1 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
3 * config/rs6000/rs6000.c (rs6000_builtin_valid_without_lhs): Add
4 vec_xst variants to the list.
5 (rs6000_gimple_fold_builtin): Add support for folding unaligned
6 vector loads and stores.
8 2018-08-13 David Edelsohn <dje.gcc@gmail.com>
10 * config.gcc (rs6000-ibm-aix4.x): Delete.
11 (rs6000-ibm-aix5.1): Delete.
12 (rs6000-ibm-aix5.2): Delete.
13 (rs6000-ibm-aix5.3): Delete.
14 * config/rs6000/aix43.h: Delete.
15 * config/rs6000/aix51.h: Delete.
16 * config/rs6000/aix52.h: Delete.
17 * config/rs6000/t-aix43: Delete.
19 2018-08-13 Ilya Leoshkevich <iii@linux.ibm.com>
21 * config/s390/s390.c (s390_decompose_constant_pool_ref):
23 (s390_decompose_address): Factor out constant pool ref
26 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
28 * config/nds32/nds32-predicates.c
29 (nds32_can_use_bclr_p): Change return type as bool.
30 (nds32_can_use_bset_p): Ditto.
31 (nds32_can_use_btgl_p): Ditto.
32 (nds32_can_use_bitci_p): Ditto.
33 * config/nds32/nds32-protos.h
34 (nds32_can_use_bclr_p): Change declaration.
35 (nds32_can_use_bset_p): Ditto.
36 (nds32_can_use_btgl_p): Ditto.
37 (nds32_can_use_bitci_p): Ditto.
39 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
41 * config/nds32/nds32.c (nds32_expand_prologue, nds32_expand_epilogue):
42 Support -msched-prolog-epilog option.
43 * config/nds32/nds32.opt (msched-prolog-epilog): New option.
45 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
47 * common/config/nds32/nds32-common.c
48 (nds32_option_optimization_table): Enalbe -malways-align.
50 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
52 * config.gcc (nds32*): Add nds32_isr.h and nds32_init.inc in
54 * common/config/nds32/nds32-common.c (nds32_handle_option): Handle
55 OPT_misr_secure_ case.
56 * config/nds32/nds32-isr.c: Implementation of backward compatibility.
57 * config/nds32/nds32-protos.h (nds32_isr_function_critical_p): New.
58 * config/nds32/nds32.c (nds32_attribute_table): Add critical and
60 * config/nds32/nds32.h (nds32_isr_nested_type): Add NDS32_CRITICAL.
61 (nds32_isr_info): New field security_level.
62 (TARGET_ISR_VECTOR_SIZE_4_BYTE): New macro.
63 * config/nds32/nds32.md (return_internal): Consider critical attribute.
64 * config/nds32/nds32.opt (misr-secure): New option.
65 * config/nds32/nds32_init.inc: New file.
66 * config/nds32/nds32_isr.h: New file.
68 2018-08-11 John David Anglin <danglin@gcc.gnu.org>
70 * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
71 Update comment for atomic instructions.
72 (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
73 atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
75 (atomic_loaddi): Revise fence expansion to only emit fence prior to
76 load for __ATOMIC_SEQ_CST model.
77 (atomic_loaddi_1): Remove float register target.
78 (atomic_storedi): Handle CONST_INT values.
79 (atomic_storedi_1): Remove float register source. Add special case
81 (memory_barrier): New expander and insn.
83 2018-08-11 Jakub Jelinek <jakub@redhat.com>
85 PR tree-optimization/86835
86 * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
87 new_stmt after def_gsi, make sure to insert new_square_stmt after
88 that stmt, not 2 stmts before it.
90 2018-08-10 Alexander Monakov <amonakov@ispras.ru>
93 * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
96 2018-08-10 Martin Liska <mliska@suse.cz>
99 * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
101 * builtins.c (expand_builtin_expect_with_probability):
103 (expand_builtin_expect_with_probability): New function.
104 (build_builtin_expect_predicate): Add new argumnet probability
105 for BUILT_IN_EXPECT_WITH_PROBABILITY.
106 (fold_builtin_expect):
109 * builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY):
110 * builtins.h (fold_builtin_expect): Set new argument.
111 * doc/extend.texi: Document __builtin_expect_with_probability.
112 * doc/invoke.texi: Likewise.
113 * gimple-fold.c (gimple_fold_call): Pass new argument.
114 * ipa-fnsummary.c (find_foldable_builtin_expect): Handle
115 also BUILT_IN_EXPECT_WITH_PROBABILITY.
116 * predict.c (get_predictor_value): New function.
117 (expr_expected_value): Add new argument probability. Assume
118 that predictor and probability are always non-null.
119 (expr_expected_value_1): Likewise. For __builtin_expect and
120 __builtin_expect_with_probability set probability. Handle
121 combination in binary expressions.
122 (tree_predict_by_opcode): Simplify code by simply calling
124 (pass_strip_predict_hints::execute): Add handling of
125 BUILT_IN_EXPECT_WITH_PROBABILITY.
126 * predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add
128 * tree.h (DECL_BUILT_IN_P): New function.
130 2018-08-10 Martin Liska <mliska@suse.cz>
132 PR tree-optimization/85799
133 * passes.def: Add argument for pass_strip_predict_hints.
134 * predict.c (class pass_strip_predict_hints): Add new argument
136 (strip_predictor_early): New function.
137 (pass_strip_predict_hints::execute): Call the function to
139 (strip_predict_hints): New function.
140 * predict.def: Fix comment.
142 2018-08-10 Thomas Preud'homme <thomas.preudhomme@linaro.org>
144 * Makefile.in: Clarify which tm.texi to copy over to assert the
145 right to grant a GFDL license for all.
147 2018-08-09 Jeff Law <law@redhat.com>
149 * config/m68k/m68k.c (m68k_adjust_decorated_operand): Remove
152 2018-08-09 Andreas Schwab <schwab@linux-m68k.org>
154 * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
157 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
159 * tree-vect-loop.c (vectorizable_reduction): Allow inner-loop
160 reductions for variable-length vectors.
162 2018-08-09 David Malcolm <dmalcolm@redhat.com>
165 * common.opt (fdiagnostics-show-line-numbers): New option.
166 * diagnostic-show-locus.c (class layout): Add fields
167 "m_show_line_numbers_p" and "m_linenum_width";
168 (num_digits): New function.
169 (test_num_digits): New function.
170 (layout::layout): Initialize new fields. Update m_x_offset
171 logic to handle any left margin.
172 (layout::print_source_line): Print line number when requested.
173 (layout::start_annotation_line): New member function.
174 (layout::print_annotation_line): Call it.
175 (layout::print_leading_fixits): Likewise.
176 (layout::print_trailing_fixits): Likewise. Update calls to
177 move_to_column for new parameter.
178 (layout::get_x_bound_for_row): Add "add_left_margin" param and use
179 it to potentially call start_annotation_line.
180 (layout::show_ruler): Call start_annotation_line.
181 (selftest::test_line_numbers_multiline_range): New selftest.
182 (selftest::diagnostic_show_locus_c_tests): Call test_num_digits
183 and selftest::test_line_numbers_multiline_range.
184 * diagnostic.c (diagnostic_initialize): Initialize
186 * diagnostic.h (struct diagnostic_context): Add field
187 "show_line_numbers_p".
188 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
189 -fno-diagnostics-show-line-numbers.
190 * dwarf2out.c (gen_producer_string): Add
191 OPT_fdiagnostics_show_line_numbers to the ignored options.
192 * lto-wrapper.c (merge_and_complain): Likewise to the "pick
193 one setting" options.
194 (append_compiler_options): Likewise to the dropped options.
195 (append_diag_options): Likewise to the passed-on options.
196 * opts.c (common_handle_option): Handle the new option.
197 * toplev.c (general_init): Set up global_dc->show_line_numbers_p.
199 2018-08-09 Kelvin Nilsen <kelvin@gcc.gnu.org>
201 * doc/extend.texi (PowerPC AltiVec Built-in Functions Available on
202 ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub. Add
203 third argument of type "const signed char" to descriptions of
204 __builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq,
205 __builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub,
206 __builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt,
207 __builtin_bcdsub_ov functions.
209 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
211 PR tree-optimization/86858
212 * tree-vect-loop.c (vect_is_simple_reduction): Restore
213 flow_bb_inside_loop_p calls.
215 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
217 PR tree-optimization/86871
218 * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
219 instead of gimple_assign_lhs.
221 2018-08-09 Richard Earnshaw <rearnsha@arm.com>
224 * config/aarch64/aarch64.md (add<mode>3_carryinC_zero): Add missing
225 register constraint to operand 0.
226 (add<mode>3_carryinC): Likewise.
227 (add<mode>3_carryinV_zero, add<mode>3_carryinV): Likewise.
229 2018-08-09 Martin Liska <mliska@suse.cz>
232 * common.opt: Remove extra line.
234 2018-08-09 Martin Liska <mliska@suse.cz>
236 * params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots
237 at the end of a line, make first letter capital and end up
238 a sentence with a dot.
239 (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise.
240 (PARAM_LOOP_BLOCK_TILE_SIZE): Likewise.
241 (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise.
