that for the skeleton for
12394 (dwarf2out_early_finish): Likewise.
12396 2018-04-12 Jakub Jelinek <jakub@redhat.com>
12399 * config/i386/sse.md
12400 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
12401 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
12402 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
12403 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
12404 and output is a reg, avoid creating invalid lowpart subreg, but
12405 instead split into a 512-bit move. Don't split if not AVX512VL,
12406 input is xmm16+ reg and output is a mem.
12407 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
12408 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
12409 xmm16+ reg and output is a mem.
12411 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
12413 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
12414 also for flag_dwarf2_cfi_asm.
12416 2018-04-12 Jakub Jelinek <jakub@redhat.com>
12418 PR rtl-optimization/85342
12419 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
12420 a bool scalar var inside of the loop instead. Don't try to update
12421 recog_data.operand after failed apply_change_group.
12423 2018-04-12 Tom de Vries <tom@codesourcery.com>
12426 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
12427 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
12428 array with flexible array member as array without given dimension.
12429 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
12430 argument for undefined param to true.
12432 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
12435 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
12436 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
12437 from PowerPC section.
12438 * config/rs6000/sysv4.opt (mcall-): Improve help text.
12439 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
12440 help text that is too long.
12441 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
12442 help text that is too long.
12443 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
12444 help text that is too long.
12446 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
12448 * config/alpha/alpha.md (stack_probe_internal): Rename
12449 from "probe_stack". Update all callers.
12451 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
12453 PR rtl-optimization/84566
12454 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
12455 sched_macro_fuse_insns.
12457 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
12460 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
12461 (compute_block_dependences): ... from here.
12463 2018-04-11 Jakub Jelinek <jakub@redhat.com>
12465 PR tree-optimization/85331
12466 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
12467 from int to HOST_WIDE_INT.
12469 2018-04-11 Martin Jambor <mjambor@suse.cz>
12472 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
12473 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
12474 not the same as the source val.
12475 (cgraph_edge_brings_value_p): New parameter.
12476 (gather_edges_for_value): Pass destination value to
12477 cgraph_edge_brings_value_p.
12478 (perhaps_add_new_callers): Likewise.
12479 (get_info_about_necessary_edges): Likewise and exclude values brought
12480 only by self-recursive edges.
12481 (create_specialized_node): Redirect only clones of self-calling edges.
12482 (+self_recursive_pass_through_p): New function.
12483 (find_more_scalar_values_for_callers_subset): Use it.
12484 (find_aggregate_values_for_callers_subset): Likewise.
12485 (known_aggs_to_agg_replacement_list): Removed.
12486 (decide_whether_version_node): Re-calculate known constants for all
12487 remaining context clones.
12489 2018-04-11 Richard Biener <rguenther@suse.de>
12492 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
12493 from early DWARF output.
12494 (dwarf2out_early_finish): Output line info unconditionally into
12495 early DWARF and add reference to it.
12497 2018-04-11 Jakub Jelinek <jakub@redhat.com>
12500 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
12501 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
12502 other than V2DFmode using iptr mode attribute.
12503 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
12505 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
12507 PR rtl-optimization/84659
12508 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
12510 2018-04-11 Jakub Jelinek <jakub@redhat.com>
12513 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
12515 (output_loc_list): Pass address of a dummy size variable even in the
12516 locview handling loop.
12517 (index_location_lists): Add comment on why skip_loc_list_entry can't
12520 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
12523 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
12526 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
12529 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
12530 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
12531 and -mstring-compare-inline-limit.
12533 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
12536 * config/rs6000/rs6000.md (allocate_stack): Put the residual size
12537 for stack clash protection in a register whenever we need it to be in
12540 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
12542 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
12543 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
12545 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
12548 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
12550 (mlong-double-): Ditto.
12551 * config/rs6000/sysv4.opt (msdata=): Ditto.
12552 (mtls-size=): Ditto.
12554 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
12556 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
12557 erroneous entries for
12558 "vector int vec_ldl (int, long int *)", and
12559 "vector unsigned int vec_ldl (int, unsigned long int *)".
12560 Add comments and entries for
12561 "vector bool char vec_ldl (int, bool char *)",
12562 "vector bool short vec_ldl (int, bool short *)",
12563 "vector bool int vec_ldl (int, bool int *)",
12564 "vector bool long long vec_ldl (int, bool long long *)",
12565 "vector pixel vec_ldl (int, pixel *)",
12566 "vector long long vec_ldl (int, long long *)",
12567 "vector unsigned long long vec_ldl (int, unsigned long long *)".
12568 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
12569 type tree bool_long_long_type_node and correct definition of
12570 bool_V2DI_type_node to make reference to this new type tree.
12571 (rs6000_mangle_type): Replace erroneous reference to
12572 bool_long_type_node with bool_long_long_type_node.
12573 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
12574 comments to emphasize sign distinctions for char and int types and
12575 replace RS6000_BTI_bool_long constant with
12576 RS6000_BTI_bool_long_long constant. Also add comment to restrict
12577 use of RS6000_BTI_pixel.
12578 (bool_long_type_node): Remove this macro definition.
12579 (bool_long_long_type_node): New macro definition
12581 2018-04-10 Jakub Jelinek <jakub@redhat.com>
12583 PR rtl-optimization/85300
12584 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
12585 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
12586 simplify_unary_operation fails.
12588 2018-04-10 Martin Liska <mliska@suse.cz>
12590 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
12591 cgraph_edge and ipa_ref.
12593 2018-04-10 Jakub Jelinek <jakub@redhat.com>
12597 * config/i386/sse.md
12598 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
12599 computation of the VEC_MERGE selector from mask.
12600 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
12601 Fix decoding of the VEC_MERGE selector into mask.
12603 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
12605 PR tree-optimization/85286
12606 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
12608 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
12610 * final.c (final_1): Set insn_last_address as well as
12611 insn_current_address.
12613 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12616 * explow.c (emit_stack_probe): Call validize_mem on memory location
12617 before passing it to gen_probe_stack. Create address operand and
12618 legitimize it for the probe_stack_address case.
12620 2018-04-09 Jan Hubicka <jh@suse.cz>
12623 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
12624 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
12625 * tree.c (free_lang_data_in_type): Fix handling of binfos;
12627 (free_lang_data): Rebuild type inheritance graph.
12629 2018-04-09 Martin Sebor <msebor@redhat.com>
12631 * invoke.texi (-finline-small-functions): Mention other optimization
12633 (-findirect-inlining, -fpartial-inlining): Same.
12634 (-finline-functions-called-once): Same.
12635 (-freorder-blocks-and-partition): Same.
12637 2018-04-09 Jan Hubicka <jh@suse.cz>
12640 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
12641 jumps; choose last target that matches the criteria (i.e.
12642 no partition changes for non-crossing jumps).
12643 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
12644 support for redirecting crossing jumps to non-crossing.
12646 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
12648 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
12649 also for naked functions.
12651 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
12653 * config/arc/arc.md (add_shift): New pattern.
12654 (add_shift2): Likewise.
12655 (sub_shift): Likewise.
12656 (sub_shift_cmp0_noout): Likewise.
12657 (compare_si_ashiftsi): Likewise.
12658 (xbfu_cmp0_noout): New combine pattern.
12659 (xbfu_cmp0"): Likewise.
12660 (movsi_set_cc_insn): Place the predicable variant first.
12661 (commutative_binary_cmp0_noout): Remove clobber.
12662 (commutative_binary_cmp0): New pattern.
12663 (noncommutative_binary_cmp0): Likewise.
12664 (noncommutative_binary_cmp0_noout): Likewise.
12665 (noncommutative_binary_comparison_result_used): Removed.
12666 (rsub_cmp0): New pattern.
12667 (rsub_cmp0_noout): Likewise.
12668 (extzvsi): Changed, keep only meaningful variants.
12669 (SQH, SEZ): New iterators.
12670 (SQH_postfix): New mode attribute.
12671 (SEZ_prefix): New code attribute.
12672 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
12673 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
12674 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
12675 of numerical value.
12676 (noncommutative_operator): Check the availability of barrel
12679 2018-04-09 Richard Biener <rguenther@suse.de>
12681 PR tree-optimization/85284
12682 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
12683 Only use the niter constraining form of simple_iv when the exit
12684 is always executed.
12686 2018-04-09 Tom de Vries <tom@codesourcery.com>
12689 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
12690 (define_expand "*memory_barrier"): New define_expand.
12691 (define_insn "memory_barrier"): New insn.
12693 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
12695 PR rtl-optimization/80463
12696 PR rtl-optimization/83972
12697 PR rtl-optimization/83480
12699 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
12700 correct producer for the insn.
12701 (tidy_control_flow): Fixup seqnos in case of debug insns.
12703 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
12705 PR rtl-optimization/83913
12707 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
12708 different sched-times when merging exprs.
12710 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
12712 PR rtl-optimization/83962
12714 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
12715 tidy_fallthru_edge and tidy_control_flow.
12717 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
12719 PR rtl-optimization/83530
12721 * sel-sched.c (force_next_insn): New global variable.
12722 (remove_insn_for_debug): When force_next_insn is true, also leave only
12723 next insn in the ready list.
12724 (sel_sched_region): When the region wasn't scheduled, make another pass
12725 over it with force_next_insn set to 1.
12727 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
12729 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
12731 * config/nds32/constants.md (unspec_volatile_element): Add enum values
12732 for interrupt control.
12733 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
12734 functions for interrupt control.
12735 * config/nds32/nds32-intrinsic.md: Likewise.
12736 * config/nds32/nds32_intrinsic.h: Likewise.
12737 * config/nds32/nds32.h (nds32_builtins): Likewise.
12739 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
12741 * config/nds32/nds32.c (nds32_init_machine_status,
12742 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
12743 strict_aligned_p field.
12744 (nds32_expand_to_rtl_hook): New function.
12745 (TARGET_EXPAND_TO_RTL_HOOK): Define.
12746 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
12748 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
12749 Chung-Ju Wu <jasonwucj@gmail.com>
12751 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
12752 * config/nds32/nds32-n7.md: New file.
12753 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
12754 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
12756 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
12757 * config/nds32/nds32.md (pipeline_model): Add n7.
12758 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
12759 * config/nds32/pipelines.md: Include n7 settings.
12761 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
12762 Chung-Ju Wu <jasonwucj@gmail.com>
12764 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
12765 * config/nds32/nds32-e8.md: New file.
12766 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
12767 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
12769 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
12770 * config/nds32/nds32.md (pipeline_model): Add e8.
12771 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
12772 * config/nds32/pipelines.md: Include e8 settings.
12774 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
12775 Chung-Ju Wu <jasonwucj@gmail.com>
12777 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
12778 * config/nds32/nds32-n8.md: New file.
12779 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
12780 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
12782 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
12783 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
12784 * config/nds32/nds32.md (pipeline_model): Add n8.
12785 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
12786 * config/nds32/pipelines.md: Include n8 settings.
12788 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
12789 Chung-Ju Wu <jasonwucj@gmail.com>
12791 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
12792 * config/nds32/nds32-n9-2r1w.md: New file.
12793 * config/nds32/nds32-n9-3r2w.md: New file.
12794 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
12795 nds32_register_ports): New or modify for cpu n9.
12796 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
12798 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
12799 * config/nds32/nds32-utils.c: New file.
12800 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
12801 TARGET_MUL_SLOW): Define.
12802 * config/nds32/nds32.md (pipeline_model): New attribute.
12803 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
12804 New options that support cpu n9.
12805 * config/nds32/pipelines.md: Include n9 settings.
12806 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
12808 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
12810 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
12811 information if necessary.
12812 (output_cond_branch_compare_zero): Likewise.
12813 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
12814 (nds32_target_alignment): Refine for alignment.
12815 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
12816 (FUNCTION_BOUNDARY): Modify.
12817 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
12819 * config/nds32/nds32.opt (malways-align, malign-functions): New.
12821 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
12823 * config/nds32/constants.md (unspec_volatile_element): Add values for
12824 TLB operation and data prefetch.
12825 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
12826 functions for TLB operation and data prefetch.
12827 * config/nds32/nds32-intrinsic.md: Likewise.
12828 * config/nds32/nds32_intrinsic.h: Likewise.
12829 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
12830 (nds32_print_operand): Likewise.
12831 * config/nds32/nds32.h (nds32_builtins): Likewise.
12833 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
12834 Andrew Pinski <pinsika@gcc.gnu.org>
12836 PR middle-end/82976
12837 * match.pd: Use constant_boolean_node of correct type instead of
12838 boolean_true_node or boolean_false_node for simplifying
12839 pointer comparisons to zero.
12841 2018-04-07 Jakub Jelinek <jakub@redhat.com>
12843 PR tree-optimization/80021
12844 * tree.c (verify_type_variant): Make error call in verify_variant_match
12845 translatable and remove final full stop.
12847 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
12849 * config/nds32/constants.md (unspec_volatile_element): Add
12850 UNSPEC_VOLATILE_EH_RETURN.
12851 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
12852 nds32_output_stack_pop): Support dwarf exception handling process.
12853 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
12854 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
12855 exception handling process.
12856 (nds32_compute_stack_frame): Likewise.
12857 (nds32_return_addr_rtx): Likewise.
12858 (nds32_initial_elimination_offset): Likewise.
12859 (nds32_expand_prologue): Likewise.
12860 (nds32_expand_epilogue): Likewise.
12861 (nds32_dynamic_chain_address): New function.
12862 * config/nds32/nds32.h (machine_function): Add fields for dwarf
12863 exception handling.
12864 (DYNAMIC_CHAIN_ADDRESS): Define.
12865 (EH_RETURN_DATA_REGNO): Define.
12866 (EH_RETURN_STACKADJ_RTX): Define.
12867 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
12868 patterns for dwarf exception handling.
12870 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
12872 * config/nds32/nds32.h: Clean up obsolete macros.
12874 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
12876 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
12877 Add enum values for particular instructions.
12878 * config/nds32/nds32-intrinsic.c: Implementation of expanding
12879 particular intrinsic functions.
12880 * config/nds32/nds32-intrinsic.md: Likewise.
12881 * config/nds32/nds32_intrinsic.h: Likewise.
12882 * config/nds32/nds32.h (nds32_builtins): Likewise.
12883 * config/nds32/nds32.md (type): Add pbsad and pbsada.
12884 (btst, ave): New patterns for particular instructions.
12886 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
12888 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
12889 Add enum values for atomic load/store and memory sync.
12890 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
12892 * config/nds32/nds32-intrinsic.md: Likewise.
12893 * config/nds32/nds32_intrinsic.h: Likewise.
12894 * config/nds32/nds32.h (nds32_builtins): Likewise.
12896 2018-04-07 Jakub Jelinek <jakub@redhat.com>
12898 PR tree-optimization/85257
12899 * fold-const.c (native_encode_vector): If not all elts could fit
12900 and off is -1, return 0 rather than offset.
12901 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
12902 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
12903 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
12904 adjust buffer in native_interpret_expr call.
12906 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
12908 * config/nds32/constants.md (unspec_volatile_element): Add cache
12909 control enum values.
12910 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
12911 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
12912 * config/nds32/nds32.c (nds32_cctl_names): New.
12913 (nds32_print_operand): Handle cache control register names.
12914 * config/nds32/nds32.h (nds32_builtins): New enum values.
12915 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
12917 * config/nds32/nds32.md (type): Add mmu.
12918 * config/nds32/pipelines.md (simple_insn): Add mmu.
12920 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
12922 * config/nds32/nds32.md (type): Remove call.
12923 * config/nds32/pipelines.md (simple_insn): Likewise.
12925 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
12927 * config/nds32/constants.md (unspec_volatile_element): Add
12928 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
12929 UNSPEC_VOLATILE_FMFCFG.
12930 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
12931 description for fmfcfg and fmfcsr.
12932 (bdesc_1arg): Add fmtcsr.
12933 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
12934 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
12935 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
12936 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
12937 unspec_fmfcfg): New patterns.
12938 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
12939 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
12940 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
12941 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
12942 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
12943 __nds32__fmfcfg): Define.
12945 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
12947 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
12948 intrinsic register names.
12949 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
12950 intrinsic register enum values and macros.
12952 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
12954 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
12955 for load/store addressing form.
12956 (nds32_print_operand_address): Likewise.
12958 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
12961 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
12962 based on LABEL_REF. Remove useless assertion.
12963 (pic_address_needs_scratch): Fix formatting.
12964 (sparc_legitimize_pic_address): Minor tweaks.
12965 (sparc_delegitimize_address): Adjust assertion accordingly.
12966 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
12967 into symbolic_operand.
12968 (movsi_high_pic_label_ref): Likewise.
12969 (movsi_lo_sum_pic_label_ref): Likewise.
12970 (movdi_pic_label_ref): Likewise.
12971 (movdi_high_pic_label_ref): Likewise.
12972 (movdi_lo_sum_pic_label_ref): Likewise.
12974 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
12976 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
12977 custom LIB_SPEC setup.
12979 2018-04-06 Ruslan Bukin <br@bsdpad.com>
12980 Kito Cheng <kito.cheng@gmail.com>
12982 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
12983 * config/riscv/freebsd.h: New.
12985 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
12987 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
12988 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
12991 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
12992 Kito Cheng <kito.cheng@gmail.com>
12994 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
12995 nds32_output_call, nds32_symbol_binds_local_p): New functions.
12996 * config/nds32/nds32-protos.h (nds32_output_call,
12997 nds32_output_return): Declare.
12998 * config/nds32/nds32.md: Refine all the call and return patterns.
13000 2018-04-06 Jakub Jelinek <jakub@redhat.com>
13003 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
13004 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
13006 PR rtl-optimization/84872
13007 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
13008 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
13009 EDGE_CROSSING edge.
13011 2018-04-06 Tamar Christina <tamar.christina@arm.com>
13013 * expr.c (copy_blkmode_to_reg): Revert 254862.
13014 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
13016 2018-04-06 Richard Biener <rguenther@suse.de>
13018 PR middle-end/85244
13019 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
13020 after seeing a component reference with an adjacent field. Treat
13021 refs to arrays at struct end of external decls similar to
13022 refs to unconstrained commons.
13024 2018-04-06 Jakub Jelinek <jakub@redhat.com>
13027 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
13028 look through SAVE_EXPRs with non-side-effects argument. Adjust
13030 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
13033 2018-04-06 Richard Biener <rguenther@suse.de>
13035 PR middle-end/85180
13036 * alias.c (find_base_term): New wrapper around find_base_term
13037 unwinding CSELIB_VAL_PTR changes.
13038 (find_base_term): Do not restore CSELIB_VAL_PTR during the
13041 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
13043 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
13045 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
13046 constant definitions.
13047 ("nop"): lr 0,0 -> nopr r0
13048 ("nop_lr0", "nop_lr1"): New insn definitions.
13050 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
13052 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
13053 NDS32_V3PUSH_AVAILABLE_P macro.
13055 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
13056 Chung-Ju Wu <jasonwucj@gmail.com>
13058 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
13059 (nds32*-*-*): Add float and fpu_config into supported_defaults.
13060 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
13061 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
13062 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
13063 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
13064 * config/nds32/constraints.md: New constraints and checking for hard
13065 float configuration.
13066 * config/nds32/iterators.md: New mode iterator and attribute for hard
13067 float configuration.
13068 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
13070 * config/nds32/nds32-fpu.md: New file.
13071 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
13072 deal with hard float code generation.
13073 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
13075 (abi_type, float_reg_number): New enum type.
13076 * config/nds32/nds32-predicates.c: New predicates for hard float.
13077 * config/nds32/nds32-protos.h: Declare functions for hard float.
13078 * config/nds32/nds32.c: Implementation for hard float configuration.
13079 * config/nds32/nds32.h: Definitions for hard float configuration.
13080 * config/nds32/nds32.md: Include hard float machine description and
13081 modify patterns for hard float configuration.
13082 * config/nds32/nds32.opt: New options for hard float configuration.
13083 * config/nds32/predicates.md: New predicates for hard float
13086 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
13088 * common/config/nds32/nds32-common.c
13089 (nds32_option_optimization_table): Enable -mreleax-hint by default.
13091 2018-04-05 Jakub Jelinek <jakub@redhat.com>
13093 PR middle-end/85195
13094 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
13095 CONSTRUCTOR_ELT (ctor, ...)->value.
13097 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
13100 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
13102 2018-04-05 Tom de Vries <tom@codesourcery.com>
13105 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
13108 2018-04-05 Shiva Chen <shiva0217@gmail.com>
13109 Kito Cheng <kito.cheng@gmail.com>
13111 * config/nds32/constraints.md (U33): Fine-tune checking condition.
13112 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
13113 * config/nds32/nds32.h (nds32_16bit_address_type): Add
13114 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
13116 2018-04-05 Shiva Chen <shiva0217@gmail.com>
13117 Kito Cheng <kito.cheng@gmail.com>
13119 * config/nds32/constraints.md (Ufe): New memory constraint.
13120 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
13121 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
13122 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
13124 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
13125 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
13127 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
13129 * config/nds32/nds32.md: Use optimize_size in the condition for
13130 alu-shift instructions.
13132 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
13134 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
13136 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
13138 * config/nds32/nds32.md (negsi2): Refine pattern.
13140 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
13141 Chung-Ju Wu <jasonwucj@gmail.com>
13143 * config/nds32/iterators.md (shift_rotate): New code iterator.
13144 (shift): New code attribute.
13145 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
13146 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
13147 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
13148 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
13149 bit-wise operations.
13150 (andsi3, *andsi3): Ditto.
13151 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
13152 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
13153 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
13154 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
13155 nds32_ior_operand, nds32_xor_operand): New predicates.
13157 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
13159 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
13160 (addsi3, subsi3): ... this.
13162 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
13164 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
13166 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
13168 * config/nds32/nds32.md: Adjust indention.
13170 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
13172 * config/nds32/nds32.md (feature): New attribute.
13174 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
13176 * config/nds32/nds32.md (subtype): New attribute.
13178 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
13181 * config/arm/arm-builtins.c (arm_expand_builtin): Change
13182 expansion to perform a bitwise AND of the argument followed by a
13183 boolean negation of the result.
13185 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
13187 PR rtl-optimization/84878
13188 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
13189 the basic block. Assert the use reference is not artificial and that
13190 it has an associated insn.
13192 2018-04-04 Michael Matz <matz@suse.de>
13194 * builtins.c (compute_objsize): Pass correct operand
13195 to array_at_struct_end_p.
13197 2018-04-04 Richard Biener <rguenther@suse.de>
13200 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
13201 from contexts for DINFO_LEVEL_TERSE and below.
13203 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
13205 * config/nds32/nds32-doubleword.md (move_<mode>): Require
13206 resiter_operand condition.
13207 * config/nds32/nds32.md (*move<mode>): Ditto.
13209 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
13210 Monk Chiang <sh.chiang04@gmail.com>
13212 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
13214 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
13216 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
13218 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
13219 Kito Cheng <kito.cheng@gmail.com>
13221 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
13222 nds32_cond_code_str, output_cond_branch,
13223 output_cond_branch_compare_zero, nds32_expand_cbranch,
13224 nds32_expand_cstore, nds32_expand_movcc,
13225 nds32_output_cbranchsi4_equality_zero,
13226 nds32_output_cbranchsi4_equality_reg,
13227 nds32_output_cbranchsi4_equality_reg_or_const_int,
13228 nds32_output_cbranchsi4_greater_less_zero: New functions.
13229 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
13230 nds32_expand_cstore, nds32_expand_movcc,
13231 nds32_output_cbranchsi4_equality_zero,
13232 nds32_output_cbranchsi4_equality_reg,
13233 nds32_output_cbranchsi4_equality_reg_or_const_int,
13234 nds32_output_cbranchsi4_greater_less_zero): Declare.
13235 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
13236 nds32_rimm11s_operand): New predicates.
13237 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
13238 * config/nds32/nds32.md: Rewrite all the branch and conditional move
13241 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
13243 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
13244 * config/nds32/nds32.md: Ditto.
13245 * config/nds32/pipelines.md: Ditto.
13247 2018-04-04 Richard Biener <rguenther@suse.de>
13249 PR tree-optimization/85168
13250 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
13251 propagating abnormals.
13253 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
13255 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
13257 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
13258 Kito Cheng <kito.cheng@gmail.com>
13260 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
13261 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
13262 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
13263 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
13264 * config/nds32/nds32.md (sibcall_internal): New.
13265 (sibcall_register): Remove.
13266 (sibcall_immediate): Remove.
13267 (sibcall_value_internal): New.
13268 (sibcall_value_register): Remove.
13269 (sibcall_value_immediate): Remove.
13270 * config/nds32/predicates.md (nds32_general_register_operand): New.
13271 (nds32_call_address_operand): New.
13273 2018-04-03 Jakub Jelinek <jakub@redhat.com>
13275 PR rtl-optimization/85167
13276 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
13277 bb_defs if *split_p, instead preinitialize it to NULL.
13279 PR tree-optimization/85156
13280 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
13281 evaluating the argument multiple times.
13283 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
13285 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
13287 (_mm_cvtpd_ps): Likewise.
13288 (_mm_cvttpd_epi32): Likewise.
13289 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
13290 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
13291 vector, pixel, and bool following altivec.h include.
13293 2018-04-03 Martin Sebor <msebor@redhat.com>
13295 * doc/extend.texi (Common Function Attributes): Clarify.
13296 (const attribute): Likewise.
13297 (pure attribute): Likewise.
13299 2018-04-03 Jakub Jelinek <jakub@redhat.com>
13302 * config/i386/i386.c (ix86_expand_vector_set): Use
13303 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
13305 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
13307 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
13308 instructions when changing rounding bits to preserve precision bits
13309 in the x87 control word.
13311 2018-04-03 Martin Liska <mliska@suse.cz>
13313 PR tree-optimization/82491
13314 * rtl.h (strip_offset_and_add): Replace += suboffset with
13315 poly_uint64 () + suboffset.
13317 2018-03-29 Martin Liska <mliska@suse.cz>
13318 Martin Jambor <mjambor@suse.cz>
13321 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
13322 param_type is not an integral or pointer type.
13324 2018-04-03 Richard Biener <rguenther@suse.de>
13326 * sese.h (recompute_all_dominators): Remove.
13328 2018-04-02 Martin Sebor <msebor@redhat.com>
13330 * doc/invoke.texi (-Wrestrict): Fix typos.
13332 2018-04-02 Jim Wilson <jimw@sifive.com>
13334 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
13335 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
13336 (<optab>di3, <optab>si3_extend): Likewise.
13337 (<optab>si3_mask, <optab>si3_mask_1): New.
13338 (<optab>di3_mask, <optab>di3_mask_1): New.
13339 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
13340 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
13341 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
13343 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
13345 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
13348 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
13350 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
13351 (nds32_canonicalize_comparison): New function.
13353 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
13354 Kito Cheng <kito.cheng@gmail.com>
13355 Kuan-Lin Chen <kuanlinchentw@gmail.com>
13357 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
13358 * config/nds32/constants.md (unspec_volatile_element): Add
13359 UNSPEC_VOLATILE_RELAX_GROUP.
13360 * config/nds32/nds32-relax-opt.c: New file.
13361 * config/nds32/nds32-predicates.c
13362 (nds32_symbol_load_store_p): New function.
13363 * config/nds32/nds32-protos.h
13364 (nds32_symbol_load_store_p): Declare function.
13365 (make_pass_nds32_relax_opt): Declare new rtl pass function.
13366 * config/nds32/nds32.c
13367 (nds32_register_pass): New function to register pass.
13368 (nds32_register_passes): New function to register passes.
13369 * config/nds32/nds32.md (relax_group): New pattern.
13370 * config/nds32/nds32.opt (mrelax-hint): New option.
13371 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
13373 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
13375 * config/nds32/t-nds32: Modify files dependency.
13377 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
13379 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
13380 (PROFILE_HOOK): Define its implementation.
13382 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
13384 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
13385 type and 32-bit size.
13387 2018-04-01 Jakub Jelinek <jakub@redhat.com>
13389 PR middle-end/85090
13390 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
13391 (V_128_256): New mode iterator.
13392 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
13393 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
13394 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
13396 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
13397 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
13399 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
13402 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
13403 NaN inputs correctly.
13405 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
13408 * config/rs6000/vsx.md (??r): New mode attribute.
13409 (*vsx_mov<mode>_64bit): Use it.
13410 (*vsx_mov<mode>_32bit): Likewise.
13412 2018-03-30 Martin Sebor <msebor@redhat.com>
13414 PR tree-optimization/84818
13415 * builtins.c (check_access): Use warning_n.
13417 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13420 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
13422 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
13425 2018-03-30 Julia Koval <julia.koval@intel.com>
13428 * x86-tune.def (movx, partial_reg_dependency): Enable for
13431 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
13433 PR inline-asm/84985
13434 * lra-constraints.c (process_alt_operands): Move setting
13435 this_alternative_matches below.
13437 2018-03-29 Martin Liska <mliska@suse.cz>
13440 * doc/invoke.texi: Document how LTO works with debug info.
13441 Describe auto-load support of binutils. Mention 'x86-64'
13442 as valid option value of -march option.
13444 2018-03-29 Jakub Jelinek <jakub@redhat.com>
13446 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
13449 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
13450 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
13451 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
13454 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
13457 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
13458 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
13459 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
13460 for RS6000_BTM_POWERPC64.
13461 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
13462 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
13463 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
13466 (DIVDEU): Likewise.
13468 2018-03-28 Carl Love <cel@us.ibm.com>
13471 2017-09-27 Carl Love <cel@us.ibm.com>
13473 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
13474 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
13475 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
13478 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13480 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
13481 instead of __vector bool.
13482 (_mm_max_pu8): Likewise.
13483 (_mm_min_pi16): Likewise.
13485 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
13488 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
13489 (DIVWEUO): Likewise.
13490 (DIVDEO): Likewise.
13491 (DIVDEUO): Likewise.
13492 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
13493 DIVWEUO and DIVDEUO.
13494 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
13495 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
13496 (div_extend): Likewise.
13497 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
13499 (__builtin_divweuo): Likewise.
13500 (__builtin_divdeo): Likewise.
13501 (__builtin_divdeuo): Likewise.
13503 2018-03-28 Jakub Jelinek <jakub@redhat.com>
13506 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
13507 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
13509 PR tree-optimization/82004
13510 * gimple-match-head.c (optimize_pow_to_exp): New function.
13511 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
13512 Don't fold to exp if optimize_pow_to_exp is false.
13514 2018-03-28 Martin Liska <mliska@suse.cz>
13517 * calls.c (initialize_argument_information): Fix trailing space.
13518 * common.opt: Fix typo and provide better explanation for
13519 -fsanitize-coverage option.
13520 * config/i386/i386.opt: Fix typo.
13522 2018-03-28 Jakub Jelinek <jakub@redhat.com>
13523 Martin Liska <mliska@suse.cz>
13526 * gimplify.c (asan_poison_variable): Don't do the check for
13527 gimplify_omp_ctxp here.
13528 (gimplify_decl_expr): Do it here.
13529 (gimplify_target_expr): Likewise.
13531 2018-03-28 Martin Liska <mliska@suse.cz>
13534 * config/i386/i386.c (ix86_function_arg_advance): Do not call
13535 chkp_type_bounds_count if MPX is not enabled.
13537 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
13539 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
13541 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
13544 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
13545 function to create the function decl for complex long double
13546 multiply and divide for -mabi=ieeelongdouble.
13547 (init_float128_ieee): Call it.
13549 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
13552 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
13553 -fcf-protection=branch -mibt.
13554 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
13556 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
13559 * config/arm/arm.c (arm_valid_symbolic_address): Handle
13560 arm_word_relocations.
13562 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
13565 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
13566 extern array declarations.
13568 2018-03-27 Richard Biener <rguenther@suse.de>
13570 PR middle-end/84067
13571 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
13572 explicit single_use checks.
13574 2018-03-27 Richard Biener <rguenther@suse.de>
13576 PR tree-optimization/85082
13577 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
13580 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
13582 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
13583 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
13584 Turn on fasynchronous-unwind-tables and funwind-tables.
13586 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
13589 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
13590 (*bmi_blsr_<mode>_ccz): Ditto.
13592 2018-03-26 Tom de Vries <tom@codesourcery.com>
13594 PR tree-optimization/85063
13595 * omp-general.c (offloading_function_p): New function. Factor out
13597 * omp-offload.c (pass_omp_target_link::gate): ... here.
13598 * omp-general.h (offloading_function_p): Declare.
13599 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
13600 with attribute omp declare target for offloading functions.
13602 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
13604 PR tree-optimization/84005
13605 * tree-data-ref.h (get_base_for_alignment): Declare.
13606 * tree-data-ref.c (get_base_for_alignment_1): New function.
13607 (get_base_for_alignment): Likewise.
13608 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
13609 get_base_for_alignment to find a suitable base object, instead
13610 of always using drb->base_address.
13612 2018-03-23 Jakub Jelinek <jakub@redhat.com>
13614 PR inline-asm/85022
13615 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
13616 known size by default.
13618 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
13620 PR inline-asm/85030
13621 * lra-constraints.c (process_alt_operands): Don't match BLKmode
13622 and non BLKmode operands.
13624 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13627 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
13628 Clean up attributes.
13630 2018-03-23 Richard Biener <rguenther@suse.de>
13633 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
13634 we are going to emit early debug for LTO.
13636 2018-03-23 Jakub Jelinek <jakub@redhat.com>
13638 PR inline-asm/85034
13639 * function.c (match_asm_constraints_1): Don't optimize if input
13640 doesn't satisfy general_operand predicate for output's mode.
13642 PR inline-asm/85022
13643 * alias.c (write_dependence_p): Don't require for x_canonicalized
13644 non-VOIDmode if x has VOIDmode.
13647 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
13648 just don't try to optimize it rather than assert it never happens.
13650 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
13652 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
13653 macro expansions for definition of ST_INTERNAL_<mode> and
13654 LD_INTERNAL_<mode> builtins.
13655 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
13657 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
13659 (altivec_expand_st_builtin): Likewise.
13660 (altivec_expand_builtin): Remove calls to deleted functions.
13661 (rs6000_address_for_altivec): Delete this function.
13662 * config/rs6000/vector.md: Remove expands for
13663 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
13665 2018-03-22 Sudakshina Das <sudi.das@arm.com>
13668 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
13669 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
13670 re-computing once computed.
13671 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
13672 (arm_init_machine_status): Initialize
13673 machine->static_chain_stack_bytes.
13675 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
13678 * doc/extend.texi: Add four new prototypes for vec_ld.
13679 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
13680 definitions for more logical presentation.
13681 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
13682 entries for V1TI variants of __builtin_altivec_ld builtin.
13683 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
13684 handling of V1TI variant of LVX icode pattern.
13685 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
13686 (rs6000_gimple_fold_builtin): Likewise.
13687 (altivec_init_builtins): Add code to define
13688 __builtin_altivec_lvx_v1ti function.
13690 2018-03-22 Jakub Jelinek <jakub@redhat.com>
13692 PR inline-asm/84941
13693 * function.c (match_asm_constraints_1): Don't do the optimization
13694 if input isn't a REG, SUBREG, MEM or constant.
13696 2018-03-22 Tom de Vries <tom@codesourcery.com>
13698 PR tree-optimization/84956
13699 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
13700 bb_has_abnormal_pred.
13702 2018-03-22 Jakub Jelinek <jakub@redhat.com>
13705 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
13706 DECL_INITIAL (decl) to decl at the end.
13707 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
13708 adjust the comment.
13710 2018-03-21 Joseph Myers <joseph@codesourcery.com>
13712 * doc/extend.texi (__builtin_tgmath): Document when complex
13713 integer types are treated as _Complex _Float64.
13715 2018-03-21 Tom de Vries <tom@codesourcery.com>
13717 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
13719 2018-03-21 Jakub Jelinek <jakub@redhat.com>
13721 PR tree-optimization/84960
13722 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
13723 if it is ENTRY block, move them into single succ of ENTRY in that case.
13725 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
13727 PR tree-optimization/84811
13728 * poly-int.h (poly_span_traits): Remove the T3 parameter and
13729 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
13730 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
13731 (known_subrange_p): Update accordingly. Cast each value involved
13732 in the size comparison, rather than casting the result of the
13735 2018-03-21 Jakub Jelinek <jakub@redhat.com>
13737 PR tree-optimization/84982
13738 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
13739 by flipping the least significant bit rather than all bits from
13740 bitpos to bitpos + bitsize - 1.
13742 2018-03-21 Nathan Sidwell <nathan@acm.org>
13744 * doc/extend.texi (Deprecated Features): Remove mention of
13745 long-deleted deprecations.
13747 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13750 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
13751 * configure: Regenerate.
13753 2018-03-21 Tom de Vries <tom@codesourcery.com>
13755 PR tree-optimization/83126
13756 * tree-parloops.c (num_phis): New function.
13757 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
13759 2018-03-21 Nathan Sidwell <nathan@acm.org>
13761 * doc/extend.texi (Deprecated Features): Update deprecated flags,
13762 mention anon-struct/union members and trailing attributes.
13764 2018-03-21 Bin Cheng <bin.cheng@arm.com>
13766 PR tree-optimization/84969
13767 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
13768 builtin memset partitions if they set different rhs values.
13770 2018-03-21 Jakub Jelinek <jakub@redhat.com>
13772 PR rtl-optimization/84989
13773 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
13774 VEC_DUPLICATE with scalar result mode.
13776 2018-03-21 Martin Liska <mliska@suse.cz>
13779 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
13780 not intended return statement.
13782 2018-03-21 Martin Liska <mliska@suse.cz>
13785 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
13786 (chkp_find_bound_slots_1): Limit number of iterations.
13788 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
13791 * Minor grammar fixes for x86 options.
13793 2018-03-20 Jakub Jelinek <jakub@redhat.com>
13796 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
13797 holding REG_CFA_RESTORE notes, instead turn them into a USE.
13799 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
13802 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
13803 (altivec_lvx_<mode>_1op): Likewise.
13804 (altivec_stvx_<mode>_2op): Likewise.
13805 (altivec_stvx_<mode>_1op): Likewise.
13806 (altivec_lvx_<VM2:mode>): New define_expand.
13807 (altivec_stvx_<VM2:mode>): Likewise.
13808 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
13809 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
13810 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
13811 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
13812 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
13813 (rs6000_gen_lvx): Likewise.
13814 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
13815 (altivec_expand_stv_builtin): Likewise.
13816 (altivec_expand_builtin): Likewise.
13817 * config/rs6000/vector.md: Likewise.
13819 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13822 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
13825 2018-03-20 Richard Biener <rguenther@suse.de>
13828 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
13829 sign-conversions as zero, fall back to standard scalar_stmt
13832 2018-03-20 Martin Liska <mliska@suse.cz>
13835 * predict.c (rebuild_frequencies): Handle case when we have
13836 PROFILE_ABSENT, but flag_guess_branch_prob is false.
13838 2018-03-20 Jakub Jelinek <jakub@redhat.com>
13841 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
13842 flag_section_anchors.
13843 * varasm.c (use_blocks_for_decl_p): Remove hack for
13844 dw2_force_const_mem.
13847 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
13849 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
13850 be created, use lowpart_subreg of operands[0] rather than operands[0]
13852 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
13853 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
13854 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
13855 and n constraint instead of aarch64_shift_imm_di and Usd.
13856 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
13857 (*aarch64_<optab>_reg_minus<mode>3): ... this.
13859 2018-03-20 Sudakshina Das <sudi.das@arm.com>
13862 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
13863 to favor GPR over NEON registers.
13864 (<shift>di3_neon): Likewise.
13866 2018-03-20 Tom de Vries <tom@codesourcery.com>
13869 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
13870 (nvptx_process_pars): Emit bar.sync asap and alap.
13872 2018-03-20 Tom de Vries <tom@codesourcery.com>
13875 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
13876 seen_label if seen_label is already set.
13878 2018-03-20 Jakub Jelinek <jakub@redhat.com>
13881 * config/i386/i386.c (fold_builtin_cpu): For features above 31
13882 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
13883 Use 1U instead of 1. Formatting fixes.
13886 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
13887 instead of TREE_TYPE (s1) for the return value.