242 (PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise.
243 (PARAM_MAX_ISL_OPERATIONS): Likewise.
244 (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise.
245 (PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise.
246 (PARAM_INDIR_CALL_TOPN_PROFILE): Likewise.
247 (PARAM_SLP_MAX_INSNS_IN_BB): Likewise.
248 (PARAM_IPA_CP_EVAL_THRESHOLD): Likewise.
249 (PARAM_IPA_CP_RECURSION_PENALTY): Likewise.
250 (PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
251 (PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise.
252 (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise.
253 (PARAM_TREE_REASSOC_WIDTH): Likewise.
254 (PARAM_HSA_GEN_DEBUG_STORES): Likewise.
255 (PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise.
256 (PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise.
258 2018-08-09 Andreas Krebbel <krebbel@linux.ibm.com>
261 * config/s390/s390.c (s390_option_override_internal): Reduce the
262 stack-clash-protection-probe-interval param if it would be too big
265 2018-08-08 Andreas Schwab <schwab@linux-m68k.org>
268 * config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define.
269 * config/m68k/m68k.c (handle_move_double): Don't call
270 m68k_final_prescan_insn.
271 (m68k_adjust_decorated_operand): Renamed from
272 m68k_final_prescan_insn, remove first and third operand and
274 (print_operand): Call it.
275 (print_operand_address): Call it.
277 2018-08-08 Nathan Sidwell <nathan@acm.org>
279 * diagnostic.c (diagnostic_report_current_module): Use
280 linemap_included_from & linemap_included_from_linemap.
282 2018-08-08 Hongbo Zhang <hongbo.zhang@linaro.org>
284 * config/aarch64/aarch64-cores.def: Add phecda core.
285 * config/aarch64/aarch64-tune.md: Regenerate.
286 * doc/invoke.texi: Add phecda core.
288 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
291 * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
293 * config/s390/s390.md ("movti"): Add more alternatives for
294 constant to GPR copies.
296 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
298 * config/s390/s390.c: Fix whitespace damage throughout the file.
299 * config/s390/s390.h: Likewise.
300 * config/s390/tpf.h: Likewise.
302 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
304 * config/s390/s390.c (s390_loadrelative_operand_p):
305 Remove TARGET_CPU_ZARCH usages.
306 (s390_rtx_costs): Likewise.
307 (s390_legitimate_constant_p): Likewise.
308 (s390_cannot_force_const_mem): Likewise.
309 (legitimate_reload_constant_p): Likewise.
310 (s390_preferred_reload_class): Likewise.
311 (legitimize_pic_address): Likewise.
312 (legitimize_tls_address): Likewise.
313 (s390_split_branches): Removed.
314 (s390_add_execute): Removed.
315 (s390_dump_pool): Remove TARGET_CPU_ZARCH usages.
316 (s390_mainpool_start): Likewise.
317 (s390_mainpool_finish): Likewise.
318 (s390_mainpool_cancel): Removed.
319 (s390_chunkify_start): Remove TARGET_CPU_ZARCH usages.
320 (s390_chunkify_cancel): Likewise.
321 (s390_return_addr_rtx): Likewise.
322 (s390_register_info): Remove split_branches_pending_p uages.
323 (s390_optimize_register_info): Likewise.
324 (s390_init_frame_layout): Remove TARGET_CPU_ZARCH and
325 split_branches_pending_p usages.
326 (s390_can_eliminate): Remove TARGET_CPU_ZARCH usages.
327 (s390_load_got): Likewise.
328 (s390_expand_split_stack_prologue): Likewise.
329 (output_asm_nops): Likewise.
330 (s390_function_profiler): Likewise.
331 (s390_emit_call): Likewise.
332 (s390_conditional_register_usage): Likewise.
333 (s390_optimize_prologue): Likewise.
334 (s390_reorg): Remove TARGET_CPU_ZARCH and
335 split_branches_pending_p usages.
336 (s390_option_override_internal): Remove TARGET_CPU_ZARCH
338 (s390_output_indirect_thunk_function): Likewise.
339 * config/s390/s390.h (TARGET_CPU_ZARCH): Removed.
340 (TARGET_CPU_ZARCH_P): Removed.
341 (struct machine_function): Remove split_branches_pending_p.
342 * config/s390/s390.md: Remove TARGET_CPU_ZARCH usages.
344 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
346 * common/config/s390/s390-common.c (processor_flags_table):
348 * config.gcc: Remove with_arch/with_tune support.
349 * config/s390/2064.md: Remove cpu attribute comparisons.
350 * config/s390/driver-native.c (s390_host_detect_local_cpu):
352 * config/s390/linux.h (ASM_SPEC):
353 Remove -march support.
354 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal):
355 Use a table to get an arch level.
356 * config/s390/s390-opts.h (enum processor_type):
359 (processor_table): Remove entries, add arch_level values.
360 (s390_issue_rate): Remove cases.
361 (s390_option_override): Adjust
362 s390_option_override_internal() call.
363 (s390_option_override_internal): Remove deprecation warning.
364 (s390_valid_target_attribute_tree): Adjust
365 s390_option_override_internal() call.
366 * config/s390/s390.h (struct s390_processor):
367 Share with s390-c.c, add arch_level field.
368 * config/s390/s390.md:
369 Remove occurrences in cpu attribute.
370 * config/s390/s390.opt: Remove -march/-mtune support.
371 * config/s390/tpf.h (ASM_SPEC): Remove -march support.
372 * doc/invoke.texi: Remove deprecation warning.
374 2018-08-08 Luis Machado <luis.machado@linaro.org>
376 * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static
378 (qdf24xx_tunings): Set vector cost structure to
381 * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
382 <register_sextend>: Set to 3.
384 2018-08-07 Richard Sandiford <richard.sandiford@arm.com>
387 * config/aarch64/iterators.md (FRECP, frecp_suffix): Delete.
388 * config/aarch64/aarch64-simd.md
389 (aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into...
390 (@aarch64_frecpe<mode>): ...here and the move FRECPX to...
391 (aarch64_frecpx<mode>): ...this new pattern.
392 * config/aarch64/aarch64-simd-builtins.def: Remove comment
393 about aarch64_frecp<FRECP:frecp_suffix><mode>.
395 2018-08-07 Martin Liska <mliska@suse.cz>
398 * predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC,
399 BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW.
400 * predict.def (PRED_MALLOC_NONNULL): New predictor.
401 * doc/extend.texi: Document that malloc attribute adds
404 2018-08-06 John David Anglin <danglin@gcc.gnu.org>
407 * config/pa/pa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
408 Define to speculation_safe_value_not_needed.
410 2018-08-06 Jeff Law <law@redhat.com>
412 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down
413 the vr_values instance to cprop_into_stmt.
414 (cprop_into_stmt): Pass vr_values instance down to cprop_operand.
415 (cprop_operand): Also query EVRP to determine if OP is a constant.
417 2018-08-06 Nathan Sidwell <nathan@acm.org>
419 * diagnostic.c (diagnostic_report_current_module): Reroll
420 included-at loop. Translate text.
422 2018-08-06 David Malcolm <dmalcolm@redhat.com>
424 * function-tests.c (selftest::test_expansion_to_rtl): Call
425 free_after_compilation.
427 2018-08-06 Alan Hayward <alan.hayward@arm.com>
429 * config/aarch64/aarch64.md: Add clobber highs to tls_desc.
431 2018-08-06 Andreas Krebbel <krebbel@linux.ibm.com>
433 * config/s390/s390.c (s390_loop_unroll_adjust): Prevent small
434 loops with memory block operations from getting unrolled.
436 2018-08-06 Ulrich Weigand <uweigand@de.ibm.com>
439 * config/spu/spu.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
440 Define to speculation_safe_value_not_needed.
442 2018-08-06 Jeff Law <law@redhat.com>
444 * reload1.c (forget_old_reloads_1): Adjust CLOBBER_HIGH
447 2018-08-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
450 * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
451 with all enabled __intN types.
453 * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
455 2018-08-06 Alan Hayward <alan.hayward@arm.com>
457 * alias.c (record_set): Check for clobber high.
458 * cfgexpand.c (expand_gimple_stmt): Likewise.
459 * combine-stack-adj.c (single_set_for_csa): Likewise.
460 * combine.c (find_single_use_1): Likewise.
461 (set_nonzero_bits_and_sign_copies): Likewise.
462 (get_combine_src_dest): Likewise.
463 (is_parallel_of_n_reg_sets): Likewise.
464 (try_combine): Likewise.
465 (record_dead_and_set_regs_1): Likewise.
466 (reg_dead_at_p_1): Likewise.
467 (reg_dead_at_p): Likewise.
468 * dce.c (deletable_insn_p): Likewise.
469 (mark_nonreg_stores_1): Likewise.
470 (mark_nonreg_stores_2): Likewise.
471 * df-scan.c (df_find_hard_reg_defs): Likewise.
472 (df_uses_record): Likewise.
473 (df_get_call_refs): Likewise.
474 * dwarf2out.c (mem_loc_descriptor): Likewise.
475 * haifa-sched.c (haifa_classify_rtx): Likewise.
476 * ira-build.c (create_insn_allocnos): Likewise.
477 * ira-costs.c (scan_one_insn): Likewise.
478 * ira.c (equiv_init_movable_p): Likewise.
479 (rtx_moveable_p): Likewise.
480 (interesting_dest_for_shprep): Likewise.
481 * jump.c (mark_jump_label_1): Likewise.
482 * postreload-gcse.c (record_opr_changes): Likewise.
483 * postreload.c (reload_cse_simplify): Likewise.
484 (struct reg_use): Add source expr.
485 (reload_combine): Check for clobber high.
486 (reload_combine_note_use): Likewise.
487 (reload_cse_move2add): Likewise.
488 (move2add_note_store): Likewise.
489 * print-rtl.c (print_pattern): Likewise.
490 * recog.c (decode_asm_operands): Likewise.
491 (store_data_bypass_p): Likewise.
492 (if_test_bypass_p): Likewise.
493 * regcprop.c (kill_clobbered_value): Likewise.
494 (kill_set_value): Likewise.
495 * reginfo.c (reg_scan_mark_refs): Likewise.
496 * reload1.c (maybe_fix_stack_asms): Likewise.
497 (eliminate_regs_1): Likewise.
498 (elimination_effects): Likewise.
499 (mark_not_eliminable): Likewise.
500 (scan_paradoxical_subregs): Likewise.
501 (forget_old_reloads_1): Likewise.
502 * reorg.c (find_end_label): Likewise.
503 (try_merge_delay_insns): Likewise.
504 (redundant_insn): Likewise.
505 (own_thread_p): Likewise.
506 (fill_simple_delay_slots): Likewise.
507 (fill_slots_from_thread): Likewise.
508 (dbr_schedule): Likewise.
509 * resource.c (update_live_status): Likewise.
510 (mark_referenced_resources): Likewise.
511 (mark_set_resources): Likewise.
512 * rtl.c (copy_rtx): Likewise.
513 * rtlanal.c (reg_referenced_p): Likewise.
514 (single_set_2): Likewise.
515 (noop_move_p): Likewise.
516 (note_stores): Likewise.
517 * sched-deps.c (sched_analyze_reg): Likewise.
518 (sched_analyze_insn): Likewise.
520 2018-08-06 Alan Hayward <alan.hayward@arm.com>
522 * cse.c (invalidate_reg): New function extracted from...
523 (invalidate): ...here.
524 (canonicalize_insn): Check for clobber high.
525 (invalidate_from_clobbers): invalidate clobber highs.
526 (invalidate_from_sets_and_clobbers): Likewise.
527 (count_reg_usage): Check for clobber high.
528 (insn_live_p): Likewise.
529 * cselib.c (cselib_expand_value_rtx_1):Likewise.
530 (cselib_invalidate_regno): Check for clobber in setter.
531 (cselib_invalidate_rtx): Pass through setter.
532 (cselib_invalidate_rtx_note_stores):
533 (cselib_process_insn): Check for clobber high.
534 * cselib.h (cselib_invalidate_rtx): Add operand.
536 2018-08-06 Alan Hayward <alan.hayward@arm.com>
538 * lra-eliminations.c (lra_eliminate_regs_1): Check for clobber high.
539 (mark_not_eliminable): Likewise.
540 * lra-int.h (struct lra_insn_reg): Add clobber high marker.
541 * lra-lives.c (process_bb_lives): Check for clobber high.
542 * lra.c (new_insn_reg): Remember clobber highs.
543 (collect_non_operand_hard_regs): Check for clobber high.
544 (lra_set_insn_recog_data): Likewise.
545 (add_regs_to_insn_regno_info): Likewise.
546 (lra_update_insn_regno_info): Likewise.
548 2018-08-06 Alan Hayward <alan.hayward@arm.com>
550 * rtl.h (reg_is_clobbered_by_clobber_high): Add declarations.
551 * rtlanal.c (reg_is_clobbered_by_clobber_high): Add function.
553 2018-08-06 Alan Hayward <alan.hayward@arm.com>
555 * emit-rtl.c (verify_rtx_sharing): Check for CLOBBER_HIGH.
556 (copy_insn_1): Likewise.
557 (gen_hard_reg_clobber_high): New gen function.
558 * genconfig.c (walk_insn_part): Check for CLOBBER_HIGH.
559 * genemit.c (gen_exp): Likewise.
560 (gen_emit_seq): Pass through info.
561 (gen_insn): Check for CLOBBER_HIGH.
562 (gen_expand): Pass through info.
563 (gen_split): Likewise.
564 (output_add_clobbers): Likewise.
565 * genrecog.c (validate_pattern): Check for CLOBBER_HIGH.
566 (remove_clobbers): Likewise.
567 * rtl.h (gen_hard_reg_clobber_high): New declaration.
569 2018-08-06 Alan Hayward <alan.hayward@arm.com>
571 * doc/rtl.texi (clobber_high): Add.
572 (parallel): Add in clobber high
573 * rtl.c (rtl_check_failed_code3): Add function.
574 * rtl.def (CLOBBER_HIGH): Add expression.
575 * rtl.h (RTL_CHECKC3): Add macro.
576 (rtl_check_failed_code3): Add declaration.
579 2018-08-05 H.J. Lu <hongjiu.lu@intel.com>
582 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
583 cfun->machine->max_used_stack_alignment if needed.
585 2018-08-04 Martin Sebor <msebor@redhat.com>
587 PR tree-optimization/86571
588 * gimple-ssa-sprintf.c (format_floating): Extend upper bound of
591 2018-08-03 Sandra Loosemore <sandra@codesourcery.com>
594 * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
597 2018-08-03 Jeff Law <law@redhat.com>
600 * config/mn10300/mn10300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
601 Define to speculation_safe_value_not_needed.
603 2018-08-03 David Malcolm <dmalcolm@redhat.com>
605 * doc/gcov.texi (-x): Remove duplicate "to".
606 * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
607 (-Wif-not-aligned): Remove duplicate "is".
608 (-flto): Remove duplicate "the".
609 (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
611 (MSP430 Options): Remove duplicate "and" from the description
612 of "-mgprel-sec=regexp".
613 (x86 Options): Remove duplicate copies of "vmldLog102" and
614 vmlsLog104 from description of "-mveclibabi=type".
616 2018-08-03 Richard Sandiford <richard.sandiford@arm.com>
618 * internal-fn.h (first_commutative_argument): Declare.
619 * internal-fn.c (first_commutative_argument): New function.
620 * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove extra
621 restrictions for pattern statements. Use first_commutative_argument
622 to look for commutative operands in calls to internal functions.
624 2018-08-03 Aldy Hernandez <aldyh@redhat.com>
626 * Makefile.in (wide-int-range.o): New.
627 * tree-vrp.c: Move all the wide_int_* functions to...
628 * wide-int-range.cc: ...here.
629 * tree-vrp.h: Move all the wide_int_* prototypes to...
630 * wide-int-range.h: ...here.
632 2018-08-03 Tom de Vries <tdevries@suse.de>
634 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): Return
636 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Remove define.
637 * except.c (output_function_exception_table): Do early exit if
638 targetm_common.except_unwind_info (&global_options) == UI_NONE.
640 2018-08-03 Martin Liska <mliska@suse.cz>
642 * predict.c (dump_prediction): Change to 2 digits
645 2018-08-03 Siddhesh Poyarekar <siddhesh@sourceware.org>
647 * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
649 (falkor_am_1_gtov_gtov): ... a new insn reservation.
651 2018-07-19 Ilya Leoshkevich <iii@linux.ibm.com>
653 * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Replace > with >=.
654 * df-problems.c (df_remove_dead_eq_notes): Replace > with >=.
655 * dwarf2out.c (mem_loc_descriptor): Replace > with >=.
656 * lra-constraints.c (spill_hard_reg_in_range): Replace <= with <.
657 * lra-remat.c (call_used_input_regno_present_p): Replace <= with <.
659 2018-08-02 David Malcolm <dmalcolm@redhat.com>
661 * diagnostic-show-locus.c (diagnostic_show_locus): Use
662 pp_take_prefix when saving the existing prefix.
663 * diagnostic.c (diagnostic_append_note): Likewise.
664 * langhooks.c (lhd_print_error_function): Likewise.
665 * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
666 param's type. Free the existing prefix.
667 (pp_take_prefix): New function.
668 (pretty_printer::pretty_printer): Drop the prefix parameter.
669 Rename the length parameter to match the comment.
670 (pretty_printer::~pretty_printer): Free the prefix.
671 * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
673 (struct pretty_printer): Drop the "const" from "prefix" field's
674 type and clarify memory management.
675 (pp_set_prefix): Drop the "const" from the 2nd param.
676 (pp_take_prefix): New decl.
678 2018-08-02 Aaron Sawdey <acsawdey@linux.ibm.com>
680 * config/rs6000/rs6000-string.c (select_block_compare_mode): Move test
681 for word_mode_ok here instead of passing as argument.
682 (expand_block_compare): Change select_block_compare_mode() call.
683 (expand_strncmp_gpr_sequence): New function.
684 (expand_strn_compare): Make use of expand_strncmp_gpr_sequence.