13889 2018-03-19 Jakub Jelinek <jakub@redhat.com>
13891 PR tree-optimization/84946
13892 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
13893 bitsize + bitsize in poly_uint64 rather than poly_int64.
13896 * dwarf2asm.c: Include fold-const.c.
13897 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
13898 of decl rather than decl itself.
13900 PR rtl-optimization/84643
13901 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
13903 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
13906 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
13907 calling assemble_variable.
13909 2018-03-19 Sudakshina Das <sudi.das@arm.com>
13912 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
13913 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
13915 2018-03-19 Jim Wilson <jimw@sifive.com>
13918 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
13919 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
13920 (riscv_first_stack_step): Likewise.
13921 (riscv_option_override): Use STACK_BOUNDARY instead of
13922 MIN_STACK_BOUNDARY.
13923 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
13924 MIN_STACK_BOUNDARY.
13925 (BIGGEST_ALIGNMENT): Set to 128.
13926 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
13927 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
13930 2018-03-19 Richard Biener <rguenther@suse.de>
13932 PR tree-optimization/84933
13933 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
13934 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
13936 2018-03-19 Richard Biener <rguenther@suse.de>
13938 PR tree-optimization/84859
13939 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
13940 (cond_if_else_store_replacement): Perform sinking operation on
13941 single-store BBs regardless of MAX_STORES_TO_SINK setting.
13942 Generalize what a BB with a single eligible store is.
13944 2018-03-19 Richard Biener <rguenther@suse.de>
13946 PR tree-optimization/84929
13947 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
13948 chrec_is_positive against non-chrec arg.
13950 2018-03-19 Tamar Christina <tamar.christina@arm.com>
13953 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
13955 2018-03-18 Martin Liska <mliska@suse.cz>
13957 PR rtl-optimization/84635
13958 * regrename.c (build_def_use): Use matches_mode only when
13961 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
13963 PR tree-optimization/84913
13964 * tree-vect-loop.c (vectorizable_reduction): Don't try to
13965 vectorize chains of COND_EXPRs.
13967 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
13969 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
13971 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
13973 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
13975 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
13977 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
13979 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
13980 Kito Cheng <kito.cheng@gmail.com>
13982 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
13983 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
13984 (nds32_adjust_reg_alloc_order): New function.
13985 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
13987 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
13989 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
13990 nds32_print_operand, nds32_print_operand_address): Use
13991 HOST_WIDE_INT_PRINT_DEC instead.
13993 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
13995 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
13997 2018-03-17 Jakub Jelinek <jakub@redhat.com>
14000 * config/i386/i386.c (initial_ix86_tune_features,
14001 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
14002 unsigned long long.
14003 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
14004 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
14005 rather than 1u << ix86_tune. Formatting fix.
14006 (ix86_option_override_internal): Change ix86_arch_mask from
14007 unsigned int to unsigned HOST_WIDE_INT, initialize to
14008 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
14009 (ix86_function_specific_restore): Likewise.
14011 2018-03-16 Jakub Jelinek <jakub@redhat.com>
14014 * postreload.c (reload_combine_recognize_pattern): Perform
14015 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
14016 truncate_int_for_mode the result for the destination's mode.
14019 * hsa-gen.c (mem_type_for_type): Fix comment typo.
14020 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
14022 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
14025 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
14028 * lra-assigns.c (lra_split_hard_reg_for): Don't use
14029 regno_allocno_class_array and sorted_pseudos.
14030 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
14031 insns where regno is used.
14033 2018-03-16 Martin Liska <mliska@suse.cz>
14036 * multiple_target.c (create_dispatcher_calls): Redirect
14037 reference in the symbol table.
14039 2018-03-16 Martin Liska <mliska@suse.cz>
14042 * multiple_target.c (create_dispatcher_calls): Redirect also
14045 2018-03-16 Jakub Jelinek <jakub@redhat.com>
14049 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
14050 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
14051 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
14053 2018-03-16 Julia Koval <julia.koval@intel.com>
14055 * doc/invoke.texi (Skylake Server): Add CLWB.
14056 Cannonlake): Remove CLWB.
14058 2018-03-16 Jakub Jelinek <jakub@redhat.com>
14060 PR tree-optimization/84841
14061 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
14063 (FLOAT_ONE_CONST_TYPE): Define.
14064 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
14065 (sort_by_operand_rank): Put entries with higher constant_type last
14066 rather than first to match comments.
14068 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
14070 * config/nios2/nios2.md (movsi_internal): Fix thinko in
14073 2018-03-15 Jakub Jelinek <jakub@redhat.com>
14076 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
14077 check and use address of target always.
14079 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
14082 * config/i386/i386.c (indirect_thunk_needed): Update comments.
14083 (indirect_thunk_bnd_needed): Likewise.
14084 (indirect_thunks_used): Likewise.
14085 (indirect_thunks_bnd_used): Likewise.
14086 (indirect_return_needed): New.
14087 (indirect_return_bnd_needed): Likewise.
14088 (output_indirect_thunk_function): Add a bool argument for
14090 (output_indirect_thunk_function): Don't generate alias for
14091 function return thunk.
14092 (ix86_code_end): Call output_indirect_thunk_function to generate
14093 function return thunks.
14094 (ix86_output_function_return): Set indirect_return_bnd_needed
14095 and indirect_return_needed instead of indirect_thunk_bnd_needed
14096 and indirect_thunk_needed.
14098 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
14100 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
14101 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
14102 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
14104 2018-03-15 David Malcolm <dmalcolm@redhat.com>
14105 Paul Hua <paul.hua.gm@gmail.com>
14108 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
14110 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
14112 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
14113 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
14114 resp. SFmode cases.
14116 2018-03-15 Tamar Christina <tamar.christina@arm.com>
14119 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
14120 instead of GET_MODE_SIZE when comparing Units.
14122 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
14125 * varasm.c (hash_section): Return an unchangeble hash value
14126 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
14127 Return !aarch64_can_use_per_function_literal_pools_p ().
14129 2018-03-15 Jakub Jelinek <jakub@redhat.com>
14132 * optabs.c (emit_conditional_move): Pass address of cmode's copy
14133 rather than address of cmode as last argument to prepare_cmp_insn.
14135 2018-03-15 Julia Koval <julia.koval@intel.com>
14137 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
14138 F_AVX512VNNI, F_AVX512BITALG): New.
14140 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
14143 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
14144 insn for floating-point loads and stores.
14146 2018-03-14 Carl Love <cel@us.ibm.com>
14148 * config/rs6000/rs6000-c.c: Add macro definitions for
14149 ALTIVEC_BUILTIN_VEC_PERMXOR.
14150 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
14151 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
14152 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
14153 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
14155 * config/doc/extend.texi: Add prototypes for vec_permxor.
14157 2018-03-14 David Malcolm <dmalcolm@redhat.com>
14160 * diagnostic-show-locus.c (class layout_point): Convert m_line
14161 from int to linenum_type.
14162 (line_span::comparator): Use linenum "compare" function when
14163 comparing line numbers.
14164 (test_line_span): New function.
14165 (layout_range::contains_point): Convert param "row" from int to
14167 (layout_range::intersects_line_p): Likewise.
14168 (layout::will_show_line_p): Likewise.
14169 (layout::print_source_line): Likewise.
14170 (layout::should_print_annotation_line_p): Likewise.
14171 (layout::print_annotation_line): Likewise.
14172 (layout::print_leading_fixits): Likewise.
14173 (layout::annotation_line_showed_range_p): Likewise.
14174 (struct line_corrections): Likewise for field m_row.
14175 (line_corrections::line_corrections): Likewise for param "row".
14176 (layout::print_trailing_fixits): Likewise.
14177 (layout::get_state_at_point): Likewise.
14178 (layout::get_x_bound_for_row): Likewise.
14179 (layout::print_line): Likewise.
14180 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
14181 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
14182 * input.c (selftest::test_linenum_comparisons): New function.
14183 (selftest::input_c_tests): Call it.
14184 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
14185 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
14186 * selftest.h (ASSERT_GT): New macro.
14187 (ASSERT_GT_AT): New macro.
14188 (ASSERT_LT): New macro.
14189 (ASSERT_LT_AT): New macro.
14191 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
14193 PR rtl-optimization/84780
14194 * combine.c (distribute_links): Don't make a link based on pc_rtx.
14196 2018-03-14 Martin Liska <mliska@suse.cz>
14198 * tree.c (record_node_allocation_statistics): Use
14199 get_stats_node_kind.
14200 (get_stats_node_kind): New function extracted from
14201 record_node_allocation_statistics.
14202 (free_node): Use get_stats_node_kind.
14204 2018-03-14 Richard Biener <rguenther@suse.de>
14206 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
14207 that the value-set of ANTIC_IN doesn't grow.
14210 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
14212 (BB_VISITED_WITH_VISITED_SUCCS): New define.
14213 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
14215 2018-03-14 Julia Koval <julia.koval@intel.com>
14217 * config.gcc (icelake-client, icelake-server): New.
14219 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
14220 (initial_ix86_arch_features): Ditto.
14221 (PTA_SKYLAKE): Add SGX.
14222 (PTA_ICELAKE): Remove.
14223 (PTA_ICELAKE_CLIENT): New.
14224 (PTA_ICELAKE_SERVER): New.
14225 (ix86_option_override_internal): Split up icelake on icelake client and
14227 (get_builtin_code_for_version): Ditto.
14228 (fold_builtin_cpu): Ditto.
14229 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
14230 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
14231 * config/i386/i386.h (processor_type): Ditto.
14232 * doc/invoke.texi: Ditto.
14234 2018-03-14 Jakub Jelinek <jakub@redhat.com>
14237 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
14238 INTEGER_CST offset, add it together with bitpos / 8 and
14239 sign extend based on POINTER_SIZE.
14243 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
14246 * config/i386/constraints.md (Yc): New register constraint.
14247 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
14248 Use Yc constraint for alternative 2 of operand 0. Remove
14249 preferred_for_speed attribute.
14251 2018-03-14 Richard Biener <rguenther@suse.de>
14253 PR tree-optimization/84830
14254 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
14255 with the old one to avoid oscillations.
14257 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
14260 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
14262 (assign_by_spills): Return a flag of reload assignment failure.
14263 Do not process the reload assignment failures. Do not spill other
14264 reload pseudos if they has the same reg class. Update n if
14266 (lra_assign): Add a return arg. Set up from the result of
14267 assign_by_spills call.
14268 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
14269 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
14270 usage_insns if it is not NULL.
14271 (spill_hard_reg_in_range): New function.
14272 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
14273 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
14274 function prototypes.
14275 (lra_assign): Change prototype.
14276 * lra.c (lra): Add code to deal with fails by splitting hard reg
14279 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
14281 * config/riscv/riscv.opt (mrelax): New option.
14282 * config/riscv/riscv.c (riscv_file_start): Emit ".option
14283 "norelax" when riscv_mrelax is disabled.
14284 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
14286 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14289 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
14290 reassociation for int modes.
14292 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
14294 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
14295 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
14297 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
14298 * config/aarch64/aarch64-sve.md
14299 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
14300 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
14301 (*extend<mode><Vwide>2): Rename to...
14302 (aarch64_sve_extend<mode><Vwide>2): ...this.
14303 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
14304 renaming the old pattern to...
14305 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
14307 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
14308 define_expand, renaming the old pattern to...
14309 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
14310 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
14311 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
14312 account when deciding which SVE instruction the optab should use.
14313 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
14315 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
14317 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
14318 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
14319 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
14320 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
14321 (tlsdesc_small_<mode>): Turn a define_expand and use
14322 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
14323 (tlsdesc_small_advsimd_<mode>): ...this.
14324 (tlsdesc_small_sve_<mode>): New pattern.
14326 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
14328 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
14329 (UNSPEC_UMUL_HIGHPART): New constants.
14330 (MUL_HIGHPART): New int iteraor.
14331 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
14332 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
14334 (*<su>mul<mode>3_highpart): New define_insn.
14336 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
14339 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
14342 2018-03-13 Martin Liska <mliska@suse.cz>
14345 * (sem_item_optimizer::sem_item_optimizer): Initialize new
14347 (sem_item_optimizer::~sem_item_optimizer): Release it.
14348 (sem_item_optimizer::merge_classes): Register variable aliases.
14349 (sem_item_optimizer::fixup_pt_set): New function.
14350 (sem_item_optimizer::fixup_points_to_sets): Likewise.
14351 * ipa-icf.h: Declare new variables and functions.
14353 2018-03-13 Jakub Jelinek <jakub@redhat.com>
14355 PR middle-end/84834
14356 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
14357 integer_pow2p@2 and test integer_pow2p in condition.
14358 (A < 0 ? C : 0): Similarly for @1.
14360 PR middle-end/84831
14361 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
14362 characters starting at p contain '\0' character, don't look beyond
14366 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
14367 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
14370 * reg-stack.c (change_stack): Change update_end var from int to
14371 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
14372 also call set_block_for_insn on the newly added insns and rescan.
14375 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
14376 on the last operand.
14379 * tree.c (stabilize_reference_1): Return save_expr (e) for
14380 STATEMENT_LIST even if it doesn't have side-effects.
14382 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
14384 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
14386 2018-03-12 Renlin Li <renlin.li@arm.com>
14388 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
14389 aarch64_output_scalar_simd_mov_immediate.
14391 2018-03-12 Martin Sebor <msebor@redhat.com>
14393 PR tree-optimization/83456
14394 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
14395 for perfectly overlapping calls to memcpy.
14396 (gimple_fold_builtin_memory_chk): Same.
14397 (gimple_fold_builtin_strcpy): Handle no-warning.
14398 (gimple_fold_builtin_stxcpy_chk): Same.
14399 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
14401 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
14403 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
14404 parameter. Use it for SFmode.
14405 (rs6000_function_arg_advance_1): Adjust.
14406 (rs6000_function_arg): Adjust.
14407 (rs6000_gimplify_va_arg): Pass false for that new parameter.
14409 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
14411 PR rtl-optimization/84169
14412 PR rtl-optimization/84780
14413 * combine.c (can_combine_p): Check for a 2-insn combination whether
14414 the destination register is used between the two insns, too.
14416 2018-03-12 Richard Biener <rguenther@suse.de>
14418 PR tree-optimization/84803
14419 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
14420 for refs DR analysis didn't process.
14422 2018-03-12 Richard Biener <rguenther@suse.de>
14424 PR tree-optimization/84777
14425 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
14426 force-vectorize loops ignore whether we are optimizing for size.
14428 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
14430 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
14431 (TARGET_MD_ASM_ADJUST): Define.
14433 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
14434 Kito Cheng <kito.cheng@gmail.com>
14435 Chung-Ju Wu <jasonwucj@gmail.com>
14437 * config/nds32/nds32.c (nds32_compute_stack_frame,
14438 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
14439 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
14440 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
14441 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
14442 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
14443 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
14444 * config/nds32/nds32.md (prologue, epilogue): Use macro
14445 NDS32_V3PUSH_AVAILABLE_P to do checking.
14447 2018-03-11 Jakub Jelinek <jakub@redhat.com>
14450 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
14451 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
14452 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
14453 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
14454 addition of most attributes on !orig_type_die or the attribute not
14455 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
14457 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
14458 Chung-Ju Wu <jasonwucj@gmail.com>
14460 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
14461 __NDS32_VH__ macro.
14462 * config/nds32/nds32.opt (mvh): New option.
14464 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
14465 Chung-Ju Wu <jasonwucj@gmail.com>
14467 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
14469 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
14470 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
14473 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
14474 Chung-Ju Wu <jasonwucj@gmail.com>
14476 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
14478 * config/nds32/nds32-multiple.md (strlensi): New pattern.
14479 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
14481 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
14482 Kito Cheng <kito.cheng@gmail.com>
14483 Chung-Ju Wu <jasonwucj@gmail.com>
14485 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
14486 UNSPEC_FFMISM and UNSPEC_FLMISM.
14487 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
14488 for ffb, ffmism and flmism.
14489 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
14490 (unspec_ffmism): Ditto.
14491 (unspec_flmism): Ditto.
14492 (nds32_expand_builtin_impl): Check if string extension is available.
14493 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
14494 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
14496 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
14499 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
14502 * lra-assigns.c (assign_by_spills): Return a flag of reload
14503 assignment failure. Do not process the reload assignment
14504 failures. Do not spill other reload pseudos if they has the same
14506 (lra_assign): Add a return arg. Set up from the result of
14507 assign_by_spills call.
14508 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
14509 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
14510 usage_insns if it is not NULL.
14511 (spill_hard_reg_in_range): New function.
14512 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
14513 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
14514 function prototypes.
14515 (lra_assign): Change prototype.
14516 * lra.c (lra): Add code to deal with fails by splitting hard reg
14519 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
14522 * config/i386/i386.opt: Replace Enforcment with Enforcement.
14524 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
14527 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
14528 (dw_val_node): Add val_symbolic_view.
14529 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
14530 (symview_upper_bound): New.
14531 (new_line_info_table): Initialize symviews_since_reset.
14532 (dwarf2out_source_line): Count symviews_since_reset and set
14533 symview_upper_bound.
14534 (dw_val_equal_p): Handle symview.
14535 (add_AT_symview): New.
14536 (print_dw_val): Handle symview.
14537 (attr_checksum, attr_checksum_ordered): Likewise.
14538 (same_dw_val_p, size_of_die): Likewise.
14539 (value_format, output_die): Likewise.
14540 (add_high_low_attributes): Use add_AT_symview for entry_view.
14541 (dwarf2out_finish): Reset symview_upper_bound, clear
14544 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
14547 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
14548 Add strict argument and use it.
14549 (rs6000_split_multireg_move): Update for new strict argument.
14550 (mem_operand_gpr): Disallow all non-offsettable addresses.
14551 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
14553 2018-03-09 Jakub Jelinek <jakub@redhat.com>
14556 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
14557 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
14558 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
14561 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
14562 decl, use remap_type if we want to use the type.
14564 2018-03-09 Martin Sebor <msebor@redhat.com>
14566 PR tree-optimization/84526
14567 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
14569 (builtin_access::generic_overlap): Be prepared to handle non-array
14572 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
14574 PR rtl-optimization/84682
14575 * lra-constraints.c (process_address_1): Check is_address flag
14576 for address constraints.
14577 (process_alt_operands): Likewise.
14578 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
14579 preprocess_constraints.
14580 * recog.h (preprocess_constraints): Add oploc parameter.
14582 * recog.c (preprocess_constraints): Test address_operand for
14583 CT_ADDRESS constraints.
14585 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
14588 * lra-assigns.c (assign_by_spills): Return a flag of reload
14589 assignment failure. Do not process the reload assignment
14590 failures. Do not spill other reload pseudos if they has the same
14592 (lra_assign): Add a return arg. Set up from the result of
14593 assign_by_spills call.
14594 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
14595 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
14596 usage_insns if it is not NULL.
14597 (spill_hard_reg_in_range): New function.
14598 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
14599 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
14600 function prototypes.
14601 (lra_assign): Change prototype.
14602 * lra.c (lra): Add code to deal with fails by splitting hard reg
14605 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14608 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
14609 Accept complain bool parameter. Only emit errors if it is true.
14610 (arm_parse_cpu_option_name): Likewise.
14611 (arm_target_thumb_only): Adjust callers of the above.
14612 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
14613 prototype to take a default true bool parameter.
14614 (arm_parse_arch_option_name): Likewise.
14616 2018-03-09 David Malcolm <dmalcolm@redhat.com>
14617 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
14621 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
14622 * configure: Regenerate.
14623 * configure.ac ("linker --version-script option"): New.
14624 ("linker soname option"): New.
14626 2018-03-09 Richard Biener <rguenther@suse.de>
14628 PR tree-optimization/84775
14629 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
14630 immediate uses of predicate stmts and mark them modified.
14633 PR tree-optimization/84178
14634 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
14636 (version_loop_for_if_conversion): Delay update_ssa call.
14637 (tree_if_conversion): Delay update_ssa until after predicate
14640 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
14643 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
14644 when the function accesses prior frames.
14646 2018-03-08 Jakub Jelinek <jakub@redhat.com>
14649 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
14650 gen_llsym, otherwise call maybe_gen_llsym.
14652 PR inline-asm/84742
14653 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
14654 has ',' character inside of it.
14656 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14659 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
14660 as clobbering CC_REGNUM.
14662 2018-03-08 Richard Biener <rguenther@suse.de>
14664 PR middle-end/84552
14665 * tree-scalar-evolution.c: Include tree-into-ssa.h.
14666 (follow_copies_to_constant): Do not follow SSA names registered
14669 2018-03-08 Richard Biener <rguenther@suse.de>
14671 PR tree-optimization/84178
14672 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
14674 (version_loop_for_if_conversion): Delay update_ssa call.
14675 (tree_if_conversion): Delay update_ssa until after predicate
14678 2018-03-08 David Malcolm <dmalcolm@redhat.com>
14680 PR tree-optimization/84178
14681 * tree-if-conv.c (release_bb_predicate): Remove the
14682 the assertion that the stmts have NULL use_ops.
14683 Discard the statements, asserting that they haven't
14684 yet been added to a BB.
14686 2018-03-08 Richard Biener <rguenther@suse.de>
14688 PR tree-optimization/84746
14689 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
14690 (phi_translate): Pass in destination ANTIC_OUT set.
14691 (phi_translate_1): Likewise. For a simplified result lookup
14692 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
14693 (phi_translate_set): Adjust.
14694 (do_pre_regular_insertion): Likewise.
14695 (do_pre_partial_partial_insertion): Likewise.
14697 2018-03-08 Martin Liska <mliska@suse.cz>
14699 PR gcov-profile/84735
14700 * doc/gcov.texi: Document usage of profile files.
14701 * gcov-io.h: Document changes in the format.
14703 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
14707 * dwarf2out.c (struct dw_line_info_table): Update comments for
14709 (FORCE_RESET_NEXT_VIEW): New.
14710 (FORCE_RESETTING_VIEW_P): New.
14711 (RESETTING_VIEW_P): Check for -1 too.
14712 (ZERO_VIEW_P): Likewise.
14713 (new_line_info_table): Force-reset next view.
14714 (dwarf2out_begin_function): Likewise.
14715 (dwarf2out_source_line): Simplify zero_view_p initialization.
14716 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
14717 view directly. Omit view when omitting .loc at line 0.
14719 2018-03-08 Jakub Jelinek <jakub@redhat.com>
14721 PR tree-optimization/84740
14722 * tree-switch-conversion.c (process_switch): Call build_constructors
14723 only if info.phi_count is non-zero.
14725 PR tree-optimization/84739
14726 * tree-tailcall.c (find_tail_calls): Check call arguments against
14727 DECL_ARGUMENTS (current_function_decl) rather than
14728 DECL_ARGUMENTS (func) when checking for tail recursion.
14730 2018-03-07 Jakub Jelinek <jakub@redhat.com>
14732 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
14733 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
14734 Volker Reichelt's entry and add entries for people that perform
14735 GCC fuzzy testing and report numerous bugs.
14737 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
14740 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
14741 readonly data in sdata, if that is disabled.
14742 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
14743 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
14744 -mreadonly-in-sdata option.
14746 2018-03-07 Martin Sebor <msebor@redhat.com>
14748 PR tree-optimization/84468
14749 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
14750 basic block when looking for nul assignment.
14752 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
14755 * except.h (output_function_exception_table): Adjust prototype.
14756 * except.c (output_function_exception_table): Remove FNNAME parameter
14757 and add SECTION parameter. Ouput one part of the table at a time.
14758 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
14759 the first part of the exception table and emit unwind directives.
14760 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
14761 (i386_pe_seh_cold_init): Likewise.
14762 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
14763 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
14764 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
14765 (ix86_output_call_insn): Emit a nop in one more case for SEH.
14766 * config/i386/winnt.c: Include except.h.
14767 (struct seh_frame_state): Add reg_offset, after_prologue and
14768 in_cold_section fields.
14769 (i386_pe_seh_end_prologue): Set seh->after_prologue.
14770 (i386_pe_seh_cold_init): New function.
14771 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
14772 to seh->in_cold_section.
14773 (seh_emit_push): Record the offset of the push.
14774 (seh_emit_save): Record the offet of the save.
14775 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
14776 Test seh->after_prologue to disregard the epilogue.
14777 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
14778 (i386_pe_end_cold_function): New function.
14780 2018-03-07 Jakub Jelinek <jakub@redhat.com>
14783 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
14784 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
14787 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
14789 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
14790 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
14792 PR middle-end/84723
14793 * multiple_target.c: Include tree-inline.h and intl.h.
14794 (expand_target_clones): Diagnose and fail if node->definition and
14795 !tree_versionable_function_p (node->decl).
14797 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
14799 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
14801 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
14802 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
14803 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
14805 2018-03-06 Jakub Jelinek <jakub@redhat.com>
14808 * combine.c (try_combine): Use reg_or_subregno instead of handling
14809 just paradoxical SUBREGs and REGs.
14811 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
14813 * config/arc/arc.c (arc_finalize_pic): Remove function.
14814 (arc_must_save_register): We use single base PIC register, remove
14815 checks to save/restore the PIC register.
14816 (arc_expand_prologue): Likewise.
14817 * config/arc/arc-protos.h (arc_set_default_type_attributes):
14819 (arc_verify_short): Likewise.
14820 (arc_attr_type): Likewise.
14821 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
14822 (walk_stores): Likewise.
14823 (arc_address_cost): Make it static.
14824 (arc_verify_short): Likewise.
14825 (branch_dest): Likewise.
14826 (arc_attr_type): Likewise.
14827 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
14828 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
14829 (arc_final_prescan_insn): Remove inserting the nops due to
14830 hardware hazards. It is done in reorg step.
14831 (insn_length_variant_t): Remove.
14832 (insn_length_parameters_t): Likewise.
14833 (arc_insn_length_parameters): Likewise.
14834 (arc_get_insn_variants): Likewise.
14835 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
14837 2018-03-06 Jakub Jelinek <jakub@redhat.com>
14839 PR inline-asm/84683
14840 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
14843 PR tree-optimization/84687
14844 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
14846 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
14848 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14850 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
14851 Rename to ppc_speculation_barrier.
14852 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
14853 __builtin_ppc_speculation_barrier.
14855 2018-03-05 Jakub Jelinek <jakub@redhat.com>
14858 * combine.c (combine_simplify_rtx): Don't try to simplify if
14859 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
14862 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
14864 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
14865 to 32 bytes when compiling for POWER9.
14867 2018-03-05 Jakub Jelinek <jakub@redhat.com>
14870 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
14871 regparm >= 3 with no arg reg available also for calls with
14872 flag_force_indirect_call. Pass decl to ix86_function_regparm.
14875 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
14877 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
14879 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
14882 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
14884 2018-03-05 Richard Biener <rguenther@suse.de>
14886 PR tree-optimization/84486
14887 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
14888 When inserting a __builtin_assume_aligned call set the LHS
14889 SSA name alignment info accordingly.
14891 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
14893 PR tree-optimization/84114
14894 * config/aarch64/aarch64.c (aarch64_reassociation_width)
14895 Avoid reassociation of FLOAT_MODE addition.
14897 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
14899 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
14900 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
14901 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
14902 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
14903 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
14904 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
14905 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
14907 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
14908 __builtin_ia32_wbinvd): New builtins.
14909 (SPECIAL_ARGS2): New.
14910 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
14911 (SPECIAL_ARGS2): New.
14912 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
14913 (ix86_valid_target_attribute_inner_p): Ditto.
14914 (ix86_init_mmx_sse_builtins): Add special_args2.
14915 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
14916 TARGET_WBNOINVD_P): New.
14917 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
14918 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
14919 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
14920 * config/i386/immintrin.h (_wbinvd): New intrinsic.
14921 * config/i386/pconfigintrin.h: New file.
14922 * config/i386/wbnoinvdintrin.h: Ditto.
14923 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
14925 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
14927 2018-03-05 Richard Biener <rguenther@suse.de>
14929 PR tree-optimization/84670
14930 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
14932 (BB_VISITED_WITH_VISITED_SUCCS): New define.
14933 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
14934 (compute_antic_aux): Only assert the number of values in ANTIC_IN
14935 doesn't grow if all successors (recursively) were visited at least
14938 2018-03-05 Richard Biener <rguenther@suse.de>
14940 PR tree-optimization/84650
14941 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
14942 if executed in the loop pipeline.
14944 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
14946 * doc/configfiles.texi (Configuration Files): Move info about
14947 conditionalizing $target-protos.h to...
14948 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
14949 differs from $target-protos.h.
14951 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
14952 Chung-Ju Wu <jasonwucj@gmail.com>
14954 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
14955 * config/nds32/nds32-multiple.md (setmemsi): Define.
14956 * config/nds32/nds32-memory-manipulation.c
14957 (nds32_gen_dup_4_byte_to_word_value): New.
14958 (emit_setmem_word_loop): New.
14959 (emit_setmem_byte_loop): New.
14960 (nds32_expand_setmem_loop): New.
14961 (nds32_expand_setmem_loop_v3m): New.
14962 (nds32_expand_setmem_unroll): New.
14963 (nds32_expand_setmem): New.
14965 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
14966 Chung-Ju Wu <jasonwucj@gmail.com>
14968 * config/nds32/nds32-memory-manipulation.c
14969 (nds32_emit_load_store): New.
14970 (nds32_emit_post_inc_load_store): New.
14971 (nds32_emit_mem_move): New.
14972 (nds32_emit_mem_move_block): New.
14973 (nds32_expand_movmemsi_loop_unknown_size): New.
14974 (nds32_expand_movmemsi_loop_known_size): New.
14975 (nds32_expand_movmemsi_loop): New.
14976 (nds32_expand_movmemsi_unroll): New.
14977 (nds32_expand_movmemqi): Rename ...
14978 (nds32_expand_movmemsi): ... to this.
14979 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
14980 (movmemsi): ... to this.
14981 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
14982 (nds32_expand_movmemsi): ... to this.
14984 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
14985 Monk Chiang <sh.chiang04@gmail.com>
14986 Chung-Ju Wu <jasonwucj@gmail.com>
14988 * config/nds32/nds32-protos.h
14989 (nds32_expand_load_multiple): New arguments.
14990 (nds32_expand_store_multiple): Ditto.
14991 (nds32_valid_multiple_load_store): Rename ...
14992 (nds32_valid_multiple_load_store_p): ... to this.
14993 * config/nds32/nds32-memory-manipulation.c
14994 (nds32_expand_load_multiple): Refine implementation.
14995 (nds32_expand_store_multiple): Ditto.
14996 * config/nds32/nds32-multiple.md
14997 (load_multiple): Update nds32_expand_load_multiple interface.
14998 (store_multiple): Update nds32_expand_store_multiple interface.
14999 * config/nds32/nds32-predicates.c
15000 (nds32_valid_multiple_load_store): Rename ...
15001 (nds32_valid_multiple_load_store_p): ... to this and refine
15003 * config/nds32/predicates.md
15004 (nds32_load_multiple_and_update_address_operation): New predicate.
15005 (nds32_store_multiple_and_update_address_operation): New predicate.
15007 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
15008 Chung-Ju Wu <jasonwucj@gmail.com>
15010 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
15011 (combo): New attribute.
15012 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
15014 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
15016 * config/nds32/nds32.opt: Change -mcmodel= default value.
15018 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
15019 Monk Chiang <sh.chiang04@gmail.com>
15020 Chung-Ju Wu <jasonwucj@gmail.com>
15022 * config/nds32/constants.md (unspec_element): New enum.
15023 * config/nds32/constraints.md (Umw): New constraint.
15024 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
15025 * config/nds32/nds32-intrinsic.md: Likewise.
15026 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
15027 (nds32_valid_smw_lwm_base_p): New.
15028 (nds32_output_smw_single_word): New.
15029 (nds32_output_lmw_single_word): New.
15030 (nds32_expand_unaligned_load): New.
15031 (nds32_expand_unaligned_store): New.
15032 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
15033 (nds32_output_smw_single_word): Declare.
15034 (nds32_output_lmw_single_word): Declare.
15035 (nds32_expand_unaligned_load): Declare.
15036 (nds32_expand_unaligned_store): Declare.
15037 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
15038 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
15039 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
15040 NDS32_BUILTIN_UASTORE_DW.
15041 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
15044 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
15045 Kito Cheng <kito.cheng@gmail.com>
15046 Chung-Ju Wu <jasonwucj@gmail.com>
15048 * config/nds32/nds32-intrinsic.c
15049 (nds32_expand_builtin_null_ftype_reg): Delete.
15050 (nds32_expand_builtin_reg_ftype_imm): Ditto.
15051 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
15052 (nds32_read_argument): New.
15053 (nds32_legitimize_target): Ditto.
15054 (nds32_legitimize_argument): Ditto.
15055 (nds32_check_constant_argument): Ditto.
15056 (nds32_expand_unop_builtin): Ditto.
15057 (nds32_expand_unopimm_builtin): Ditto.
15058 (nds32_expand_binop_builtin): Ditto.
15059 (nds32_builtin_decl_impl): Ditto.
15060 (builtin_description): Ditto.
15061 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
15062 (nds32_init_builtins_impl): Ditto.
15063 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
15064 (nds32_builtin_decl): New.
15065 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
15066 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
15068 2018-03-02 Jeff Law <law@redhat.com>
15070 * reorg.c (stop_search_p): Handle DEBUG_INSN.
15071 (redundant_insn, fill_simple_delay_slots): Likewise.
15072 (fill_slots_from_thread): Likewise.
15073 * resource.c (mark_referenced_resources): Likewise.
15074 (mark_set_resources, find_dead_or_set_registers): Likewise.
15076 2018-03-02 Jakub Jelinek <jakub@redhat.com>
15078 * substring-locations.h (format_warning_va): Formatting fix for
15079 ATTRIBUTE_GCC_DIAG.
15080 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
15082 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
15083 * substring-locations.c: Include intl.h.
15084 (format_warning_va): Turned into small wrapper around
15085 format_warning_n_va, renamed to ...
15086 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
15087 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
15089 (format_warning_at_substring_n): New function.
15090 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
15091 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
15092 format_warning_at_substring with just a shorter name instead of
15093 const function pointer.
15094 (fmtwarn_n): New function.
15095 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
15096 appropriate, get rid of all the fmtstr temporaries, move conditionals
15097 with G_() wrapped string literals directly into fmtwarn arguments,
15098 cast dir.len to (int), formatting fixes.
15100 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
15102 * doc/invoke.texi: Remove "Cilk Plus" references.
15104 2018-03-02 Jakub Jelinek <jakub@redhat.com>
15105 Richard Biener <rguenther@suse.de>
15108 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
15109 for error or warning attributes if CALL_FROM_THUNK_P is set.
15112 2018-03-02 Jakub Jelinek <jakub@redhat.com>
15115 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
15116 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
15119 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
15120 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
15122 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
15123 instead of -1U in last predictors element's probability member.
15125 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
15128 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
15129 arguments if they are comparable.
15131 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
15133 PR tree-optimization/84634
15134 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
15135 masks and masked_loop_p with a single loop_masks, making sure it's
15136 null for bb vectorization.
15138 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
15140 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
15141 (vect_analyze_data_ref_access): Use loop->safe_len rather than
15142 loop->force_vectorize to check whether there is no alias.
15144 2018-03-02 Jakub Jelinek <jakub@redhat.com>
15147 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
15149 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
15151 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
15152 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
15153 instead of a loop around prev_real_insn.
15154 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
15157 PR inline-asm/84625
15158 * config/i386/i386.c (ix86_print_operand): Use conditional
15159 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
15162 2018-03-02 Richard Biener <rguenther@suse.de>
15164 PR tree-optimization/84427
15165 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
15166 (bitmap_set_subtract_values): Rewrite to handle multiple
15169 (prune_clobbered_mems): Likewise.
15170 (phi_translate): Take edge instead of pred/phiblock.
15171 (phi_translate_1): Likewise.
15172 (phi_translate_set): Likewise. Insert all translated
15173 exprs for a value into the set, keeping possibly multiple
15174 expressions per value.
15175 (compute_antic_aux): Adjust for phi_translate changes.
15176 When intersecting union the expressions and prune those
15177 not in the final value set, keeping possibly multiple
15178 expressions per value. Do not use value-insertion
15179 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
15180 all expressions. Add verification that the value-sets
15181 only shrink during iteration.
15182 (compute_partial_antic_aux): Adjust for the phi_translate changes.
15183 (do_pre_regular_insertion): Likewise.
15184 (do_pre_partial_partial_insertion): Likewise.
15186 2018-03-02 Richard Biener <rguenther@suse.de>
15189 * config/darwin.c (saved_debug_info_level): New static global.
15190 (darwin_asm_lto_start): Disable debug info generation for LTO out.
15191 (darwin_asm_lto_end): Restore debug info generation settings.
15193 2018-03-01 Martin Liska <mliska@suse.cz>
15196 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
15197 volatile arguments.
15199 2018-03-01 Richard Biener <rguenther@suse.de>
15202 * dwarf2out.c (gen_variable_die): Properly handle late VLA
15203 type annotation with LTO when debug was disabled at compile-time.
15205 2018-03-01 Matthew Fortune <mfortune@gmail.com>
15207 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
15209 (mips_final_postscan_insn): Likewise.
15211 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
15213 PR rtl-optimization/84528
15214 * alias.c (init_alias_target): Add commentary.
15215 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
15216 a unique base value if the frame pointer is not eliminated
15217 to the stack pointer.
15219 2018-03-01 Tom de Vries <tom@codesourcery.com>
15221 PR rtl-optimization/83327
15222 * lra-int.h (hard_regs_spilled_into): Declare.
15223 * lra.c (hard_regs_spilled_into): Define.
15224 (init_reg_info): Init hard_regs_spilled_into.
15225 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
15226 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
15227 (process_bb_lives): Handle hard_regs_spilled_into.
15228 (lra_create_live_ranges_1): Before doing liveness propagation, clear
15229 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
15231 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
15233 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
15234 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
15235 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
15236 * config/rs6000/aix72.h: New file.
15238 2018-02-28 Jakub Jelinek <jakub@redhat.com>
15240 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
15241 instead of warning_at with conditional singular and plural messages
15245 * stor-layout.c (update_alignment_for_field): For
15246 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
15247 && !DECL_PACKED (field), do the alignment update, just use
15248 only desired_align instead of MAX (type_align, desired_align)
15250 (place_field): Don't do known_align < desired_align handling
15251 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
15252 is non-NULL, instead do it after rli->prev_field handling and
15253 only if not within a bitfield word. For DECL_PACKED (field)
15254 use type_align of BITS_PER_UNIT.
15256 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
15258 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
15259 superfluous parentheses and trailing spaces.
15261 2018-02-28 Richard Biener <rguenther@suse.de>
15263 PR tree-optimization/84584
15264 * graphite-scop-detection.c (scop_detection::add_scop): Discard
15265 SCoPs with fake exit edge.
15267 2018-02-28 Martin Liska <mliska@suse.cz>
15270 * timevar.c (timer::print): Fix format to properly print 100%
15273 2018-02-28 Richard Biener <rguenther@suse.de>
15275 PR middle-end/84607
15276 * genmatch.c (capture_info::walk_match): Do not mark
15277 captured expressions without operands as expr_p given
15278 they act more like predicates and should be subject to
15279 "lost tail" side-effect preserving.
15281 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
15283 PR rtl-optimization/81611
15284 * auto-inc-dec.c (attempt_change): Move dead note from
15285 mem_insn if it's the next use of regno
15286 (find_address): Take address use of reg holding
15287 non-incremented value. Add parm to limit search to the named
15289 (merge_in_block): Attempt to use a mem insn that is the next
15290 use of the original regno.
15292 2018-02-27 Martin Sebor <msebor@redhat.com>
15295 * doc/invoke.texi (-Wmissing-attributes): New option.
15296 * print-tree.c (print_node): Handle DECL_UNINLINABLE.
15298 2018-02-27 Martin Sebor <msebor@redhat.com>
15300 PR translation/84207
15301 * diagnostic-core.h (warning_n, error_n, inform_n): Change
15302 n argument to unsigned HOST_WIDE_INT.