686 2018-08-02 Jeff Law <law@redhat.com>
689 * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
690 Define to speculation_safe_value_not_needed.
693 * config/h8300/h8300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
694 Define to speculation_safe_value_not_needed.
696 2018-08-02 Tom de Vries <tdevries@suse.de>
699 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): New
700 function. Return UI_TARGET unconditionally.
701 (TARGET_EXCEPT_UNWIND_INFO): Redefine to nvptx_except_unwind_info.
702 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Emit commented out '.byte'.
704 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
706 * genemit.c (print_overload_test): Fix typo.
708 2018-08-02 Richard Biener <rguenther@suse.de>
710 PR tree-optimization/86816
711 * tree-ssa-tail-merge.c (tail_merge_valueize): New function
712 which checks for value availability before querying it.
713 (gvn_uses_equal): Use it.
714 (same_succ_hash): Likewise.
715 (gimple_equal_p): Likewise.
717 2018-08-02 Nick Clifton <nickc@redhat.com>
720 * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
721 Define to speculation_safe_value_not_needed.
724 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
725 Define to speculation_safe_value_not_needed.
728 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
729 Define to speculation_safe_value_not_needed.
732 * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
733 Define to speculation_safe_value_not_needed.
736 * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
737 Define to speculation_safe_value_not_needed.
740 * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
741 Define to speculation_safe_value_not_needed.
744 * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
745 Define to speculation_safe_value_not_needed.
748 * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
749 Define to speculation_safe_value_not_needed.
752 * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
753 speculation_safe_value_not_needed.
756 * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define
757 to speculation_safe_value_not_needed.
759 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
761 * doc/md.texi: Expand the documentation of instruction names
762 to mention port-local uses. Document '@' in pattern names.
763 * read-md.h (overloaded_instance, overloaded_name): New structs.
765 (md_reader::handle_overloaded_name): New member function.
766 (md_reader::get_overloads): Likewise.
767 (md_reader::m_first_overload): New member variable.
768 (md_reader::m_next_overload_ptr): Likewise.
769 (md_reader::m_overloads_htab): Likewise.
770 * read-md.c (md_reader::md_reader): Initialize m_first_overload,
771 m_next_overload_ptr and m_overloads_htab.
772 * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields.
773 (get_mode_token, get_code_token, get_int_token): New functions.
774 (map_attr_string): Add an optional argument that passes back
775 the associated iterator.
776 (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p):
777 (md_reader::handle_overloaded_name, add_overload_instance): New
779 (apply_iterators): Handle '@' names. Report an error if '@'
780 is used without iterators.
781 (initialize_iterators): Initialize the new iterator_group fields.
782 * genopinit.c (handle_overloaded_code_for)
783 (handle_overloaded_gen): New functions.
784 (main): Use them to print declarations of maybe_code_for_* and
785 maybe_gen_* functions, and inline definitions of code_for_* and gen_*.
786 * genemit.c (print_overload_arguments, print_overload_test)
787 (handle_overloaded_code_for, handle_overloaded_gen): New functions.
788 (main): Use it to print definitions of maybe_code_for_* and
789 maybe_gen_* functions.
790 * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use
791 gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high}
792 instead of explicit mode checks.
793 (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine.
794 (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov.
795 (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive.
796 (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive.
797 (aarch64_expand_compare_and_swap): Likewise
798 gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse
799 (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas.
800 (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp.
801 (aarch64_constant_pool_reload_icode): Delete.
802 (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp
803 instead of aarch64_constant_pool_reload_icode. Use
804 code_for_aarch64_reload_mov instead of explicit mode checks.
805 (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete.
806 (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of
807 get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type.
808 (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete.
809 (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of
810 get_recpe_type and gen_aarch64_frecps instead of get_recps_type.
811 (aarch64_atomic_load_op_code): Delete.
812 (aarch64_emit_atomic_load_op): Likewise.
813 (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of
814 aarch64_atomic_load_op_code. Use gen_aarch64_atomic_load
815 instead of aarch64_emit_atomic_load_op.
816 * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>)
817 (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>)
818 (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high)
819 (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@'
820 character before the pattern name.
821 * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
822 (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>)
823 (aarch64_simd_combine<mode>, aarch64_frecpe<mode>)
824 (aarch64_frecps<mode>): Likewise.
825 * config/aarch64/atomics.md (atomic_compare_and_swap<mode>)
826 (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse)
827 (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>)
828 (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>)
829 (aarch64_atomic_load<atomic_ldop><mode>): Likewise.
831 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
833 * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
834 Allow HFmode constants if TARGET_FP_F16INST.
836 2018-08-02 Jackson Woodruff <jackson.woodruff@arm.com>
839 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp):
840 No longer check last store for clobber of address register.
842 2018-08-02 Martin Liska <mliska@suse.cz>
844 PR gcov-profile/86817
845 * gcov.c (process_all_functions): New function.
847 (process_file): Move functions processing to
848 process_all_functions.
850 2018-08-02 David Malcolm <dmalcolm@redhat.com>
852 * dumpfile.c (dump_user_location_t::dump_user_location_t): Add
853 "const" to the "gimple *" and "rtx_insn *" parameters.
854 * dumpfile.h (dump_user_location_t::dump_user_location_t):
856 (dump_location_t::dump_location_t): Likewise.
858 2018-08-01 Martin Sebor <msebor@redhat.com>
860 PR tree-optimization/86650
861 * gimple-pretty-print.c (percent_G_format): Accept a "gimple *"
862 rather than a "gcall *". Directly pass the data of interest
863 to percent_K_format, rather than building a temporary CALL_EXPR
865 * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust.
866 (gimple_fold_builtin_strncat): Adjust.
867 * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace
868 gcall* argument with gimple*.
869 * gimple-ssa-warn-restrict.c (check_call): Same.
870 (wrestrict_dom_walker::before_dom_children): Same.
871 (builtin_access::builtin_access): Same.
872 (check_bounds_or_overlap): Same
873 (maybe_diag_overlap): Same.
874 (maybe_diag_offset_bounds): Same.
875 * tree-diagnostic.c (default_tree_printer): Move usage of
876 EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format
878 * tree-pretty-print.c (percent_K_format): Add argument.
879 * tree-pretty-print.h: Add argument.
880 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust.
881 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust.
882 (maybe_diag_stxncpy_trunc): Same.
883 (handle_builtin_stxncpy): Same.
884 (handle_builtin_strcat): Same.
886 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
888 * match.pd: Optimise pointer range checks.
890 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
892 PR tree-optimization/86758
893 * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try
894 to remove pattern statements.
896 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
898 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the
899 result of dfs_enumerate_from when constructing stmt_vec_infos,
900 instead of additionally calling get_loop_body.
902 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
904 * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p
906 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise.
907 When creating an iv, assert that the step is not known to be zero.
908 (vect_setup_realignment): Update call accordingly.
909 * tree-vect-stmts.c (vectorizable_store): Likewise.
910 (vectorizable_load): Likewise. Handle VMAT_INVARIANT separately.
912 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
914 * tree-vectorizer.h (vect_stmt_to_vectorize): New function.
915 * tree-vect-loop.c (vect_update_vf_for_slp): Use it.
916 (vectorizable_reduction): Likewise.
917 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
918 (vect_detect_hybrid_slp_stmts): Likewise.
919 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
921 2018-08-01 Aldy Hernandez <aldyh@redhat.com>
923 * tree-vrp (zero_nonzero_bits_from_bounds): Rename to...
924 (wide_int_set_zero_nonzero_bits): ...this.
925 (zero_nonzero_bits_from_vr): Rename to...
926 (vrp_set_zero_nonzero_bits): ...this.
927 (extract_range_from_multiplicative_op_1): Abstract wide int
929 (wide_int_range_multiplicative_op): ...here.
930 (extract_range_from_binary_expr_1): Extract wide int binary
931 operations into their own functions.
932 (wide_int_range_lshift): New.
933 (wide_int_range_can_optimize_bit_op): New.
934 (wide_int_range_shift_undefined_p): New.
935 (wide_int_range_bit_xor): New.
936 (wide_int_range_bit_ior): New.
937 (wide_int_range_bit_and): New.
938 (wide_int_range_trunc_mod): New.
939 (extract_range_into_wide_ints): New.
940 (vrp_shift_undefined_p): New.
941 (extract_range_from_multiplicative_op): New.
942 (vrp_can_optimize_bit_op): New.
943 * tree-vrp.h (value_range::dump): New.
944 (wide_int_range_multiplicative_op): New.
945 (wide_int_range_lshift):New.
946 (wide_int_range_shift_undefined_p): New.
947 (wide_int_range_bit_xor): New.
948 (wide_int_range_bit_ior): New.
949 (wide_int_range_bit_and): New.
950 (wide_int_range_trunc_mod): New.
951 (zero_nonzero_bits_from_bounds): Rename to...
952 (wide_int_set_zero_nonzero_bits): ...this.
953 (zero_nonzero_bits_from_vr): Rename to...
954 (vrp_set_zero_nonzero_bits): ...this.
955 (range_easy_mask_min_max): Rename to...
956 (wide_int_range_can_optimize_bit_op): this.
957 * vr-values.c (simplify_bit_ops_using_ranges): Rename
958 zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits.
960 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
962 * tree-vectorizer.h (vect_orig_stmt): New function.
963 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it.
964 * tree-vect-loop.c (vect_model_reduction_cost): Likewise.
965 (vect_create_epilog_for_reduction): Likewise.
966 (vectorizable_live_operation): Likewise.
967 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
968 (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise.
969 * tree-vect-stmts.c (vectorizable_call): Likewise.
970 (vectorizable_simd_clone_call, vect_remove_stores): Likewise.
972 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
974 * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store
976 * tree-vect-stmts.c (vect_transform_stmt): Likewise.
977 * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly.
978 (vect_transform_loop): Likewise.
979 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
981 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
983 * tree-vectorizer.h (vect_schedule_slp): Return void.
984 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
985 (vect_schedule_slp): Likewise.
987 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
989 * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled
991 (vect_transform_loop): Update calls accordingly. Schedule SLP
992 instances before the main loop, if any exist.
994 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
996 PR tree-optimization/86749
997 * tree-vect-patterns.c (vect_determine_min_output_precision_1):
998 If the lhs is used in a COND_EXPR, check that it is being used
999 as the "then" or "else" value.
1001 2018-08-01 Tom de Vries <tdevries@suse.de>
1004 * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
1005 speculation_safe_value_not_needed.
1007 2018-08-01 Richard Biener <rguenther@suse.de>
1009 * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses
1012 2018-08-01 Martin Liska <mliska@suse.cz>
1014 * value-prof.c (gimple_divmod_fixed_value_transform): Unify
1015 format how successful transformation is dumped.
1016 (gimple_mod_pow2_value_transform): Likewise.
1017 (gimple_mod_subtract_transform): Likewise.
1018 (gimple_stringops_transform): Likewise.
1020 2018-08-01 Martin Liska <mliska@suse.cz>
1023 * value-prof.c (interesting_stringop_to_profile_p):
1024 Simplify the code and add BUILT_IN_MEMMOVE.
1025 (gimple_stringops_transform): Likewise.
1027 2018-08-01 Sam Tebbs <sam.tebbs@arm.com>
1029 * config/aarch64/aarch64-simd.md
1030 (*aarch64_get_lane_zero_extendsi<mode>): Rename to...
1031 (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and
1032 use GPI iterator instead of SI mode.
1034 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
1036 * config/rs6000/rs6000.md (speculation_barrier): Renamed from
1037 rs6000_speculation_barrier.
1038 * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for
1039 new barrier pattern name.
1041 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
1043 * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
1044 (speculation_barrier): New insn.
1046 2018-08-01 Richard Biener <rguenther@suse.de>
1049 * graphite.h: Include isl/id.h and isl/space.h to allow build
1052 2018-08-01 Jan Willem Jagersma <jwjagersma@gmail.com>
1055 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
1056 mode for COFF targets.
1057 * defaults.h (TARGET_COFF): Define.
1058 * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
1059 TARGET_COFF): Define.
1060 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
1061 * config/i386/djgpp.c (saved_debug_info_level): New static variable.
1062 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
1064 2018-07-31 Alexandre Oliva <oliva@adacore.com>
1066 * gimple-streamer-in.c (input_bb): Restore BB discriminator.
1067 * gimple-streamer-out.c (output_bb): Save it.
1068 * lto-streamer-in.c (input_struct_function_base): Restore
1069 instance discriminator if available. Create map on demand.
1070 * lto-streamer-out.c (output_struct_function_base): Save it if
1072 * final.c (decl_to_instance_map): Document LTO strategy.
1074 2018-07-31 Alexandre Oliva <oliva@adacore.com>
1075 Olivier Hainque <hainque@adacore.com>
1077 * debug.h (decl_to_instance_map_t): New type.
1078 (decl_to_instance_map): Declare.
1079 (maybe_create_decl_to_instance_map): New inline function.
1080 * final.c (bb_discriminator, last_bb_discriminator): New statics,
1081 to track basic block discriminators.
1082 (final_start_function_1): Initialize them.
1083 (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
1085 (decl_to_instance_map): New variable.
1086 (map_decl_to_instance, maybe_set_discriminator): New functions.
1087 (notice_source_line): Set discriminator.
1089 2018-07-31 Ian Lance Taylor <iant@golang.org>
1091 * targhooks.c (default_have_speculation_safe_value): Add
1094 2018-07-31 David Malcolm <dmalcolm@redhat.com>
1096 * dump-context.h: Include "pretty-print.h".
1097 (dump_context::refresh_dumps_are_enabled): New decl.
1098 (dump_context::emit_item): New decl.
1099 (class dump_context): Add fields "m_test_pp" and
1101 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
1102 (temp_dump_context::get_dumped_text): New decl.
1103 (class temp_dump_context): Add field "m_pp".
1104 * dumpfile.c (refresh_dumps_are_enabled): Convert to...
1105 (dump_context::refresh_dumps_are_enabled): ...and add a test for
1107 (set_dump_file): Update for above change.
1108 (set_alt_dump_file): Likewise.
1109 (dump_loc): New overload, taking a pretty_printer *.
1110 (dump_context::dump_loc): Call end_any_optinfo. Dump the location
1111 to any test pretty-printer.
1112 (make_item_for_dump_gimple_stmt): New function, adapted from
1113 optinfo::add_gimple_stmt.
1114 (dump_context::dump_gimple_stmt): Call it, and use the result,
1115 eliminating the direct usage of dump_file and alt_dump_file in
1116 favor of indirectly using them via emit_item.
1117 (make_item_for_dump_gimple_expr): New function, adapted from
1118 optinfo::add_gimple_expr.
1119 (dump_context::dump_gimple_expr): Call it, and use the result,
1120 eliminating the direct usage of dump_file and alt_dump_file in
1121 favor of indirectly using them via emit_item.
1122 (make_item_for_dump_generic_expr): New function, adapted from
1124 (dump_context::dump_generic_expr): Call it, and use the result,
1125 eliminating the direct usage of dump_file and alt_dump_file in
1126 favor of indirectly using them via emit_item.
1127 (make_item_for_dump_printf_va): New function, adapted from
1128 optinfo::add_printf_va.
1129 (make_item_for_dump_printf): New function.
1130 (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
1131 and use the result, eliminating the direct usage of dump_file and
1132 alt_dump_file in favor of indirectly using them via emit_item.
1133 (make_item_for_dump_dec): New function.
1134 (dump_context::dump_dec): Call it, and use the result,
1135 eliminating the direct usage of dump_file and alt_dump_file in
1136 favor of indirectly using them via emit_item.
1137 (make_item_for_dump_symtab_node): New function, adapted from
1138 optinfo::add_symtab_node.
1139 (dump_context::dump_symtab_node): Call it, and use the result,
1140 eliminating the direct usage of dump_file and alt_dump_file in
1141 favor of indirectly using them via emit_item.
1142 (dump_context::begin_scope): Reimplement, avoiding direct usage
1143 of dump_file and alt_dump_file in favor of indirectly using them
1145 (dump_context::emit_item): New member function.
1146 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
1147 Set up test pretty-printer on the underlying context. Call
1148 refresh_dumps_are_enabled.
1149 (temp_dump_context::~temp_dump_context): Call
1150 refresh_dumps_are_enabled.
1151 (temp_dump_context::get_dumped_text): New member function.
1152 (selftest::verify_dumped_text): New function.
1153 (ASSERT_DUMPED_TEXT_EQ): New macro.
1154 (selftest::test_capture_of_dump_calls): Run all tests twice, with
1155 and then without optinfo enabled. Add uses of
1156 ASSERT_DUMPED_TEXT_EQ to all tests. Add test of nested scopes.
1157 * dumpfile.h: Update comment for the dump_* API.
1158 * optinfo-emit-json.cc
1159 (selftest::test_building_json_from_dump_calls): Update for new
1160 param for temp_dump_context ctor.
1161 * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
1162 and "m_owned" field.
1163 (optinfo_item::~optinfo_item): Likewise.
1164 (optinfo::add_item): New member function.
1165 (optinfo::emit): Update comment.
1166 (optinfo::add_string): Delete.
1167 (optinfo::add_printf): Delete.
1168 (optinfo::add_printf_va): Delete.
1169 (optinfo::add_gimple_stmt): Delete.
1170 (optinfo::add_gimple_expr): Delete.
1171 (optinfo::add_tree): Delete.
1172 (optinfo::add_symtab_node): Delete.
1173 (optinfo::add_dec): Delete.
1174 * optinfo.h (class dump_context): New forward decl.
1175 (optinfo::add_item): New decl.
1176 (optinfo::add_string): Delete.
1177 (optinfo::add_printf): Delete.
1178 (optinfo::add_printf_va): Delete.
1179 (optinfo::add_gimple_stmt): Delete.
1180 (optinfo::add_gimple_expr): Delete.
1181 (optinfo::add_tree): Delete.
1182 (optinfo::add_symtab_node): Delete.
1183 (optinfo::add_dec): Delete.
1184 (optinfo::add_poly_int): Delete.
1185 (optinfo_item::optinfo_item): Remove "owned" param.
1186 (class optinfo_item): Remove field "m_owned".
1188 2018-07-31 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1191 * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
1192 MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
1193 requested variable alignment.