15303 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
15304 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
15305 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
15306 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
15308 2018-02-27 Richard Biener <rguenther@suse.de>
15310 PR tree-optimization/84512
15311 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
15312 Do not use the estimate returned from record_stmt_cost for
15313 the scalar iteration cost but sum properly using add_stmt_cost.
15315 2018-02-27 Richard Biener <rguenther@suse.de>
15317 PR tree-optimization/84466
15318 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
15319 Adjust last change to less strictly validate use operands.
15321 2018-02-27 Martin Liska <mliska@suse.cz>
15323 PR gcov-profile/84548
15324 * gcov.c (process_file): Allow partial overlap and consider it
15325 also as group functions.
15326 (output_lines): Properly calculate range of lines for a group.
15328 2018-02-27 Martin Liska <mliska@suse.cz>
15330 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
15331 'ggc' suffixes. Change first column width.
15332 (timer::print): Fix formatting of the column.
15334 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
15336 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
15337 preserve inline entry blocks for the sake of debug inline
15338 entry point markers alone.
15339 (remove_unused_locals): Suggest in comments a better place to
15340 force the preservation of inline entry blocks that are
15341 otherwise unused, but do not preserve them.
15343 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
15345 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
15347 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
15350 * config/i386/constraints.md (Bs): Replace
15351 ix86_indirect_branch_register with
15352 TARGET_INDIRECT_BRANCH_REGISTER.
15354 * config/i386/i386.md (indirect_jump): Likewise.
15355 (tablejump): Likewise.
15356 (*sibcall_memory): Likewise.
15357 (*sibcall_value_memory): Likewise.
15358 Peepholes of indirect call and jump via memory: Likewise.
15359 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
15360 (*sibcall_value_GOT_32): Likewise.
15361 * config/i386/predicates.md (indirect_branch_operand): Likewise.
15362 (GOT_memory_operand): Likewise.
15363 (call_insn_operand): Likewise.
15364 (sibcall_insn_operand): Likewise.
15365 (GOT32_symbol_operand): Likewise.
15366 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
15368 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
15370 PR rtl-optimization/83496
15371 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
15372 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
15373 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
15374 redundant insn, if any.
15375 (relax_delay_slots): Likewise.
15376 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
15378 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
15380 PR tree-optimization/83965
15381 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
15382 that grouped statements are part of a reduction chain. Return
15383 true if the statement is not marked as a reduction itself but
15384 is part of a group.
15385 (vect_recog_dot_prod_pattern): Don't check whether the statement
15386 is part of a group here.
15387 (vect_recog_sad_pattern): Likewise.
15388 (vect_recog_widen_sum_pattern): Likewise.
15390 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
15393 * final.c (rest_of_clean_state): Also look for calls inside sequences.
15395 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
15398 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
15400 (ix86_output_indirect_function_return): New prototype.
15401 (ix86_split_simple_return_pop_internal): Likewise.
15402 * config/i386/i386.c (indirect_return_via_cx): New.
15403 (indirect_return_via_cx_bnd): Likewise.
15404 (indirect_thunk_name): Handle return va CX_REG.
15405 (output_indirect_thunk_function): Create alias for
15406 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
15407 (ix86_output_indirect_jmp): Remove the bool argument.
15408 (ix86_output_indirect_function_return): New function.
15409 (ix86_split_simple_return_pop_internal): Likewise.
15410 * config/i386/i386.md (*indirect_jump): Don't pass false
15411 to ix86_output_indirect_jmp.
15412 (*tablejump_1): Likewise.
15413 (simple_return_pop_internal): Change it to define_insn_and_split.
15414 Call ix86_split_simple_return_pop_internal to split it for
15415 -mfunction-return=.
15416 (simple_return_indirect_internal): Call
15417 ix86_output_indirect_function_return instead of
15418 ix86_output_indirect_jmp.
15420 2018-02-26 Jakub Jelinek <jakub@redhat.com>
15423 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
15424 memset and value initialization afterwards.
15426 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
15428 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
15430 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
15433 * common/config/aarch64/aarch64-common.c
15434 (aarch_option_optimization_table[]): Switch
15435 off fomit-frame-pointer
15437 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
15438 Chung-Ju Wu <jasonwucj@gmail.com>
15440 * config/nds32/nds32-multiple.md (load_multiple): Disallow
15442 (store_multiple): Ditto.
15444 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
15446 * config.gcc: Add --with-cpu support for nds32 target.
15447 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
15448 * config/nds32/nds32.opt: Add -mcpu= option.
15450 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
15452 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
15453 isel=yes): Warn for these deprecated options.
15455 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
15457 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
15458 ISA_2_5_MASKS_EMBEDDED.
15460 2018-02-23 Jakub Jelinek <jakub@redhat.com>
15462 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
15463 p->max as pointers rather than using iterative_hash_expr.
15465 2018-02-23 Carl Love <cel@us.ibm.com>
15467 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
15468 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
15470 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
15471 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
15472 P8V_BUILTIN_VEC_VUNSIGNED2.
15474 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
15477 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
15478 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
15480 * lra-constraints.c (curr_insn_transform): Set up
15481 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
15483 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
15484 * lra-eliminations.c (spill_pseudos): Ditto.
15485 (process_insn_for_elimination): Ditto.
15486 * lra-lives.c (reg_early_clobber_p): Use the new macros.
15487 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
15488 LRA_NON_CLOBBERED_ALT.
15490 2018-02-22 Martin Sebor <msebor@redhat.com>
15492 PR tree-optimization/84480
15493 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
15494 to maybe_diag_stxncpy_trunc. Call it.
15495 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
15496 from gimple_fold_builtin_strcpy. Print inlining stack.
15497 (handle_builtin_stxncpy): Print inlining stack.
15498 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
15500 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
15503 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
15504 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
15505 and -fcheck-pointer-bounds are used together.
15506 (indirect_thunk_prefix): New enum.
15507 (indirect_thunk_need_prefix): New function.
15508 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
15509 "_nt" instead of "_bnd" for NOTRACK prefix.
15510 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
15511 (output_indirect_thunk_function): Likewise.
15513 (ix86_code_end): Update output_indirect_thunk_function calls.
15514 (ix86_output_indirect_branch_via_reg): Replace
15515 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
15516 (ix86_output_indirect_branch_via_push): Likewise.
15517 (ix86_output_function_return): Likewise.
15518 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
15519 incompatible with -fcf-protection=branch and
15520 -fcheck-pointer-bounds.
15522 2018-02-22 Steve Ellcey <sellcey@cavium.com>
15525 * config/aarch64/aarch64.c (aarch64_print_address_internal):
15526 Change gcc_assert call to output_operand_lossage.
15528 2018-02-22 Steve Ellcey <sellcey@cavium.com>
15530 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
15532 2018-02-22 DJ Delorie <dj@redhat.com>
15533 Sebastian Perta <sebastian.perta@renesas.com>
15534 Oleg Endo <olegendo@gcc.gnu.org>
15536 * config/rx/rx.c (rx_rtx_costs): New function.
15537 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
15539 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
15541 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
15543 2018-02-22 Martin Liska <mliska@suse.cz>
15546 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
15547 Add "native" as a possible value.
15549 2018-02-22 Martin Liska <mliska@suse.cz>
15552 * config/i386/i386.c (ix86_option_override_internal):
15553 Add "native" as a possible value for -march and -mtune.
15555 2018-02-22 Jakub Jelinek <jakub@redhat.com>
15558 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
15559 to all type variants.
15561 PR tree-optimization/84503
15562 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
15563 width as info->bitpos + info->bitsize - start.
15564 (merged_store_group::merge_overlapping): Simplify width computation.
15565 (check_no_overlap): New function.
15566 (imm_store_chain_info::try_coalesce_bswap): Compute expected
15567 start + width and last_order of the group, fail if check_no_overlap
15569 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
15570 to group if check_no_overlap fails.
15572 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
15574 * config/rs6000/altivec.md: Delete contraint arguments to
15575 define_expand, define_split, and define_peephole2, and in
15576 define_insn_and_split if always unused.
15577 * config/rs6000/darwin.md: Ditto.
15578 * config/rs6000/dfp.md: Ditto.
15579 * config/rs6000/rs6000.md: Ditto.
15580 * config/rs6000/sync.md: Ditto.
15581 * config/rs6000/vector.md: Ditto.
15582 * config/rs6000/vsx.md: Ditto.
15584 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
15586 * config/rs6000/altivec.md: Write output control strings as braced
15587 blocks instead of double-quoted strings.
15588 * config/rs6000/darwin.md: Ditto.
15589 * config/rs6000/rs6000.md: Ditto.
15590 * config/rs6000/vector.md: Ditto.
15591 * config/rs6000/vsx.md: Ditto.
15593 2018-02-21 Jason Merrill <jason@redhat.com>
15595 PR c++/84314 - ICE with templates and fastcall attribute.
15596 * attribs.c (build_type_attribute_qual_variant): Remove assert.
15598 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
15600 * ipa-cp.c (determine_versionability): Fix comment typos.
15602 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
15605 * ipa-cp.c (determine_versionability): Do not version functions caling
15608 2018-02-21 Martin Liska <mliska@suse.cz>
15611 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
15612 Add "native" as a possible value.
15613 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
15614 the macro when native cpu detection is available.
15616 2018-02-21 Martin Liska <mliska@suse.cz>
15619 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
15620 Add "native" as a possible value.
15621 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
15622 when native cpu detection is available.
15624 2018-02-21 Jakub Jelinek <jakub@redhat.com>
15625 Martin Sebor <msebor@redhat.com>
15627 PR tree-optimization/84478
15628 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
15630 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
15631 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
15632 support which is conservatively correct, for 2 only stay conservative
15633 for maxlen. Formatting and comment capitalization fixes. Add STRICT
15634 argument to the 2 argument get_range_strlen, adjust 6 arg
15635 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
15637 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
15638 (gimple_fold_builtin_strlen): Pass true as last argument to
15641 2018-02-20 Martin Sebor <msebor@redhat.com>
15643 PR middle-end/84095
15644 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
15645 (builtin_memref::set_base_and_offset): Same. Handle inner references.
15646 (builtin_memref::builtin_memref): Factor out parts into
15647 set_base_and_offset and call it.
15649 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
15651 PR middle-end/84406
15652 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
15653 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
15654 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
15655 search at the associated MODE_INT.
15657 2018-02-20 Jeff Law <law@redhat.com>
15659 PR middle-end/82123
15660 PR tree-optimization/81592
15661 PR middle-end/79257
15662 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
15663 for range data rather than using global data.
15664 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
15665 range data rather than using global data.
15666 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
15667 pass it to children as needed.
15668 (struct directive::fmtresult): Similarly.
15669 (struct directive::set_width): Similarly.
15670 (struct directive::set_precision): Similarly.
15671 (format_integer, format_directive, parse_directive): Similarly.
15672 (format_none): Accept unnamed vr_values parameter.
15673 (format_percent, format_floating, format_character): Similarly.
15674 (format_string, format_plain): Similarly.
15675 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
15676 the EVRP range analyzer for range data rather than using global data.
15677 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
15678 gimple-ssa-evrp-analyze.h
15679 (class sprintf_dom_walker): Add after_dom_children member function.
15680 Add evrp_range_analyzer member.
15681 (sprintf_dom_walker::before_dom_children): Call into the EVRP
15682 range analyzer as needed.
15683 (sprintf_dom_walker::after_dom_children): New member function.
15684 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
15686 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
15687 (evrp_range_analyzer::pop_to_marker): Likewise.
15689 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
15691 PR tree-optimization/84419
15692 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
15693 with the required type if its current type is compatible but
15696 2018-02-20 Jakub Jelinek <jakub@redhat.com>
15698 PR middle-end/82004
15699 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
15700 after vectorization.
15702 2018-02-20 Martin Liska <mliska@suse.cz>
15705 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
15706 possible values if we don't have a hint.
15708 2018-02-20 Martin Liska <mliska@suse.cz>
15712 * final.c (shorten_branches): Build align_tab array with one
15714 * opts.c (finish_options): Add alignment option limit check.
15715 (MAX_CODE_ALIGN): Likewise.
15716 (MAX_CODE_ALIGN_VALUE): Likewise.
15717 * doc/invoke.texi: Document maximum allowed option value for
15718 all -falign-* options.
15720 2018-02-19 Jakub Jelinek <jakub@redhat.com>
15723 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
15724 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
15725 * var-tracking.c (emit_note_insn_var_location): Remove all references
15726 to NOTE_INSN_CALL_ARG_LOCATION.
15727 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
15728 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
15729 Use copy_rtx_if_shared.
15730 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
15731 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
15732 (dwarf2out_var_location): Remove handling of
15733 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
15735 * final.c (final_scan_insn): Remove all references to
15736 NOTE_INSN_CALL_ARG_LOCATION.
15737 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
15738 before dumping final insns.
15739 * except.c (emit_note_eh_region_end): Remove all references to
15740 NOTE_INSN_CALL_ARG_LOCATION.
15741 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
15742 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
15743 * config/arc/arc.c (hwloop_optimize): Likewise.
15744 * config/arm/arm.c (create_fix_barrier): Likewise.
15745 * config/s390/s390.c (s390_chunkify_start): Likewise.
15746 * config/sh/sh.c (find_barrier): Likewise.
15747 * config/i386/i386.c (rest_of_insert_endbranch,
15748 ix86_seh_fixup_eh_fallthru): Likewise.
15749 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
15750 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
15751 * config/frv/frv.c (frv_function_prologue): Likewise.
15752 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
15754 (note_outside_basic_block_p): Remove all references to
15755 NOTE_INSN_CALL_ARG_LOCATION.
15756 * gengtype.c (adjust_field_rtx_def): Likewise.
15757 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
15759 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
15760 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
15763 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
15766 PR tree-optimization/84452
15767 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
15768 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
15771 2018-02-19 Martin Liska <mliska@suse.cz>
15774 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
15776 2018-02-19 Martin Liska <mliska@suse.cz>
15777 Richard Sandiford <richard.sandiford@linaro.org>
15779 PR tree-optimization/82491
15780 * gimple-fold.c (get_base_constructor): Make earlier bail out
15783 2018-02-19 Carl Love <cel@us.ibm.com>
15785 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
15786 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
15788 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
15789 P8V_BUILTIN_VEC_NEG.
15791 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
15793 * config/rl78/rl78.md (movdf): New define expand.
15795 2018-02-19 Martin Liska <mliska@suse.cz>
15798 * doc/invoke.texi: Fix typo.
15799 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
15801 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
15803 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
15804 handle rs6000_single_float and rs6000_double_float specially for
15807 2018-02-16 Jeff Law <law@redhat.com>
15809 * config/rx/rx.c (add_pop_cfi_notes): New function.;
15810 (pop_regs): Use it.
15812 2018-02-16 Jakub Jelinek <jakub@redhat.com>
15815 * ipa-inline.c (inline_small_functions): Fix a typo.
15817 2018-02-16 Nathan Sidwell <nathan@acm.org>
15819 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
15821 2018-02-16 Carl Love <cel@us.ibm.com>
15823 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
15824 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
15825 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
15826 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
15827 expansion to P8V_BUILTIN_VEC_FLOAT2.
15829 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
15831 PR rtl-optimization/70023
15832 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
15833 src_regno into account.
15835 2018-02-16 Carl Love <cel@us.ibm.com>
15837 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
15838 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
15839 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
15840 * config/rs6000/rs6000.c: Remove case statements for
15841 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
15842 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
15843 and P9V_BUILTIN_VEC_VINSERT4B.
15844 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
15845 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
15846 * config/rs6000/vsx.md:
15847 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
15850 2018-02-16 Carl Love <cel@us.ibm.com>
15852 * config/rs6000/altivec.h: Add builtin names vec_extract4b
15854 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
15856 * config/rs6000/rs6000-c.c: Add the definitions for
15857 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
15858 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
15859 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
15860 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
15861 definition for insert4b and define insn *insert3b_internal.
15862 * doc/extend.texi: Add documentation for vec_extract4b.
15864 2018-02-16 Nathan Sidwell <nathan@acm.org>
15866 * doc/extend.texi (Backwards Compatibility): Mention friend
15867 injection. Note for-scope is deprecated.
15868 * doc/invoke.texi (-ffriend-injection): Deprecate.
15870 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
15872 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
15873 that moved to I2, also allow destinations that are a paradoxical
15874 subreg (instead of a normal reg).
15876 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
15879 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
15882 2018-02-16 Richard Biener <rguenther@suse.de>
15884 PR tree-optimization/84037
15885 PR tree-optimization/84016
15887 * config/i386/i386.c (ix86_builtin_vectorization_cost):
15888 Adjust vec_construct for the fact we need additional higher latency
15889 128bit inserts for AVX256 and AVX512 vector builds.
15890 (ix86_add_stmt_cost): Scale vector construction cost for
15893 2018-02-16 Richard Biener <rguenther@suse.de>
15895 PR tree-optimization/84417
15896 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
15897 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
15898 (non_rewritable_lvalue_p): Likewise, use poly-ints.
15900 2018-02-16 Martin Liska <mliska@suse.cz>
15903 * internal-fn.def (ASAN_CHECK): Set proper flags.
15904 (ASAN_MARK): Likewise.
15906 2018-02-16 Julia Koval <julia.koval@intel.com>
15908 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
15909 from PTA_CANNONLAKE.
15911 2018-02-16 Jakub Jelinek <jakub@redhat.com>
15914 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
15915 Use ++iter rather than iter++ for std::list iterators.
15916 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
15917 defer deleting them until all nodes in the forest are processed. Do
15918 free even leaf nodes. Change to_process into auto_vec.
15921 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
15922 * vec.h (vec_default_construct): Use memset instead of placement new
15923 if BROKEN_VALUE_INITIALIZATION is defined.
15924 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
15925 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
15928 PR rtl-optimization/83723
15929 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
15930 * lra.c (lra_substitute_pseudo): Likewise. If true, use
15931 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
15933 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
15935 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
15937 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
15939 PR rtl-optimization/81443
15940 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
15941 from inner REGs to paradoxical SUBREGs.
15943 2018-02-16 Richard Biener <rguenther@suse.de>
15945 PR tree-optimization/84399
15946 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
15947 For operands we can analyze at their definition make sure we can
15948 analyze them at each use as well.
15950 2018-02-16 Richard Biener <rguenther@suse.de>
15952 PR tree-optimization/84190
15953 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
15954 volatile accesses if the decl isn't volatile.
15956 2018-02-15 Jason Merrill <jason@redhat.com>
15958 PR c++/84314 - ICE with templates and fastcall attribute.
15959 * attribs.c (build_type_attribute_qual_variant): Don't clobber
15960 TYPE_CANONICAL on an existing type.
15962 2018-02-15 Jakub Jelinek <jakub@redhat.com>
15964 PR tree-optimization/84383
15965 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
15966 dstoff nor call operand_equal_p if dstbase is NULL.
15968 PR tree-optimization/84334
15969 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
15970 also a CONSTANT_CLASS_P, punt.
15972 2018-02-14 Jim Wilson <jimw@sifive.com>
15974 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
15975 first SMALL_OPERAND check. New local min_second_step. Move assert
15976 to where locals are set. Add TARGET_RVC support.
15977 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
15979 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
15981 * doc/invoke.texi: Correct -Wformat-overflow code sample.
15983 2018-02-14 Martin Sebor <msebor@redhat.com>
15985 PR tree-optimization/83698
15986 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
15987 arrays constrain the offset range to their bounds.
15988 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
15989 (builtin_access::overlap): Avoid setting the size of overlap if it's
15991 (maybe_diag_overlap): Also consider arrays when deciding what values
15992 of offsets to include in diagnostics.
15994 2018-02-14 Martin Sebor <msebor@redhat.com>
15997 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
15998 that correspond to the kind of a declaration.
16000 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
16003 * config/pa/pa.md: Load address of PIC label using the linkage table
16004 if the label is nonlocal.
16006 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
16008 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
16009 warning message if user requests -maltivec=be.
16010 * doc/invoke.texi: Document deprecation of -maltivec=be.
16012 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
16015 * config/rs6000/rs6000-c.c: Update definitions for
16016 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
16017 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
16019 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
16022 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
16023 add _get_ssp intrinsics. Remove argument from
16024 __builtin_ia32_rdssp[d|q].
16025 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
16026 * config/i386/i386-builtin.def: Remove argument from
16027 __builtin_ia32_rdssp[d|q].
16028 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
16029 ix86_expand_special_args_builtin for _rdssp[d|q].
16030 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
16031 Clear register before usage.
16032 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
16033 Add documentation for new _get_ssp and _inc_ssp intrinsics.
16035 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
16037 PR tree-optimization/84357
16038 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
16039 operand 1 of an ARRAY_REF too.
16041 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
16044 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
16045 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
16047 (set_of_reg): New struct.
16048 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
16049 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
16050 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
16052 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
16053 Split into bitclr, bitset, bitinvert patterns if appropriate.
16054 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
16055 use rx_fuse_in_memory_bitop.
16056 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
16057 to named insn, correct maximum insn length.
16059 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
16062 * machmode.def: Define a complex mode for PARTIAL_INT.
16063 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
16065 * doc/rtl.texi: Document CSPImode.
16066 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
16068 (msp430_hard_regno_nregs_with_padding): Likewise.
16070 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
16073 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
16075 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
16077 PR rtl-optimization/84169
16078 * combine.c (try_combine): New variable split_i2i3. Set it to true if
16079 we generated a parallel as new i3 and we split that to new i2 and i3
16080 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
16081 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
16082 those to i2, not i1. Partially rewrite this scan code.
16084 2018-02-13 Jakub Jelinek <jakub@redhat.com>
16087 * stor-layout.c (place_field): For variable length fields, adjust
16088 offset_align afterwards not just based on the field's alignment,
16089 but also on the size.
16091 PR middle-end/84309
16092 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
16093 of exps and logs in the use_exp2 case.
16095 2018-02-13 Jeff Law <law@redhat.com>
16097 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
16098 entry for "vector".
16100 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
16103 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
16107 * common.opt (gas-loc-support, gas-locview-support): New.
16108 (ginline-points, ginternal-reset-location-views): New.
16109 * doc/invoke.texi: Document them. Use @itemx where intended.
16110 (gvariable-location-views): Adjust.
16111 * target.def (reset_location_view): New.
16112 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
16113 (TARGET_RESET_LOCATION_VIEW): New.
16114 * doc/tm.texi: Rebuilt.
16115 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
16116 (dwarf2out_default_as_locview_support): New.
16117 (output_asm_line_debug_info): Use option variables.
16118 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
16119 (output_loc_list): Likewise.
16120 (add_high_low_attributes): Check option variables.
16121 Don't output entry view attribute in strict mode.
16122 (gen_inlined_subroutine_die): Check option variables.
16123 (dwarf2out_inline_entry): Likewise.
16124 (init_sections_and_labels): Likewise.
16125 (dwarf2out_early_finish): Likewise.
16126 (maybe_reset_location_view): New, from...
16127 (dwarf2out_var_location): ... here. Call it.
16128 * debug.h (dwarf2out_default_as_loc_support): Declare.
16129 (dwarf2out_default_as_locview_support): Declare.
16130 * hooks.c (hook_int_rtx_insn_0): New.
16131 * hooks.h (hook_int_rtx_insn_0): Declare.
16132 * toplev.c (process_options): Take -gas-loc-support and
16133 -gas-locview-support from dwarf2out. Enable
16134 -gvariable-location-views by default only with locview
16135 assembler support. Enable -ginternal-reset-location-views by
16136 default only if the target defines the corresponding hook.
16137 Enable -ginline-points by default if location views are
16138 enabled; force it disabled if statement frontiers are
16140 * tree-inline.c (expand_call_inline): Check option variables.
16141 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
16143 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
16145 PR tree-optimization/84321
16146 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
16147 handling. Also check whether the anti-range contains any values
16148 that satisfy the mask; switch to a VR_RANGE if not.
16150 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
16153 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
16155 2018-02-13 Martin Jambor <mjambor@suse.cz>
16158 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
16159 of call statements, also set location of a load to a temporary.
16161 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
16163 * config/rl78/rl78.c (add_vector_labels): New function.
16164 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
16165 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
16166 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
16167 which checks that no arguments are passed.
16168 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
16169 * doc/extend.texi: Documentation for the new attribute.
16171 2018-02-13 Andreas Schwab <schwab@suse.de>
16173 * config/riscv/linux.h (CPP_SPEC): Define.
16175 2018-02-13 Jakub Jelinek <jakub@redhat.com>
16178 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
16179 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
16180 OPTION_MASK_ISA_AES as first argument to def_builtin_const
16181 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
16182 instead of OPTION_MASK_ISA_PCLMUL as first argument to
16183 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
16184 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
16185 temporarily for AES and PCLMUL builtins.
16187 PR tree-optimization/84339
16188 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
16189 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
16192 PR middle-end/84309
16193 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
16194 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
16195 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
16197 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
16199 * omp-simd-clone.h: New file.
16200 * omp-simd-clone.c: Include omp-simd-clone.h.
16201 (expand_simd_clones): No longer static.
16202 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
16203 cgraph.h and omp-simd-clone.h.
16204 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
16205 (vect_recog_widen_shift_pattern): Formatting fix.
16206 (vect_pattern_recog_1): Don't check optab for calls.
16209 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
16210 operands[2] into a REG before using gen_lowpart on it.
16212 2018-02-12 Jeff Law <law@redhat.com>
16215 * config/sh/sh.c (find_barrier): Consider a sibling call
16218 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
16219 successfully back substituting a reg.
16221 2018-02-12 Richard Biener <rguenther@suse.de>
16223 PR tree-optimization/84037
16224 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
16225 parameter, move visited init to caller.
16226 (vect_slp_analyze_operations): Separate cost from validity
16227 check, initialize visited once for all instances.
16228 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
16230 * tree-vect-stmts.c (vect_model_simple_cost): Make early
16232 (vect_model_promotion_demotion_cost): Likewise.
16233 (vectorizable_bswap): Guard cost modeling with !slp_node
16234 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
16236 (vectorizable_call): Likewise.
16237 (vectorizable_conversion): Likewise.
16238 (vectorizable_assignment): Likewise.
16239 (vectorizable_shift): Likewise.
16240 (vectorizable_operation): Likewise.
16241 (vectorizable_store): Likewise.
16242 (vectorizable_load): Likewise.
16243 (vectorizable_condition): Likewise.
16244 (vectorizable_comparison): Likewise.
16246 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
16249 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
16250 (ASAN_MARK): Fix fnspec to account for return value, change pointer
16251 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
16253 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
16255 PR middle-end/83665
16256 * params.def (inline-min-speedup): Increase from 8 to 15.
16257 (max-inline-insns-auto): Decrease from 40 to 30.
16258 * ipa-split.c (consider_split): Add some buffer for function to
16259 be considered inlining candidate.
16260 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
16263 2018-02-12 Richard Biener <rguenther@suse.de>
16265 PR tree-optimization/84037
16266 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
16267 matched stmts if we cannot swap the non-matched ones.
16269 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
16271 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
16272 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
16273 _mm_maskz_scalef_round_ss): New intrinsics.
16274 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
16275 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
16276 __builtin_ia32_scalefss_round): Remove.
16277 (__builtin_ia32_scalefsd_mask_round,
16278 __builtin_ia32_scalefss_mask_round): New intrinsics.
16279 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
16280 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
16281 ((match_operand:VF_128 2 "<round_nimm_predicate>"
16282 "<round_constraint>")): Changed to ...
16283 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
16284 "<round_scalar_constraint>")): ... this.
16285 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
16286 %0, %1, %2<round_op3>}"): Changed to ...
16287 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
16288 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
16289 %2<round_scalar_mask_op3>}"): ... this.
16290 * config/i386/subst.md (round_scalar_nimm_predicate): New.
16292 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
16294 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
16295 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
16296 (_mm_maskz_sqrt_round_ss): New intrinsics.
16297 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
16298 (__builtin_ia32_sqrtsd_mask_round)
16299 (__builtin_ia32_sqrtss_mask_round): New builtins.
16300 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
16301 (__builtin_ia32_sqrtss_round): Remove.
16302 (__builtin_ia32_sqrtsd_mask_round)
16303 (__builtin_ia32_sqrtss_mask_round): New builtins.
16304 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
16305 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
16306 ((match_operand:VF_128 1 "vector_operand"
16307 "xBm,<round_constraint>")): Changed to ...
16308 ((match_operand:VF_128 1 "vector_operand"
16309 "xBm,<round_scalar_constraint>")): ... this.
16310 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
16311 %0, %2, %<iptr>1<round_op3>}): Changed to ...
16312 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
16313 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
16314 %<iptr>1<round_scalar_mask_op3>}): ... this.
16315 ((set_attr "prefix" "<round_prefix>")): Changed to ...
16316 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
16318 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
16321 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
16322 Cast vec_cmpeq result to correct type.
16323 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
16324 Cast vec_cmpgt result to correct type.
16326 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
16328 * final.c (final_scan_insn_1): Renamed from...
16329 (final_scan_insn): ... this. New wrapper, to recover
16330 seen from the outermost call in recursive ones.
16331 * config/sparc/sparc.c (output_return): Drop seen from call.
16332 (output_sibcall): Likewise.
16333 * config/visium/visium.c (output_branch): Likewise.
16335 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
16337 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
16340 2018-02-10 Alan Modra <amodra@gmail.com>
16343 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
16344 Specify LR as an input.
16346 2018-02-10 Jakub Jelinek <jakub@redhat.com>
16349 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
16350 remove_member_access_dummy_vars): New functions.
16351 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
16352 lower_omp_1, execute_lower_omp): Use them.
16354 PR rtl-optimization/84308
16355 * shrink-wrap.c (spread_components): Release todo vector.
16357 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
16359 PR rtl-optimization/57193
16360 * ira-color.c (struct allocno_color_data): Add member
16361 conflict_allocno_hard_prefs.
16362 (update_conflict_allocno_hard_prefs): New.
16363 (bucket_allocno_compare_func): Add a preference based on
16364 conflict_allocno_hard_prefs.
16365 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
16366 (color_allocnos): Remove a dead code. Initiate
16367 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
16369 2018-02-09 Jakub Jelinek <jakub@redhat.com>
16372 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
16373 constraint from =wa to wa. Avoid a subreg on the output operand,
16374 instead use a pseudo and subreg it in a move.
16375 (p9_xxbrd_<mode>): Changed to ...
16376 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
16377 (p9_xxbrd_v2df): New expander.
16378 (p9_xxbrw_<mode>): Changed to ...
16379 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
16380 (p9_xxbrw_v4sf): New expander.
16382 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
16384 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
16386 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
16389 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
16390 multiply in 32-bit mode.
16391 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
16392 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
16395 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
16397 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
16398 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
16399 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
16400 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
16402 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
16405 * dwarf2out.c (is_trivial_indirect_ref): New function.
16406 (dwarf2out_late_global_decl): Do not generate a location
16407 attribute for variables that have a non-trivial DECL_VALUE_EXPR
16408 and that are not defined in the current unit.
16410 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
16412 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
16413 instead of a libcall for UNORDERED.
16415 2018-02-09 Tamar Christina <tamar.christina@arm.com>
16418 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
16419 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
16421 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
16424 * config/s390/s390.c (s390_set_current_function): Invoke
16425 s390_indirect_branch_settings also if fndecl didn't change.
16427 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
16429 * config/rs6000/rs6000.md (blockage): Set length to zero.
16431 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
16433 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
16435 2018-02-09 Jakub Jelinek <jakub@redhat.com>
16438 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
16439 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
16443 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
16444 PARALLEL incoming that failed vt_get_decl_and_offset check.
16446 PR middle-end/84237
16447 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
16448 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
16450 (get_variable_section): For decls in named .bss* sections pass true as
16451 second argument to bss_initializer_p.
16453 2018-02-09 Marek Polacek <polacek@redhat.com>
16454 Jakub Jelinek <jakub@redhat.com>
16457 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
16458 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
16459 Sync some changes from cxx_fold_indirect_ref.
16461 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
16463 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
16465 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
16466 (BLOCK_INLINE_ENTRY_LABEL): New.
16467 (dwarf2out_var_location): Disregard inline entry markers.
16468 (inline_entry_data): New struct.
16469 (inline_entry_data_hasher): New hashtable type.
16470 (inline_entry_data_hasher::hash): New.
16471 (inline_entry_data_hasher::equal): New.
16472 (inline_entry_data_table): New variable.
16473 (add_high_low_attributes): Add DW_AT_entry_pc and
16474 DW_AT_GNU_entry_view attributes if a pending entry is found
16475 in inline_entry_data_table. Add old entry_pc attribute only
16476 if debug nonbinding markers are disabled.
16477 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
16478 markers are enabled.
16479 (block_within_block_p, dwarf2out_inline_entry): New.
16480 (dwarf2out_finish): Check that no entries remained in
16481 inline_entry_data_table.
16482 * final.c (reemit_insn_block_notes): Handle inline entry notes.
16483 (final_scan_insn, notice_source_line): Likewise.
16484 (rest_of_clean_state): Skip inline entry markers.
16485 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
16487 * gimple.c (gimple_build_debug_inline_entry): New.
16488 * gimple.h (enum gimple_debug_subcode): Add
16489 GIMPLE_DEBUG_INLINE_ENTRY.
16490 (gimple_build_debug_inline_entry): Declare.
16491 (gimple_debug_inline_entry_p): New.
16492 (gimple_debug_nonbind_marker_p): Adjust.
16493 * insn-notes.def (INLINE_ENTRY): New.
16494 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
16495 inline entry marker notes.
16496 (print_insn): Likewise.
16497 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
16498 (INSN_DEBUG_MARKER_KIND): Likewise.
16499 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
16500 * tree-inline.c (expand_call_inline): Build and insert
16501 debug_inline_entry stmt.
16502 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
16503 inline entry blocks early, if nonbind markers are enabled.
16504 (dump_scope_block): Dump fragment info.
16505 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
16506 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
16507 (gimple_build_debug_inline_entry): New.
16508 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
16509 Enable/disable inline entry points too.
16510 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
16511 (DEBUG_INSN): Describe inline entry markers.
16513 * common.opt (gvariable-location-views): New.
16514 (gvariable-location-views=incompat5): New.
16515 * config.in: Rebuilt.
16516 * configure: Rebuilt.
16517 * configure.ac: Test assembler for view support.
16518 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
16519 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
16520 * dwarf2out.c (var_loc_view): New typedef.
16521 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
16522 (dwarf2out_locviews_in_attribute): New.
16523 (dwarf2out_locviews_in_loclist): New.
16524 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
16525 (enum dw_line_info_opcode): Add LI_adv_address.
16526 (struct dw_line_info_table): Add view.
16527 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
16528 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
16529 (zero_view_p): New variable.
16530 (ZERO_VIEW_P): New macro.
16531 (output_asm_line_debug_info): New.
16532 (struct var_loc_node): Add view.
16533 (add_AT_view_list, AT_loc_list): New.
16534 (add_var_loc_to_decl): Add view param. Test it against last.
16535 (new_loc_list): Add view params. Record them.
16536 (AT_loc_list_ptr): Handle loc and view lists.
16537 (view_list_to_loc_list_val_node): New.
16538 (print_dw_val): Handle dw_val_class_view_list.
16539 (size_of_die): Likewise.
16540 (value_format): Likewise.
16541 (loc_list_has_views): New.
16542 (gen_llsym): Set vl_symbol too.
16543 (maybe_gen_llsym, skip_loc_list_entry): New.
16544 (dwarf2out_maybe_output_loclist_view_pair): New.
16545 (output_loc_list): Output view list or entries too.
16546 (output_view_list_offset): New.
16547 (output_die): Handle dw_val_class_view_list.
16548 (output_dwarf_version): New.
16549 (output_compilation_unit_header): Use it.
16550 (output_skeleton_debug_sections): Likewise.
16551 (output_rnglists, output_line_info): Likewise.
16552 (output_pubnames, output_aranges): Update version comments.
16553 (output_one_line_info_table): Output view numbers in asm comments.
16554 (dw_loc_list): Determine current endview, pass it to new_loc_list.
16555 Call maybe_gen_llsym.
16556 (loc_list_from_tree_1): Adjust.
16557 (add_AT_location_description): Create view list attribute if
16558 needed, check it's absent otherwise.
16559 (convert_cfa_to_fb_loc_list): Adjust.
16560 (maybe_emit_file): Call output_asm_line_debug_info for test.
16561 (dwarf2out_var_location): Reset views as needed. Precompute
16562 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
16563 attribute. Set view.
16564 (new_line_info_table): Reset next view.
16565 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
16566 (dwarf2out_source_line): Likewise. Output view resets and labels to
16567 the assembler, or select appropriate line info opcodes.
16568 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
16569 (optimize_string_length): Catch it. Adjust.
16570 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
16571 dw_val_class_view_list, and remove it if no longer needed.
16572 (hash_loc_list): Hash view numbers.
16573 (loc_list_hasher::equal): Compare them.
16574 (optimize_location_lists): Check whether a view list symbol is
16575 needed, and whether the locview attribute is present, and
16576 whether they match. Remove the locview attribute if no longer
16578 (index_location_lists): Call skip_loc_list_entry for test.
16579 (dwarf2out_finish): Call output_asm_line_debug_info for test.
16580 Use output_dwarf_version.
16581 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
16582 (struct dw_val_node): Add val_view_list.
16583 * final.c (SEEN_NEXT_VIEW): New.
16584 (set_next_view_needed): New.
16585 (clear_next_view_needed): New.
16586 (maybe_output_next_view): New.
16587 (final_start_function): Rename to...
16588 (final_start_function_1): ... this. Take pointer to FIRST,
16589 add SEEN parameter. Emit param bindings in the initial view.
16590 (final_start_function): Reintroduce SEEN-less interface.
16591 (final): Rename to...
16592 (final_1): ... this. Take SEEN parameter. Output final pending
16593 next view at the end.
16594 (final): Reintroduce seen-less interface.
16595 (final_scan_insn): Output pending next view before switching
16596 sections or ending a block. Mark the next view as needed when
16597 outputting variable locations. Notify debug backend of section
16598 changes, and of location view changes.
16599 (rest_of_handle_final): Adjust.
16600 * toplev.c (process_options): Autodetect value for debug variable
16601 location views option. Warn on incompat5 without -gdwarf-5.
16602 * doc/invoke.texi (gvariable-location-views): New.
16603 (gvariable-location-views=incompat5): New.
16604 (gno-variable-location-views): New.
16606 2018-02-08 David Malcolm <dmalcolm@redhat.com>
16608 PR tree-optimization/84136
16609 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
16610 that the result of find_edge is non-NULL.
16612 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
16615 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
16616 storing integer register in SImode. Fix cost of 256 and 512
16617 byte aligned SSE register store.
16619 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
16621 * config/i386/i386.c (ix86_multiplication_cost): Fix
16622 multiplication cost for TARGET_AVX512DQ.
16624 2018-02-08 Marek Polacek <polacek@redhat.com>
16626 PR tree-optimization/84238
16627 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
16630 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
16632 PR tree-optimization/84265
16633 * tree-vect-stmts.c (vectorizable_store): Don't treat
16634 VMAT_CONTIGUOUS accesses as grouped.
16635 (vectorizable_load): Likewise.
16637 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
16639 PR tree-optimization/81635
16640 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
16641 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
16642 (test_round_for_mask): New functions.
16643 (wide_int_cc_tests): Call test_round_for_mask.
16644 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
16645 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
16646 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
16647 range returned by get_range_info.
16649 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
16652 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
16653 * symtab.c: Include builtins.h
16654 (symtab_node::output_to_lto_symbol_table_p): Move here
16655 from lto-streamer-out.c:output_symbol_p.
16656 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
16657 (output_symbol_p): Move all logic to symtab.c
16658 (produce_symtab): Update.
16660 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
16662 * config/s390/s390-opts.h (enum indirect_branch): Define.
16663 * config/s390/s390-protos.h (s390_return_addr_from_memory)
16664 (s390_indirect_branch_via_thunk)
16665 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
16666 (enum s390_indirect_branch_type): Define.
16667 * config/s390/s390.c (struct s390_frame_layout, struct
16668 machine_function): Remove.
16669 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
16670 (indirect_branch_table_label_no, indirect_branch_table_name):
16672 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
16673 (enum s390_indirect_branch_option): Define.
16674 (s390_return_addr_from_memory): New function.
16675 (s390_handle_string_attribute): New function.
16676 (s390_attribute_table): Add new attribute handler.
16677 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
16678 (s390_indirect_branch_via_thunk): New function.
16679 (s390_indirect_branch_via_inline_thunk): New function.
16680 (s390_function_ok_for_sibcall): When jumping via thunk disallow
16681 sibling call optimization for non z10 compiles.
16682 (s390_emit_call): Force indirect branch target to be a single
16683 register. Add r1 clobber for non-z10 compiles.
16684 (s390_emit_epilogue): Emit return jump via return_use expander.
16685 (s390_reorg): Handle JUMP_INSNs as execute targets.
16686 (s390_option_override_internal): Perform validity checks for the
16687 new command line options.
16688 (s390_indirect_branch_attrvalue): New function.
16689 (s390_indirect_branch_settings): New function.
16690 (s390_set_current_function): Invoke s390_indirect_branch_settings.
16691 (s390_output_indirect_thunk_function): New function.
16692 (s390_code_end): Implement target hook.
16693 (s390_case_values_threshold): Implement target hook.
16694 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
16696 * config/s390/s390.h (struct s390_frame_layout)
16697 (struct machine_function): Move here from s390.c.
16698 (TARGET_INDIRECT_BRANCH_NOBP_RET)
16699 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
16700 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
16701 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
16702 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
16703 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
16704 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
16705 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
16706 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
16707 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
16708 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
16709 (mnemonic attribute): Add values which aren't recognized
16711 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
16712 pattern for branch conversion. Fix mnemonic attribute.
16713 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
16714 indirect branch via thunk if requested.
16715 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
16716 ("*indirect_jump"): Disable for branch conversion using out of
16718 ("indirect_jump_via_thunk<mode>_z10")
16719 ("indirect_jump_via_thunk<mode>")
16720 ("indirect_jump_via_inlinethunk<mode>_z10")
16721 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
16722 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
16723 ("casesi_jump_via_inlinethunk<mode>_z10")
16724 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
16725 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
16726 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
16727 ("*indirect2_jump"): Disable for branch conversion.
16728 ("casesi_jump"): Turn into expander and expand patterns for branch
16730 ("return_use"): New expander.
16731 ("*return"): Emit return via thunk and rename it to ...
16732 ("*return<mode>"): ... this one.
16733 * config/s390/s390.opt: Add new options and and enum for the
16736 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
16738 * lra-constraints.c (match_reload): Unconditionally use
16739 gen_lowpart_SUBREG, rather than selecting between that
16740 and equivalent gen_rtx_SUBREG code.
16742 2018-02-08 Richard Biener <rguenther@suse.de>
16744 PR tree-optimization/84233
16745 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
16746 changed flag instead of boguously re-using phi_inserted.
16748 2018-02-08 Martin Jambor <mjambor@suse.cz>
16750 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
16751 static local variables.
16753 2018-02-08 Richard Biener <rguenther@suse.de>
16755 PR tree-optimization/84278
16756 * tree-vect-stmts.c (vectorizable_store): When looking for
16757 smaller vector types to perform grouped strided loads/stores
16758 make sure the mode is supported by the target.
16759 (vectorizable_load): Likewise.
16761 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
16763 * config/aarch64/aarch64.c (aarch64_components_for_bb):
16764 Increase LDP/STP opportunities by adding adjacent callee-saves.
16766 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
16768 PR rtl-optimization/84068
16769 PR rtl-optimization/83459
16770 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
16772 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
16774 PR tree-optimization/84224
16775 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
16776 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
16777 non-zero arguments.
16779 2018-02-07 Iain Sandoe <iain@codesourcery.com>
16782 * config/rs6000/altivec.md (*restore_world): Remove LR use.
16783 * config/rs6000/predicates.md (restore_world_operation): Adjust op
16784 count, remove one USE.
16786 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
16788 * doc/install.texi (Configuration): Document the
16789 --with-long-double-format={ibm,ieee} PowerPC configuration
16793 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
16794 Convert from define_expand to be define_insn_and_split. Rework
16795 float/double/_Float128 conversions to QI/HI/SImode to work with
16796 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
16797 conversions to QI/HImode types did a store and then a load to
16798 truncate the value. For conversions to VSX registers, don't split
16799 the insn, instead emit the code directly. Use the code iterator
16800 any_fix to combine signed and unsigned conversions.
16801 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
16802 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
16803 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
16804 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
16805 (fix_<mode>di2_hw): Likewise.
16806 (fixuns_<mode>di2_hw): Likewise.
16807 (fix_<mode>si2_hw): Likewise.
16808 (fixuns_<mode>si2_hw): Likewise.
16809 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
16810 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
16811 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
16812 fix<uns>_trunc<SFDF:mode>si2_p8.
16813 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
16814 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
16815 (fix<uns>_<mode>_mem): Likewise.
16816 (fctiw<u>z_<mode>_mem): Likewise.
16817 (fix<uns>_<mode>_mem): Likewise.
16818 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
16819 the register allocator from doing a direct move to the GPRs to do
16820 a store, and instead use the ISA 3.0 store byte/half-word from
16821 vector register instruction. For IEEE 128-bit floating point,
16822 also optimize stores of 32-bit ints.
16823 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
16825 2018-02-07 Alan Hayward <alan.hayward@arm.com>
16827 * genextract.c (push_pathstr_operand): New function to support
16829 (walk_rtx): Call push_pathstr_operand.
16830 (print_path): Support [a-zA-Z].
16832 2018-02-07 Richard Biener <rguenther@suse.de>
16834 PR tree-optimization/84037
16835 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
16836 (cse_and_gimplify_to_preheader): Declare.
16837 (vect_get_place_in_interleaving_chain): Likewise.
16838 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
16840 (_loop_vec_info::~_loop_vec_info): Delete it.
16841 (cse_and_gimplify_to_preheader): New function.
16842 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
16843 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
16844 (vectorizable_load): Likewise. For grouped stores always base
16845 the IV on the first element.
16846 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
16847 condition before gimplifying.
16849 2018-02-07 Jakub Jelinek <jakub@redhat.com>
16851 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
16852 *DIV_EXPR and *MOD_EXPR.
16854 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
16857 * config/i386/i386.c (ix86_option_override_internal): Mask out
16858 the CF_SET bit when checking -fcf-protection.
16860 2018-02-07 Tom de Vries <tom@codesourcery.com>
16863 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
16866 2018-02-07 Richard Biener <rguenther@suse.de>
16868 PR tree-optimization/84204
16869 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
16872 PR tree-optimization/84205
16873 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
16874 special-case isl_ast_op_zdiv_r.
16876 PR tree-optimization/84223
16877 * graphite-scop-detection.c (gather_bbs::before_dom_children):
16878 Only add conditions from within the region.
16879 (gather_bbs::after_dom_children): Adjust.
16881 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
16884 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
16885 * config/avr/avr.md: Only post-reload split REG-REG moves if
16886 either register is GENERAL_REG_P.
16888 2018-02-07 Jakub Jelinek <jakub@redhat.com>
16890 PR tree-optimization/84235
16891 * tree-ssa-scopedtables.c
16892 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
16893 if the subtraction is performed in floating point type where NaNs are
16894 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
16895 build 1. Formatting fix.
16897 2018-02-06 Jakub Jelinek <jakub@redhat.com>
16900 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
16901 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
16902 and skip it regardless of bb boundaries. Use CALL_P macro,
16903 don't test INSN_P (insn) together with CALL_P or JUMP_P check
16904 unnecessarily, formatting fix.
16906 2018-02-06 Michael Collison <michael.collison@arm.com>
16908 * config/arm/thumb2.md:
16909 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
16910 (*thumb_mov_notscc): Ditto.
16912 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
16915 * config/rs6000/rs6000.md (su code attribute): Use "u" for
16916 unsigned_fix, not "s".
16918 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
16920 * configure.ac (gcc_fn_eh_frame_ro): New function.
16921 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
16922 correct .eh_frame permissions.
16923 * configure: Regenerate.
16925 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
16927 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
16928 irrelevant options.
16930 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16932 * config/rs6000/rs6000.c (rs6000_option_override_internal):
16933 Display warning message for -mno-speculate-indirect-jumps.
16935 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
16937 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
16939 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
16941 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
16943 PR tree-optimization/84225
16944 * tree-eh.c (find_trapping_overflow): Only call
16945 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
16947 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
16950 * config/i386/i386.c: Reimplement the check of possible options
16951 -mibt/-mshstk conbination. Change error messages.
16952 * doc/invoke.texi: Fix a typo: remove extra '='.
16954 2018-02-06 Marek Polacek <polacek@redhat.com>
16956 PR tree-optimization/84228
16957 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
16959 2018-02-06 Tamar Christina <tamar.christina@arm.com>
16962 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
16963 emitted arch directives.
16964 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
16965 __ARM_FEATURE_COPROC before changing architectures.
16967 2018-02-06 Richard Biener <rguenther@suse.de>
16969 * config/i386/i386.c (print_reg): Fix typo.
16970 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
16972 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
16974 * configure: Regenerate.
16976 2018-02-05 Martin Sebor <msebor@redhat.com>
16978 PR tree-optimization/83369
16979 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
16982 2018-02-05 Martin Liska <mliska@suse.cz>
16984 * doc/invoke.texi: Cherry-pick upstream r323995.
16986 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
16988 * ira.c (ira_init_register_move_cost): Adjust comment.
16990 2018-02-05 Martin Liska <mliska@suse.cz>
16992 PR gcov-profile/84137
16993 * doc/gcov.texi: Fix typo in documentation.
16995 2018-02-05 Martin Liska <mliska@suse.cz>
16997 PR gcov-profile/83879
16998 * doc/gcov.texi: Document necessity of --dynamic-list-data when
16999 using dlopen functionality.
17001 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
17003 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
17004 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
17005 _mm_maskz_range_ss, _mm_mask_range_round_ss,
17006 _mm_maskz_range_round_ss): New intrinsics.
17007 (__builtin_ia32_rangesd128_round)
17008 (__builtin_ia32_rangess128_round): Remove.
17009 (__builtin_ia32_rangesd128_mask_round,
17010 __builtin_ia32_rangess128_mask_round): New builtins.
17011 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
17012 __builtin_ia32_rangess128_round): Remove.
17013 (__builtin_ia32_rangesd128_mask_round,
17014 __builtin_ia32_rangess128_mask_round): New builtins.
17015 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
17016 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
17017 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
17018 "<round_saeonly_constraint>")): Changed to ...
17019 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
17020 "<round_saeonly_scalar_constraint>")): ... this.
17021 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
17022 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
17023 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
17024 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
17025 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
17027 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
17029 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
17031 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
17032 Remove all values except native, 8540 and 8548.
17034 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
17036 * config/i386/i386.c (ix86_output_function_return): Pass
17037 INVALID_REGNUM, instead of -1, as invalid register number to
17038 indirect_thunk_name and output_indirect_thunk.
17040 2018-02-02 Julia Koval <julia.koval@intel.com>
17042 * config.gcc: Add -march=icelake.
17043 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
17044 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
17045 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
17046 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
17047 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
17048 (processor_target_table): Add icelake.
17049 (ix86_option_override_internal): Handle new PTAs.
17050 (get_builtin_code_for_version): Handle icelake.
17051 (M_INTEL_COREI7_ICELAKE): New.
17052 (fold_builtin_cpu): Handle icelake.
17053 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
17054 * doc/invoke.texi: Add -march=icelake.
17056 2018-02-02 Julia Koval <julia.koval@intel.com>
17058 * config/i386/i386.c (ix86_option_override_internal): Change flags type
17059 to wide_int_bitmask.
17060 * wide-int-bitmask.h: New.
17062 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
17065 * config/i386/i386.md: Replace Pmode with word_mode in
17066 builtin_setjmp_setup and builtin_longjmp to support x32.
17068 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
17072 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
17074 (rs6000_supported_cpu_names): New static variable.
17075 (linux_cpu_translation_table): Likewise.
17076 (elf_platform) <cpu>: Define new static variable and use it.
17077 Translate kernel AT_PLATFORM name to canonical name if needed.
17078 Error if platform name is unknown.
17080 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
17083 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
17085 2018-02-01 Jeff Law <law@redhat.com>
17088 * config/i386/i386.c (release_scratch_register_on_entry): Add new
17089 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
17090 the scratch if RELEASE_VIA_POP is false.
17091 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
17092 If we have to save a temporary register, decrement SIZE appropriately.
17093 Pass new arguments to release_scratch_register_on_entry.
17094 (ix86_adjust_stack_and_probe): Likewise.
17095 (ix86_emit_probe_stack_range): Pass new arguments to
17096 release_scratch_register_on_entry.
17098 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
17100 PR rtl-optimization/84157
17101 * combine.c (change_zero_ext): Use REG_P predicate in
17102 front of HARD_REGISTER_P predicate.
17104 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
17106 * config/avr/avr.c (avr_option_override): Move disabling of
17107 -fdelete-null-pointer-checks to...
17108 * common/config/avr/avr-common.c (avr_option_optimization_table):
17111 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
17113 PR tree-optimization/81635
17114 * tree-data-ref.c (split_constant_offset_1): For types that
17115 wrap on overflow, try to use range info to prove that wrapping
17118 2018-02-01 Renlin Li <renlin.li@arm.com>
17121 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
17122 TAILCALL_ADDR_REGS.
17123 (aarch64_register_move_cost): Likewise.
17124 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
17125 TAILCALL_ADDR_REGS.
17126 (REG_CLASS_NAMES): Likewise.
17127 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
17128 TAILCALL_ADDR_REGS. Remove IP registers.
17129 * config/aarch64/aarch64.md (Ucs): Update register constraint.
17131 2018-02-01 Richard Biener <rguenther@suse.de>
17133 * domwalk.h (dom_walker::dom_walker): Add additional constructor
17134 for specifying RPO order and allow NULL for that.
17135 * domwalk.c (dom_walker::dom_walker): Likewise.
17136 (dom_walker::walk): Handle NULL RPO order.
17137 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
17139 (rewrite_update_dom_walker): Likewise.
17140 (mark_def_dom_walker): Likewise.
17142 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
17144 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
17145 (aarch64_maybe_expand_sve_subreg_move): Declare.
17146 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
17147 * config/aarch64/predicates.md (aarch64_any_register_operand): New
17149 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
17150 that are semantically a reverse operation.
17151 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
17152 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
17153 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
17155 (aarch64_can_change_mode_class): For big-endian, forbid changes
17156 between two SVE modes if they have different element sizes.
17158 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
17160 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
17161 the TImode handling for big-endian targets.
17163 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
17165 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
17166 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
17168 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
17169 Remove BSWAP handing for big-endian targets and use the form of
17170 LD1RQ appropariate for the mode.
17172 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
17174 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
17175 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
17176 duplicated element.
17178 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
17181 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
17182 check for operands that need to go through aarch64_sve_reload_be.
17184 2018-02-01 Jakub Jelinek <jakub@redhat.com>
17186 PR tree-optimization/81661
17187 PR tree-optimization/84117
17188 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
17189 * tree-eh.c: Include gimplify.h.
17190 (find_trapping_overflow, replace_trapping_overflow,
17191 rewrite_to_non_trapping_overflow): New functions.
17192 * tree-vect-loop.c: Include tree-eh.h.
17193 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
17194 * tree-data-ref.c: Include tree-eh.h.
17195 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
17197 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
17199 PR rtl-optimization/84123
17200 * combine.c (change_zero_ext): Check if hard register satisfies
17201 can_change_dest_mode before calling gen_lowpart_SUBREG.
17203 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
17206 * ira.c (ira_init_register_move_cost): Remove assert.
17208 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
17210 PR rtl-optimization/84071
17211 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
17212 * doc/tm.texi: Regenerate.
17214 2018-01-31 Richard Biener <rguenther@suse.de>
17216 PR tree-optimization/84132
17217 * tree-data-ref.c (analyze_miv_subscript): Properly
17218 check whether evolution_function_is_affine_multivariate_p
17219 before calling gcd_of_steps_may_divide_p.
17221 2018-01-31 Julia Koval <julia.koval@intel.com>
17224 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
17225 * config/i386/i386.md (rdpid_rex64) New.
17226 (rdpid): Make 32bit only.
17228 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
17231 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
17232 an IDENTIFIER_NODE for FUNCTION_TYPE's.
17234 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
17237 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
17239 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
17241 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
17243 PR rtl-optimization/84071
17244 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
17245 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
17247 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
17249 * config/arc/arc.c (arc_handle_aux_attribute): New function.
17250 (arc_attribute_table): Add 'aux' attribute.
17251 (arc_in_small_data_p): Consider aux like variables.
17252 (arc_is_aux_reg_p): New function.
17253 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
17254 (arc_get_aux_arg): New function.
17255 (prepare_move_operands): Handle aux-register access.
17256 (arc_handle_aux_attribute): New function.
17257 * doc/extend.texi (ARC Variable attributes): Add subsection.
17259 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
17261 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
17262 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
17263 (arc_attribute_table): Add 'uncached' attribute.
17264 (arc_print_operand): Print '.di' flag for uncached memory
17266 (arc_in_small_data_p): Do not consider for small data the uncached
17268 (arc_is_uncached_mem_p): New function.
17269 * config/arc/predicates.md (compact_store_memory_operand): Check
17270 for uncached memory accesses.
17271 (nonvol_nonimm_operand): Likewise.
17272 * doc/extend.texi (ARC Type Attribute): New subsection.
17274 2018-01-31 Jakub Jelinek <jakub@redhat.com>
17277 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
17278 falign-loops=): Add Optimization flag.
17280 2018-01-30 Jeff Law <law@redhat.com>
17283 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
17284 INT_REGISTERS_SAVED. Check it prior to calling
17285 get_scratch_register_on_entry.
17286 (ix86_adjust_stack_and_probe): Similarly.
17287 (ix86_emit_probe_stack_range): Similarly.
17288 (ix86_expand_prologue): Corresponding changes.
17290 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
17293 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
17294 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
17296 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
17299 * lra-constraints.c (curr_insn_transform): Process AND in the
17302 2018-01-30 Jakub Jelinek <jakub@redhat.com>
17304 PR rtl-optimization/83986
17305 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
17306 dependence against last_pending_memory_flush in addition to
17307 pending_jump_insns.
17309 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
17311 PR tree-optimization/81611
17312 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
17315 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17318 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
17321 2018-01-30 Richard Biener <rguenther@suse.de>
17322 Jakub Jelinek <jakub@redhat.com>
17324 PR tree-optimization/84111
17325 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
17326 inner loops added during recursion, as they don't have up-to-date
17329 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
17332 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
17333 (can_inline_edge_by_limits_p): ... here.
17334 (can_early_inline_edge_p, check_callers,
17335 update_caller_keys, update_callee_keys, recursive_inlining,
17336 add_new_edges_to_heap, speculation_useful_p,
17337 inline_small_functions,
17338 inline_small_functions, flatten_function,
17339 inline_to_all_callers_1): Update.
17341 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
17343 * profile-count.c (profile_count::combine_with_ipa_count): Handle
17346 2018-01-30 Richard Biener <rguenther@suse.de>
17348 PR tree-optimization/83008
17349 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
17350 invariant and constant vector uses in stmts when they need
17351 more than one stmt.
17353 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
17356 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
17357 * configure: Regenerate.
17359 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
17361 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
17363 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
17364 Use gen_rtx_REG rather than gen_lowpart.
17366 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
17368 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
17369 rather than 0 when creating partial subregs.
17371 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
17373 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
17376 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
17379 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
17380 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
17381 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
17382 flags. This restores the settings used before the 2017-07-24.
17383 Turning off pre increment/decrement/modify allows IVOPTS to
17384 optimize DF/SF loops where the index is an int.
17386 2018-01-29 Richard Biener <rguenther@suse.de>
17387 Kelvin Nilsen <kelvin@gcc.gnu.org>
17390 * tree-vect-stmts.c (vectorizable_call): Don't call
17391 targetm.vectorize_builtin_md_vectorized_function if callee is
17394 2018-01-22 Carl Love <cel@us.ibm.com>
17396 * doc/extend.tex: Fix typo in second arg in
17397 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
17399 2018-01-29 Richard Biener <rguenther@suse.de>
17401 PR tree-optimization/84086
17402 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
17403 (flush_ssaname_freelist): When SSA names were released reset
17404 the SCEV hash table.
17406 2018-01-29 Richard Biener <rguenther@suse.de>
17408 PR tree-optimization/84057
17409 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
17410 removed paths when removing edges.
17412 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
17414 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
17415 -mfunction-return=@var{choice}.
17417 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
17419 PR diagnostic/84034
17420 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
17421 Handle CR like TAB.
17422 (layout::print_source_line): Likewise.
17423 (test_get_line_width_without_trailing_whitespace): Add test cases.
17425 2018-01-27 Jakub Jelinek <jakub@redhat.com>
17427 PR middle-end/84040
17428 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
17431 2018-01-26 Jim Wilson <jimw@sifive.com>
17433 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
17435 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
17438 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17440 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
17441 and CMP + SUB-immediate -> SUBS.
17443 2018-01-26 Martin Sebor <msebor@redhat.com>
17445 PR tree-optimization/83896
17446 * tree-ssa-strlen.c (get_string_len): Rename...
17447 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
17448 Avoid assuming length is constant.
17449 (handle_char_store): Use HOST_WIDE_INT for string length.
17451 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
17454 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
17455 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
17457 2018-01-26 Richard Biener <rguenther@suse.de>
17459 PR rtl-optimization/84003
17460 * dse.c (record_store): Only record redundant stores when
17461 the earlier store aliases at least all accesses the later one does.
17463 2018-01-26 Jakub Jelinek <jakub@redhat.com>
17465 PR rtl-optimization/83985
17466 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
17467 REG_CFA_RESTORE insns.
17468 (delete_unmarked_insns): Don't ignore separate shrink wrapping
17469 REG_CFA_RESTORE insns here.
17472 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
17473 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
17475 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
17477 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
17478 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
17479 (arc_init): Likewise.
17480 (arc_override_options): Likewise.
17481 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
17483 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
17485 * config/arc/arc.h (TARGET_DBNZ): Define.
17486 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
17487 properly set the tune attribute.
17488 (dbnz): Use TARGET_DBNZ guard.
17489 * config/arc/arc.opt (mtune): Add core3 option.
17491 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
17493 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
17494 recognize new pic like addresses.
17495 (arc_delegitimize_address): Clean up.
17497 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
17499 * config/arc/arc-arches.def: Option mrf16 valid for all
17501 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
17502 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
17503 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
17504 * config/arc/arc-tables.opt: Regenerate.
17505 * config/arc/arc.c (arc_conditional_register_usage): Handle
17506 reduced register file case.
17507 (arc_file_start): Set must have build attributes.
17508 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
17509 mrf16 option value.
17510 * config/arc/arc.opt (mrf16): Add new option.
17511 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
17512 * config/arc/genmultilib.awk: Handle new mrf16 option.
17513 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
17514 * config/arc/t-multilib: Regenerate.
17515 * doc/invoke.texi (ARC Options): Document mrf16 option.
17517 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
17519 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
17520 * config/arc/arc.c (arc_handle_secure_attribute): New function.
17521 (arc_attribute_table): Add 'secure_call' attribute.
17522 (arc_print_operand): Print secure call operand.
17523 (arc_function_ok_for_sibcall): Don't optimize tail calls when
17525 (arc_is_secure_call_p): New function. * config/arc/arc.md
17526 (call_i): Add support for sjli instruction.
17527 (call_value_i): Likewise.
17528 * config/arc/constraints.md (Csc): New constraint.
17530 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
17531 John Eric Martin <John.Martin@emmicro-us.com>
17533 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
17534 * config/arc/arc.c (_arc_jli_section): New struct.
17535 (arc_jli_section): New type.
17536 (rc_jli_sections): New static variable.
17537 (arc_handle_jli_attribute): New function.
17538 (arc_attribute_table): Add jli_always and jli_fixed attribute.
17539 (arc_file_end): New function.
17540 (TARGET_ASM_FILE_END): Define.
17541 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
17542 (arc_add_jli_section): New function.
17543 (jli_call_scan): Likewise.
17544 (arc_reorg): Call jli_call_scan.
17545 (arc_output_addsi): Remove 'S' from printing asm operand.
17546 (arc_is_jli_call_p): New function.
17547 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
17549 (movhi_insn): Likewise.
17550 (movsi_insn): Likewise.
17551 (movsi_set_cc_insn): Likewise.
17552 (loadqi_update): Likewise.
17553 (load_zeroextendqisi_update): Likewise.
17554 (load_signextendqisi_update): Likewise.
17555 (loadhi_update): Likewise.
17556 (load_zeroextendhisi_update): Likewise.
17557 (load_signextendhisi_update): Likewise.
17558 (loadsi_update): Likewise.
17559 (loadsf_update): Likewise.
17560 (movsicc_insn): Likewise.
17561 (bset_insn): Likewise.
17562 (bxor_insn): Likewise.
17563 (bclr_insn): Likewise.
17564 (bmsk_insn): Likewise.
17565 (bicsi3_insn): Likewise.
17566 (cmpsi_cc_c_insn): Likewise.
17567 (movsi_ne): Likewise.
17568 (movsi_cond_exec): Likewise.
17569 (clrsbsi2): Likewise.
17570 (norm_f): Likewise.
17580 (call_i): Remove 'S' asm letter, add jli instruction.
17581 (call_value_i): Likewise.
17582 * config/arc/arc.op (mjli-always): New option.
17583 * config/arc/constraints.md (Cji): New constraint.
17584 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
17586 (subsf3_fpx): Likewise.
17587 (mulsf3_fpx): Likewise.
17588 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
17590 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
17591 function attrbutes.
17592 * doc/invoke.texi (ARC): Document mjli-always option.
17594 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
17596 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
17597 avoid addition with 0 and use incw and decw where possible.
17599 2018-01-26 Richard Biener <rguenther@suse.de>
17601 PR tree-optimization/81082
17602 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
17603 association if it requires casting to unsigned.
17604 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
17605 from fold_plusminus_mult_expr to catch important cases late when
17606 range info is available.
17608 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
17610 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
17611 * configure.ac (hidden_linkonce): New test.
17612 * configure: Regenerate.
17613 * config.in: Regenerate.
17615 2018-01-26 Julia Koval <julia.koval@intel.com>
17617 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
17618 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
17619 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
17620 _mm_mask_bitshuffle_epi64_mask): Fix type.
17621 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
17622 USI_FTYPE_V4DI_V4DI_USI): Remove.
17623 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
17624 __builtin_ia32_vpshufbitqmb256_mask,
17625 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
17626 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
17627 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
17629 2018-01-26 Alan Modra <amodra@gmail.com>
17632 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
17633 UNSPEC_VBPERMQ. Sort other unspecs.
17635 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
17637 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
17639 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
17641 PR middle-end/83055
17642 * predict.c (drop_profile): Do not push/pop cfun; update also
17644 (handle_missing_profiles): Fix logic looking for zero profiles.
17646 2018-01-25 Jakub Jelinek <jakub@redhat.com>
17648 PR middle-end/83977
17649 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
17650 on functions with #pragma omp declare simd or functions with simd
17652 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
17653 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
17654 Remove trailing \n from warning_at calls.
17656 2018-01-25 Tom de Vries <tom@codesourcery.com>
17659 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
17660 for neutered workers.
17662 2018-01-24 Joseph Myers <joseph@codesourcery.com>
17665 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
17666 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
17668 2018-01-24 Jeff Law <law@redhat.com>
17671 * i386.c (get_probe_interval): Move to earlier point.
17672 (ix86_compute_frame_layout): If -fstack-clash-protection and
17673 the frame is larger than the probe interval, then use pushes
17674 to save registers rather than reg->mem moves.
17675 (ix86_expand_prologue): Remove conditional for int_registers_saved
17678 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
17681 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
17682 min/max for never referenced object.
17684 2018-01-24 Jakub Jelinek <jakub@redhat.com>
17686 PR middle-end/83977
17687 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
17689 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
17690 attributes from DECL_ATTRIBUTES (decl) without affecting
17691 DECL_ATTRIBUTES (current_function_decl).
17692 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
17693 functions with non-NULL DECL_ABSTRACT_ORIGIN.
17695 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
17697 PR tree-optimization/83979
17698 * fold-const.c (fold_comparison): Use constant_boolean_node
17699 instead of boolean_{true,false}_node.
17701 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
17703 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
17706 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
17708 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
17709 Simplify the clause that sets the length attribute.
17710 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
17711 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
17712 clause that sets the length attribute.
17713 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
17715 2018-01-24 Tom de Vries <tom@codesourcery.com>
17718 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
17719 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
17720 Add strict parameter.
17721 (prevent_branch_around_nothing): Insert dummy insn between branch to
17722 label and label with no ptx insn inbetween.
17723 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
17725 2018-01-24 Tom de Vries <tom@codesourcery.com>
17728 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
17729 for neutered threads in warp.
17730 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
17732 2018-01-24 Richard Biener <rguenther@suse.de>
17734 PR tree-optimization/83176
17735 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
17738 2018-01-24 Richard Biener <rguenther@suse.de>
17740 PR tree-optimization/82819
17741 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
17742 code generating pluses that are no-ops in the target precision.
17744 2018-01-24 Richard Biener <rguenther@suse.de>
17746 PR middle-end/84000
17747 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
17749 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
17751 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
17752 to merge probabilities.
17753 * predict.c (probably_never_executed): Also mark as cold functions
17754 with global 0 profile and guessed local profile.
17755 * profile-count.c (profile_probability::combine_with_count): New
17757 * profile-count.h (profile_probability::operator*,
17758 profile_probability::operator*=, profile_probability::operator/,
17759 profile_probability::operator/=): Reduce precision to adjusted
17760 and set value to guessed on contradictory divisions.
17761 (profile_probability::combine_with_freq): Remove.
17762 (profile_probability::combine_wiht_count): Declare.
17763 (profile_count::force_nonzero):: Set to adjusted.
17764 (profile_count::probability_in):: Set quality to adjusted.
17765 * tree-ssa-tail-merge.c (replace_block_by): Use
17766 combine_with_count.
17768 2018-01-23 Andrew Waterman <andrew@sifive.com>
17769 Jim Wilson <jimw@sifive.com>
17771 * config/riscv/riscv.c (riscv_stack_boundary): New.
17772 (riscv_option_override): Set riscv_stack_boundary. Handle
17773 riscv_preferred_stack_boundary_arg.
17774 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
17775 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
17776 (STACK_BOUNDARY): Set to riscv_stack_boundary.
17777 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
17778 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
17779 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
17781 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
17784 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
17785 of struct ix86_frame.
17786 (ix86_expand_epilogue): Likewise. Add a local variable for
17787 the reg_save_offset field in struct ix86_frame.
17789 2018-01-23 Bin Cheng <bin.cheng@arm.com>
17791 PR tree-optimization/82604
17792 * tree-loop-distribution.c (enum partition_kind): New enum item
17793 PKIND_PARTIAL_MEMSET.
17794 (partition_builtin_p): Support above new enum item.
17795 (generate_code_for_partition): Ditto.
17796 (compute_access_range): Differentiate cases that equality can be
17797 proven at all loops, the innermost loops or no loops.
17798 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
17799 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
17800 (finalize_partitions, distribute_loop): Don't fuse partition of
17801 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
17802 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
17803 parloop is enabled.
17805 2018-01-23 Martin Liska <mliska@suse.cz>
17807 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
17808 order to ignore the predictor.
17809 (PRED_POLYMORPHIC_CALL): Likewise.
17810 (PRED_RECURSIVE_CALL): Likewise.
17812 2018-01-23 Martin Liska <mliska@suse.cz>
17814 * tree-profile.c (tree_profiling): Print function header to
17815 aware reader which function we are working on.
17816 * value-prof.c (gimple_find_values_to_profile): Do not print
17817 not interesting value histograms.
17819 2018-01-23 Martin Liska <mliska@suse.cz>
17821 * profile-count.h (enum profile_quality): Add
17822 profile_uninitialized as the first value. Do not number values
17823 as they are zero based.
17824 (profile_count::verify): Update sanity check.
17825 (profile_probability::verify): Likewise.
17827 2018-01-23 Nathan Sidwell <nathan@acm.org>
17829 * doc/invoke.texi (ffor-scope): Deprecate.
17831 2018-01-23 David Malcolm <dmalcolm@redhat.com>
17833 PR tree-optimization/83510
17834 * domwalk.c (set_all_edges_as_executable): New function.
17835 (dom_walker::dom_walker): Convert bool param
17836 "skip_unreachable_blocks" to enum reachability. Move setup of
17837 edge flags to set_all_edges_as_executable and only do it when
17838 reachability is REACHABLE_BLOCKS.
17839 * domwalk.h (enum dom_walker::reachability): New enum.
17840 (dom_walker::dom_walker): Convert bool param
17841 "skip_unreachable_blocks" to enum reachability.
17842 (set_all_edges_as_executable): New decl.
17843 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
17844 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
17846 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
17847 but converting true to REACHABLE_BLOCKS.
17848 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
17850 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
17851 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
17852 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
17854 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
17855 if check_all_array_refs will be called.
17857 2018-01-23 David Malcolm <dmalcolm@redhat.com>
17859 * tree.c (selftest::test_location_wrappers): Add more test
17862 2018-01-23 David Malcolm <dmalcolm@redhat.com>
17864 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
17865 (selftest::test_bit_in_range): Likewise.
17867 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
17870 * doc/sourcebuild.texi (vect_float): Say that the selector
17871 only describes the situation when -funsafe-math-optimizations is on.
17872 (vect_float_strict): Document.
17874 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
17876 PR tree-optimization/83965
17877 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
17878 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
17879 instead of checking only for a reduction.
17880 (vect_recog_widen_sum_pattern): Likewise.
17882 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
17884 * predict.c (probably_never_executed): Only use precise profile info.
17885 (compute_function_frequency): Skip after inlining hack since we now
17886 have quality checking.
17888 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
17890 * profile-count.h (profile_probability::very_unlikely,
17891 profile_probability::unlikely, profile_probability::even): Set
17892 precision to guessed.
17894 2018-01-23 Richard Biener <rguenther@suse.de>
17896 PR tree-optimization/83963
17897 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
17898 Properly terminate dominator walk when crossing the exit edge not
17899 when visiting its source block.
17901 2018-01-23 Jakub Jelinek <jakub@redhat.com>
17904 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
17905 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
17907 2018-01-22 Jakub Jelinek <jakub@redhat.com>
17909 PR tree-optimization/83957
17910 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
17911 semicolon after for body surrounded by braces.
17913 PR tree-optimization/83081
17914 * profile-count.h (profile_probability::split): New method.
17915 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
17916 Use profile_probability::split.
17917 (do_compare_rtx_and_jump): Fix adjustment of probabilities
17918 when splitting a single conditional jump into 2.
17920 2018-01-22 David Malcolm <dmalcolm@redhat.com>
17922 PR tree-optimization/69452
17923 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
17926 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
17928 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
17929 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
17930 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
17932 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
17934 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
17936 * config/rl78/rl78.md (movdi): New define_expand.
17937 * config/rl78/rl78.c (rl78_split_movdi): New function.
17939 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
17942 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
17944 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
17945 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
17946 128-bit to produce an UNSPEC move to get the double word with the
17947 signbit and then a shift directly to do signbit.
17948 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
17949 implementation with a new version that just does either a direct
17950 move or a regular move. Move memory interface to separate insns.
17951 Move insns so they are next to the expander.
17952 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
17953 with signbit move. Split big and little endian case.
17954 (signbit<mode>2_dm_mem_le): Likewise.
17955 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
17956 (signbit<mode>2_dm2): Likewise.
17958 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
17960 * config/rl78/rl78.md (anddi3): New define_expand.
17962 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
17964 * config/rl78/rl78.md (umindi3): New define_expand.
17966 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
17968 * config/rl78/rl78.md (smindi3): New define_expand.
17970 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
17972 * config/rl78/rl78.md (smaxdi3): New define_expand.
17974 2018-01-22 Carl Love <cel@us.ibm.com>
17976 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
17977 LVX_V1TI): Add macro expansion.
17978 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
17979 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
17980 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
17981 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
17982 Change check to determine if the instruction is a byte reversing
17983 entry. Fix typo in comment.
17984 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
17985 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
17986 Add def_builtin calls for new builtins.
17987 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
17988 Add define_insn expansion.
17990 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
17992 * config/rl78/rl78.md (umaxdi3): New define_expand.
17994 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
17996 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
17997 for non-QImode registers.
17999 2018-01-22 Richard Biener <rguenther@suse.de>
18001 PR tree-optimization/83963
18002 * graphite-scop-detection.c (scop_detection::get_sese): Delay
18003 including the loop exit block.
18004 (scop_detection::merge_sese): Likewise.
18005 (scop_detection::add_scop): Do it here instead.
18007 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18009 * doc/sourcebuild.texi (arm_softfloat): Document.
18011 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
18014 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
18015 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
18016 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
18018 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
18019 David Edelsohn <dje.gcc@gmail.com>
18022 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
18023 Change "crset eq" to "crset 2".
18024 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
18025 (*call_indirect_aix<mode>_nospec): Likewise.
18026 (*call_value_indirect_aix<mode>_nospec): Likewise.
18027 (*call_indirect_elfv2<mode>_nospec): Likewise.
18028 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
18029 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
18030 change assembly output from . to $.
18031 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
18032 (indirect_jump<mode>_nospec): Change assembly output from . to $.
18033 (*tablejump<mode>_internal1_nospec): Likewise.
18035 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
18038 * config/sh/sh_optimize_sett_clrt.cc:
18039 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
18041 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
18043 PR tree-optimization/83940
18044 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
18045 offset_dt to vect_constant_def rather than vect_unknown_def_type.
18046 (vect_check_load_store_mask): Add a mask_dt_out parameter and
18047 use it to pass back the definition type.
18048 (vect_check_store_rhs): Likewise rhs_dt_out.
18049 (vect_build_gather_load_calls): Add a mask_dt argument and use
18050 it instead of a call to vect_is_simple_use.
18051 (vectorizable_store): Update calls to vect_check_load_store_mask
18052 and vect_check_store_rhs. Use the dt returned by the latter instead
18053 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
18054 instead of calls to vect_is_simple_use. Pass the scalar rather
18055 than the vector operand to vect_is_simple_use when handling
18056 second and subsequent copies of an rhs value.
18057 (vectorizable_load): Update calls to vect_check_load_store_mask
18058 and vect_build_gather_load_calls. Use the cached mask_dt and
18059 gs_info.offset_dt instead of calls to vect_is_simple_use.
18061 2018-01-20 Jakub Jelinek <jakub@redhat.com>
18063 PR middle-end/83945
18064 * tree-emutls.c: Include gimplify.h.
18065 (lower_emutls_2): New function.
18066 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
18067 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
18068 it before further processing.
18071 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
18072 UINTVAL (trueop1) instead of INTVAL (op1).
18074 2018-01-19 Jakub Jelinek <jakub@redhat.com>
18078 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
18079 INCOMING_FRAME_SP_OFFSET if not defined.
18080 (scan_trace): Add ENTRY argument. If true and
18081 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
18082 emit a note to adjust the CFA offset.
18083 (create_cfi_notes): Adjust scan_trace callers.
18084 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
18085 INCOMING_FRAME_SP_OFFSET in the CIE.
18086 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
18087 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
18089 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
18090 * doc/tm.texi: Regenerated.
18092 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
18094 PR rtl-optimization/83147
18095 * lra-constraints.c (remove_inheritance_pseudos): Use
18096 lra_substitute_pseudo_within_insn.
18098 2018-01-19 Tom de Vries <tom@codesourcery.com>
18099 Cesar Philippidis <cesar@codesourcery.com>
18102 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
18104 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
18107 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
18108 spaces for function labels.
18110 2018-01-19 Martin Liska <mliska@suse.cz>
18112 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
18113 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
18114 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
18115 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
18116 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
18117 (PRED_CONST_RETURN): Change from 69 to 65.