1194 (expand_one_ssa_partition): Likewise.
1195 (expand_one_var): Likewise.
1197 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1199 * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
1200 to speculation_safe_value_not_needed.
1202 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1204 * targhooks.h (speculation_safe_value_not_needed): New prototype.
1205 * targhooks.c (speculation_safe_value_not_needed): New function.
1206 * target.def (have_speculation_safe_value): Update documentation.
1207 * doc/tm.texi: Regenerated.
1209 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1211 * config/aarch64/iterators.md (ALLI_TI): New iterator.
1212 * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
1214 (despeculate_copy<ALLI:mode>_insn): New insn.
1215 (despeculate_copyti_insn): New insn.
1216 (despeculate_simple<ALLI:mode>): New insn
1217 (despeculate_simpleti): New insn.
1218 * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
1220 (TARGET_SPECULATION_SAFE_VALUE): Redefine to
1221 aarch64_speculation_safe_value.
1222 (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
1224 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1226 * config/aarch64/aarch64-speculation.cc: New file.
1227 * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
1228 before pass_reorder_blocks.
1229 * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
1231 * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
1232 X14 and X15 when tracking speculation.
1233 * config/aarch64/aarch64.md (register name constants): Add
1234 SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
1235 (unspec): Add UNSPEC_SPECULATION_TRACKER.
1236 (speculation_barrier): New insn attribute.
1237 (cmp<mode>): Allow SP in comparisons.
1238 (speculation_tracker): New insn.
1239 (speculation_barrier): Add speculation_barrier attribute.
1240 * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
1241 * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
1242 * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
1244 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1246 * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
1247 aarch64_track_speculation is true.
1248 (tb<optab><mode>1): Likewise.
1249 * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
1250 generate CB[N]Z when tracking speculation.
1251 (aarch64_split_compare_and_swap): Likewise.
1252 (aarch64_split_atomic_op): Likewise.
1254 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1256 * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
1258 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1260 * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
1261 (speculation_barrier): New insn.
1263 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1265 * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
1266 * config/arm/arm.md (speculation_barrier): New expand.
1267 (speculation_barrier_insn): New pattern.
1269 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1271 * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
1272 (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
1273 (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
1274 * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
1276 * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
1277 (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
1278 (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
1279 (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
1280 (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
1281 (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
1282 (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
1283 * builtins.c (expand_speculation_safe_value): New function.
1284 (expand_builtin): Call it.
1285 * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
1286 * doc/extend.texi: Document __builtin_speculation_safe_value.
1287 * doc/md.texi: Document "speculation_barrier" pattern.
1288 * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
1289 TARGET_HAVE_SPECULATION_SAFE_VALUE.
1290 * doc/tm.texi: Regenerated.
1291 * target.def (have_speculation_safe_value, speculation_safe_value): New
1293 * targhooks.c (default_have_speculation_safe_value): New function.
1294 (default_speculation_safe_value): New function.
1295 * targhooks.h (default_have_speculation_safe_value): Add prototype.
1296 (default_speculation_safe_value): Add prototype.
1298 2018-07-31 David Malcolm <dmalcolm@redhat.com>
1300 * dump-context.h (dump_context::dump_loc): New decl.
1301 * dumpfile.c (dump_context::dump_loc): New member function.
1302 (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
1303 and dump_gimple_stmt.
1304 (dump_context::dump_gimple_expr_loc): Likewise, using
1306 (dump_context::dump_generic_expr_loc): Likewise, using
1308 (dump_context::dump_printf_loc_va): Likewise, using
1310 (dump_context::begin_scope): Explicitly using the global function
1311 "dump_loc", rather than the member function.
1313 2018-07-31 Martin Sebor <msebor@redhat.com>
1315 PR tree-optimization/86741
1316 * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
1318 2018-07-31 Andreas Krebbel <krebbel@linux.ibm.com>
1320 * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
1321 depend on whether prefetch instructions will be emitted or not.
1322 Use TARGET_SETMEM_PFD for checking whether prefetch instructions
1323 will be emitted or not.
1324 * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
1325 (TARGET_SETMEM_PFD): New macros.
1327 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1329 * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
1330 (NULL_STMT_VEC_INFO): Delete.
1331 (stmt_vec_info::operator*): Likewise.
1332 (stmt_vec_info::operator gimple *): Likewise.
1333 * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
1334 of NULL_STMT_VEC_INFO.
1335 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
1336 (vect_reassociating_reduction_p): Likewise.
1337 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
1338 (vectorizable_store): Likewise.
1339 * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
1340 (vec_info::free_stmt_vec_infos): Likewise.
1342 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1344 * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
1345 * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
1347 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1349 * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
1350 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
1351 (vec_info::free_stmt_vec_info): New private member functions.
1352 (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
1353 (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
1354 * tree-parloops.c (gather_scalar_reductions): Remove calls to
1355 set_stmt_vec_info_vec and free_stmt_vec_infos.
1356 * tree-vect-loop.c (_loop_vec_info): Remove call to
1357 set_stmt_vec_info_vec.
1358 * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
1359 (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
1360 * tree-vectorizer.c (vec_info::new_stmt_vec_info)
1361 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
1362 (vec_info::free_stmt_vec_info): ...these new functions. Remove
1363 assignments in {vec_info::,}new_stmt_vec_info that are redundant
1364 with the clearing in the xcalloc.
1365 (stmt_vec_info_vec): Delete.
1366 (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
1367 (vectorize_loops): Likewise.
1368 (vec_info::~vec_info): Remove argument from call to
1369 free_stmt_vec_infos.
1370 (vec_info::add_stmt): Remove vinfo argument from call to
1373 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1375 * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
1376 rather than a gimple stmt.
1377 * tree-vect-stmts.c (free_stmt_vec_info): Likewise. Don't free
1378 information for pattern statements when passed the original
1379 statement; instead wait to be passed the pattern statement itself.
1380 Don't call set_vinfo_for_stmt here.
1381 (free_stmt_vec_infos): Update call to free_stmt_vec_info.
1382 * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
1383 stmt_vec_infos here.
1384 * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
1385 * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
1386 stmt_vec_infos entry.
1388 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1390 * tree-vectorizer.h (vec_info::replace_stmt): Declare.
1391 * tree-vectorizer.c (vec_info::replace_stmt): New function.
1392 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
1393 * tree-vect-stmts.c (vectorizable_call): Likewise.
1394 (vectorizable_simd_clone_call): Likewise.
1396 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1398 * tree-vectorizer.h (vec_info::remove_stmt): Declare.
1399 * tree-vectorizer.c (vec_info::remove_stmt): New function.
1400 * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
1401 * tree-vect-loop.c (vect_transform_loop): Likewise.
1402 * tree-vect-slp.c (vect_schedule_slp): Likewise.
1403 * tree-vect-stmts.c (vect_remove_stores): Likewise.
1405 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1407 * tree-vectorizer.h (vec_info::lookup_dr): New member function.
1408 (vect_dr_stmt): Delete.
1409 * tree-vectorizer.c (vec_info::lookup_dr): New function.
1410 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
1412 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
1413 (vect_analyze_data_ref_dependence, vect_record_base_alignments)
1414 (vect_verify_datarefs_alignment, vect_peeling_supportable)
1415 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
1416 (vect_analyze_data_refs): Likewise.
1417 (vect_slp_analyze_data_ref_dependence): Likewise. Take a vec_info
1419 (vect_find_same_alignment_drs): Likewise.
1420 (vect_slp_analyze_node_dependences): Update calls accordingly.
1421 (vect_analyze_data_refs_alignment): Likewise. Use vec_info::lookup_dr
1422 instead of DR_VECT_AUX.
1423 (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
1424 of a vector data references. Use vec_info::lookup_dr instead of
1426 (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
1427 (vect_enhance_data_refs_alignment): Likewise. Use vec_info::lookup_dr
1428 instead of DR_VECT_AUX.
1430 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1432 * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
1434 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
1436 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
1437 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
1438 (vect_gen_prolog_loop_niters): Likewise.
1440 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1442 * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
1443 (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
1444 (vect_known_alignment_in_bytes, vect_dr_behavior)
1445 (vect_get_scalar_dr_size): Take references as dr_vec_infos
1446 instead of data_references. Update calls to other routines for
1447 which the same change has been made.
1448 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
1449 dr_vec_infos instead of stmt_vec_infos.
1450 (vect_analyze_data_ref_dependence): Update call accordingly.
1451 (vect_slp_analyze_data_ref_dependence)
1452 (vect_record_base_alignments): Use DR_VECT_AUX.
1453 (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
1454 (vect_update_misalignment_for_peel, verify_data_ref_alignment)
1455 (vector_alignment_reachable_p, vect_get_data_access_cost)
1456 (vect_peeling_supportable, vect_analyze_group_access_1)
1457 (vect_analyze_group_access, vect_analyze_data_ref_access)
1458 (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
1459 (vect_compile_time_alias, vect_small_gap_p)
1460 (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
1461 (vect_supportable_dr_alignment): Take references as dr_vec_infos
1462 instead of data_references. Update calls to other routines for
1463 which the same change has been made.