18118 (PRED_NULL_RETURN): Change from 91 to 71.
18119 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
18120 (PRED_LOOP_GUARD): Change from 66 to 73.
18122 2018-01-19 Martin Liska <mliska@suse.cz>
18124 * predict.c (predict_insn_def): Add new assert.
18125 (struct branch_predictor): Change type to signed integer.
18126 (test_prediction_value_range): Amend test to cover
18127 PROB_UNINITIALIZED.
18128 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
18129 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
18130 (PRED_LOOP_ITERATIONS_MAX): Likewise.
18131 (PRED_LOOP_IV_COMPARE): Likewise.
18132 * predict.h (PROB_UNINITIALIZED): Define new constant.
18134 2018-01-19 Martin Liska <mliska@suse.cz>
18136 * predict.c (dump_prediction): Add new format for
18137 analyze_brprob.py script which is enabled with -details
18139 * profile-count.h (precise_p): New function.
18141 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
18143 PR tree-optimization/83922
18144 * tree-vect-loop.c (vect_verify_full_masking): Return false if
18145 there are no statements that need masking.
18146 (vect_active_double_reduction_p): New function.
18147 (vect_analyze_loop_operations): Use it when handling phis that
18148 are not in the loop header.
18150 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
18152 PR tree-optimization/83914
18153 * tree-vect-loop.c (vectorizable_induction): Don't convert
18154 init_expr or apply the peeling adjustment for inductions
18155 that are nested within the vectorized loop.
18157 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18159 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
18162 2018-01-18 Jakub Jelinek <jakub@redhat.com>
18166 * function.h (gimplify_parameters): Add gimple_seq * argument.
18167 * function.c: Include gimple.h and options.h.
18168 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
18169 for the added local temporaries if needed.
18170 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
18171 if there are any parameter cleanups, wrap whole body into a
18172 try/finally with the cleanups.
18174 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
18177 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
18178 Use GET_MODE_CLASS for scalar floating point.
18180 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
18184 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
18185 Fix call of call_cgraph_insertion_hooks.
18187 2018-01-18 Martin Sebor <msebor@redhat.com>
18189 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
18191 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
18194 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
18197 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
18200 * common.opt: (-ffile-prefix-map): New option.
18201 * opts.c (common_handle_option): Defer it.
18202 * opts-global.c (handle_common_deferred_options): Handle it.
18203 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
18204 * file-prefix-map.h: New file.
18205 (remap_debug_filename, add_debug_prefix_map): ...here.
18206 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
18207 * final.c (debug_prefix_map, add_debug_prefix_map
18208 remap_debug_filename): Move to...
18209 * file-prefix-map.c: New file.
18210 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
18211 generalize, get rid of alloca(), use strrchr() instead of strchr().
18212 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
18213 Implement in terms of add_prefix_map().
18214 (remap_macro_filename, remap_debug_filename): Implement in term of
18216 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
18217 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
18218 * dbxout.c: Include file-prefix-map.h.
18219 * varasm.c: Likewise.
18220 * vmsdbgout.c: Likewise.
18221 * xcoffout.c: Likewise.
18222 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
18223 * doc/cppopts.texi (-fmacro-prefix-map): Document.
18224 * doc/invoke.texi (-ffile-prefix-map): Document.
18225 (-fdebug-prefix-map): Update description.
18227 2018-01-18 Martin Liska <mliska@suse.cz>
18229 * config/i386/i386.c (indirect_thunk_name): Document that also
18231 (output_indirect_thunk): Document why both instructions
18232 (pause and lfence) are generated.
18234 2018-01-18 Richard Biener <rguenther@suse.de>
18236 PR tree-optimization/83887
18237 * graphite-scop-detection.c
18238 (scop_detection::get_nearest_dom_with_single_entry): Remove.
18239 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
18240 (scop_detection::merge_sese): Re-implement with a flood-fill
18241 algorithm that properly finds a SESE region if it exists.
18243 2018-01-18 Jakub Jelinek <jakub@redhat.com>
18246 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
18247 pointer_diff optimizations use view_convert instead of convert.
18249 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
18251 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
18252 Generate different code for -mno-speculate-indirect-jumps.
18253 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
18254 (*call_indirect_aix<mode>): Disable for
18255 -mno-speculate-indirect-jumps.
18256 (*call_indirect_aix<mode>_nospec): New define_insn.
18257 (*call_value_indirect_aix<mode>): Disable for
18258 -mno-speculate-indirect-jumps.
18259 (*call_value_indirect_aix<mode>_nospec): New define_insn.
18260 (*sibcall_nonlocal_sysv<mode>): Generate different code for
18261 -mno-speculate-indirect-jumps.
18262 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
18264 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
18266 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
18267 long double type, set the flags for noting the default long double
18268 type, even if we don't pass or return a long double type.
18270 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
18273 * ipa-inline.c (flatten_function): Do not overwrite final inlining
18276 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
18278 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
18279 support for merge[hl].
18280 (fold_mergehl_helper): New helper function.
18281 (tree-vector-builder.h): New #include for tree_vector_builder usage.
18282 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
18283 (altivec_vmrglw_direct): Add xxmrglw insn.
18285 2018-01-17 Andrew Waterman <andrew@sifive.com>
18287 * config/riscv/riscv.c (riscv_conditional_register_usage): If
18288 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
18290 2018-01-17 David Malcolm <dmalcolm@redhat.com>
18293 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
18294 call the lto_location_cache before reading the
18295 DECL_SOURCE_LOCATION of the types.
18297 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
18298 Richard Sandiford <richard.sandiford@linaro.org>
18300 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
18301 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
18302 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
18303 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
18304 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
18306 * config/aarch64/constraints.md (aarch64_movti_operand):
18308 * config/aarch64/predicates.md (Uti): Add new constraint.
18310 2018-01-17 Carl Love <cel@us.ibm.com>
18312 * config/rs6000/vsx.md (define_expand xl_len_r,
18313 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
18314 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
18316 (define_expand, define_insn): Move the shift left from the
18317 define_insn to the define_expand for lxvl and stxvl instructions.
18318 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
18319 and XL_LEN_R definitions to PURE.
18321 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
18323 * config/i386/i386.c (indirect_thunk_name): Declare regno
18324 as unsigned int. Compare regno with INVALID_REGNUM.
18325 (output_indirect_thunk): Ditto.
18326 (output_indirect_thunk_function): Ditto.
18327 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
18328 in the call to output_indirect_thunk_function.
18330 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
18332 PR middle-end/83884
18333 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
18334 rather than the size of inner_type to determine the stack slot size
18335 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
18337 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
18340 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
18343 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
18345 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
18346 endian Linux systems to optionally enable multilibs for selecting
18347 the long double type if the user configured an explicit type.
18348 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
18349 have no long double multilibs if not defined.
18350 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
18351 warn if the user used -mabi={ieee,ibm}longdouble and we built
18352 multilibs for long double.
18353 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
18354 appropriate multilib option.
18355 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
18357 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
18358 for building long double multilibs.
18359 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
18361 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
18363 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
18366 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
18368 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
18371 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
18374 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
18377 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
18379 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
18380 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
18382 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
18384 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
18385 different rtl trees depending on TARGET_64BIT.
18386 (rs6000_gen_lvx): Likewise.
18388 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
18390 * config/visium/visium.md (nop): Tweak comment.
18391 (hazard_nop): Likewise.
18393 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
18395 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
18396 -mspeculate-indirect-jumps.
18397 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
18398 for -mno-speculate-indirect-jumps.
18399 (*call_indirect_elfv2<mode>_nospec): New define_insn.
18400 (*call_value_indirect_elfv2<mode>): Disable for
18401 -mno-speculate-indirect-jumps.
18402 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
18403 (indirect_jump): Emit different RTL for
18404 -mno-speculate-indirect-jumps.
18405 (*indirect_jump<mode>): Disable for
18406 -mno-speculate-indirect-jumps.
18407 (*indirect_jump<mode>_nospec): New define_insn.
18408 (tablejump): Emit different RTL for
18409 -mno-speculate-indirect-jumps.
18410 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
18411 (tablejumpsi_nospec): New define_expand.
18412 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
18413 (tablejumpdi_nospec): New define_expand.
18414 (*tablejump<mode>_internal1): Disable for
18415 -mno-speculate-indirect-jumps.
18416 (*tablejump<mode>_internal1_nospec): New define_insn.
18417 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
18420 2018-01-16 Artyom Skrobov tyomitch@gmail.com
18422 * caller-save.c (insert_save): Drop unnecessary parameter. All
18425 2018-01-16 Jakub Jelinek <jakub@redhat.com>
18426 Richard Biener <rguenth@suse.de>
18429 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
18430 return early, inline manually is_gimple_sizepos. Make sure if we
18431 call gimplify_expr we don't end up with a gimple constant.
18432 * tree.c (variably_modified_type_p): Don't return true for
18433 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
18434 * gimplify.h (is_gimple_sizepos): Remove.
18436 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
18438 PR tree-optimization/83857
18439 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
18440 vectorizable_live_operation for pure SLP statements.
18441 (vectorizable_live_operation): Handle PHIs.
18443 2018-01-16 Richard Biener <rguenther@suse.de>
18445 PR tree-optimization/83867
18446 * tree-vect-stmts.c (vect_transform_stmt): Precompute
18447 nested_in_vect_loop_p since the scalar stmt may get invalidated.
18449 2018-01-16 Jakub Jelinek <jakub@redhat.com>
18452 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
18453 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
18454 If off is not INTEGER_CST, issue a may not be aligned warning
18455 rather than isn't aligned. Use isn%'t rather than isn't.
18456 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
18458 <case MULT_EXPR>: Improve the case when bottom and one of the
18459 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
18460 operand, in that case check if the other operand is multiple of
18461 bottom divided by the INTEGER_CST operand.
18463 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
18466 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
18467 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
18468 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
18469 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
18470 * config/pa/pa.c (pa_function_arg_advance): Likewise.
18471 (pa_function_arg, pa_arg_partial_bytes): Likewise.
18472 (pa_function_arg_size): New function.
18474 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
18476 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
18477 in a separate statement.
18479 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
18481 PR tree-optimization/83847
18482 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
18483 group gathers and scatters.
18485 2018-01-16 Jakub Jelinek <jakub@redhat.com>
18487 PR rtl-optimization/86620
18488 * params.def (max-sched-ready-insns): Bump minimum value to 1.
18490 PR rtl-optimization/83213
18491 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
18492 to last if both are JUMP_INSNs.
18494 PR tree-optimization/83843
18495 * gimple-ssa-store-merging.c
18496 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
18497 store_immediate_info for bswap/nop orig_stores.
18499 2018-01-15 Andrew Waterman <andrew@sifive.com>
18501 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
18503 <UDIV>: Increase cost if !TARGET_DIV.
18505 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
18507 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
18508 (define_attr "cr_logical_3op"): New.
18509 (cceq_ior_compare): Adjust.
18510 (cceq_ior_compare_complement): Adjust.
18511 (*cceq_rev_compare): Adjust.
18512 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
18513 (is_cracked_insn): Adjust.
18514 (insn_must_be_first_in_group): Adjust.
18515 * config/rs6000/40x.md: Adjust.
18516 * config/rs6000/440.md: Adjust.
18517 * config/rs6000/476.md: Adjust.
18518 * config/rs6000/601.md: Adjust.
18519 * config/rs6000/603.md: Adjust.
18520 * config/rs6000/6xx.md: Adjust.
18521 * config/rs6000/7450.md: Adjust.
18522 * config/rs6000/7xx.md: Adjust.
18523 * config/rs6000/8540.md: Adjust.
18524 * config/rs6000/cell.md: Adjust.
18525 * config/rs6000/e300c2c3.md: Adjust.
18526 * config/rs6000/e500mc.md: Adjust.
18527 * config/rs6000/e500mc64.md: Adjust.
18528 * config/rs6000/e5500.md: Adjust.
18529 * config/rs6000/e6500.md: Adjust.
18530 * config/rs6000/mpc.md: Adjust.
18531 * config/rs6000/power4.md: Adjust.
18532 * config/rs6000/power5.md: Adjust.
18533 * config/rs6000/power6.md: Adjust.
18534 * config/rs6000/power7.md: Adjust.
18535 * config/rs6000/power8.md: Adjust.
18536 * config/rs6000/power9.md: Adjust.
18537 * config/rs6000/rs64.md: Adjust.
18538 * config/rs6000/titan.md: Adjust.
18540 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
18542 * config/i386/predicates.md (indirect_branch_operand): Rewrite
18543 ix86_indirect_branch_register logic.
18545 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
18547 * config/i386/constraints.md (Bs): Update
18548 ix86_indirect_branch_register check. Don't check
18549 ix86_indirect_branch_register with GOT_memory_operand.
18551 * config/i386/predicates.md (GOT_memory_operand): Don't check
18552 ix86_indirect_branch_register here.
18553 (GOT32_symbol_operand): Likewise.
18555 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
18557 * config/i386/predicates.md (constant_call_address_operand):
18558 Rewrite ix86_indirect_branch_register logic.
18559 (sibcall_insn_operand): Likewise.
18561 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
18563 * config/i386/constraints.md (Bs): Replace
18564 ix86_indirect_branch_thunk_register with
18565 ix86_indirect_branch_register.
18567 * config/i386/i386.md (indirect_jump): Likewise.
18568 (tablejump): Likewise.
18569 (*sibcall_memory): Likewise.
18570 (*sibcall_value_memory): Likewise.
18571 Peepholes of indirect call and jump via memory: Likewise.
18572 * config/i386/i386.opt: Likewise.
18573 * config/i386/predicates.md (indirect_branch_operand): Likewise.
18574 (GOT_memory_operand): Likewise.
18575 (call_insn_operand): Likewise.
18576 (sibcall_insn_operand): Likewise.
18577 (GOT32_symbol_operand): Likewise.
18579 2018-01-15 Jakub Jelinek <jakub@redhat.com>
18581 PR middle-end/83837
18582 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
18583 type rather than type addr's type points to.
18584 (expand_omp_atomic_mutex): Likewise.
18585 (expand_omp_atomic): Likewise.
18587 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
18590 * config/i386/i386.c (output_indirect_thunk_function): Use
18591 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
18592 for __x86_return_thunk.
18594 2018-01-15 Richard Biener <rguenther@suse.de>
18596 PR middle-end/83850
18597 * expmed.c (extract_bit_field_1): Fix typo.
18599 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18602 * config/arm/iterators.md (VF): New mode iterator.
18603 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
18604 Remove integer-related logic from pattern.
18605 (neon_vabd<mode>_3): Likewise.
18607 2018-01-15 Jakub Jelinek <jakub@redhat.com>
18609 PR middle-end/82694
18610 * common.opt (fstrict-overflow): No longer an alias.
18611 (fwrapv-pointer): New option.
18612 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
18613 also for pointer types based on flag_wrapv_pointer.
18614 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
18615 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
18616 opts->x_flag_wrapv got set.
18617 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
18618 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
18619 POINTER_TYPE_OVERFLOW_UNDEFINED.
18620 * match.pd: Likewise in address comparison pattern.
18621 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
18623 2018-01-15 Richard Biener <rguenther@suse.de>
18626 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
18627 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
18628 Reset type names to their identifier if their TYPE_DECL doesn't
18629 have linkage (and thus is used for ODR and devirt).
18630 (save_debug_info_for_decl): Remove.
18631 (save_debug_info_for_type): Likewise.
18632 (add_tree_to_fld_list): Adjust.
18633 * tree-pretty-print.c (dump_generic_node): Make dumping of
18634 type names more robust.
18636 2018-01-15 Richard Biener <rguenther@suse.de>
18638 * BASE-VER: Bump to 8.0.1.
18640 2018-01-14 Martin Sebor <msebor@redhat.com>
18643 * builtins.c (check_access): Avoid warning when the no-warning bit
18646 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
18648 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
18649 * ira-color (allocno_hard_regs_compare): Likewise.
18651 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
18654 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
18655 Use .pushsection/.popsection.
18657 2018-01-14 Martin Sebor <msebor@redhat.com>
18660 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
18662 2018-01-14 Jakub Jelinek <jakub@redhat.com>
18664 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
18665 entry from extra_headers.
18666 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
18667 extra_headers, make the list bitwise identical to the i?86-*-* one.
18669 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
18671 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
18672 -mcmodel=large with -mindirect-branch=thunk,
18673 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
18674 -mfunction-return=thunk-extern.
18675 * doc/invoke.texi: Document -mcmodel=large is incompatible with
18676 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
18677 -mfunction-return=thunk and -mfunction-return=thunk-extern.
18679 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
18681 * config/i386/i386.c (print_reg): Print the name of the full
18682 integer register without '%'.
18683 (ix86_print_operand): Handle 'V'.
18684 * doc/extend.texi: Document 'V' modifier.
18686 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
18688 * config/i386/constraints.md (Bs): Disallow memory operand for
18689 -mindirect-branch-register.
18691 * config/i386/predicates.md (indirect_branch_operand): Likewise.
18692 (GOT_memory_operand): Likewise.
18693 (call_insn_operand): Likewise.
18694 (sibcall_insn_operand): Likewise.
18695 (GOT32_symbol_operand): Likewise.
18696 * config/i386/i386.md (indirect_jump): Call convert_memory_address
18697 for -mindirect-branch-register.
18698 (tablejump): Likewise.
18699 (*sibcall_memory): Likewise.
18700 (*sibcall_value_memory): Likewise.
18701 Disallow peepholes of indirect call and jump via memory for
18702 -mindirect-branch-register.
18703 (*call_pop): Replace m with Bw.
18704 (*call_value_pop): Likewise.
18705 (*sibcall_pop_memory): Replace m with Bs.
18706 * config/i386/i386.opt (mindirect-branch-register): New option.
18707 * doc/invoke.texi: Document -mindirect-branch-register option.
18709 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
18711 * config/i386/i386-protos.h (ix86_output_function_return): New.
18712 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
18713 set function_return_type.
18714 (indirect_thunk_name): Add ret_p to indicate thunk for function
18716 (output_indirect_thunk_function): Pass false to
18717 indirect_thunk_name.
18718 (ix86_output_indirect_branch_via_reg): Likewise.
18719 (ix86_output_indirect_branch_via_push): Likewise.
18720 (output_indirect_thunk_function): Create alias for function
18721 return thunk if regno < 0.
18722 (ix86_output_function_return): New function.
18723 (ix86_handle_fndecl_attribute): Handle function_return.
18724 (ix86_attribute_table): Add function_return.
18725 * config/i386/i386.h (machine_function): Add
18726 function_return_type.
18727 * config/i386/i386.md (simple_return_internal): Use
18728 ix86_output_function_return.
18729 (simple_return_internal_long): Likewise.
18730 * config/i386/i386.opt (mfunction-return=): New option.
18731 (indirect_branch): Mention -mfunction-return=.
18732 * doc/extend.texi: Document function_return function attribute.
18733 * doc/invoke.texi: Document -mfunction-return= option.
18735 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
18737 * config/i386/i386-opts.h (indirect_branch): New.
18738 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
18739 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
18740 with local indirect jump when converting indirect call and jump.
18741 (ix86_set_indirect_branch_type): New.
18742 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
18743 (indirectlabelno): New.
18744 (indirect_thunk_needed): Likewise.
18745 (indirect_thunk_bnd_needed): Likewise.
18746 (indirect_thunks_used): Likewise.
18747 (indirect_thunks_bnd_used): Likewise.
18748 (INDIRECT_LABEL): Likewise.
18749 (indirect_thunk_name): Likewise.
18750 (output_indirect_thunk): Likewise.
18751 (output_indirect_thunk_function): Likewise.
18752 (ix86_output_indirect_branch_via_reg): Likewise.
18753 (ix86_output_indirect_branch_via_push): Likewise.
18754 (ix86_output_indirect_branch): Likewise.
18755 (ix86_output_indirect_jmp): Likewise.
18756 (ix86_code_end): Call output_indirect_thunk_function if needed.
18757 (ix86_output_call_insn): Call ix86_output_indirect_branch if
18759 (ix86_handle_fndecl_attribute): Handle indirect_branch.
18760 (ix86_attribute_table): Add indirect_branch.
18761 * config/i386/i386.h (machine_function): Add indirect_branch_type
18762 and has_local_indirect_jump.
18763 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
18765 (tablejump): Likewise.
18766 (*indirect_jump): Use ix86_output_indirect_jmp.
18767 (*tablejump_1): Likewise.
18768 (simple_return_indirect_internal): Likewise.
18769 * config/i386/i386.opt (mindirect-branch=): New option.
18770 (indirect_branch): New.
18773 (thunk-inline): Likewise.
18774 (thunk-extern): Likewise.
18775 * doc/extend.texi: Document indirect_branch function attribute.
18776 * doc/invoke.texi: Document -mindirect-branch= option.
18778 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
18781 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
18783 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
18785 * ipa-inline.c (want_inline_small_function_p): Return false if
18786 inlining has already failed with CIF_FINAL_ERROR.
18787 (update_caller_keys): Call want_inline_small_function_p before
18789 (update_callee_keys): Likewise.
18791 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
18793 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
18795 (rs6000_quadword_masked_address_p): Likewise.
18796 (quad_aligned_load_p): Likewise.
18797 (quad_aligned_store_p): Likewise.
18798 (const_load_sequence_p): Add comment to describe the outer-most loop.
18799 (mimic_memory_attributes_and_flags): New function.
18800 (rs6000_gen_stvx): Likewise.
18801 (replace_swapped_aligned_store): Likewise.
18802 (rs6000_gen_lvx): Likewise.
18803 (replace_swapped_aligned_load): Likewise.
18804 (replace_swapped_load_constant): Capitalize argument name in
18805 comment describing this function.
18806 (rs6000_analyze_swaps): Add a third pass to search for vector loads
18807 and stores that access quad-word aligned addresses and replace
18808 with stvx or lvx instructions when appropriate.
18809 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
18810 New function prototype.
18811 (rs6000_quadword_masked_address_p): Likewise.
18812 (rs6000_gen_lvx): Likewise.
18813 (rs6000_gen_stvx): Likewise.
18814 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
18815 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
18816 when memory address is aligned.
18817 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
18818 this split to select lvx instruction when memory address is aligned.
18819 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
18820 instruction when memory address is aligned.
18821 (*vsx_le_perm_load_v16qi): Likewise.
18822 (four unnamed splitters): Modify to select the stvx instruction
18823 when memory is aligned.
18825 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
18827 * predict.c (determine_unlikely_bbs): Handle correctly BBs
18828 which appears in the queue multiple times.
18830 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18831 Alan Hayward <alan.hayward@arm.com>
18832 David Sherwood <david.sherwood@arm.com>
18834 * tree-vectorizer.h (vec_lower_bound): New structure.
18835 (_loop_vec_info): Add check_nonzero and lower_bounds.
18836 (LOOP_VINFO_CHECK_NONZERO): New macro.
18837 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
18838 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
18839 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
18840 fields. Make seg_len the distance travelled, not including the
18842 (dr_direction_indicator): Declare.
18843 (dr_zero_step_indicator): Likewise.
18844 (dr_known_forward_stride_p): Likewise.
18845 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
18847 (runtime_alias_check_p): Allow runtime alias checks with
18849 (operator ==): Compare access_size and align.
18850 (prune_runtime_alias_test_list): Rework for new distinction between
18851 the access_size and seg_len.
18852 (create_intersect_range_checks_index): Likewise. Cope with polynomial
18854 (get_segment_min_max): New function.
18855 (create_intersect_range_checks): Use it.
18856 (dr_step_indicator): New function.
18857 (dr_direction_indicator): Likewise.
18858 (dr_zero_step_indicator): Likewise.
18859 (dr_known_forward_stride_p): Likewise.
18860 * tree-loop-distribution.c (data_ref_segment_size): Return
18861 DR_STEP * (niters - 1).
18862 (compute_alias_check_pairs): Update call to the dr_with_seg_len
18864 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
18865 (vect_preserves_scalar_order_p): New function, split out from...
18866 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
18867 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
18868 (vect_vfa_access_size): New function.
18869 (vect_vfa_align): Likewise.
18870 (vect_compile_time_alias): Take access_size_a and access_b arguments.
18871 (dump_lower_bound): New function.
18872 (vect_check_lower_bound): Likewise.
18873 (vect_small_gap_p): Likewise.
18874 (vectorizable_with_step_bound_p): Likewise.
18875 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
18876 depencies if the vectorization factor is 1. Convert the checks
18877 for nonzero steps into checks on the bounds of DR_STEP. Try using
18878 a bunds check for variable steps if the minimum required step is
18879 relatively small. Update calls to the dr_with_seg_len
18880 constructor and to vect_compile_time_alias.
18881 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
18883 (vect_loop_versioning): Call it.
18884 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
18886 (vect_estimate_min_profitable_iters): Account for any bounds checks.
18888 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18889 Alan Hayward <alan.hayward@arm.com>
18890 David Sherwood <david.sherwood@arm.com>
18892 * doc/sourcebuild.texi (vect_scatter_store): Document.
18893 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
18895 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
18897 * genopinit.c (main): Add supports_vec_scatter_store and
18898 supports_vec_scatter_store_cached to target_optabs.
18899 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
18900 IFN_MASK_SCATTER_STORE.
18901 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
18903 * internal-fn.h (internal_store_fn_p): Declare.
18904 (internal_fn_stored_value_index): Likewise.
18905 * internal-fn.c (scatter_store_direct): New macro.
18906 (expand_scatter_store_optab_fn): New function.
18907 (direct_scatter_store_optab_supported_p): New macro.
18908 (internal_store_fn_p): New function.
18909 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
18910 IFN_MASK_SCATTER_STORE.
18911 (internal_fn_mask_index): Likewise.
18912 (internal_fn_stored_value_index): New function.
18913 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
18914 for scatter stores.
18915 * optabs-query.h (supports_vec_scatter_store_p): Declare.
18916 * optabs-query.c (supports_vec_scatter_store_p): New function.
18917 * tree-vectorizer.h (vect_get_store_rhs): Declare.
18918 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
18919 true for scatter stores.
18920 (vect_gather_scatter_fn_p): Handle scatter stores too.
18921 (vect_check_gather_scatter): Consider using scatter stores if
18922 supports_vec_scatter_store_p.
18923 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
18924 scatter stores too.
18925 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
18926 internal_fn_stored_value_index.
18927 (check_load_store_masking): Handle scatter stores too.
18928 (vect_get_store_rhs): Make public.
18929 (vectorizable_call): Use internal_store_fn_p.
18930 (vectorizable_store): Handle scatter store internal functions.
18931 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
18932 when deciding whether the end of the group has been reached.
18933 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
18934 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
18935 (mask_scatter_store<mode>): New insns.
18937 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18938 Alan Hayward <alan.hayward@arm.com>
18939 David Sherwood <david.sherwood@arm.com>
18941 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
18942 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
18943 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
18945 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
18946 Use vect_truncate_gather_scatter_offset if we can't treat the
18947 operation as a normal gather load or scatter store.
18948 (get_group_load_store_type): Take the gather_scatter_info
18949 as argument. Try using a gather load or scatter store for
18950 single-element groups.
18951 (get_load_store_type): Update calls to get_group_load_store_type
18952 and vect_use_strided_gather_scatters_p.
18954 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18955 Alan Hayward <alan.hayward@arm.com>
18956 David Sherwood <david.sherwood@arm.com>
18958 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
18959 optional tree argument.
18960 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
18962 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
18963 but continue to use the current value as a fallback.
18964 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
18965 to compare the updates.
18966 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
18967 (get_load_store_type): Use it when handling a strided access.
18968 (vect_get_strided_load_store_ops): New function.
18969 (vect_get_data_ptr_increment): Likewise.
18970 (vectorizable_load): Handle strided gather loads. Always pass
18971 a step to vect_create_data_ref_ptr and bump_vector_ptr.
18973 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18974 Alan Hayward <alan.hayward@arm.com>
18975 David Sherwood <david.sherwood@arm.com>
18977 * doc/md.texi (gather_load@var{m}): Document.
18978 (mask_gather_load@var{m}): Likewise.
18979 * genopinit.c (main): Add supports_vec_gather_load and
18980 supports_vec_gather_load_cached to target_optabs.
18981 * optabs-tree.c (init_tree_optimization_optabs): Use
18982 ggc_cleared_alloc to allocate target_optabs.
18983 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
18984 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
18986 * internal-fn.h (internal_load_fn_p): Declare.
18987 (internal_gather_scatter_fn_p): Likewise.
18988 (internal_fn_mask_index): Likewise.
18989 (internal_gather_scatter_fn_supported_p): Likewise.
18990 * internal-fn.c (gather_load_direct): New macro.
18991 (expand_gather_load_optab_fn): New function.
18992 (direct_gather_load_optab_supported_p): New macro.
18993 (direct_internal_fn_optab): New function.
18994 (internal_load_fn_p): Likewise.
18995 (internal_gather_scatter_fn_p): Likewise.
18996 (internal_fn_mask_index): Likewise.
18997 (internal_gather_scatter_fn_supported_p): Likewise.
18998 * optabs-query.c (supports_at_least_one_mode_p): New function.
18999 (supports_vec_gather_load_p): Likewise.
19000 * optabs-query.h (supports_vec_gather_load_p): Declare.
19001 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
19002 and memory_type field.
19003 (NUM_PATTERNS): Bump to 15.
19004 * tree-vect-data-refs.c: Include internal-fn.h.
19005 (vect_gather_scatter_fn_p): New function.
19006 (vect_describe_gather_scatter_call): Likewise.
19007 (vect_check_gather_scatter): Try using internal functions for
19008 gather loads. Recognize existing calls to a gather load function.
19009 (vect_analyze_data_refs): Consider using gather loads if
19010 supports_vec_gather_load_p.
19011 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
19012 (vect_get_gather_scatter_offset_type): Likewise.
19013 (vect_convert_mask_for_vectype): Likewise.
19014 (vect_add_conversion_to_patterm): Likewise.
19015 (vect_try_gather_scatter_pattern): Likewise.
19016 (vect_recog_gather_scatter_pattern): New pattern recognizer.
19017 (vect_vect_recog_func_ptrs): Add it.
19018 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
19019 internal_fn_mask_index and internal_gather_scatter_fn_p.
19020 (check_load_store_masking): Take the gather_scatter_info as an
19021 argument and handle gather loads.
19022 (vect_get_gather_scatter_ops): New function.
19023 (vectorizable_call): Check internal_load_fn_p.
19024 (vectorizable_load): Likewise. Handle gather load internal
19026 (vectorizable_store): Update call to check_load_store_masking.
19027 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
19028 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
19029 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
19030 (aarch64_gather_scale_operand_d): New predicates.
19031 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
19032 (mask_gather_load<mode>): New insns.
19034 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19035 Alan Hayward <alan.hayward@arm.com>
19036 David Sherwood <david.sherwood@arm.com>
19038 * optabs.def (fold_left_plus_optab): New optab.
19039 * doc/md.texi (fold_left_plus_@var{m}): Document.
19040 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
19041 * internal-fn.c (fold_left_direct): Define.
19042 (expand_fold_left_optab_fn): Likewise.
19043 (direct_fold_left_optab_supported_p): Likewise.
19044 * fold-const-call.c (fold_const_fold_left): New function.
19045 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
19046 * tree-parloops.c (valid_reduction_p): New function.
19047 (gather_scalar_reductions): Use it.
19048 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
19049 (vect_finish_replace_stmt): Declare.
19050 * tree-vect-loop.c (fold_left_reduction_fn): New function.
19051 (needs_fold_left_reduction_p): New function, split out from...
19052 (vect_is_simple_reduction): ...here. Accept reductions that
19053 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
19054 (vect_force_simple_reduction): Also store the reduction type in
19055 the assignment's STMT_VINFO_REDUC_TYPE.
19056 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
19057 (merge_with_identity): New function.
19058 (vect_expand_fold_left): Likewise.
19059 (vectorize_fold_left_reduction): Likewise.
19060 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
19061 scalar phi in place for it. Check for target support and reject
19062 cases that would reassociate the operation. Defer the transform
19063 phase to vectorize_fold_left_reduction.
19064 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
19065 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
19066 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
19068 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19070 * tree-if-conv.c (predicate_mem_writes): Remove redundant
19071 call to ifc_temp_var.
19073 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19074 Alan Hayward <alan.hayward@arm.com>
19075 David Sherwood <david.sherwood@arm.com>
19077 * target.def (legitimize_address_displacement): Take the original
19078 offset as a poly_int.
19079 * targhooks.h (default_legitimize_address_displacement): Update
19081 * targhooks.c (default_legitimize_address_displacement): Likewise.
19082 * doc/tm.texi: Regenerate.
19083 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
19084 as an argument, moving assert of ad->disp == ad->disp_term to...
19085 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
19086 Try calling targetm.legitimize_address_displacement before expanding
19087 the address rather than afterwards, and adjust for the new interface.
19088 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
19089 Match the new hook interface. Handle SVE addresses.
19090 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
19091 new hook interface.
19093 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19095 * Makefile.in (OBJS): Add early-remat.o.
19096 * target.def (select_early_remat_modes): New hook.
19097 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
19098 * doc/tm.texi: Regenerate.
19099 * targhooks.h (default_select_early_remat_modes): Declare.
19100 * targhooks.c (default_select_early_remat_modes): New function.
19101 * timevar.def (TV_EARLY_REMAT): New timevar.
19102 * passes.def (pass_early_remat): New pass.
19103 * tree-pass.h (make_pass_early_remat): Declare.
19104 * early-remat.c: New file.
19105 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
19107 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
19109 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19110 Alan Hayward <alan.hayward@arm.com>
19111 David Sherwood <david.sherwood@arm.com>
19113 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
19114 vfm1 with a bound_epilog parameter.
19115 (vect_do_peeling): Update calls accordingly, and move the prologue
19116 call earlier in the function. Treat the base bound_epilog as 0 for
19117 fully-masked loops and retain vf - 1 for other loops. Add 1 to
19118 this base when peeling for gaps.
19119 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
19120 with fully-masked loops.
19121 (vect_estimate_min_profitable_iters): Handle the single peeled
19122 iteration in that case.
19124 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19125 Alan Hayward <alan.hayward@arm.com>
19126 David Sherwood <david.sherwood@arm.com>
19128 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
19129 single-element interleaving even if the size is not a power of 2.
19130 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
19131 accesses for single-element interleaving if the group size is
19134 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19135 Alan Hayward <alan.hayward@arm.com>
19136 David Sherwood <david.sherwood@arm.com>
19138 * doc/md.texi (fold_extract_last_@var{m}): Document.
19139 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
19140 * optabs.def (fold_extract_last_optab): New optab.
19141 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
19142 * internal-fn.c (fold_extract_direct): New macro.
19143 (expand_fold_extract_optab_fn): Likewise.
19144 (direct_fold_extract_optab_supported_p): Likewise.
19145 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
19146 * tree-vect-loop.c (vect_model_reduction_cost): Handle
19147 EXTRACT_LAST_REDUCTION.
19148 (get_initial_def_for_reduction): Do not create an initial vector
19149 for EXTRACT_LAST_REDUCTION reductions.
19150 (vectorizable_reduction): Leave the scalar phi in place for
19151 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
19152 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
19153 epilogue code for EXTRACT_LAST_REDUCTION and defer the
19154 transform phase to vectorizable_condition.
19155 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
19157 (vect_finish_stmt_generation): ...here.
19158 (vect_finish_replace_stmt): New function.
19159 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
19160 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
19162 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
19164 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19165 Alan Hayward <alan.hayward@arm.com>
19166 David Sherwood <david.sherwood@arm.com>
19168 * doc/md.texi (extract_last_@var{m}): Document.
19169 * optabs.def (extract_last_optab): New optab.
19170 * internal-fn.def (EXTRACT_LAST): New internal function.
19171 * internal-fn.c (cond_unary_direct): New macro.
19172 (expand_cond_unary_optab_fn): Likewise.
19173 (direct_cond_unary_optab_supported_p): Likewise.
19174 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
19175 loops using EXTRACT_LAST.
19176 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
19177 (extract_last_<mode>): ...this optab.
19178 (vec_extract<mode><Vel>): Update accordingly.
19180 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19181 Alan Hayward <alan.hayward@arm.com>
19182 David Sherwood <david.sherwood@arm.com>
19184 * target.def (empty_mask_is_expensive): New hook.
19185 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
19186 * doc/tm.texi: Regenerate.
19187 * targhooks.h (default_empty_mask_is_expensive): Declare.
19188 * targhooks.c (default_empty_mask_is_expensive): New function.
19189 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
19190 if the target says that empty masks are expensive.
19191 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
19193 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
19195 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19196 Alan Hayward <alan.hayward@arm.com>
19197 David Sherwood <david.sherwood@arm.com>
19199 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
19200 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
19201 (vect_use_loop_mask_for_alignment_p): New function.
19202 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
19203 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
19204 niters_skip argument. Make sure that the first niters_skip elements
19205 of the first iteration are inactive.
19206 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
19207 Update call to vect_set_loop_masks_directly.
19208 (get_misalign_in_elems): New function, split out from...
19209 (vect_gen_prolog_loop_niters): ...here.
19210 (vect_update_init_of_dr): Take a code argument that specifies whether
19211 the adjustment should be added or subtracted.
19212 (vect_update_init_of_drs): Likewise.
19213 (vect_prepare_for_masked_peels): New function.
19214 (vect_do_peeling): Skip prologue peeling if we're using a mask
19215 instead. Update call to vect_update_inits_of_drs.
19216 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
19218 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
19219 alignment. Do not include the number of peeled iterations in
19220 the minimum threshold in that case.
19221 (vectorizable_induction): Adjust the start value down by
19222 LOOP_VINFO_MASK_SKIP_NITERS iterations.
19223 (vect_transform_loop): Call vect_prepare_for_masked_peels.
19224 Take the number of skipped iterations into account when calculating
19226 * tree-vect-stmts.c (vect_gen_while_not): New function.
19228 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19229 Alan Hayward <alan.hayward@arm.com>
19230 David Sherwood <david.sherwood@arm.com>
19232 * doc/sourcebuild.texi (vect_fully_masked): Document.
19233 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
19234 default value to 0.
19235 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
19237 (vect_analyze_loop_2): ...here. Don't check the vectorization
19238 factor against the number of loop iterations if the loop is
19241 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19242 Alan Hayward <alan.hayward@arm.com>
19243 David Sherwood <david.sherwood@arm.com>
19245 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
19246 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
19247 (dump_groups): Update accordingly.
19248 (iv_use::mem_type): New member variable.
19249 (address_p): New function.
19250 (record_use): Add a mem_type argument and initialize the new
19252 (record_group_use): Add a mem_type argument. Use address_p.
19253 Remove obsolete null checks of base_object. Update call to record_use.
19254 (find_interesting_uses_op): Update call to record_group_use.
19255 (find_interesting_uses_cond): Likewise.
19256 (find_interesting_uses_address): Likewise.
19257 (get_mem_type_for_internal_fn): New function.
19258 (find_address_like_use): Likewise.
19259 (find_interesting_uses_stmt): Try find_address_like_use before
19260 calling find_interesting_uses_op.
19261 (addr_offset_valid_p): Use the iv mem_type field as the type
19262 of the addressed memory.
19263 (add_autoinc_candidates): Likewise.
19264 (get_address_cost): Likewise.
19265 (split_small_address_groups_p): Use address_p.
19266 (split_address_groups): Likewise.
19267 (add_iv_candidate_for_use): Likewise.
19268 (autoinc_possible_for_pair): Likewise.
19269 (rewrite_groups): Likewise.
19270 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
19271 (determine_group_iv_cost): Update after split of USE_ADDRESS.
19272 (get_alias_ptr_type_for_ptr_address): New function.
19273 (rewrite_use_address): Rewrite address uses in calls that were
19274 identified by find_address_like_use.
19276 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19277 Alan Hayward <alan.hayward@arm.com>
19278 David Sherwood <david.sherwood@arm.com>
19280 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
19282 * gimple-expr.h (is_gimple_addressable: Likewise.
19283 * gimple-expr.c (is_gimple_address): Likewise.
19284 * internal-fn.c (expand_call_mem_ref): New function.
19285 (expand_mask_load_optab_fn): Use it.
19286 (expand_mask_store_optab_fn): Likewise.
19288 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19289 Alan Hayward <alan.hayward@arm.com>
19290 David Sherwood <david.sherwood@arm.com>
19292 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
19293 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
19294 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
19295 (cond_umax@var{mode}): Document.
19296 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
19297 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
19298 (cond_umin_optab, cond_umax_optab): New optabs.
19299 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
19300 (COND_IOR, COND_XOR): New internal functions.
19301 * internal-fn.h (get_conditional_internal_fn): Declare.
19302 * internal-fn.c (cond_binary_direct): New macro.
19303 (expand_cond_binary_optab_fn): Likewise.
19304 (direct_cond_binary_optab_supported_p): Likewise.
19305 (get_conditional_internal_fn): New function.
19306 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
19307 Cope with reduction statements that are vectorized as calls rather
19309 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
19310 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
19311 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
19312 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
19313 (UNSPEC_COND_EOR): New unspecs.
19314 (optab): Add mappings for them.
19315 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
19316 (sve_int_op, sve_fp_op): New int attributes.
19318 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19319 Alan Hayward <alan.hayward@arm.com>
19320 David Sherwood <david.sherwood@arm.com>
19322 * optabs.def (while_ult_optab): New optab.
19323 * doc/md.texi (while_ult@var{m}@var{n}): Document.
19324 * internal-fn.def (WHILE_ULT): New internal function.
19325 * internal-fn.h (direct_internal_fn_supported_p): New override
19326 that takes two types as argument.
19327 * internal-fn.c (while_direct): New macro.
19328 (expand_while_optab_fn): New function.
19329 (convert_optab_supported_p): Likewise.
19330 (direct_while_optab_supported_p): New macro.
19331 * wide-int.h (wi::udiv_ceil): New function.
19332 * tree-vectorizer.h (rgroup_masks): New structure.
19333 (vec_loop_masks): New typedef.
19334 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
19335 and fully_masked_p.
19336 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
19337 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
19338 (vect_max_vf): New function.
19339 (slpeel_make_loop_iterate_ntimes): Delete.
19340 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
19341 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
19342 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
19343 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
19344 internal-fn.h, stor-layout.h and optabs-query.h.
19345 (vect_set_loop_mask): New function.
19346 (add_preheader_seq): Likewise.
19347 (add_header_seq): Likewise.
19348 (interleave_supported_p): Likewise.
19349 (vect_maybe_permute_loop_masks): Likewise.
19350 (vect_set_loop_masks_directly): Likewise.
19351 (vect_set_loop_condition_masked): Likewise.
19352 (vect_set_loop_condition_unmasked): New function, split out from
19353 slpeel_make_loop_iterate_ntimes.
19354 (slpeel_make_loop_iterate_ntimes): Rename to..
19355 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
19356 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
19357 (vect_do_peeling): Update call accordingly.
19358 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
19360 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
19361 mask_compare_type, can_fully_mask_p and fully_masked_p.
19362 (release_vec_loop_masks): New function.
19363 (_loop_vec_info): Use it to free the loop masks.
19364 (can_produce_all_loop_masks_p): New function.
19365 (vect_get_max_nscalars_per_iter): Likewise.
19366 (vect_verify_full_masking): Likewise.
19367 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
19368 retries, and free the mask rgroups before retrying. Check loop-wide
19369 reasons for disallowing fully-masked loops. Make the final decision
19370 about whether use a fully-masked loop or not.
19371 (vect_estimate_min_profitable_iters): Do not assume that peeling
19372 for the number of iterations will be needed for fully-masked loops.
19373 (vectorizable_reduction): Disable fully-masked loops.
19374 (vectorizable_live_operation): Likewise.
19375 (vect_halve_mask_nunits): New function.
19376 (vect_double_mask_nunits): Likewise.
19377 (vect_record_loop_mask): Likewise.
19378 (vect_get_loop_mask): Likewise.
19379 (vect_transform_loop): Handle the case in which the final loop
19380 iteration might handle a partial vector. Call vect_set_loop_condition
19381 instead of slpeel_make_loop_iterate_ntimes.
19382 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
19383 (check_load_store_masking): New function.
19384 (prepare_load_store_mask): Likewise.
19385 (vectorizable_store): Handle fully-masked loops.
19386 (vectorizable_load): Likewise.
19387 (supportable_widening_operation): Use vect_halve_mask_nunits for
19389 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
19390 (vect_gen_while): New function.
19391 * config/aarch64/aarch64.md (umax<mode>3): New expander.
19392 (aarch64_uqdec<mode>): New insn.
19394 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19395 Alan Hayward <alan.hayward@arm.com>
19396 David Sherwood <david.sherwood@arm.com>
19398 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
19399 (reduc_xor_scal_optab): New optabs.
19400 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
19401 (reduc_xor_scal_@var{m}): Document.
19402 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
19403 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
19404 internal functions.
19405 * fold-const-call.c (fold_const_call): Handle them.
19406 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
19407 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
19408 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
19409 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
19410 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
19411 (UNSPEC_XORV): New unspecs.
19412 (optab): Add entries for them.
19413 (BITWISEV): New int iterator.
19414 (bit_reduc_op): New int attributes.
19416 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19417 Alan Hayward <alan.hayward@arm.com>
19418 David Sherwood <david.sherwood@arm.com>
19420 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
19421 * internal-fn.def (VEC_SHL_INSERT): New internal function.
19422 * optabs.def (vec_shl_insert_optab): New optab.
19423 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
19424 (duplicate_and_interleave): Likewise.
19425 * tree-vect-loop.c: Include internal-fn.h.
19426 (neutral_op_for_slp_reduction): New function, split out from
19427 get_initial_defs_for_reduction.
19428 (get_initial_def_for_reduction): Handle option 2 for variable-length
19429 vectors by loading the neutral value into a vector and then shifting
19430 the initial value into element 0.
19431 (get_initial_defs_for_reduction): Replace the code argument with
19432 the neutral value calculated by neutral_op_for_slp_reduction.
19433 Use gimple_build_vector for constant-length vectors.
19434 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
19435 but the first group_size elements have a neutral value.
19436 Use duplicate_and_interleave otherwise.
19437 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
19438 Update call to get_initial_defs_for_reduction. Handle SLP
19439 reductions for variable-length vectors by creating one vector
19440 result for each scalar result, with the elements associated
19441 with other scalar results stubbed out with the neutral value.
19442 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
19443 Require IFN_VEC_SHL_INSERT for double reductions on
19444 variable-length vectors, or SLP reductions that have
19445 a neutral value. Require can_duplicate_and_interleave_p
19446 support for variable-length unchained SLP reductions if there
19447 is no neutral value, such as for MIN/MAX reductions. Also require
19448 the number of vector elements to be a multiple of the number of
19449 SLP statements when doing variable-length unchained SLP reductions.
19450 Update call to vect_create_epilog_for_reduction.
19451 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
19452 and remove initial values.
19453 (duplicate_and_interleave): Make public.
19454 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
19455 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
19457 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19458 Alan Hayward <alan.hayward@arm.com>
19459 David Sherwood <david.sherwood@arm.com>
19461 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
19462 (can_duplicate_and_interleave_p): New function.
19463 (vect_get_and_check_slp_defs): Take the vector of statements
19464 rather than just the current one. Remove excess parentheses.
19465 Restriction rejectinon of vect_constant_def and vect_external_def
19466 for variable-length vectors to boolean types, or types for which
19467 can_duplicate_and_interleave_p is false.
19468 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
19469 (duplicate_and_interleave): New function.
19470 (vect_get_constant_vectors): Use gimple_build_vector for
19471 constant-length vectors and suitable variable-length constant
19472 vectors. Use duplicate_and_interleave for other variable-length
19473 vectors. Don't defer the update when inserting new statements.
19475 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19476 Alan Hayward <alan.hayward@arm.com>
19477 David Sherwood <david.sherwood@arm.com>
19479 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
19480 min_profitable_iters doesn't go negative.
19482 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19483 Alan Hayward <alan.hayward@arm.com>
19484 David Sherwood <david.sherwood@arm.com>
19486 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
19487 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
19488 * optabs.def (vec_mask_load_lanes_optab): New optab.
19489 (vec_mask_store_lanes_optab): Likewise.
19490 * internal-fn.def (MASK_LOAD_LANES): New internal function.
19491 (MASK_STORE_LANES): Likewise.
19492 * internal-fn.c (mask_load_lanes_direct): New macro.
19493 (mask_store_lanes_direct): Likewise.
19494 (expand_mask_load_optab_fn): Handle masked operations.
19495 (expand_mask_load_lanes_optab_fn): New macro.
19496 (expand_mask_store_optab_fn): Handle masked operations.
19497 (expand_mask_store_lanes_optab_fn): New macro.
19498 (direct_mask_load_lanes_optab_supported_p): Likewise.
19499 (direct_mask_store_lanes_optab_supported_p): Likewise.
19500 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
19502 (vect_load_lanes_supported): Likewise.
19503 * tree-vect-data-refs.c (strip_conversion): New function.
19504 (can_group_stmts_p): Likewise.
19505 (vect_analyze_data_ref_accesses): Use it instead of checking
19506 for a pair of assignments.
19507 (vect_store_lanes_supported): Take a masked_p parameter.
19508 (vect_load_lanes_supported): Likewise.
19509 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
19510 vect_store_lanes_supported and vect_load_lanes_supported.
19511 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
19512 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
19513 parameter. Don't allow gaps for masked accesses.
19514 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
19515 and vect_load_lanes_supported.
19516 (get_load_store_type): Take a masked_p parameter and update
19517 call to get_group_load_store_type.
19518 (vectorizable_store): Update call to get_load_store_type.
19519 Handle IFN_MASK_STORE_LANES.
19520 (vectorizable_load): Update call to get_load_store_type.
19521 Handle IFN_MASK_LOAD_LANES.
19523 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19524 Alan Hayward <alan.hayward@arm.com>
19525 David Sherwood <david.sherwood@arm.com>
19527 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
19529 * config/aarch64/aarch64-protos.h
19530 (aarch64_sve_struct_memory_operand_p): Declare.
19531 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
19532 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
19533 (VPRED, vpred): Handle SVE structure modes.
19534 * config/aarch64/constraints.md (Utx): New constraint.
19535 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
19536 (aarch64_sve_struct_nonimmediate_operand): New predicates.
19537 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
19538 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
19539 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
19540 structure modes. Split into pieces after RA.
19541 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
19542 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
19544 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
19545 SVE structure modes.
19546 (aarch64_classify_address): Likewise.
19547 (sizetochar): Move earlier in file.
19548 (aarch64_print_operand): Handle SVE register lists.
19549 (aarch64_array_mode): New function.
19550 (aarch64_sve_struct_memory_operand_p): Likewise.
19551 (TARGET_ARRAY_MODE): Redefine.
19553 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19554 Alan Hayward <alan.hayward@arm.com>
19555 David Sherwood <david.sherwood@arm.com>
19557 * target.def (array_mode): New target hook.
19558 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
19559 * doc/tm.texi: Regenerate.
19560 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
19561 * hooks.c (hook_optmode_mode_uhwi_none): New function.
19562 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
19563 targetm.array_mode.
19564 * stor-layout.c (mode_for_array): Likewise. Support polynomial
19567 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19568 Alan Hayward <alan.hayward@arm.com>
19569 David Sherwood <david.sherwood@arm.com>
19571 * fold-const.c (fold_binary_loc): Check the argument types
19572 rather than the result type when testing for a vector operation.
19574 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19576 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
19577 * doc/tm.texi: Regenerate.
19579 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19580 Alan Hayward <alan.hayward@arm.com>
19581 David Sherwood <david.sherwood@arm.com>
19583 * doc/invoke.texi (-msve-vector-bits=): Document new option.
19584 (sve): Document new AArch64 extension.
19585 * doc/md.texi (w): Extend the description of the AArch64
19586 constraint to include SVE vectors.
19587 (Upl, Upa): Document new AArch64 predicate constraints.
19588 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
19590 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
19591 (msve-vector-bits=): New option.
19592 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
19593 SVE when these are disabled.
19594 (sve): New extension.
19595 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
19596 modes. Adjust their number of units based on aarch64_sve_vg.
19597 (MAX_BITSIZE_MODE_ANY_MODE): Define.
19598 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
19599 aarch64_addr_query_type.
19600 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
19601 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
19602 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
19603 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
19604 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
19605 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
19606 (aarch64_simd_imm_zero_p): Delete.
19607 (aarch64_check_zero_based_sve_index_immediate): Declare.
19608 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
19609 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
19610 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
19611 (aarch64_sve_float_mul_immediate_p): Likewise.
19612 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
19613 rather than an rtx.
19614 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
19615 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
19616 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
19617 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
19618 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
19619 (aarch64_regmode_natural_size): Likewise.
19620 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
19621 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
19623 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
19624 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
19625 for VG and the SVE predicate registers.
19626 (V_ALIASES): Add a "z"-prefixed alias.
19627 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
19628 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
19629 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
19630 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
19631 (REG_CLASS_NAMES): Add entries for them.
19632 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
19633 and the predicate registers.
19634 (aarch64_sve_vg): Declare.
19635 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
19636 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
19637 (REGMODE_NATURAL_SIZE): Define.
19638 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
19640 * config/aarch64/aarch64.c: Include cfgrtl.h.
19641 (simd_immediate_info): Add a constructor for series vectors,
19642 and an associated step field.
19643 (aarch64_sve_vg): New variable.
19644 (aarch64_dbx_register_number): Handle VG and the predicate registers.
19645 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
19646 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
19647 (VEC_ANY_DATA, VEC_STRUCT): New constants.
19648 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
19649 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
19650 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
19651 (aarch64_get_mask_mode): New functions.
19652 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
19653 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
19654 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
19655 predicate modes and predicate registers. Explicitly restrict
19656 GPRs to modes of 16 bytes or smaller. Only allow FP registers
19657 to store a vector mode if it is recognized by
19658 aarch64_classify_vector_mode.
19659 (aarch64_regmode_natural_size): New function.
19660 (aarch64_hard_regno_caller_save_mode): Return the original mode
19662 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
19663 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
19664 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
19665 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
19667 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
19668 does not overlap dest if the function is frame-related. Handle
19670 (aarch64_split_add_offset): New function.
19671 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
19672 them aarch64_add_offset.
19673 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
19674 and update call to aarch64_sub_sp.
19675 (aarch64_add_cfa_expression): New function.
19676 (aarch64_expand_prologue): Pass extra temporary registers to the
19677 functions above. Handle the case in which we need to emit new
19678 DW_CFA_expressions for registers that were originally saved
19679 relative to the stack pointer, but now have to be expressed
19680 relative to the frame pointer.
19681 (aarch64_output_mi_thunk): Pass extra temporary registers to the
19683 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
19684 IP0 and IP1 values for SVE frames.
19685 (aarch64_expand_vec_series): New function.
19686 (aarch64_expand_sve_widened_duplicate): Likewise.
19687 (aarch64_expand_sve_const_vector): Likewise.
19688 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
19689 Handle SVE constants. Use emit_move_insn to move a force_const_mem
19690 into the register, rather than emitting a SET directly.
19691 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
19692 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
19693 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
19694 (offset_9bit_signed_scaled_p): New functions.
19695 (aarch64_replicate_bitmask_imm): New function.
19696 (aarch64_bitmask_imm): Use it.
19697 (aarch64_cannot_force_const_mem): Reject expressions involving
19698 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
19699 (aarch64_classify_index): Handle SVE indices, by requiring
19700 a plain register index with a scale that matches the element size.
19701 (aarch64_classify_address): Handle SVE addresses. Assert that
19702 the mode of the address is VOIDmode or an integer mode.
19703 Update call to aarch64_classify_symbol.
19704 (aarch64_classify_symbolic_expression): Update call to
19705 aarch64_classify_symbol.
19706 (aarch64_const_vec_all_in_range_p): New function.
19707 (aarch64_print_vector_float_operand): Likewise.
19708 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
19709 "vN" for FP registers with SVE modes. Handle (const ...) vectors
19710 and the FP immediates 1.0 and 0.5.
19711 (aarch64_print_address_internal): Handle SVE addresses.
19712 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
19713 (aarch64_regno_regclass): Handle predicate registers.
19714 (aarch64_secondary_reload): Handle big-endian reloads of SVE
19716 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
19717 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
19718 (aarch64_convert_sve_vector_bits): New function.
19719 (aarch64_override_options): Use it to handle -msve-vector-bits=.
19720 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
19721 rather than an rtx.
19722 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
19723 Handle SVE vector and predicate modes. Accept VL-based constants
19724 that need only one temporary register, and VL offsets that require
19725 no temporary registers.
19726 (aarch64_conditional_register_usage): Mark the predicate registers
19727 as fixed if SVE isn't available.
19728 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
19729 Return true for SVE vector and predicate modes.
19730 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
19731 rather than an unsigned int. Handle SVE modes.
19732 (aarch64_preferred_simd_mode): Update call accordingly. Handle
19734 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
19736 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
19737 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
19738 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
19739 (aarch64_sve_float_mul_immediate_p): New functions.
19740 (aarch64_sve_valid_immediate): New function.
19741 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
19742 Explicitly reject structure modes. Check for INDEX constants.
19743 Handle PTRUE and PFALSE constants.
19744 (aarch64_check_zero_based_sve_index_immediate): New function.
19745 (aarch64_simd_imm_zero_p): Delete.
19746 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
19747 vector modes. Accept constants in the range of CNT[BHWD].
19748 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
19749 ask for an Advanced SIMD mode.
19750 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
19751 (aarch64_simd_vector_alignment): Handle SVE predicates.
19752 (aarch64_vectorize_preferred_vector_alignment): New function.
19753 (aarch64_simd_vector_alignment_reachable): Use it instead of
19755 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
19756 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
19758 (MAX_VECT_LEN): Delete.
19759 (expand_vec_perm_d): Add a vec_flags field.
19760 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
19761 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
19762 (aarch64_evpc_ext): Don't apply a big-endian lane correction
19764 (aarch64_evpc_rev): Rename to...
19765 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
19766 (aarch64_evpc_rev_global): New function.
19767 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
19768 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
19770 (aarch64_evpc_sve_tbl): New function.
19771 (aarch64_expand_vec_perm_const_1): Update after rename of
19772 aarch64_evpc_rev. Handle SVE permutes too, trying
19773 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
19774 than aarch64_evpc_tbl.
19775 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
19776 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
19777 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
19778 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
19779 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
19780 (aarch64_expand_sve_vcond): New functions.
19781 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
19782 of aarch64_vector_mode_p.
19783 (aarch64_dwarf_poly_indeterminate_value): New function.
19784 (aarch64_compute_pressure_classes): Likewise.
19785 (aarch64_can_change_mode_class): Likewise.
19786 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
19787 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
19788 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
19789 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
19790 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
19791 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
19792 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
19793 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
19795 (Dn, Dl, Dr): Accept const as well as const_vector.
19796 (Dz): Likewise. Compare against CONST0_RTX.
19797 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
19798 of "vector" where appropriate.
19799 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
19800 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
19801 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
19802 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
19803 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
19804 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
19805 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
19806 (v_int_equiv): Extend to SVE modes.
19807 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
19809 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
19810 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
19811 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
19812 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
19813 (SVE_COND_FP_CMP): New int iterators.
19814 (perm_hilo): Handle the new unpack unspecs.
19815 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
19817 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
19818 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
19819 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
19820 (aarch64_equality_operator, aarch64_constant_vector_operand)
19821 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
19822 (aarch64_sve_nonimmediate_operand): Likewise.
19823 (aarch64_sve_general_operand): Likewise.
19824 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
19825 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
19826 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
19827 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
19828 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
19829 (aarch64_sve_float_arith_immediate): Likewise.
19830 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
19831 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
19832 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
19833 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
19834 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
19835 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
19836 (aarch64_sve_float_arith_operand): Likewise.
19837 (aarch64_sve_float_arith_with_sub_operand): Likewise.
19838 (aarch64_sve_float_mul_operand): Likewise.
19839 (aarch64_sve_vec_perm_operand): Likewise.
19840 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
19841 (aarch64_mov_operand): Accept const_poly_int and const_vector.
19842 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
19843 as well as const_vector.
19844 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
19845 in file. Use CONST0_RTX and CONSTM1_RTX.
19846 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
19847 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
19848 Use aarch64_simd_imm_zero.
19849 * config/aarch64/aarch64-sve.md: New file.
19850 * config/aarch64/aarch64.md: Include it.
19851 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
19852 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
19853 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
19854 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
19855 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
19856 (sve): New attribute.
19857 (enabled): Disable instructions with the sve attribute unless
19859 (movqi, movhi): Pass CONST_POLY_INT operaneds through
19860 aarch64_expand_mov_immediate.
19861 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
19862 CNT[BHSD] immediates.
19863 (movti): Split CONST_POLY_INT moves into two halves.
19864 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
19865 Split additions that need a temporary here if the destination
19866 is the stack pointer.
19867 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
19868 (*add<mode>3_poly_1): New instruction.
19869 (set_clobber_cc): New expander.
19871 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19873 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
19874 parameter and use it instead of GET_MODE_SIZE (innermode). Use
19875 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
19876 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
19877 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
19878 Change innermode from fixed_mode_size to machine_mode.
19879 (simplify_subreg): Update call accordingly. Handle a constant-sized
19880 subreg of a variable-length CONST_VECTOR.
19882 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19883 Alan Hayward <alan.hayward@arm.com>
19884 David Sherwood <david.sherwood@arm.com>
19886 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
19887 (add_offset_to_base): New function, split out from...
19888 (create_mem_ref): ...here. When handling a scale other than 1,
19889 check first whether the address is valid without the offset.
19890 Add it into the base if so, leaving the index and scale as-is.
19892 2018-01-12 Jakub Jelinek <jakub@redhat.com>
19895 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
19896 fold_for_warn before checking if arg2 is INTEGER_CST.
19898 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
19900 * config/rs6000/predicates.md (load_multiple_operation): Delete.
19901 (store_multiple_operation): Delete.
19902 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
19903 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
19904 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
19905 guarded by TARGET_STRING.
19906 (rs6000_output_load_multiple): Delete.
19907 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
19908 OPTION_MASK_STRING / TARGET_STRING handling.
19909 (print_operand) <'N', 'O'>: Add comment that these are unused now.
19910 (const rs6000_opt_masks) <"string">: Change mask to 0.
19911 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
19912 (MASK_STRING): Delete.
19913 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
19915 (load_multiple): Delete.
19922 (store_multiple): Delete.
19929 (movmemsi_8reg): Delete.
19930 (corresponding unnamed define_insn): Delete.
19931 (movmemsi_6reg): Delete.
19932 (corresponding unnamed define_insn): Delete.
19933 (movmemsi_4reg): Delete.
19934 (corresponding unnamed define_insn): Delete.
19935 (movmemsi_2reg): Delete.
19936 (corresponding unnamed define_insn): Delete.
19937 (movmemsi_1reg): Delete.
19938 (corresponding unnamed define_insn): Delete.
19939 * config/rs6000/rs6000.opt (mno-string): New.
19940 (mstring): Replace by deprecation warning stub.
19941 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
19943 2018-01-12 Jakub Jelinek <jakub@redhat.com>
19945 * regrename.c (regrename_do_replace): If replacing the same
19946 reg multiple times, try to reuse last created gen_raw_REG.
19949 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
19950 main to workaround a bug in GDB.
19952 2018-01-12 Tom de Vries <tom@codesourcery.com>
19955 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
19957 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
19959 PR rtl-optimization/80481
19960 * ira-color.c (get_cap_member): New function.
19961 (allocnos_conflict_by_live_ranges_p): Use it.
19962 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
19963 (setup_slot_coalesced_allocno_live_ranges): Ditto.
19965 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
19968 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
19969 (*saddl_se_1): Ditto.
19970 (*ssubsi_1): Ditto.
19971 (*ssubl_se_1): Ditto.
19973 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
19975 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
19976 rather than wi::to_widest for DR_INITs.
19977 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
19978 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
19979 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
19981 (vect_analyze_group_access_1): Note that here.
19983 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
19985 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
19986 polynomial type sizes.
19988 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
19990 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
19991 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
19992 (gimple_add_tmp_var): Likewise.
19994 2018-01-12 Martin Liska <mliska@suse.cz>
19996 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
19997 (gimple_alloc_sizes): Likewise.
19998 (dump_gimple_statistics): Use PRIu64 in printf format.
19999 * gimple.h: Change uint64_t to int.
20001 2018-01-12 Martin Liska <mliska@suse.cz>
20003 * tree-core.h: Use uint64_t instead of int.
20004 * tree.c (tree_node_counts): Likewise.
20005 (tree_node_sizes): Likewise.
20006 (dump_tree_statistics): Use PRIu64 in printf format.
20008 2018-01-12 Martin Liska <mliska@suse.cz>
20010 * Makefile.in: As qsort_chk is implemented in vec.c, add
20011 vec.o to linkage of gencfn-macros.
20012 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
20013 passing the info to record_node_allocation_statistics.
20014 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
20016 * ggc-common.c (struct ggc_usage): Add operator== and use
20017 it in operator< and compare function.
20018 * mem-stats.h (struct mem_usage): Likewise.
20019 * vec.c (struct vec_usage): Remove operator< and compare
20020 function. Can be simply inherited.
20022 2018-01-12 Martin Jambor <mjambor@suse.cz>
20025 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
20026 * tree-ssa-math-opts.c: Include domwalk.h.
20027 (convert_mult_to_fma_1): New function.
20028 (fma_transformation_info): New type.
20029 (fma_deferring_state): Likewise.
20030 (cancel_fma_deferring): New function.
20031 (result_of_phi): Likewise.
20032 (last_fma_candidate_feeds_initial_phi): Likewise.
20033 (convert_mult_to_fma): Added deferring logic, split actual
20034 transformation to convert_mult_to_fma_1.
20035 (math_opts_dom_walker): New type.
20036 (math_opts_dom_walker::after_dom_children): New method, body moved
20037 here from pass_optimize_widening_mul::execute, added deferring logic
20039 (pass_optimize_widening_mul::execute): Moved most of code to
20040 math_opts_dom_walker::after_dom_children.
20041 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
20042 * config/i386/i386.c (ix86_option_override_internal): Added
20043 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
20045 2018-01-12 Richard Biener <rguenther@suse.de>
20048 * dwarf2out.c (gen_variable_die): Do not reset old_die for
20049 inline instance vars.
20051 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
20054 * config/rx/rx.c (rx_is_restricted_memory_address):
20055 Handle SUBREG case.
20057 2018-01-12 Richard Biener <rguenther@suse.de>
20059 PR tree-optimization/80846
20060 * target.def (split_reduction): New target hook.
20061 * targhooks.c (default_split_reduction): New function.
20062 * targhooks.h (default_split_reduction): Declare.
20063 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
20064 target requests first reduce vectors by combining low and high
20066 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
20067 (get_vectype_for_scalar_type_and_size): Export.
20068 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
20069 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
20070 * doc/tm.texi: Regenerate.
20071 * config/i386/i386.c (ix86_split_reduction): Implement
20072 TARGET_VECTORIZE_SPLIT_REDUCTION.
20074 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
20077 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
20078 in PIC mode except for TARGET_VXWORKS_RTP.
20079 * config/sparc/sparc.c: Include cfgrtl.h.
20080 (TARGET_INIT_PIC_REG): Define.
20081 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
20082 (sparc_pic_register_p): New predicate.
20083 (sparc_legitimate_address_p): Use it.
20084 (sparc_legitimize_pic_address): Likewise.
20085 (sparc_delegitimize_address): Likewise.
20086 (sparc_mode_dependent_address_p): Likewise.
20087 (gen_load_pcrel_sym): Remove 4th parameter.
20088 (load_got_register): Adjust call to above. Remove obsolete stuff.
20089 (sparc_expand_prologue): Do not call load_got_register here.
20090 (sparc_flat_expand_prologue): Likewise.
20091 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
20092 (sparc_use_pseudo_pic_reg): New function.
20093 (sparc_init_pic_reg): Likewise.
20094 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
20095 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
20097 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
20099 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
20100 Add item for branch_cost.
20102 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
20104 PR rtl-optimization/83565
20105 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
20106 not extend the result to a larger mode for rotate operations.
20107 (num_sign_bit_copies1): Likewise.
20109 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
20112 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
20114 Use values-Xc.o for -pedantic.
20115 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
20117 2018-01-12 Martin Liska <mliska@suse.cz>
20120 * ipa-devirt.c (final_warning_record::grow_type_warnings):
20122 (possible_polymorphic_call_targets): Use it.
20123 (ipa_devirt): Likewise.
20125 2018-01-12 Martin Liska <mliska@suse.cz>
20127 * profile-count.h (enum profile_quality): Use 0 as invalid
20128 enum value of profile_quality.
20130 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
20132 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
20133 -mext-string options.
20135 2018-01-12 Richard Biener <rguenther@suse.de>
20137 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
20138 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
20139 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
20141 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
20143 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
20145 * configure.ac (--with-long-double-format): Add support for the
20146 configuration option to change the default long double format on
20148 * config.gcc (powerpc*-linux*-*): Likewise.
20149 * configure: Regenerate.
20150 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
20151 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
20152 used without modification.
20154 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
20156 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
20157 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
20158 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
20159 MISC_BUILTIN_SPEC_BARRIER.
20160 (rs6000_init_builtins): Likewise.
20161 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
20163 (speculation_barrier): New define_insn.
20164 * doc/extend.texi: Document __builtin_speculation_barrier.
20166 2018-01-11 Jakub Jelinek <jakub@redhat.com>
20169 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
20170 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
20171 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
20173 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
20174 integral modes instead of "ss" and "sd".
20175 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
20176 vectors with 32-bit and 64-bit elements.
20177 (vecdupssescalarmodesuffix): New mode attribute.
20178 (vec_dup<mode>): Use it.
20180 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
20183 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
20184 frame if argument is passed on stack.
20186 2018-01-11 Jakub Jelinek <jakub@redhat.com>
20189 * ree.c (combine_reaching_defs): Optimize also
20190 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
20191 reg2=any_extend(exp); reg1=reg2;, formatting fix.
20193 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
20195 PR middle-end/83189
20196 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
20198 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
20200 PR middle-end/83718
20201 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
20202 after they are computed.
20204 2018-01-11 Bin Cheng <bin.cheng@arm.com>
20206 PR tree-optimization/83695
20207 * gimple-loop-linterchange.cc
20208 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
20209 reset cached scev information after interchange.
20210 (pass_linterchange::execute): Remove call to scev_reset_htab.
20212 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20214 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
20215 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
20216 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
20217 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
20218 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
20219 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
20220 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
20221 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
20222 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
20223 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
20224 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
20225 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
20226 (V_lane_reg): Likewise.
20227 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
20229 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
20230 (vfmal_lane_low<mode>_intrinsic,
20231 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
20232 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
20233 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
20234 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
20235 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
20236 vfmsl_lane_high<mode>_intrinsic): New define_insns.
20238 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20240 * config/arm/arm-cpus.in (fp16fml): New feature.
20241 (ALL_SIMD): Add fp16fml.
20242 (armv8.2-a): Add fp16fml as an option.
20243 (armv8.3-a): Likewise.
20244 (armv8.4-a): Add fp16fml as part of fp16.
20245 * config/arm/arm.h (TARGET_FP16FML): Define.
20246 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
20248 * config/arm/arm-modes.def (V2HF): Define.
20249 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
20250 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
20251 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
20252 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
20253 vfmsl_low, vfmsl_high): New set of builtins.
20254 * config/arm/iterators.md (PLUSMINUS): New code iterator.
20255 (vfml_op): New code attribute.
20256 (VFMLHALVES): New int iterator.
20257 (VFML, VFMLSEL): New mode attributes.
20258 (V_reg): Define mapping for V2HF.
20259 (V_hi, V_lo): New mode attributes.
20260 (VF_constraint): Likewise.
20261 (vfml_half, vfml_half_selector): New int attributes.
20262 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
20264 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
20265 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
20267 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
20268 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
20269 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
20270 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
20272 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
20273 Document new effective target and option set.
20275 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20277 * config/arm/arm-cpus.in (armv8_4): New feature.
20278 (ARMv8_4a): New fgroup.
20279 (armv8.4-a): New arch.
20280 * config/arm/arm-tables.opt: Regenerate.
20281 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
20282 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
20283 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
20284 Add matching rules for -march=armv8.4-a and extensions.
20285 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
20287 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
20290 * config/rx/rx.md (BW): New mode attribute.
20291 (sync_lock_test_and_setsi): Add mode suffix to insn output.
20293 2018-01-11 Richard Biener <rguenther@suse.de>
20295 PR tree-optimization/83435
20296 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
20297 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
20298 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
20300 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
20301 Alan Hayward <alan.hayward@arm.com>
20302 David Sherwood <david.sherwood@arm.com>
20304 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
20306 (aarch64_classify_address): Initialize it. Track polynomial offsets.
20307 (aarch64_print_address_internal): Use it to check for a zero offset.
20309 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
20310 Alan Hayward <alan.hayward@arm.com>
20311 David Sherwood <david.sherwood@arm.com>
20313 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
20314 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
20315 Return a poly_int64 rather than a HOST_WIDE_INT.
20316 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
20317 rather than a HOST_WIDE_INT.
20318 * config/aarch64/aarch64.h (aarch64_frame): Protect with
20319 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
20320 hard_fp_offset, frame_size, initial_adjust, callee_offset and
20321 final_offset from HOST_WIDE_INT to poly_int64.
20322 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
20323 to_constant when getting the number of units in an Advanced SIMD
20325 (aarch64_builtin_vectorized_function): Check for a constant number
20327 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
20329 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
20330 attribute instead of GET_MODE_NUNITS.
20331 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
20332 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
20333 GET_MODE_SIZE for fixed-size registers.
20334 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
20335 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
20336 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
20337 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
20338 (aarch64_print_operand, aarch64_print_address_internal)
20339 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
20340 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
20341 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
20342 Handle polynomial GET_MODE_SIZE.
20343 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
20344 wider than SImode without modification.
20345 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
20346 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
20347 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
20348 passing and returning SVE modes.
20349 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
20350 rather than GEN_INT.
20351 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
20352 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
20353 (aarch64_allocate_and_probe_stack_space): Likewise.
20354 (aarch64_layout_frame): Cope with polynomial offsets.
20355 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
20356 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
20357 polynomial offsets.
20358 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
20359 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
20360 poly_int64 rather than a HOST_WIDE_INT.
20361 (aarch64_get_separate_components, aarch64_process_components)
20362 (aarch64_expand_prologue, aarch64_expand_epilogue)
20363 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
20364 (aarch64_anchor_offset): New function, split out from...
20365 (aarch64_legitimize_address): ...here.
20366 (aarch64_builtin_vectorization_cost): Handle polynomial
20367 TYPE_VECTOR_SUBPARTS.
20368 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
20370 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
20371 number of elements from the PARALLEL rather than the mode.
20372 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
20373 rather than GET_MODE_BITSIZE.
20374 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
20375 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
20376 (aarch64_expand_vec_perm_const_1): Handle polynomial
20377 d->perm.length () and d->perm elements.
20378 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
20379 Apply to_constant to d->perm elements.
20380 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
20381 polynomial CONST_VECTOR_NUNITS.
20382 (aarch64_move_pointer): Take amount as a poly_int64 rather
20384 (aarch64_progress_pointer): Avoid temporary variable.
20385 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
20386 the mode attribute instead of GET_MODE.
20388 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
20389 Alan Hayward <alan.hayward@arm.com>
20390 David Sherwood <david.sherwood@arm.com>
20392 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
20393 x exists before using it.
20394 (aarch64_add_constant_internal): Rename to...
20395 (aarch64_add_offset_1): ...this. Replace regnum with separate
20396 src and dest rtxes. Handle the case in which they're different,
20397 including when the offset is zero. Replace scratchreg with an rtx.
20398 Use 2 additions if there is no spare register into which we can
20399 move a 16-bit constant.
20400 (aarch64_add_constant): Delete.
20401 (aarch64_add_offset): Replace reg with separate src and dest
20402 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
20403 Use aarch64_add_offset_1.
20404 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
20405 an rtx rather than an int. Take the delta as a poly_int64
20406 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
20407 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
20408 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
20409 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
20410 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
20411 and aarch64_add_sp.
20412 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
20413 aarch64_add_constant.
20415 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
20417 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
20418 Use scalar_float_mode.
20420 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
20422 * config/aarch64/aarch64-simd.md
20423 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
20424 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
20425 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
20426 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
20427 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
20428 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
20429 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
20430 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
20431 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
20432 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
20434 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
20437 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
20438 targ_options->x_arm_arch_string is non NULL.
20440 2018-01-11 Tamar Christina <tamar.christina@arm.com>
20442 * config/aarch64/aarch64.h
20443 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
20445 2018-01-11 Sudakshina Das <sudi.das@arm.com>
20448 * expmed.c (emit_store_flag_force): Swap if const op0
20449 and change VOIDmode to mode of op0.
20451 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
20453 PR rtl-optimization/83761
20454 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
20455 than bytes to mode_for_size.
20457 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
20459 PR middle-end/83189
20460 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
20461 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
20464 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
20466 PR middle-end/83575
20467 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
20468 when in layout mode.
20469 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
20470 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
20473 2018-01-10 Michael Collison <michael.collison@arm.com>
20475 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
20476 * config/aarch64/aarch64-option-extension.def: Add
20477 AARCH64_OPT_EXTENSION of 'fp16fml'.
20478 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
20479 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
20480 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
20481 * config/aarch64/constraints.md (Ui7): New constraint.
20482 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
20483 (VFMLA_SEL_W): Ditto.
20486 (VFMLA16_LOW): New int iterator.
20487 (VFMLA16_HIGH): Ditto.
20488 (UNSPEC_FMLAL): New unspec.
20489 (UNSPEC_FMLSL): Ditto.
20490 (UNSPEC_FMLAL2): Ditto.
20491 (UNSPEC_FMLSL2): Ditto.
20492 (f16mac): New code attribute.
20493 * config/aarch64/aarch64-simd-builtins.def
20494 (aarch64_fmlal_lowv2sf): Ditto.
20495 (aarch64_fmlsl_lowv2sf): Ditto.
20496 (aarch64_fmlalq_lowv4sf): Ditto.
20497 (aarch64_fmlslq_lowv4sf): Ditto.
20498 (aarch64_fmlal_highv2sf): Ditto.
20499 (aarch64_fmlsl_highv2sf): Ditto.
20500 (aarch64_fmlalq_highv4sf): Ditto.
20501 (aarch64_fmlslq_highv4sf): Ditto.
20502 (aarch64_fmlal_lane_lowv2sf): Ditto.
20503 (aarch64_fmlsl_lane_lowv2sf): Ditto.
20504 (aarch64_fmlal_laneq_lowv2sf): Ditto.
20505 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
20506 (aarch64_fmlalq_lane_lowv4sf): Ditto.
20507 (aarch64_fmlsl_lane_lowv4sf): Ditto.
20508 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
20509 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
20510 (aarch64_fmlal_lane_highv2sf): Ditto.
20511 (aarch64_fmlsl_lane_highv2sf): Ditto.
20512 (aarch64_fmlal_laneq_highv2sf): Ditto.
20513 (aarch64_fmlsl_laneq_highv2sf): Ditto.
20514 (aarch64_fmlalq_lane_highv4sf): Ditto.
20515 (aarch64_fmlsl_lane_highv4sf): Ditto.
20516 (aarch64_fmlalq_laneq_highv4sf): Ditto.
20517 (aarch64_fmlsl_laneq_highv4sf): Ditto.
20518 * config/aarch64/aarch64-simd.md:
20519 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
20520 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
20521 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
20522 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
20523 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
20524 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
20525 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
20526 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
20527 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
20528 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
20529 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
20530 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
20531 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
20532 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
20533 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
20534 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
20535 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
20536 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
20537 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
20538 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
20539 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
20540 (vfmlsl_low_u32): Ditto.
20541 (vfmlalq_low_u32): Ditto.