1464 (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
1465 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
1466 (vect_slp_analyze_and_verify_node_alignment)
1467 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
1468 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
1469 (vect_setup_realignment): Use dr_vec_infos. Update calls after
1471 (_vect_peel_info::dr): Replace with...
1472 (_vect_peel_info::dr_info): ...this new field.
1473 (vect_peeling_hash_get_most_frequent)
1474 (vect_peeling_hash_choose_best_peeling): Update accordingly.
1475 (vect_peeling_hash_get_lowest_cost):
1476 (vect_enhance_data_refs_alignment): Likewise. Update calls to other
1477 routines for which the same change has been made.
1478 (vect_peeling_hash_insert): Likewise. Take a dr_vec_info instead of a
1480 * tree-vect-loop-manip.c (get_misalign_in_elems)
1481 (vect_gen_prolog_loop_niters): Use dr_vec_infos. Update calls after
1483 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
1484 * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
1485 (vect_truncate_gather_scatter_offset, compare_step_with_zero)
1486 (get_group_load_store_type, get_negative_load_store_type)
1487 (vect_get_data_ptr_increment, vectorizable_store)
1488 (vectorizable_load): Likewise.
1489 (ensure_base_align): Take a dr_vec_info instead of a data_reference.
1490 Update calls to other routines for which the same change has been made.
1492 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1494 * tree-vectorizer.h (vec_info::move_dr): New member function.
1495 (dataref_aux): Rename to...
1496 (dr_vec_info): ...this and add "dr" and "stmt" fields.
1497 (_stmt_vec_info::dr_aux): Update accordingly.
1498 (_stmt_vec_info::data_ref_info): Delete.
1499 (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
1500 (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
1501 (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
1502 (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
1504 (STMT_VINFO_DATA_REF): Likewise. Turn into an lvalue.
1505 (STMT_VINFO_DR_INFO): New macro.
1506 (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
1507 (set_dr_misalignment): Update after rename of dataref_aux.
1508 (vect_dr_stmt): Move earlier in file. Return dr_aux.stmt.
1509 * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
1510 initialization of STMT_VINFO_DATA_REF.
1511 * tree-vectorizer.c (vec_info::move_dr): New function.
1512 * tree-vect-patterns.c (vect_recog_bool_pattern)
1513 (vect_recog_mask_conversion_pattern)
1514 (vect_recog_gather_scatter_pattern): Use it.
1515 * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
1516 the "dr" and "stmt" fields of dr_vec_info instead of
1517 STMT_VINFO_DATA_REF.
1519 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1521 * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
1522 (is_pattern_stmt_p): Use it.
1523 * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
1524 on pattern statements.
1526 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1528 * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
1529 original stmt as a stmt_vec_info rather than a gimple stmt.
1530 (vect_pattern_recog_1): Take the statement directly as a
1531 stmt_vec_info, rather than via a gimple_stmt_iterator.
1532 Update call to vect_mark_pattern_stmts.
1533 (vect_pattern_recog): Update calls accordingly.
1535 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1537 * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
1538 (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
1539 a vect_def_type for the first argument.
1540 * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
1541 (vect_get_vec_def_for_stmt_copy): Likewise. Return the original
1542 operand if it isn't defined by a vectorized statement.
1543 (vect_build_gather_load_calls): Remove the mask_dt argument and
1544 update calls to vect_get_vec_def_for_stmt_copy.
1545 (vectorizable_bswap): Likewise the dt argument.
1546 (vectorizable_call): Update calls to vectorizable_bswap and
1547 vect_get_vec_def_for_stmt_copy.
1548 (vectorizable_simd_clone_call, vectorizable_assignment)
1549 (vectorizable_shift, vectorizable_operation, vectorizable_condition)
1550 (vectorizable_comparison): Update calls to
1551 vect_get_vec_def_for_stmt_copy.
1552 (vectorizable_store): Likewise. Remove now-unnecessary calls to
1554 (vect_get_loop_based_defs): Remove dt argument and update call
1555 to vect_get_vec_def_for_stmt_copy.
1556 (vectorizable_conversion): Update calls to vect_get_loop_based_defs
1557 and vect_get_vec_def_for_stmt_copy.
1558 (vectorizable_load): Update calls to vect_build_gather_load_calls
1559 and vect_get_vec_def_for_stmt_copy.
1560 * tree-vect-loop.c (vect_create_epilog_for_reduction)
1561 (vectorizable_reduction, vectorizable_live_operation): Update calls
1562 to vect_get_vec_def_for_stmt_copy.
1564 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1566 * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
1567 and gimple stmt arguments with a stmt_vec_info.
1568 (vect_record_base_alignments): Update calls accordingly.
1569 * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
1570 and gimple stmt arguments with a stmt_vec_info.
1571 (vect_build_slp_tree_1): Remove vinfo argument and update call
1572 to vect_record_max_nunits.
1573 (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
1574 and vect_record_max_nunits.
1576 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1578 * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
1579 file and take a stmt_vec_info instead of a gimple stmt.
1580 (supportable_widening_operation, vect_finish_replace_stmt)
1581 (vect_finish_stmt_generation, vect_get_store_rhs)
1582 (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
1583 (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
1584 (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
1585 (vect_get_smallest_scalar_type, vect_check_gather_scatter)
1586 (vect_create_data_ref_ptr, bump_vector_ptr)
1587 (vect_permute_store_chain, vect_setup_realignment)
1588 (vect_transform_grouped_load, vect_record_grouped_load_vectors)
1589 (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
1590 (vectorizable_reduction, vectorizable_induction)
1591 (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
1592 (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
1593 than gimple stmts as arguments.
1594 * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
1595 (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
1596 (can_group_stmts_p, vect_check_gather_scatter)
1597 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
1598 (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
1599 (vect_permute_load_chain, vect_shift_permute_load_chain)
1600 (vect_transform_grouped_load)
1601 (vect_record_grouped_load_vectors): Likewise.
1602 * tree-vect-loop.c (vect_fixup_reduc_chain)
1603 (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
1604 (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
1605 (vectorizable_reduction, vectorizable_induction)
1606 (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
1607 * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
1608 (vect_get_load_store_mask): Likewise.
1609 * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
1610 (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
1611 * tree-vect-stmts.c (vect_mark_relevant)
1612 (is_simple_and_all_uses_invariant)
1613 (exist_non_indexing_operands_for_use_p, process_use)
1614 (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
1615 (vect_get_vec_def_for_operand, vect_get_vec_defs)
1616 (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
1617 (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
1618 (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
1619 (get_negative_load_store_type, get_load_store_type)
1620 (vect_check_load_store_mask, vect_check_store_rhs)
1621 (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
1622 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
1623 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
1624 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
1625 (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
1626 (vectorizable_load, vectorizable_condition, vectorizable_comparison)
1627 (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
1628 (supportable_widening_operation): Likewise.
1630 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1632 * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
1633 a stmt_vec_info instead of a gcall.
1634 (vect_check_gather_scatter): Update call accordingly.
1635 * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
1637 (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
1638 (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
1639 * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
1640 instead of a gimple stmt.
1641 (vect_transform_loop): Update calls accordingly.
1642 * tree-vect-slp.c (vect_split_slp_store_group): Take and return
1643 stmt_vec_infos instead of gimple stmts.
1644 (vect_analyze_slp_instance): Update use accordingly.
1645 * tree-vect-stmts.c (read_vector_array, write_vector_array)
1646 (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
1647 (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
1648 (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
1649 (vect_gen_widened_results_half, vect_get_loop_based_defs)
1650 (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
1651 Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
1652 down to subroutines.
1654 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1656 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
1657 of the worklist from a vector of gimple stmts to a vector of
1659 * tree-vect-stmts.c (vect_mark_relevant, process_use)
1660 (vect_mark_stmts_to_be_vectorized): Likewise
1662 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1664 * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
1665 statement before passing it to vect_analyze_stmt.
1666 (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
1667 the chain of phi vector definitions. Track the exit phi via its
1669 (vectorizable_reduction): Set cond_stmt_vinfo directly from the
1670 STMT_VINFO_REDUC_DEF.
1671 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
1672 stmt_vec_infos to handle the statement chains.
1673 (vect_get_slp_defs): Record the first statement in the node
1674 using a stmt_vec_info.
1675 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
1676 statements here and pass their stmt_vec_info down to subroutines.
1677 (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
1678 down to vect_finish_stmt_generation.
1679 (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
1680 (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
1681 stmt_vec_infos to subroutines.
1682 (vect_remove_stores): Use stmt_vec_infos to handle the statement
1685 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1687 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
1688 (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
1689 (vect_permute_store_chain, vect_setup_realignment)
1690 (vect_permute_load_chain, vect_shift_permute_load_chain)
1691 (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
1692 stmts internally, and when passing values to other vectorizer routines.
1693 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
1694 * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
1695 (vect_analyze_loop_operations, get_initial_def_for_reduction)
1696 (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
1697 (vectorizable_reduction, vectorizable_induction)
1698 (vectorizable_live_operation, vect_transform_loop_stmt)
1699 (vect_transform_loop): Likewise.
1700 * tree-vect-patterns.c (vect_reassociating_reduction_p)
1701 (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
1702 (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
1703 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
1704 (vect_slp_analyze_node_operations_1): Likewise.