20542 (vfmlslq_low_u32): Ditto.
20543 (vfmlal_high_u32): Ditto.
20544 (vfmlsl_high_u32): Ditto.
20545 (vfmlalq_high_u32): Ditto.
20546 (vfmlslq_high_u32): Ditto.
20547 (vfmlal_lane_low_u32): Ditto.
20548 (vfmlsl_lane_low_u32): Ditto.
20549 (vfmlal_laneq_low_u32): Ditto.
20550 (vfmlsl_laneq_low_u32): Ditto.
20551 (vfmlalq_lane_low_u32): Ditto.
20552 (vfmlslq_lane_low_u32): Ditto.
20553 (vfmlalq_laneq_low_u32): Ditto.
20554 (vfmlslq_laneq_low_u32): Ditto.
20555 (vfmlal_lane_high_u32): Ditto.
20556 (vfmlsl_lane_high_u32): Ditto.
20557 (vfmlal_laneq_high_u32): Ditto.
20558 (vfmlsl_laneq_high_u32): Ditto.
20559 (vfmlalq_lane_high_u32): Ditto.
20560 (vfmlslq_lane_high_u32): Ditto.
20561 (vfmlalq_laneq_high_u32): Ditto.
20562 (vfmlslq_laneq_high_u32): Ditto.
20563 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
20564 (AARCH64_FL_FOR_ARCH8_4): New.
20565 (AARCH64_ISA_F16FML): New ISA flag.
20566 (TARGET_F16FML): New feature flag for fp16fml.
20567 (doc/invoke.texi): Document new fp16fml option.
20569 2018-01-10 Michael Collison <michael.collison@arm.com>
20571 * config/aarch64/aarch64-builtins.c:
20572 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
20573 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
20574 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
20575 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
20576 (AARCH64_ISA_SHA3): New ISA flag.
20577 (TARGET_SHA3): New feature flag for sha3.
20578 * config/aarch64/iterators.md (sha512_op): New int attribute.
20579 (CRYPTO_SHA512): New int iterator.
20580 (UNSPEC_SHA512H): New unspec.
20581 (UNSPEC_SHA512H2): Ditto.
20582 (UNSPEC_SHA512SU0): Ditto.
20583 (UNSPEC_SHA512SU1): Ditto.
20584 * config/aarch64/aarch64-simd-builtins.def
20585 (aarch64_crypto_sha512hqv2di): New builtin.
20586 (aarch64_crypto_sha512h2qv2di): Ditto.
20587 (aarch64_crypto_sha512su0qv2di): Ditto.
20588 (aarch64_crypto_sha512su1qv2di): Ditto.
20589 (aarch64_eor3qv8hi): Ditto.
20590 (aarch64_rax1qv2di): Ditto.
20591 (aarch64_xarqv2di): Ditto.
20592 (aarch64_bcaxqv8hi): Ditto.
20593 * config/aarch64/aarch64-simd.md:
20594 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
20595 (aarch64_crypto_sha512su0qv2di): Ditto.
20596 (aarch64_crypto_sha512su1qv2di): Ditto.
20597 (aarch64_eor3qv8hi): Ditto.
20598 (aarch64_rax1qv2di): Ditto.
20599 (aarch64_xarqv2di): Ditto.
20600 (aarch64_bcaxqv8hi): Ditto.
20601 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
20602 (vsha512h2q_u64): Ditto.
20603 (vsha512su0q_u64): Ditto.
20604 (vsha512su1q_u64): Ditto.
20605 (veor3q_u16): Ditto.
20606 (vrax1q_u64): Ditto.
20607 (vxarq_u64): Ditto.
20608 (vbcaxq_u16): Ditto.
20609 * config/arm/types.md (crypto_sha512): New type attribute.
20610 (crypto_sha3): Ditto.
20611 (doc/invoke.texi): Document new sha3 option.
20613 2018-01-10 Michael Collison <michael.collison@arm.com>
20615 * config/aarch64/aarch64-builtins.c:
20616 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
20617 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
20618 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
20619 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
20620 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
20621 (AARCH64_ISA_SM4): New ISA flag.
20622 (TARGET_SM4): New feature flag for sm4.
20623 * config/aarch64/aarch64-simd-builtins.def
20624 (aarch64_sm3ss1qv4si): Ditto.
20625 (aarch64_sm3tt1aq4si): Ditto.
20626 (aarch64_sm3tt1bq4si): Ditto.
20627 (aarch64_sm3tt2aq4si): Ditto.
20628 (aarch64_sm3tt2bq4si): Ditto.
20629 (aarch64_sm3partw1qv4si): Ditto.
20630 (aarch64_sm3partw2qv4si): Ditto.
20631 (aarch64_sm4eqv4si): Ditto.
20632 (aarch64_sm4ekeyqv4si): Ditto.
20633 * config/aarch64/aarch64-simd.md:
20634 (aarch64_sm3ss1qv4si): Ditto.
20635 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
20636 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
20637 (aarch64_sm4eqv4si): Ditto.
20638 (aarch64_sm4ekeyqv4si): Ditto.
20639 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
20640 (sm3part_op): Ditto.
20641 (CRYPTO_SM3TT): Ditto.
20642 (CRYPTO_SM3PART): Ditto.
20643 (UNSPEC_SM3SS1): New unspec.
20644 (UNSPEC_SM3TT1A): Ditto.
20645 (UNSPEC_SM3TT1B): Ditto.
20646 (UNSPEC_SM3TT2A): Ditto.
20647 (UNSPEC_SM3TT2B): Ditto.
20648 (UNSPEC_SM3PARTW1): Ditto.
20649 (UNSPEC_SM3PARTW2): Ditto.
20650 (UNSPEC_SM4E): Ditto.
20651 (UNSPEC_SM4EKEY): Ditto.
20652 * config/aarch64/constraints.md (Ui2): New constraint.
20653 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
20654 * config/arm/types.md (crypto_sm3): New type attribute.
20655 (crypto_sm4): Ditto.
20656 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
20657 (vsm3tt1aq_u32): Ditto.
20658 (vsm3tt1bq_u32): Ditto.
20659 (vsm3tt2aq_u32): Ditto.
20660 (vsm3tt2bq_u32): Ditto.
20661 (vsm3partw1q_u32): Ditto.
20662 (vsm3partw2q_u32): Ditto.
20663 (vsm4eq_u32): Ditto.
20664 (vsm4ekeyq_u32): Ditto.
20665 (doc/invoke.texi): Document new sm4 option.
20667 2018-01-10 Michael Collison <michael.collison@arm.com>
20669 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
20670 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
20671 (AARCH64_FL_FOR_ARCH8_4): New.
20672 (AARCH64_FL_V8_4): New flag.
20673 (doc/invoke.texi): Document new armv8.4-a option.
20675 2018-01-10 Michael Collison <michael.collison@arm.com>
20677 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
20678 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
20679 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
20680 * config/aarch64/aarch64-option-extension.def: Add
20681 AARCH64_OPT_EXTENSION of 'sha2'.
20682 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
20683 (crypto): Disable sha2 and aes if crypto disabled.
20684 (crypto): Enable aes and sha2 if enabled.
20685 (simd): Disable sha2 and aes if simd disabled.
20686 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
20688 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
20689 (TARGET_SHA2): New feature flag for sha2.
20690 (TARGET_AES): New feature flag for aes.
20691 * config/aarch64/aarch64-simd.md:
20692 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
20693 conditional on TARGET_AES.
20694 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
20695 (aarch64_crypto_sha1hsi): Make pattern conditional
20697 (aarch64_crypto_sha1hv4si): Ditto.
20698 (aarch64_be_crypto_sha1hv4si): Ditto.
20699 (aarch64_crypto_sha1su1v4si): Ditto.
20700 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
20701 (aarch64_crypto_sha1su0v4si): Ditto.
20702 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
20703 (aarch64_crypto_sha256su0v4si): Ditto.
20704 (aarch64_crypto_sha256su1v4si): Ditto.
20705 (doc/invoke.texi): Document new aes and sha2 options.
20707 2018-01-10 Martin Sebor <msebor@redhat.com>
20709 PR tree-optimization/83781
20710 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
20713 2018-01-11 Martin Sebor <msebor@gmail.com>
20714 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
20716 PR tree-optimization/83501
20717 PR tree-optimization/81703
20719 * tree-ssa-strlen.c (get_string_cst): Rename...
20720 (get_string_len): ...to this. Handle global constants.
20721 (handle_char_store): Adjust.
20723 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
20724 Jim Wilson <jimw@sifive.com>
20726 * config/riscv/riscv-protos.h (riscv_output_return): New.
20727 * config/riscv/riscv.c (struct machine_function): New naked_p field.
20728 (riscv_attribute_table, riscv_output_return),
20729 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
20730 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
20731 (riscv_compute_frame_info): Only compute frame->mask if not a naked
20733 (riscv_expand_prologue): Add early return for naked function.
20734 (riscv_expand_epilogue): Likewise.
20735 (riscv_function_ok_for_sibcall): Return false for naked function.
20736 (riscv_set_current_function): New.
20737 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
20738 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
20739 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
20740 * doc/extend.texi (RISC-V Function Attributes): New.
20742 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
20744 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
20745 check for 128-bit long double before checking TCmode.
20746 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
20747 128-bit long doubles before checking TFmode or TCmode.
20748 (FLOAT128_IBM_P): Likewise.
20750 2018-01-10 Martin Sebor <msebor@redhat.com>
20752 PR tree-optimization/83671
20753 * builtins.c (c_strlen): Unconditionally return zero for the empty
20755 Use -Warray-bounds for warnings.
20756 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
20757 for non-constant array indices with COMPONENT_REF, arrays of
20758 arrays, and pointers to arrays.
20759 (gimple_fold_builtin_strlen): Determine and set length range for
20760 non-constant character arrays.
20762 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
20764 PR middle-end/81897
20765 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
20768 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
20770 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
20772 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
20775 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
20776 VECTOR_MEM_ALTIVEC_OR_VSX_P.
20777 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
20778 indexed_or_indirect_operand predicate.
20779 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
20780 (*vsx_le_perm_load_v8hi): Likewise.
20781 (*vsx_le_perm_load_v16qi): Likewise.
20782 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
20783 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
20784 (*vsx_le_perm_store_v8hi): Likewise.
20785 (*vsx_le_perm_store_v16qi): Likewise.
20786 (eight unnamed splitters): Likewise.
20788 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
20790 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
20791 * config/rs6000/emmintrin.h: Likewise.
20792 * config/rs6000/mmintrin.h: Likewise.
20793 * config/rs6000/xmmintrin.h: Likewise.
20795 2018-01-10 David Malcolm <dmalcolm@redhat.com>
20798 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
20800 * tree.c (tree_nop_conversion): Return true for location wrapper
20802 (maybe_wrap_with_location): New function.
20803 (selftest::check_strip_nops): New function.
20804 (selftest::test_location_wrappers): New function.
20805 (selftest::tree_c_tests): Call it.
20806 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
20807 (maybe_wrap_with_location): New decl.
20808 (EXPR_LOCATION_WRAPPER_P): New macro.
20809 (location_wrapper_p): New inline function.
20810 (tree_strip_any_location_wrapper): New inline function.
20812 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
20815 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
20816 stack_realign_offset for the largest alignment of stack slot
20818 (ix86_find_max_used_stack_alignment): New function.
20819 (ix86_finalize_stack_frame_flags): Use it. Set
20820 max_used_stack_alignment if we don't realign stack.
20821 * config/i386/i386.h (machine_function): Add
20822 max_used_stack_alignment.
20824 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
20826 * config/arm/arm.opt (-mbranch-cost): New option.
20827 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
20830 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
20833 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
20834 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
20836 2018-01-10 Richard Biener <rguenther@suse.de>
20839 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
20840 early out so it also covers the case where we have a non-NULL
20843 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
20845 PR tree-optimization/83753
20846 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
20847 for non-strided grouped accesses if the number of elements is 1.
20849 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
20852 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
20853 * i386.h (TARGET_USE_GATHER): Define.
20854 * x86-tune.def (X86_TUNE_USE_GATHER): New.
20856 2018-01-10 Martin Liska <mliska@suse.cz>
20859 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
20860 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
20862 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
20863 CLEANUP_NO_PARTITIONING is not set.
20865 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
20867 * doc/rtl.texi: Remove documentation of (const ...) wrappers
20868 for vectors, as a partial revert of r254296.
20869 * rtl.h (const_vec_p): Delete.
20870 (const_vec_duplicate_p): Don't test for vector CONSTs.
20871 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
20872 * expmed.c (make_tree): Likewise.
20875 * common.md (E, F): Use CONSTANT_P instead of checking for
20877 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
20878 checking for CONST_VECTOR.
20880 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
20882 PR middle-end/83575
20883 * predict.c (force_edge_cold): Handle in more sane way edges
20884 with no prediction.
20886 2018-01-09 Carl Love <cel@us.ibm.com>
20888 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
20890 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
20891 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
20892 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
20893 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
20894 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
20895 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
20896 * config/rs6000/rs6000-protos.h: Add extern defition for
20897 rs6000_generate_float2_double_code.
20898 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
20900 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
20901 (float2_v2df): Add define_expand.
20903 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
20906 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
20907 op_mode in the force_to_mode call.
20909 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
20911 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
20912 instead of checking each element individually.
20913 (aarch64_evpc_uzp): Likewise.
20914 (aarch64_evpc_zip): Likewise.
20915 (aarch64_evpc_ext): Likewise.
20916 (aarch64_evpc_rev): Likewise.
20917 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
20918 instead of checking each element individually. Return true without
20920 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
20921 whether all selected elements come from the same input, instead of
20922 checking each element individually. Remove calls to gen_rtx_REG,
20923 start_sequence and end_sequence and instead assert that no rtl is
20926 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
20928 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
20929 order of HIGH and CONST checks.
20931 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
20933 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
20934 if the destination isn't an SSA_NAME.
20936 2018-01-09 Richard Biener <rguenther@suse.de>
20938 PR tree-optimization/83668
20939 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
20941 (canonicalize_loop_form): ... here, renamed from ...
20942 (canonicalize_loop_closed_ssa_form): ... this and amended to
20943 swap successor edges for loop exit blocks to make us use
20944 the RPO order we need for initial schedule generation.
20946 2018-01-09 Joseph Myers <joseph@codesourcery.com>
20948 PR tree-optimization/64811
20949 * match.pd: When optimizing comparisons with Inf, avoid
20950 introducing or losing exceptions from comparisons with NaN.
20952 2018-01-09 Martin Liska <mliska@suse.cz>
20955 * asan.c (shadow_mem_size): Add gcc_assert.
20957 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
20959 Don't save registers in main().
20962 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
20963 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
20964 * config/avr/avr.c (avr_set_current_function): Don't error if
20965 naked, OS_task or OS_main are specified at the same time.
20966 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
20968 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
20970 * common/config/avr/avr-common.c (avr_option_optimization_table):
20971 Switch on -mmain-is-OS_task for optimizing compilations.
20973 2018-01-09 Richard Biener <rguenther@suse.de>
20975 PR tree-optimization/83572
20976 * graphite.c: Include cfganal.h.
20977 (graphite_transform_loops): Connect infinite loops to exit
20978 and remove fake edges at the end.
20980 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
20982 * ipa-inline.c (edge_badness): Revert accidental checkin.
20984 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
20987 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
20988 symbols; not inline clones.
20990 2018-01-09 Jakub Jelinek <jakub@redhat.com>
20993 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
20994 hard registers. Formatting fixes.
20996 PR preprocessor/83722
20997 * gcc.c (try_generate_repro): Pass
20998 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
20999 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
21002 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
21003 Kito Cheng <kito.cheng@gmail.com>
21005 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
21006 (riscv_leaf_function_p): Delete.
21007 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
21009 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
21011 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
21013 (do_ifelse): New function.
21014 (do_isel): New function.
21015 (do_sub3): New function.
21016 (do_add3): New function.
21017 (do_load_mask_compare): New function.
21018 (do_overlap_load_compare): New function.
21019 (expand_compare_loop): New function.
21020 (expand_block_compare): Call expand_compare_loop() when appropriate.
21021 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
21022 option description.
21023 (-mblock-compare-inline-loop-limit): New option.
21025 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
21028 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
21029 Reverse order of second and third operands in first alternative.
21030 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
21031 of first and second elements in UNSPEC_VPERMR vector.
21032 (altivec_expand_vec_perm_le): Likewise.
21034 2018-01-08 Jeff Law <law@redhat.com>
21036 PR rtl-optimizatin/81308
21037 * tree-switch-conversion.c (cfg_altered): New file scoped static.
21038 (process_switch): If group_case_labels makes a change, then set
21040 (pass_convert_switch::execute): If a switch is converted, then
21041 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
21043 PR rtl-optimization/81308
21044 * recog.c (split_all_insns): Conditionally cleanup the CFG after
21047 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
21049 PR target/83663 - Revert r255946
21050 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
21051 generation for cases where splatting a value is not useful.
21052 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
21053 across a vec_duplicate and a paradoxical subreg forming a vector
21054 mode to a vec_concat.
21056 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21058 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
21059 -march=armv8.3-a variants.
21060 * config/arm/t-multilib: Likewise.
21061 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
21063 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
21065 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
21067 (cceq_ior_compare_complement): Give it a name so I can use it, and
21068 change boolean_or_operator predicate to boolean_operator so it can
21069 be used to generate a crand.
21070 (eqne): New code iterator.
21071 (bd/bd_neg): New code_attrs.
21072 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
21073 a single define_insn.
21074 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
21075 decrement (bdnzt/bdnzf/bdzt/bdzf).
21076 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
21077 with the new names of the branch decrement patterns, and added the
21078 names of the branch decrement conditional patterns.
21080 2018-01-08 Richard Biener <rguenther@suse.de>
21082 PR tree-optimization/83563
21083 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
21086 2018-01-08 Richard Biener <rguenther@suse.de>
21088 PR middle-end/83713
21089 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
21091 2018-01-08 Richard Biener <rguenther@suse.de>
21093 PR tree-optimization/83685
21094 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
21095 references to abnormals.
21097 2018-01-08 Richard Biener <rguenther@suse.de>
21100 * dwarf2out.c (output_indirect_strings): Handle empty
21101 skeleton_debug_str_hash.
21102 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
21104 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
21106 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
21107 (emit_store_direct): Likewise.
21108 (arc_trampoline_adjust_address): Likewise.
21109 (arc_asm_trampoline_template): New function.
21110 (arc_initialize_trampoline): Use asm_trampoline_template.
21111 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
21112 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
21113 * config/arc/arc.md (flush_icache): Delete pattern.
21115 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
21117 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
21118 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
21121 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
21124 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
21125 by not USED_FOR_TARGET.
21126 (make_pass_resolve_sw_modes): Likewise.
21128 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
21130 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
21133 2018-01-08 Richard Biener <rguenther@suse.de>
21135 PR middle-end/83580
21136 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
21138 2018-01-08 Richard Biener <rguenther@suse.de>
21140 PR middle-end/83517
21141 * match.pd ((t * 2) / 2) -> t): Add missing :c.
21143 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
21145 PR middle-end/81897
21146 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
21147 basic blocks with a small number of successors.
21148 (convert_control_dep_chain_into_preds): Improve handling of
21150 (dump_predicates): Split apart into...
21151 (dump_pred_chain): ...here...
21152 (dump_pred_info): ...and here.
21153 (can_one_predicate_be_invalidated_p): Add debugging printfs.
21154 (can_chain_union_be_invalidated_p): Improve check for invalidation
21156 (uninit_uses_cannot_happen): Avoid unnecessary if
21157 convert_control_dep_chain_into_preds yielded nothing.
21159 2018-01-06 Martin Sebor <msebor@redhat.com>
21161 PR tree-optimization/83640
21162 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
21163 subtracting negative offset from size.
21164 (builtin_access::overlap): Adjust offset bounds of the access to fall
21165 within the size of the object if possible.
21167 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
21169 PR rtl-optimization/83699
21170 * expmed.c (extract_bit_field_1): Restrict the vector usage of
21171 extract_bit_field_as_subreg to cases in which the extracted
21172 value is also a vector.
21174 * lra-constraints.c (process_alt_operands): Test for the equivalence
21175 substitutions when detecting a possible reload cycle.
21177 2018-01-06 Jakub Jelinek <jakub@redhat.com>
21180 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
21181 by default if flag_selective_schedling{,2}. Formatting fixes.
21183 PR rtl-optimization/83682
21184 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
21185 if it has non-VECTOR_MODE element mode.
21186 (vec_duplicate_p): Likewise.
21188 PR middle-end/83694
21189 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
21190 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
21192 2018-01-05 Jakub Jelinek <jakub@redhat.com>
21195 * config/i386/i386-builtin.def
21196 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
21197 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
21198 Require also OPTION_MASK_ISA_AVX512F in addition to
21199 OPTION_MASK_ISA_GFNI.
21200 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
21201 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
21202 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
21203 to OPTION_MASK_ISA_GFNI.
21204 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
21205 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
21206 OPTION_MASK_ISA_AVX512BW.
21207 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
21208 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
21209 addition to OPTION_MASK_ISA_GFNI.
21210 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
21211 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
21212 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
21213 to OPTION_MASK_ISA_GFNI.
21214 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
21215 a requirement for all ISAs rather than any of them with a few
21217 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
21219 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
21220 bitmasks to be enabled with 3 exceptions, instead of requiring any
21221 enabled ISA with lots of exceptions.
21222 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
21223 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
21224 Change avx512bw in isa attribute to avx512f.
21225 * config/i386/sgxintrin.h: Add license boilerplate.
21226 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
21227 to __AVX512F__ and __AVX512VL to __AVX512VL__.
21228 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
21229 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
21231 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
21232 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
21233 temporarily sse2 rather than sse if not enabled already.
21236 * config/i386/sse.md (VI248_VLBW): Rename to ...
21237 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
21238 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
21239 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
21240 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
21241 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
21242 mode iterator instead of VI248_VLBW.
21244 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
21246 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
21247 (record_modified): Skip clobbers; add debug output.
21248 (param_change_prob): Use sreal frequencies.
21250 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
21252 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
21253 punt for user-aligned variables.
21255 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
21257 * tree-chrec.c (chrec_contains_symbols): Return true for
21260 2018-01-05 Sudakshina Das <sudi.das@arm.com>
21263 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
21264 of (x|y) == x for BICS pattern.
21266 2018-01-05 Jakub Jelinek <jakub@redhat.com>
21268 PR tree-optimization/83605
21269 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
21270 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
21273 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
21275 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
21276 * config/epiphany/rtems.h: New file.
21278 2018-01-04 Jakub Jelinek <jakub@redhat.com>
21279 Uros Bizjak <ubizjak@gmail.com>
21282 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
21283 QIreg_operand instead of register_operand predicate.
21284 * config/i386/i386.c (ix86_rop_should_change_byte_p,
21285 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
21286 comments instead of -fmitigate[-_]rop.
21288 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
21291 * cgraphunit.c (symbol_table::compile): Switch to text_section
21292 before calling assembly_start debug hook.
21293 * run-rtl-passes.c (run_rtl_passes): Likewise.
21296 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
21298 * tree-vrp.c (extract_range_from_binary_expr_1): Check
21299 range_int_cst_p rather than !symbolic_range_p before calling
21300 extract_range_from_multiplicative_op_1.
21302 2018-01-04 Jeff Law <law@redhat.com>
21304 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
21305 redundant test in assertion.
21307 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
21309 * doc/rtl.texi: Document machine_mode wrapper classes.
21311 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
21313 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
21314 using tree_to_uhwi.
21316 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
21318 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
21319 the VEC_PERM_EXPR fold to fail.
21321 2018-01-04 Jakub Jelinek <jakub@redhat.com>
21324 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
21325 to switched_sections.
21327 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
21330 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
21331 test for d.testing.
21333 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
21336 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
21337 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
21339 2018-01-04 Jakub Jelinek <jakub@redhat.com>
21342 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
21343 is BLKmode and bitpos not zero or mode change is needed.
21345 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
21348 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
21351 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
21354 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
21355 instead of MULT rtx. Update all corresponding splitters.
21356 (*saddl_se): Ditto.
21357 (*ssub<modesuffix>): Ditto.
21358 (*ssubl_se): Ditto.
21359 (*cmp_sadd_di): Update split patterns.
21360 (*cmp_sadd_si): Ditto.
21361 (*cmp_sadd_sidi): Ditto.
21362 (*cmp_ssub_di): Ditto.
21363 (*cmp_ssub_si): Ditto.
21364 (*cmp_ssub_sidi): Ditto.
21365 * config/alpha/predicates.md (const23_operand): New predicate.
21366 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
21367 Look for ASHIFT, not MULT inner operand.
21368 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
21370 2018-01-04 Martin Liska <mliska@suse.cz>
21372 PR gcov-profile/83669
21373 * gcov.c (output_intermediate_file): Add version to intermediate
21375 * doc/gcov.texi: Document new field 'version' in intermediate
21376 file format. Fix location of '-k' option of gcov command.
21378 2018-01-04 Martin Liska <mliska@suse.cz>
21381 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
21383 2018-01-04 Jakub Jelinek <jakub@redhat.com>
21385 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
21387 2018-01-03 Martin Sebor <msebor@redhat.com>
21389 PR tree-optimization/83655
21390 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
21391 checking calls with invalid arguments.
21393 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21395 * tree-vect-stmts.c (vect_get_store_rhs): New function.
21396 (vectorizable_mask_load_store): Delete.
21397 (vectorizable_call): Return false for masked loads and stores.
21398 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
21399 instead of gimple_assign_rhs1.
21400 (vectorizable_load): Handle IFN_MASK_LOAD.
21401 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
21403 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21405 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
21407 (vectorizable_mask_load_store): ...here.
21408 (vectorizable_load): ...and here.
21410 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21412 * tree-vect-stmts.c (vect_build_all_ones_mask)
21413 (vect_build_zero_merge_argument): New functions, split out from...
21414 (vectorizable_load): ...here.
21416 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21418 * tree-vect-stmts.c (vect_check_store_rhs): New function,
21420 (vectorizable_mask_load_store): ...here.
21421 (vectorizable_store): ...and here.
21423 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21425 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
21427 (vectorizable_mask_load_store): ...here.
21429 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21431 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
21432 (vect_model_store_cost): Take a vec_load_store_type instead of a
21434 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
21435 (vect_model_store_cost): Take a vec_load_store_type instead of a
21437 (vectorizable_mask_load_store): Update accordingly.
21438 (vectorizable_store): Likewise.
21439 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
21441 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21443 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
21444 IFN_MASK_LOAD calls here rather than...
21445 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
21447 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21448 Alan Hayward <alan.hayward@arm.com>
21449 David Sherwood <david.sherwood@arm.com>
21451 * expmed.c (extract_bit_field_1): For vector extracts,
21452 fall back to extract_bit_field_as_subreg if vec_extract
21455 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21456 Alan Hayward <alan.hayward@arm.com>
21457 David Sherwood <david.sherwood@arm.com>
21459 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
21460 they are variable or constant sized.
21461 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
21462 slots for constant-sized data.
21464 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21465 Alan Hayward <alan.hayward@arm.com>
21466 David Sherwood <david.sherwood@arm.com>
21468 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
21469 handling COND_EXPRs with boolean comparisons, try to find a better
21470 basis for the mask type than the boolean itself.
21472 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21474 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
21475 is calculated and how it can be overridden.
21476 * genmodes.c (max_bitsize_mode_any_mode): New variable.
21477 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
21479 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
21482 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21483 Alan Hayward <alan.hayward@arm.com>
21484 David Sherwood <david.sherwood@arm.com>
21486 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
21487 Remove the mode argument.
21488 (aarch64_simd_valid_immediate): Remove the mode and inverse
21490 * config/aarch64/iterators.md (bitsize): New iterator.
21491 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
21492 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
21493 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
21494 aarch64_simd_valid_immediate.
21495 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
21496 (aarch64_reg_or_bic_imm): Likewise.
21497 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
21498 with an insn_type enum and msl with a modifier_type enum.
21499 Replace element_width with a scalar_mode. Change the shift
21500 to unsigned int. Add constructors for scalar_float_mode and
21501 scalar_int_mode elements.
21502 (aarch64_vect_float_const_representable_p): Delete.
21503 (aarch64_can_const_movi_rtx_p)
21504 (aarch64_simd_scalar_immediate_valid_for_move)
21505 (aarch64_simd_make_constant): Update call to
21506 aarch64_simd_valid_immediate.
21507 (aarch64_advsimd_valid_immediate_hs): New function.
21508 (aarch64_advsimd_valid_immediate): Likewise.
21509 (aarch64_simd_valid_immediate): Remove mode and inverse
21510 arguments. Rewrite to use the above. Use const_vec_duplicate_p
21511 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
21512 and aarch64_float_const_representable_p on the result.
21513 (aarch64_output_simd_mov_immediate): Remove mode argument.
21514 Update call to aarch64_simd_valid_immediate and use of
21515 simd_immediate_info.
21516 (aarch64_output_scalar_simd_mov_immediate): Update call
21519 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21520 Alan Hayward <alan.hayward@arm.com>
21521 David Sherwood <david.sherwood@arm.com>
21523 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
21524 (mode_nunits): Likewise CONST_MODE_NUNITS.
21525 * machmode.def (ADJUST_NUNITS): Document.
21526 * genmodes.c (mode_data::need_nunits_adj): New field.
21527 (blank_mode): Update accordingly.
21528 (adj_nunits): New variable.
21529 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
21531 (emit_mode_size_inline): Set need_bytesize_adj for all modes
21532 listed in adj_nunits.
21533 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
21534 listed in adj_nunits. Don't emit case statements for such modes.
21535 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
21536 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
21537 nothing if adj_nunits is nonnull.
21538 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
21539 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
21540 (emit_mode_fbit): Update use of print_maybe_const_decl.
21541 (emit_move_size): Likewise. Treat the array as non-const
21543 (emit_mode_adjustments): Handle adj_nunits.
21545 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21547 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
21548 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
21549 (VECTOR_MODES): Use it.
21550 (make_vector_modes): Take the prefix as an argument.
21552 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21553 Alan Hayward <alan.hayward@arm.com>
21554 David Sherwood <david.sherwood@arm.com>
21556 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
21557 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
21558 for MODE_VECTOR_BOOL.
21559 * machmode.def (VECTOR_BOOL_MODE): Document.
21560 * genmodes.c (VECTOR_BOOL_MODE): New macro.
21561 (make_vector_bool_mode): New function.
21562 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
21564 * lto-streamer-in.c (lto_input_mode_table): Likewise.
21565 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
21567 * stor-layout.c (int_mode_for_mode): Likewise.
21568 * tree.c (build_vector_type_for_mode): Likewise.
21569 * varasm.c (output_constant_pool_2): Likewise.
21570 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
21571 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
21572 for MODE_VECTOR_BOOL.
21573 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
21574 of mode class checks.
21575 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
21576 instead of a list of mode class checks.
21577 (expand_vector_scalar_condition): Likewise.
21578 (type_for_widest_vector_mode): Handle BImode as an inner mode.
21580 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21581 Alan Hayward <alan.hayward@arm.com>
21582 David Sherwood <david.sherwood@arm.com>
21584 * machmode.h (mode_size): Change from unsigned short to
21586 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
21587 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
21588 or if measurement_type is not polynomial.
21589 (fixed_size_mode::includes_p): Check for constant-sized modes.
21590 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
21591 return a poly_uint16 rather than an unsigned short.
21592 (emit_mode_size): Change the type of mode_size from unsigned short
21593 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
21594 (emit_mode_adjustments): Cope with polynomial vector sizes.
21595 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
21597 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
21599 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
21600 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
21601 * caller-save.c (setup_save_areas): Likewise.
21602 (replace_reg_with_saved_mem): Likewise.
21603 * calls.c (emit_library_call_value_1): Likewise.
21604 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
21605 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
21606 (gen_lowpart_for_combine): Likewise.
21607 * convert.c (convert_to_integer_1): Likewise.
21608 * cse.c (equiv_constant, cse_insn): Likewise.
21609 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
21610 (cselib_subst_to_values): Likewise.
21611 * dce.c (word_dce_process_block): Likewise.
21612 * df-problems.c (df_word_lr_mark_ref): Likewise.
21613 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
21614 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
21615 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
21616 (rtl_for_decl_location): Likewise.
21617 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
21618 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
21619 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
21620 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
21621 (expand_expr_real_1): Likewise.
21622 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
21623 (pad_below): Likewise.
21624 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
21625 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
21626 * ira.c (get_subreg_tracking_sizes): Likewise.
21627 * ira-build.c (ira_create_allocno_objects): Likewise.
21628 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
21629 (ira_sort_regnos_for_alter_reg): Likewise.
21630 * ira-costs.c (record_operand_costs): Likewise.
21631 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
21632 (resolve_simple_move): Likewise.
21633 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
21634 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
21635 (lra_constraints): Likewise.
21636 (CONST_POOL_OK_P): Reject variable-sized modes.
21637 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
21638 (add_pseudo_to_slot, lra_spill): Likewise.
21639 * omp-low.c (omp_clause_aligned_alignment): Likewise.
21640 * optabs-query.c (get_best_extraction_insn): Likewise.
21641 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
21642 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
21643 (expand_mult_highpart, valid_multiword_target_p): Likewise.
21644 * recog.c (offsettable_address_addr_space_p): Likewise.
21645 * regcprop.c (maybe_mode_change): Likewise.
21646 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
21647 * regrename.c (build_def_use): Likewise.
21648 * regstat.c (dump_reg_info): Likewise.
21649 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
21650 (find_reloads, find_reloads_subreg_address): Likewise.
21651 * reload1.c (eliminate_regs_1): Likewise.
21652 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
21653 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
21654 (simplify_binary_operation_1, simplify_subreg): Likewise.
21655 * targhooks.c (default_function_arg_padding): Likewise.
21656 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
21657 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
21658 (verify_gimple_assign_ternary): Likewise.
21659 * tree-inline.c (estimate_move_cost): Likewise.
21660 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
21661 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
21662 (get_address_cost_ainc): Likewise.
21663 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
21664 (vect_supportable_dr_alignment): Likewise.
21665 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
21666 (vectorizable_reduction): Likewise.
21667 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
21668 (vectorizable_operation, vectorizable_load): Likewise.
21669 * tree.c (build_same_sized_truth_vector_type): Likewise.
21670 * valtrack.c (cleanup_auto_inc_dec): Likewise.
21671 * var-tracking.c (emit_note_insn_var_location): Likewise.
21672 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
21673 (ADDR_VEC_ALIGN): Likewise.
21675 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21676 Alan Hayward <alan.hayward@arm.com>
21677 David Sherwood <david.sherwood@arm.com>
21679 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
21681 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
21682 or if measurement_type is polynomial.
21683 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
21684 * combine.c (make_extraction): Likewise.
21685 * dse.c (find_shift_sequence): Likewise.
21686 * dwarf2out.c (mem_loc_descriptor): Likewise.
21687 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
21688 (extract_bit_field, extract_low_bits): Likewise.
21689 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
21690 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
21691 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
21692 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
21693 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
21694 * reload.c (find_reloads): Likewise.
21695 * reload1.c (alter_reg): Likewise.
21696 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
21697 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
21698 * tree-if-conv.c (predicate_mem_writes): Likewise.
21699 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
21700 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
21701 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
21702 * valtrack.c (dead_debug_insert_temp): Likewise.
21703 * varasm.c (mergeable_constant_section): Likewise.
21704 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
21706 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21707 Alan Hayward <alan.hayward@arm.com>
21708 David Sherwood <david.sherwood@arm.com>
21710 * expr.c (expand_assignment): Cope with polynomial mode sizes
21711 when assigning to a CONCAT.
21713 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21714 Alan Hayward <alan.hayward@arm.com>
21715 David Sherwood <david.sherwood@arm.com>
21717 * machmode.h (mode_precision): Change from unsigned short to
21719 (mode_to_precision): Return a poly_uint16 rather than an unsigned
21721 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
21722 or if measurement_type is not polynomial.
21723 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
21724 in which the mode is already known to be a scalar_int_mode.
21725 * genmodes.c (emit_mode_precision): Change the type of mode_precision
21726 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
21728 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
21729 for GET_MODE_PRECISION.
21730 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
21731 for GET_MODE_PRECISION.
21732 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
21734 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
21735 (expand_field_assignment, make_extraction): Likewise.
21736 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
21737 (get_last_value): Likewise.
21738 * convert.c (convert_to_integer_1): Likewise.
21739 * cse.c (cse_insn): Likewise.
21740 * expr.c (expand_expr_real_1): Likewise.
21741 * lra-constraints.c (simplify_operand_subreg): Likewise.
21742 * optabs-query.c (can_atomic_load_p): Likewise.
21743 * optabs.c (expand_atomic_load): Likewise.
21744 (expand_atomic_store): Likewise.
21745 * ree.c (combine_reaching_defs): Likewise.
21746 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
21747 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
21748 * tree.h (type_has_mode_precision_p): Likewise.
21749 * ubsan.c (instrument_si_overflow): Likewise.
21751 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21752 Alan Hayward <alan.hayward@arm.com>
21753 David Sherwood <david.sherwood@arm.com>
21755 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
21756 polynomial numbers of units.
21757 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
21758 (valid_vector_subparts_p): New function.
21759 (build_vector_type): Remove temporary shim and take the number
21760 of units as a poly_uint64 rather than an int.
21761 (build_opaque_vector_type): Take the number of units as a
21762 poly_uint64 rather than an int.
21763 * tree.c (build_vector_from_ctor): Handle polynomial
21764 TYPE_VECTOR_SUBPARTS.
21765 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
21766 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
21767 (build_vector_from_val): If the number of units is variable,
21768 use build_vec_duplicate_cst for constant operands and
21769 VEC_DUPLICATE_EXPR otherwise.
21770 (make_vector_type): Remove temporary is_constant ().
21771 (build_vector_type, build_opaque_vector_type): Take the number of
21772 units as a poly_uint64 rather than an int.
21773 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
21775 * cfgexpand.c (expand_debug_expr): Likewise.
21776 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
21777 (store_constructor, expand_expr_real_1): Likewise.
21778 (const_scalar_mask_from_tree): Likewise.
21779 * fold-const-call.c (fold_const_reduction): Likewise.
21780 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
21781 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
21782 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
21783 (fold_relational_const): Likewise.
21784 (native_interpret_vector): Likewise. Change the size from an
21785 int to an unsigned int.
21786 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
21787 TYPE_VECTOR_SUBPARTS.
21788 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
21789 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
21790 duplicating a non-constant operand into a variable-length vector.
21791 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
21792 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
21793 * ipa-icf.c (sem_variable::equals): Likewise.
21794 * match.pd: Likewise.
21795 * omp-simd-clone.c (simd_clone_subparts): Likewise.
21796 * print-tree.c (print_node): Likewise.
21797 * stor-layout.c (layout_type): Likewise.
21798 * targhooks.c (default_builtin_vectorization_cost): Likewise.
21799 * tree-cfg.c (verify_gimple_comparison): Likewise.
21800 (verify_gimple_assign_binary): Likewise.
21801 (verify_gimple_assign_ternary): Likewise.
21802 (verify_gimple_assign_single): Likewise.
21803 * tree-pretty-print.c (dump_generic_node): Likewise.
21804 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
21805 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
21806 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
21807 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
21808 (vect_shift_permute_load_chain): Likewise.
21809 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
21810 (expand_vector_condition, optimize_vector_constructor): Likewise.
21811 (lower_vec_perm, get_compute_type): Likewise.
21812 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
21813 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
21814 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
21815 (vect_recog_mask_conversion_pattern): Likewise.
21816 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
21817 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
21818 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
21819 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
21820 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
21821 (vectorizable_shift, vectorizable_operation, vectorizable_store)
21822 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
21823 (supportable_widening_operation): Likewise.
21824 (supportable_narrowing_operation): Likewise.
21825 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
21827 * varasm.c (output_constant): Likewise.
21829 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21830 Alan Hayward <alan.hayward@arm.com>
21831 David Sherwood <david.sherwood@arm.com>
21833 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
21834 so that both the length == 3 and length != 3 cases set up their
21835 own permute vectors. Add comments explaining why we know the
21836 number of elements is constant.
21837 (vect_permute_load_chain): Likewise.
21839 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21840 Alan Hayward <alan.hayward@arm.com>
21841 David Sherwood <david.sherwood@arm.com>
21843 * machmode.h (mode_nunits): Change from unsigned char to
21845 (ONLY_FIXED_SIZE_MODES): New macro.
21846 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
21847 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
21848 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
21850 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
21851 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
21852 or if measurement_type is not polynomial.
21853 * genmodes.c (ZERO_COEFFS): New macro.
21854 (emit_mode_nunits_inline): Make mode_nunits_inline return a
21856 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
21857 Use ZERO_COEFFS when emitting initializers.
21858 * data-streamer.h (bp_pack_poly_value): New function.
21859 (bp_unpack_poly_value): Likewise.
21860 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
21861 for GET_MODE_NUNITS.
21862 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
21863 for GET_MODE_NUNITS.
21864 * tree.c (make_vector_type): Remove temporary shim and make
21865 the real function take the number of units as a poly_uint64
21866 rather than an int.
21867 (build_vector_type_for_mode): Handle polynomial nunits.
21868 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
21869 * emit-rtl.c (const_vec_series_p_1): Likewise.
21870 (gen_rtx_CONST_VECTOR): Likewise.
21871 * fold-const.c (test_vec_duplicate_folding): Likewise.
21872 * genrecog.c (validate_pattern): Likewise.
21873 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
21874 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
21875 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
21876 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
21877 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
21878 * rtlanal.c (subreg_get_info): Likewise.
21879 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
21880 (vect_grouped_load_supported): Likewise.
21881 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
21882 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
21883 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
21884 (simplify_const_unary_operation, simplify_binary_operation_1)
21885 (simplify_const_binary_operation, simplify_ternary_operation)
21886 (test_vector_ops_duplicate, test_vector_ops): Likewise.
21887 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
21888 instead of CONST_VECTOR_NUNITS.
21889 * varasm.c (output_constant_pool_2): Likewise.
21890 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
21891 explicit-encoded elements in the XVEC for variable-length vectors.
21893 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21895 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
21897 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21898 Alan Hayward <alan.hayward@arm.com>
21899 David Sherwood <david.sherwood@arm.com>
21901 * coretypes.h (fixed_size_mode): Declare.
21902 (fixed_size_mode_pod): New typedef.
21903 * builtins.h (target_builtins::x_apply_args_mode)
21904 (target_builtins::x_apply_result_mode): Change type to
21905 fixed_size_mode_pod.
21906 * builtins.c (apply_args_size, apply_result_size, result_vector)
21907 (expand_builtin_apply_args_1, expand_builtin_apply)
21908 (expand_builtin_return): Update accordingly.
21910 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21912 * cse.c (hash_rtx_cb): Hash only the encoded elements.
21913 * cselib.c (cselib_hash_rtx): Likewise.
21914 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
21915 CONST_VECTOR encoding.
21917 2018-01-03 Jakub Jelinek <jakub@redhat.com>
21918 Jeff Law <law@redhat.com>
21921 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
21922 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
21923 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
21924 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
21927 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
21928 explicitly probe *sp in a noreturn function if there were any callee
21929 register saves or frame pointer is needed.
21931 2018-01-03 Jakub Jelinek <jakub@redhat.com>
21934 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
21935 BLKmode for ternary, binary or unary expressions.
21938 * var-tracking.c (delete_vta_debug_insn): New inline function.
21939 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
21940 insns from get_insns () to NULL instead of each bb separately.
21941 Use delete_vta_debug_insn. No longer static.
21942 (vt_debug_insns_local, variable_tracking_main_1): Adjust
21943 delete_vta_debug_insns callers.
21944 * rtl.h (delete_vta_debug_insns): Declare.
21945 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
21946 instead of variable_tracking_main.
21948 2018-01-03 Martin Sebor <msebor@redhat.com>
21950 PR tree-optimization/83603
21951 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
21952 arguments past the endof the argument list in functions declared
21953 without a prototype.
21954 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
21955 Avoid checking when arguments are null.
21957 2018-01-03 Martin Sebor <msebor@redhat.com>
21960 * doc/extend.texi (attribute const): Fix a typo.
21961 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
21962 issuing -Wsuggest-attribute for void functions.
21964 2018-01-03 Martin Sebor <msebor@redhat.com>
21966 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
21967 offset_int::from instead of wide_int::to_shwi.
21968 (maybe_diag_overlap): Remove assertion.
21969 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
21970 * gimple-ssa-sprintf.c (format_directive): Same.
21971 (parse_directive): Same.
21972 (sprintf_dom_walker::compute_format_length): Same.
21973 (try_substitute_return_value): Same.
21975 2018-01-03 Jeff Law <law@redhat.com>
21977 PR middle-end/83654
21978 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
21979 non-constant residual for zero at runtime and avoid probing in
21980 that case. Reorganize code for trailing problem to mirror handling
21983 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
21985 PR tree-optimization/83501
21986 * tree-ssa-strlen.c (get_string_cst): New.
21987 (handle_char_store): Call get_string_cst.
21989 2018-01-03 Martin Liska <mliska@suse.cz>
21991 PR tree-optimization/83593
21992 * tree-ssa-strlen.c: Include tree-cfg.h.
21993 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
21994 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
21995 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
21997 (strlen_dom_walker::before_dom_children): Call
21998 gimple_purge_dead_eh_edges. Dump tranformation with details
22000 (strlen_dom_walker::before_dom_children): Update call by adding
22001 new argument cleanup_eh.
22002 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
22004 2018-01-03 Martin Liska <mliska@suse.cz>
22007 * cif-code.def (VARIADIC_THUNK): New enum value.
22008 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
22011 2018-01-03 Jan Beulich <jbeulich@suse.com>
22013 * sse.md (mov<mode>_internal): Tighten condition for when to use
22014 vmovdqu<ssescalarsize> for TI and OI modes.
22016 2018-01-03 Jakub Jelinek <jakub@redhat.com>
22018 Update copyright years.
22020 2018-01-03 Martin Liska <mliska@suse.cz>
22023 * ipa-visibility.c (function_and_variable_visibility): Skip
22024 functions with noipa attribure.
22026 2018-01-03 Jakub Jelinek <jakub@redhat.com>
22028 * gcc.c (process_command): Update copyright notice dates.
22029 * gcov-dump.c (print_version): Ditto.
22030 * gcov.c (print_version): Ditto.
22031 * gcov-tool.c (print_version): Ditto.
22032 * gengtype.c (create_file): Ditto.
22033 * doc/cpp.texi: Bump @copying's copyright year.
22034 * doc/cppinternals.texi: Ditto.
22035 * doc/gcc.texi: Ditto.
22036 * doc/gccint.texi: Ditto.
22037 * doc/gcov.texi: Ditto.
22038 * doc/install.texi: Ditto.
22039 * doc/invoke.texi: Ditto.
22041 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22043 * vector-builder.h (vector_builder::m_full_nelts): Change from
22044 unsigned int to poly_uint64.
22045 (vector_builder::full_nelts): Update prototype accordingly.
22046 (vector_builder::new_vector): Likewise.
22047 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
22048 (vector_builder::operator ==): Likewise.
22049 (vector_builder::finalize): Likewise.
22050 * int-vector-builder.h (int_vector_builder::int_vector_builder):
22051 Take the number of elements as a poly_uint64 rather than an
22053 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
22054 from unsigned int to poly_uint64.
22055 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
22056 (vec_perm_indices::new_vector): Likewise.
22057 (vec_perm_indices::length): Likewise.
22058 (vec_perm_indices::nelts_per_input): Likewise.
22059 (vec_perm_indices::input_nelts): Likewise.
22060 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
22061 number of elements per input as a poly_uint64 rather than an
22062 unsigned int. Use the original encoding for variable-length
22063 vectors, rather than clamping each individual element.
22064 For the second and subsequent elements in each pattern,
22065 clamp the step and base before clamping their sum.
22066 (vec_perm_indices::series_p): Handle polynomial element counts.
22067 (vec_perm_indices::all_in_range_p): Likewise.
22068 (vec_perm_indices_to_tree): Likewise.
22069 (vec_perm_indices_to_rtx): Likewise.
22070 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
22071 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
22072 (tree_vector_builder::new_binary_operation): Handle polynomial
22073 element counts. Return false if we need to know the number
22074 of elements at compile time.
22075 * fold-const.c (fold_vec_perm): Punt if the number of elements
22076 isn't known at compile time.
22078 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22080 * vec-perm-indices.h (vec_perm_builder): Change element type
22081 from HOST_WIDE_INT to poly_int64.
22082 (vec_perm_indices::element_type): Update accordingly.
22083 (vec_perm_indices::clamp): Handle polynomial element_types.
22084 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
22085 (vec_perm_indices::all_in_range_p): Likewise.
22086 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
22088 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
22089 polynomial vec_perm_indices element types.
22090 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
22091 * fold-const.c (fold_vec_perm): Likewise.
22092 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
22093 * tree-vect-generic.c (lower_vec_perm): Likewise.
22094 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
22095 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
22096 element type to HOST_WIDE_INT.
22098 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22099 Alan Hayward <alan.hayward@arm.com>
22100 David Sherwood <david.sherwood@arm.com>
22102 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
22103 rather than an int. Use plus_constant.
22104 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
22105 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
22107 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22108 Alan Hayward <alan.hayward@arm.com>
22109 David Sherwood <david.sherwood@arm.com>
22111 * calls.c (emit_call_1, expand_call): Change struct_value_size from
22112 a HOST_WIDE_INT to a poly_int64.
22114 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22115 Alan Hayward <alan.hayward@arm.com>
22116 David Sherwood <david.sherwood@arm.com>
22118 * calls.c (load_register_parameters): Cope with polynomial
22119 mode sizes. Require a constant size for BLKmode parameters
22120 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
22121 forces a parameter to be padded at the lsb end in order to
22122 fill a complete number of words, require the parameter size
22123 to be ordered wrt UNITS_PER_WORD.
22125 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22126 Alan Hayward <alan.hayward@arm.com>
22127 David Sherwood <david.sherwood@arm.com>
22129 * reload1.c (spill_stack_slot_width): Change element type
22130 from unsigned int to poly_uint64_pod.
22131 (alter_reg): Treat mode sizes as polynomial.
22133 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22134 Alan Hayward <alan.hayward@arm.com>
22135 David Sherwood <david.sherwood@arm.com>
22137 * reload.c (complex_word_subreg_p): New function.
22138 (reload_inner_reg_of_subreg, push_reload): Use it.
22140 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22141 Alan Hayward <alan.hayward@arm.com>
22142 David Sherwood <david.sherwood@arm.com>
22144 * lra-constraints.c (process_alt_operands): Reject matched
22145 operands whose sizes aren't ordered.
22146 (match_reload): Refer to this check here.
22148 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22149 Alan Hayward <alan.hayward@arm.com>
22150 David Sherwood <david.sherwood@arm.com>
22152 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
22153 that the mode size is in the set {1, 2, 4, 8, 16}.
22155 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22156 Alan Hayward <alan.hayward@arm.com>
22157 David Sherwood <david.sherwood@arm.com>
22159 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
22160 Use plus_constant instead of gen_rtx_PLUS.
22162 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22163 Alan Hayward <alan.hayward@arm.com>
22164 David Sherwood <david.sherwood@arm.com>
22166 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
22167 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
22168 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
22169 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
22170 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
22171 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
22172 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
22173 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
22174 * config/i386/i386.c (ix86_push_rounding): ...this new function.
22175 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
22177 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
22178 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
22179 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
22180 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
22181 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
22182 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
22183 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
22184 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
22185 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
22186 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
22188 * expr.c (emit_move_resolve_push): Treat the input and result
22189 of PUSH_ROUNDING as a poly_int64.
22190 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
22191 (emit_push_insn): Likewise.
22192 * lra-eliminations.c (mark_not_eliminable): Likewise.
22193 * recog.c (push_operand): Likewise.
22194 * reload1.c (elimination_effects): Likewise.
22195 * rtlanal.c (nonzero_bits1): Likewise.
22196 * calls.c (store_one_arg): Likewise. Require the padding to be
22197 known at compile time.
22199 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22200 Alan Hayward <alan.hayward@arm.com>
22201 David Sherwood <david.sherwood@arm.com>
22203 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
22204 Use plus_constant instead of gen_rtx_PLUS.
22206 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22207 Alan Hayward <alan.hayward@arm.com>
22208 David Sherwood <david.sherwood@arm.com>
22210 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
22211 rather than an int.
22213 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22214 Alan Hayward <alan.hayward@arm.com>
22215 David Sherwood <david.sherwood@arm.com>
22217 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
22218 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
22219 via stack temporaries. Treat the mode size as polynomial too.
22221 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22222 Alan Hayward <alan.hayward@arm.com>
22223 David Sherwood <david.sherwood@arm.com>
22225 * expr.c (expand_expr_real_2): When handling conversions involving
22226 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
22227 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
22228 as a poly_uint64 too.
22230 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22231 Alan Hayward <alan.hayward@arm.com>
22232 David Sherwood <david.sherwood@arm.com>
22234 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
22236 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22237 Alan Hayward <alan.hayward@arm.com>
22238 David Sherwood <david.sherwood@arm.com>
22240 * combine.c (can_change_dest_mode): Handle polynomial
22241 REGMODE_NATURAL_SIZE.
22242 * expmed.c (store_bit_field_1): Likewise.
22243 * expr.c (store_constructor): Likewise.
22244 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
22245 and polynomial REGMODE_NATURAL_SIZE.
22246 (gen_lowpart_common): Likewise.
22247 * reginfo.c (record_subregs_of_mode): Likewise.
22248 * rtlanal.c (read_modify_subreg_p): Likewise.
22250 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22251 Alan Hayward <alan.hayward@arm.com>
22252 David Sherwood <david.sherwood@arm.com>
22254 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
22255 numbers of elements.
22257 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22258 Alan Hayward <alan.hayward@arm.com>
22259 David Sherwood <david.sherwood@arm.com>
22261 * match.pd: Cope with polynomial numbers of vector elements.
22263 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22264 Alan Hayward <alan.hayward@arm.com>
22265 David Sherwood <david.sherwood@arm.com>
22267 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
22268 in a POINTER_PLUS_EXPR.
22270 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22271 Alan Hayward <alan.hayward@arm.com>
22272 David Sherwood <david.sherwood@arm.com>
22274 * omp-simd-clone.c (simd_clone_subparts): New function.
22275 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
22276 (ipa_simd_modify_function_body): Likewise.
22278 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22279 Alan Hayward <alan.hayward@arm.com>
22280 David Sherwood <david.sherwood@arm.com>
22282 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
22283 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
22284 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
22285 (expand_vector_condition, vector_element): Likewise.
22286 (subparts_gt): New function.
22287 (get_compute_type): Use subparts_gt.
22288 (count_type_subparts): Delete.
22289 (expand_vector_operations_1): Use subparts_gt instead of
22290 count_type_subparts.
22292 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22293 Alan Hayward <alan.hayward@arm.com>
22294 David Sherwood <david.sherwood@arm.com>
22296 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
22297 (vect_compile_time_alias): ...this new function. Do the calculation
22298 on poly_ints rather than trees.
22299 (vect_prune_runtime_alias_test_list): Update call accordingly.
22301 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22302 Alan Hayward <alan.hayward@arm.com>
22303 David Sherwood <david.sherwood@arm.com>
22305 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
22307 (vect_schedule_slp_instance): Likewise.
22309 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22310 Alan Hayward <alan.hayward@arm.com>
22311 David Sherwood <david.sherwood@arm.com>
22313 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
22314 constant and extern definitions for variable-length vectors.
22315 (vect_get_constant_vectors): Note that the number of units
22316 is known to be constant.
22318 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22319 Alan Hayward <alan.hayward@arm.com>
22320 David Sherwood <david.sherwood@arm.com>
22322 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
22323 of units as polynomial. Choose between WIDE and NARROW based
22326 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22327 Alan Hayward <alan.hayward@arm.com>
22328 David Sherwood <david.sherwood@arm.com>
22330 * tree-vect-stmts.c (simd_clone_subparts): New function.
22331 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
22333 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22334 Alan Hayward <alan.hayward@arm.com>
22335 David Sherwood <david.sherwood@arm.com>
22337 * tree-vect-stmts.c (vectorizable_call): Treat the number of
22338 vectors as polynomial. Use build_index_vector for
22339 IFN_GOMP_SIMD_LANE.
22341 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22342 Alan Hayward <alan.hayward@arm.com>
22343 David Sherwood <david.sherwood@arm.com>
22345 * tree-vect-stmts.c (get_load_store_type): Treat the number of
22346 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
22347 for variable-length vectors.
22348 (vectorizable_mask_load_store): Treat the number of units as
22349 polynomial, asserting that it is constant if the condition has
22350 already been enforced.
22351 (vectorizable_store, vectorizable_load): Likewise.
22353 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22354 Alan Hayward <alan.hayward@arm.com>
22355 David Sherwood <david.sherwood@arm.com>
22357 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
22358 of units as polynomial. Punt if we can't tell at compile time
22359 which vector contains the final result.
22361 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22362 Alan Hayward <alan.hayward@arm.com>
22363 David Sherwood <david.sherwood@arm.com>
22365 * tree-vect-loop.c (vectorizable_induction): Treat the number
22366 of units as polynomial. Punt on SLP inductions. Use an integer
22367 VEC_SERIES_EXPR for variable-length integer reductions. Use a
22368 cast of such a series for variable-length floating-point
22371 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22372 Alan Hayward <alan.hayward@arm.com>
22373 David Sherwood <david.sherwood@arm.com>
22375 * tree.h (build_index_vector): Declare.
22376 * tree.c (build_index_vector): New function.
22377 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
22378 of units as polynomial, forcibly converting it to a constant if
22379 vectorizable_reduction has already enforced the condition.
22380 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
22381 to create a {1,2,3,...} vector.
22382 (vectorizable_reduction): Treat the number of units as polynomial.
22383 Choose vectype_in based on the largest scalar element size rather
22384 than the smallest number of units. Enforce the restrictions
22387 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22388 Alan Hayward <alan.hayward@arm.com>
22389 David Sherwood <david.sherwood@arm.com>
22391 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
22392 number of units as polynomial.
22394 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22395 Alan Hayward <alan.hayward@arm.com>
22396 David Sherwood <david.sherwood@arm.com>
22398 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
22399 * target.def (autovectorize_vector_sizes): Return the vector sizes
22400 by pointer, using vector_sizes rather than a bitmask.
22401 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
22402 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
22403 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
22405 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
22406 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
22407 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
22408 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
22409 * omp-general.c (omp_max_vf): Likewise.
22410 * omp-low.c (omp_clause_aligned_alignment): Likewise.
22411 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
22412 * tree-vect-loop.c (vect_analyze_loop): Likewise.
22413 * tree-vect-slp.c (vect_slp_bb): Likewise.
22414 * doc/tm.texi: Regenerate.
22415 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
22417 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
22418 the vector size as a poly_uint64 rather than an unsigned int.
22419 (current_vector_size): Change from an unsigned int to a poly_uint64.
22420 (get_vectype_for_scalar_type): Update accordingly.
22421 * tree.h (build_truth_vector_type): Take the size and number of
22422 units as a poly_uint64 rather than an unsigned int.
22423 (build_vector_type): Add a temporary overload that takes
22424 the number of units as a poly_uint64 rather than an unsigned int.
22425 * tree.c (make_vector_type): Likewise.
22426 (build_truth_vector_type): Take the number of units as a poly_uint64
22427 rather than an unsigned int.
22429 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22430 Alan Hayward <alan.hayward@arm.com>
22431 David Sherwood <david.sherwood@arm.com>
22433 * target.def (get_mask_mode): Take the number of units and length
22434 as poly_uint64s rather than unsigned ints.
22435 * targhooks.h (default_get_mask_mode): Update accordingly.
22436 * targhooks.c (default_get_mask_mode): Likewise.
22437 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
22438 * doc/tm.texi: Regenerate.
22440 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22441 Alan Hayward <alan.hayward@arm.com>
22442 David Sherwood <david.sherwood@arm.com>
22444 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
22445 * omp-general.c (omp_max_vf): Likewise.
22446 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
22447 (expand_omp_simd): Handle polynomial safelen.
22448 * omp-low.c (omplow_simd_context): Add a default constructor.
22449 (omplow_simd_context::max_vf): Change from int to poly_uint64.
22450 (lower_rec_simd_input_clauses): Update accordingly.
22451 (lower_rec_input_clauses): Likewise.
22453 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22454 Alan Hayward <alan.hayward@arm.com>
22455 David Sherwood <david.sherwood@arm.com>
22457 * tree-vectorizer.h (vect_nunits_for_cost): New function.
22458 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
22459 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
22460 (vect_analyze_slp_cost): Likewise.
22461 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
22462 (vect_model_load_cost): Likewise.
22464 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22465 Alan Hayward <alan.hayward@arm.com>
22466 David Sherwood <david.sherwood@arm.com>
22468 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
22469 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
22470 from an unsigned int * to a poly_uint64_pod *.
22471 (calculate_unrolling_factor): New function.
22472 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
22474 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22475 Alan Hayward <alan.hayward@arm.com>
22476 David Sherwood <david.sherwood@arm.com>
22478 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
22479 from an unsigned int to a poly_uint64.
22480 (_loop_vec_info::slp_unrolling_factor): Likewise.
22481 (_loop_vec_info::vectorization_factor): Change from an int
22483 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
22484 (vect_get_num_vectors): New function.
22485 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
22486 (vect_get_num_copies): Use vect_get_num_vectors.
22487 (vect_analyze_data_ref_dependences): Change max_vf from an int *
22488 to an unsigned int *.
22489 (vect_analyze_data_refs): Change min_vf from an int * to a
22491 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
22492 than an unsigned HOST_WIDE_INT.
22493 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
22494 (vect_analyze_data_ref_dependence): Change max_vf from an int *
22495 to an unsigned int *.
22496 (vect_analyze_data_ref_dependences): Likewise.
22497 (vect_compute_data_ref_alignment): Handle polynomial vf.
22498 (vect_enhance_data_refs_alignment): Likewise.
22499 (vect_prune_runtime_alias_test_list): Likewise.
22500 (vect_shift_permute_load_chain): Likewise.
22501 (vect_supportable_dr_alignment): Likewise.
22502 (dependence_distance_ge_vf): Take the vectorization factor as a
22503 poly_uint64 rather than an unsigned HOST_WIDE_INT.
22504 (vect_analyze_data_refs): Change min_vf from an int * to a
22506 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
22507 vfm1 as a poly_uint64 rather than an int. Make the same change
22508 for the returned bound_scalar.
22509 (vect_gen_vector_loop_niters): Handle polynomial vf.
22510 (vect_do_peeling): Likewise. Update call to
22511 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
22512 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
22514 * tree-vect-loop.c (vect_determine_vectorization_factor)
22515 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
22516 (vect_get_known_peeling_cost): Likewise.
22517 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
22518 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
22519 (vect_transform_loop): Likewise. Use the lowest possible VF when
22520 updating the upper bounds of the loop.
22521 (vect_min_worthwhile_factor): Make static. Return an unsigned int
22522 rather than an int.
22523 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
22524 polynomial unroll factors.
22525 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
22526 (vect_make_slp_decision): Likewise.
22527 (vect_supported_load_permutation_p): Likewise, and polynomial
22529 (vect_analyze_slp_cost): Handle polynomial vf.
22530 (vect_slp_analyze_node_operations): Likewise.
22531 (vect_slp_analyze_bb_1): Likewise.
22532 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
22533 than an unsigned HOST_WIDE_INT.
22534 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
22535 (vectorizable_load): Handle polynomial vf.
22536 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
22538 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
22540 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22541 Alan Hayward <alan.hayward@arm.com>
22542 David Sherwood <david.sherwood@arm.com>
22544 * match.pd: Handle bit operations involving three constants
22545 and try to fold one pair.
22547 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22549 * tree-vect-loop-manip.c: Include gimple-fold.h.
22550 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
22551 niters_maybe_zero parameters. Handle other cases besides a step of 1.
22552 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
22553 Add a path that uses a step of VF instead of 1, but disable it
22555 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
22556 and niters_no_overflow parameters. Update calls to
22557 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
22558 Create a new SSA name if the latter choses to use a ste other
22559 than zero, and return it via niters_vector_mult_vf_var.
22560 * tree-vect-loop.c (vect_transform_loop): Update calls to
22561 vect_do_peeling, vect_gen_vector_loop_niters and
22562 slpeel_make_loop_iterate_ntimes.
22563 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
22564 (vect_gen_vector_loop_niters): Update declarations after above changes.
22566 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
22568 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
22569 128-bit round to integer instructions.
22570 (ceil<mode>2): Likewise.
22571 (btrunc<mode>2): Likewise.
22572 (round<mode>2): Likewise.
22574 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
22576 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
22577 unaligned VSX load/store on P8/P9.
22578 (expand_block_clear): Allow the use of unaligned VSX
22579 load/store on P8/P9.
22581 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
22583 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
22585 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
22586 swap associated with both a load and a store.
22588 2018-01-02 Andrew Waterman <andrew@sifive.com>
22590 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
22591 * config/riscv/riscv.md (clear_cache): Use it.
22593 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
22595 * web.c: Remove out-of-date comment.
22597 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
22599 * expr.c (fixup_args_size_notes): Check that any existing
22600 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
22601 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
22602 (emit_single_push_insn): ...here.
22604 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
22606 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
22607 (const_vector_encoded_nelts): New function.
22608 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
22609 (const_vector_int_elt, const_vector_elt): Declare.
22610 * emit-rtl.c (const_vector_int_elt_1): New function.
22611 (const_vector_elt): Likewise.
22612 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
22613 of CONST_VECTOR_ELT.
22615 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
22617 * expr.c: Include rtx-vector-builder.h.
22618 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
22619 directly on the tree encoding.
22620 (const_vector_from_tree): Likewise.
22621 * optabs.c: Include rtx-vector-builder.h.
22622 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
22623 sequence of "u" values.
22624 * vec-perm-indices.c: Include rtx-vector-builder.h.
22625 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
22626 directly on the vec_perm_indices encoding.
22628 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
22630 * doc/rtl.texi (const_vector): Describe new encoding scheme.
22631 * Makefile.in (OBJS): Add rtx-vector-builder.o.
22632 * rtx-vector-builder.h: New file.
22633 * rtx-vector-builder.c: Likewise.
22634 * rtl.h (rtx_def::u2): Add a const_vector field.
22635 (CONST_VECTOR_NPATTERNS): New macro.
22636 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
22637 (CONST_VECTOR_DUPLICATE_P): Likewise.
22638 (CONST_VECTOR_STEPPED_P): Likewise.
22639 (CONST_VECTOR_ENCODED_ELT): Likewise.
22640 (const_vec_duplicate_p): Check for a duplicated vector encoding.
22641 (unwrap_const_vec_duplicate): Likewise.
22642 (const_vec_series_p): Check for a non-duplicated vector encoding.
22643 Say that the function only returns true for integer vectors.
22644 * emit-rtl.c: Include rtx-vector-builder.h.
22645 (gen_const_vec_duplicate_1): Delete.
22646 (gen_const_vector): Call gen_const_vec_duplicate instead of
22647 gen_const_vec_duplicate_1.
22648 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
22649 (gen_const_vec_duplicate): Use rtx_vector_builder.
22650 (gen_const_vec_series): Likewise.
22651 (gen_rtx_CONST_VECTOR): Likewise.
22652 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
22653 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
22654 Build a new vector rather than modifying a CONST_VECTOR in-place.
22655 (handle_special_swappables): Update call accordingly.
22656 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
22657 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
22658 Build a new vector rather than modifying a CONST_VECTOR in-place.
22659 (handle_special_swappables): Update call accordingly.
22661 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
22663 * simplify-rtx.c (simplify_const_binary_operation): Use
22664 CONST_VECTOR_ELT instead of XVECEXP.
22666 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
22668 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
22669 the selector elements to be different from the data elements
22670 if the selector is a VECTOR_CST.
22671 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
22672 ssizetype for the selector.
22674 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
22676 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
22677 before testing each element individually.
22678 * tree-vect-generic.c (lower_vec_perm): Likewise.
22680 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
22682 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
22683 * selftest-run-tests.c (selftest::run_tests): Call it.
22684 * vector-builder.h (vector_builder::operator ==): New function.
22685 (vector_builder::operator !=): Likewise.
22686 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
22687 (vec_perm_indices::all_from_input_p): New function.
22688 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
22689 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
22690 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
22691 instead of reading the VECTOR_CST directly. Detect whether both
22692 vector inputs are the same before constructing the vec_perm_indices,
22693 and update the number of inputs argument accordingly. Use the
22694 utility functions added above. Only construct sel2 if we need to.
22696 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
22698 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
22699 the broadcast of the low byte.
22700 (expand_mult_highpart): Use an explicit encoding for the permutes.
22701 * optabs-query.c (can_mult_highpart_p): Likewise.
22702 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
22703 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
22704 (vectorizable_bswap): Likewise.
22705 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
22706 explicit encoding for the power-of-2 permutes.
22707 (vect_permute_store_chain): Likewise.
22708 (vect_grouped_load_supported): Likewise.
22709 (vect_permute_load_chain): Likewise.
22711 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
22713 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
22714 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
22715 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
22716 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
22717 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
22718 (vect_gen_perm_mask_any): Likewise.
22720 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
22722 * int-vector-builder.h: New file.
22723 * vec-perm-indices.h: Include int-vector-builder.h.
22724 (vec_perm_indices): Redefine as an int_vector_builder.
22725 (auto_vec_perm_indices): Delete.
22726 (vec_perm_builder): Redefine as a stand-alone class.
22727 (vec_perm_indices::vec_perm_indices): New function.
22728 (vec_perm_indices::clamp): Likewise.
22729 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
22730 (vec_perm_indices::new_vector): New function.
22731 (vec_perm_indices::new_expanded_vector): Update for new
22732 vec_perm_indices class.
22733 (vec_perm_indices::rotate_inputs): New function.
22734 (vec_perm_indices::all_in_range_p): Operate directly on the
22735 encoded form, without computing elided elements.
22736 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
22737 encoding. Update for new vec_perm_indices class.
22738 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
22739 the given vec_perm_builder.
22740 (expand_vec_perm_var): Update vec_perm_builder constructor.
22741 (expand_mult_highpart): Use vec_perm_builder instead of
22742 auto_vec_perm_indices.
22743 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
22744 vec_perm_indices instead of auto_vec_perm_indices. Use a single
22745 or double series encoding as appropriate.
22746 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
22747 vec_perm_indices instead of auto_vec_perm_indices.
22748 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
22749 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
22750 (vect_permute_store_chain): Likewise.
22751 (vect_grouped_load_supported): Likewise.
22752 (vect_permute_load_chain): Likewise.
22753 (vect_shift_permute_load_chain): Likewise.
22754 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
22755 (vect_transform_slp_perm_load): Likewise.
22756 (vect_schedule_slp_instance): Likewise.
22757 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
22758 (vectorizable_mask_load_store): Likewise.
22759 (vectorizable_bswap): Likewise.
22760 (vectorizable_store): Likewise.
22761 (vectorizable_load): Likewise.
22762 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
22763 vec_perm_indices instead of auto_vec_perm_indices. Use
22764 tree_to_vec_perm_builder to read the vector from a tree.
22765 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
22766 vec_perm_builder instead of a vec_perm_indices.
22767 (have_whole_vector_shift): Use vec_perm_builder and
22768 vec_perm_indices instead of auto_vec_perm_indices. Leave the
22769 truncation to calc_vec_perm_mask_for_shift.
22770 (vect_create_epilog_for_reduction): Likewise.
22771 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
22772 from auto_vec_perm_indices to vec_perm_indices.
22773 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
22774 instead of changing individual elements.
22775 (aarch64_vectorize_vec_perm_const): Use new_vector to install
22776 the vector in d.perm.
22777 * config/arm/arm.c (expand_vec_perm_d::perm): Change
22778 from auto_vec_perm_indices to vec_perm_indices.
22779 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
22780 instead of changing individual elements.
22781 (arm_vectorize_vec_perm_const): Use new_vector to install
22782 the vector in d.perm.
22783 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
22784 Update vec_perm_builder constructor.
22785 (rs6000_expand_interleave): Likewise.
22786 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
22787 (rs6000_expand_interleave): Likewise.
22789 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
22791 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
22792 to qimode could truncate the indices.
22793 * optabs.c (expand_vec_perm_var): Likewise.
22795 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
22797 * Makefile.in (OBJS): Add vec-perm-indices.o.
22798 * vec-perm-indices.h: New file.
22799 * vec-perm-indices.c: Likewise.
22800 * target.h (vec_perm_indices): Replace with a forward class
22802 (auto_vec_perm_indices): Move to vec-perm-indices.h.
22803 * optabs.h: Include vec-perm-indices.h.
22804 (expand_vec_perm): Delete.
22805 (selector_fits_mode_p, expand_vec_perm_var): Declare.
22806 (expand_vec_perm_const): Declare.
22807 * target.def (vec_perm_const_ok): Replace with...
22808 (vec_perm_const): ...this new hook.
22809 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
22810 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
22811 * doc/tm.texi: Regenerate.
22812 * optabs.def (vec_perm_const): Delete.
22813 * doc/md.texi (vec_perm_const): Likewise.
22814 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
22815 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
22816 expand_vec_perm for constant permutation vectors. Assert that
22817 the mode of variable permutation vectors is the integer equivalent
22818 of the mode that is being permuted.
22819 * optabs-query.h (selector_fits_mode_p): Declare.
22820 * optabs-query.c: Include vec-perm-indices.h.
22821 (selector_fits_mode_p): New function.
22822 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
22823 is defined, instead of checking whether the vec_perm_const_optab
22824 exists. Use targetm.vectorize.vec_perm_const instead of
22825 targetm.vectorize.vec_perm_const_ok. Check whether the indices
22826 fit in the vector mode before using a variable permute.
22827 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
22828 vec_perm_indices instead of an rtx.
22829 (expand_vec_perm): Replace with...
22830 (expand_vec_perm_const): ...this new function. Take the selector
22831 as a vec_perm_indices rather than an rtx. Also take the mode of
22832 the selector. Update call to shift_amt_for_vec_perm_mask.
22833 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
22834 Use vec_perm_indices::new_expanded_vector to expand the original
22835 selector into bytes. Check whether the indices fit in the vector
22836 mode before using a variable permute.
22837 (expand_vec_perm_var): Make global.
22838 (expand_mult_highpart): Use expand_vec_perm_const.
22839 * fold-const.c: Includes vec-perm-indices.h.
22840 * tree-ssa-forwprop.c: Likewise.
22841 * tree-vect-data-refs.c: Likewise.
22842 * tree-vect-generic.c: Likewise.
22843 * tree-vect-loop.c: Likewise.
22844 * tree-vect-slp.c: Likewise.
22845 * tree-vect-stmts.c: Likewise.
22846 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
22848 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
22849 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
22850 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
22851 (aarch64_vectorize_vec_perm_const): ...this new function.
22852 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
22853 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
22854 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
22855 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
22856 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
22857 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
22858 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
22860 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
22861 check for NEON modes.
22862 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
22863 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
22864 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
22865 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
22867 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
22868 the old VEC_PERM_CONST conditions.
22869 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
22870 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
22871 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
22872 (ia64_vectorize_vec_perm_const_ok): Merge into...
22873 (ia64_vectorize_vec_perm_const): ...this new function.
22874 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
22875 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
22876 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
22877 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
22878 * config/mips/mips.c (mips_expand_vec_perm_const)
22879 (mips_vectorize_vec_perm_const_ok): Merge into...
22880 (mips_vectorize_vec_perm_const): ...this new function.
22881 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
22882 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
22883 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
22884 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
22885 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
22886 (rs6000_expand_vec_perm_const): Delete.
22887 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
22889 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
22890 (altivec_expand_vec_perm_const_le): Take each operand individually.
22891 Operate on constant selectors rather than rtxes.
22892 (altivec_expand_vec_perm_const): Likewise. Update call to
22893 altivec_expand_vec_perm_const_le.
22894 (rs6000_expand_vec_perm_const): Delete.
22895 (rs6000_vectorize_vec_perm_const_ok): Delete.
22896 (rs6000_vectorize_vec_perm_const): New function.
22897 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
22898 an element count and rtx array.
22899 (rs6000_expand_extract_even): Update call accordingly.
22900 (rs6000_expand_interleave): Likewise.
22901 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
22902 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
22903 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
22904 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
22905 (rs6000_expand_vec_perm_const): Delete.
22906 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
22907 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
22908 (altivec_expand_vec_perm_const_le): Take each operand individually.
22909 Operate on constant selectors rather than rtxes.
22910 (altivec_expand_vec_perm_const): Likewise. Update call to
22911 altivec_expand_vec_perm_const_le.
22912 (rs6000_expand_vec_perm_const): Delete.
22913 (rs6000_vectorize_vec_perm_const_ok): Delete.
22914 (rs6000_vectorize_vec_perm_const): New function. Remove stray
22915 reference to the SPE evmerge intructions.
22916 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
22917 an element count and rtx array.
22918 (rs6000_expand_extract_even): Update call accordingly.
22919 (rs6000_expand_interleave): Likewise.
22920 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
22921 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
22923 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
22925 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
22927 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
22928 vector mode and that that mode matches the mode of the data
22930 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
22931 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
22932 directly using expand_vec_perm_1 when forcing selectors into
22934 (expand_vec_perm_var): New function, split out from expand_vec_perm.
22936 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
22938 * optabs-query.h (can_vec_perm_p): Delete.
22939 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
22940 * optabs-query.c (can_vec_perm_p): Split into...
22941 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
22942 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
22943 particular selector is valid.
22944 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
22945 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
22946 (vect_grouped_load_supported): Likewise.
22947 (vect_shift_permute_load_chain): Likewise.
22948 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
22949 (vect_transform_slp_perm_load): Likewise.
22950 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
22951 (vectorizable_bswap): Likewise.
22952 (vect_gen_perm_mask_checked): Likewise.
22953 * fold-const.c (fold_ternary_loc): Likewise. Don't take
22954 implementations of variable permutation vectors into account
22955 when deciding which selector to use.
22956 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
22957 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
22958 with a false third argument.
22959 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
22960 to test whether the constant selector is valid and can_vec_perm_var_p
22961 to test whether a variable selector is valid.
22963 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
22965 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
22966 * optabs-query.c (can_vec_perm_p): Likewise.
22967 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
22968 instead of vec_perm_indices.
22969 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
22970 (vect_gen_perm_mask_checked): Likewise,
22971 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
22972 (vect_gen_perm_mask_checked): Likewise,
22974 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
22976 * optabs-query.h (qimode_for_vec_perm): Declare.
22977 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
22978 (qimode_for_vec_perm): ...this new function.
22979 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
22981 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
22983 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
22984 does not have a conditional at the top.
22986 2018-01-02 Richard Biener <rguenther@suse.de>
22988 * ipa-inline.c (big_speedup_p): Fix expression.
22990 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
22993 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
22996 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
23000 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
23001 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
23002 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
23003 cond_taken_branch_cost 3->4.
23005 2018-01-01 Jakub Jelinek <jakub@redhat.com>
23007 PR tree-optimization/83581
23008 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
23009 TODO_cleanup_cfg if any changes have been made.
23011 PR middle-end/83608
23012 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
23013 convert_modes if target mode has the right side, but different mode
23016 PR middle-end/83609
23017 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
23018 last argument when extracting from CONCAT. If either from_real or
23019 from_imag is NULL, use expansion through memory. If result is not
23020 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
23021 the parts directly to inner mode, if even that fails, use expansion
23024 PR middle-end/83623
23025 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
23026 check for bswap in mode rather than HImode and use that in expand_unop
23029 Copyright (C) 2018 Free Software Foundation, Inc.
23031 Copying and distribution of this file, with or without modification,
23032 are permitted in any medium without royalty provided the copyright
23033 notice and this notice are preserved.