1705 * tree-vect-stmts.c (vect_mark_relevant, process_use)
1706 (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
1707 (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
1708 (vect_finish_stmt_generation_1, get_group_load_store_type)
1709 (get_load_store_type, vect_build_gather_load_calls)
1710 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
1711 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
1712 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
1713 (vectorizable_store, vectorizable_load, vectorizable_condition)
1714 (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
1715 (supportable_widening_operation): Likewise.
1716 (vect_get_vector_types_for_stmt): Likewise.
1717 * tree-vectorizer.h (vect_dr_behavior): Likewise.
1719 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1721 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
1722 (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
1723 (vect_permute_store_chain, vect_permute_load_chain)
1724 (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
1725 repeated stmt_vec_info lookups.
1726 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
1727 (vect_update_ivs_after_vectorizer): Likewise.
1728 * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
1729 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
1730 * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
1731 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
1732 (vect_bb_slp_scalar_cost): Likewise.
1733 * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
1735 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1737 * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
1738 gcall rather than the generic gimple stmt to gimple_call_internal_fn.
1739 (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
1740 to get gassigns and gcalls, rather than operating on generc gimple
1742 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
1743 (vect_mark_stmts_to_be_vectorized, vectorizable_store)
1744 (vectorizable_load, vect_analyze_stmt): Likewise.
1745 * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
1747 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1749 * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
1750 return stmt_vec_infos rather than gimple stmts. Do not accept
1752 (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
1754 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
1755 Update use of get_later_stmt.
1756 (vect_get_constant_vectors): Update call accordingly.
1757 (vect_schedule_slp_instance): Likewise
1758 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
1759 (vect_slp_analyze_instance_dependence): Likewise.
1760 (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
1762 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1764 * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
1765 (stmt_info_for_cost::stmt_info): ...this new field.
1766 (add_stmt_costs): Update accordingly.
1767 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
1768 (vect_get_known_peeling_cost): Likewise.
1769 (vect_estimate_min_profitable_iters): Likewise.
1770 * tree-vect-stmts.c (record_stmt_cost): Likewise.
1772 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1774 * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
1775 from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
1776 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
1778 * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
1780 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1782 * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
1783 a gimple stmt to a stmt_vec_info.
1784 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
1786 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1788 * tree-vectorizer.h (vec_info::grouped_stores): Change from
1789 an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
1790 (_loop_vec_info::reduction_chains): Likewise.
1791 * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
1793 * tree-vect-slp.c (vect_analyze_slp): Likewise.
1795 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1797 * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
1798 a gimple stmt to a stmt_vec_info.
1799 (_stmt_vec_info::next_element): Likewise.
1800 * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
1801 (vect_slp_analyze_and_verify_node_alignment)
1802 (vect_analyze_group_access_1, vect_analyze_group_access)
1803 (vect_small_gap_p, vect_prune_runtime_alias_test_list)
1804 (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
1805 (vect_supportable_dr_alignment): Update accordingly.
1806 * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
1807 (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
1808 (vect_is_simple_reduction, vectorizable_reduction): Likewise.
1809 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
1810 * tree-vect-slp.c (vect_build_slp_tree_1)
1811 (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
1812 (vect_split_slp_store_group, vect_analyze_slp_instance)
1813 (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
1814 * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
1815 (get_group_load_store_type, get_load_store_type)
1816 (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
1817 (vect_transform_stmt, vect_remove_stores): Likewise.
1819 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1821 * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
1823 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
1824 (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
1825 (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
1826 (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
1827 (vector_alignment_reachable_p, vect_get_data_access_cost)
1828 (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
1829 (vect_peeling_supportable, vect_enhance_data_refs_alignment)
1830 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
1831 (vect_analyze_group_access_1, vect_analyze_group_access)
1832 (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
1833 (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
1834 (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
1835 result of vect_dr_stmt and use the stmt_vec_info instead of
1836 the associated gimple stmt.
1837 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
1838 (vect_gen_prolog_loop_niters): Likewise.
1839 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
1841 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1843 * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
1844 to a vec<stmt_vec_info>.
1845 * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
1846 (vect_create_new_slp_node): Take a vec<gimple *> instead of a
1848 (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
1849 to a vec<stmt_vec_info>.
1850 (bst_traits::value_type, bst_traits::value_type): Likewise.
1851 (bst_traits::hash): Update accordingly.
1852 (vect_get_and_check_slp_defs): Change the stmts parameter from
1853 a vec<gimple *> to a vec<stmt_vec_info>.
1854 (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
1855 (vect_build_slp_tree): Likewise.
1856 (vect_build_slp_tree_2): Likewise. Update uses of
1857 SLP_TREE_SCALAR_STMTS.
1858 (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
1859 (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
1860 (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
1861 (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
1862 (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
1863 (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
1864 (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
1865 (vect_get_constant_vectors, vect_get_slp_defs)
1866 (vect_transform_slp_perm_load, vect_schedule_slp_instance)
1867 (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
1868 (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
1869 instead of gimple stmts.
1870 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
1871 the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
1872 (vect_slp_analyze_instance_dependence): Update uses of
1873 SLP_TREE_SCALAR_STMTS.
1874 (vect_slp_analyze_and_verify_node_alignment): Likewise.
1875 (vect_slp_analyze_and_verify_instance_alignment): Likewise.
1876 * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
1877 (get_initial_defs_for_reduction): Likewise.
1878 (vect_create_epilog_for_reduction): Likewise.
1879 (vectorize_fold_left_reduction): Likewise.
1880 * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
1881 (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
1882 (can_vectorize_live_stmts): Likewise.
1884 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1886 * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
1887 auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
1888 (vect_force_simple_reduction): Take and return stmt_vec_infos rather
1890 * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
1892 (gather_scalar_reductions): Update after above interface changes.
1893 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
1894 (vect_is_simple_reduction): Take and return stmt_vec_infos rather
1896 (vect_force_simple_reduction): Likewise.
1897 * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
1898 LOOP_VINFO_REDUCTIONS.
1899 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
1901 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1903 * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
1904 a gimple stmt to a stmt_vec_info.
1905 * tree-vect-loop.c (vect_active_double_reduction_p)
1906 (vect_force_simple_reduction, vectorizable_reduction): Update
1909 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1911 * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
1912 vec<gimple *> to a vec<stmt_vec_info>.
1913 * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
1914 the reduction_phis argument from a vec<gimple *> to a
1916 (vectorizable_reduction): Likewise the phis local variable that
1917 is passed to vect_create_epilog_for_reduction. Update for new type
1918 of SLP_TREE_VEC_STMTS.
1919 (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
1920 (vectorizable_live_operation): Likewise.
1921 * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
1922 (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
1924 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1926 * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
1927 a gimple stmt to a stmt_vec_info.
1928 (vectorizable_condition, vectorizable_live_operation)
1929 (vectorizable_reduction, vectorizable_induction): Pass back the
1930 vectorized statement as a stmt_vec_info.
1931 * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
1932 use of STMT_VINFO_VEC_STMT.
1933 * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
1934 accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
1935 as stmt_vec_infos rather than gimple stmts.
1936 (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
1938 (vectorizable_live_operation): Likewise.
1939 (vectorizable_reduction, vectorizable_induction): Likewise,
1940 updating use of STMT_VINFO_VEC_STMT.
1941 * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
1942 of STMT_VINFO_VEC_STMT.
1943 (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
1944 (vectorizable_simd_clone_call, vectorizable_conversion)
1945 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
1946 (vectorizable_store, vectorizable_load, vectorizable_condition)
1947 (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
1948 from a gimple stmt to a stmt_vec_info.
1949 (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT. Pass a
1950 pointer to a stmt_vec_info to the vectorizable_* routines.
1952 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1954 * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
1955 a gimple stmt to a stmt_vec_info.
1956 (is_pattern_stmt_p): Update accordingly.
1957 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
1958 (vect_record_grouped_load_vectors): Likewise.
1959 * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
1960 (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
1961 (vect_model_reduction_cost): Likewise.
1962 (vect_create_epilog_for_reduction): Likewise.
1963 (vectorizable_reduction, vectorizable_induction): Likewise.
1964 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
1965 Return the stmt_vec_info for the pattern statement.
1966 (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
1967 (vect_split_statement, vect_mark_pattern_stmts): Likewise.
1968 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
1969 (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
1970 * tree-vect-stmts.c (vect_mark_relevant): Likewise.
1971 (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
1972 (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
1973 (free_stmt_vec_info, vect_is_simple_use): Likewise.
1975 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1977 * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
1978 (vect_finish_stmt_generation): Likewise.
1979 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
1980 (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
1981 (vect_build_gather_load_calls): Use the return value of the above
1982 functions instead of a separate call to vinfo_for_stmt. Use narrow
1983 scopes for the input gimple stmt and wider scopes for the associated
1984 stmt_vec_info. Use vec_info::lookup_def when setting these
1985 stmt_vec_infos from an SSA_NAME definition.
1986 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
1987 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
1988 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
1989 (vectorizable_store, vectorizable_load, vectorizable_condition)
1990 (vectorizable_comparison): Likewise.
1991 * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
1992 (vectorizable_reduction): Likewise.
1994 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1996 * tree-vectorizer.h (vect_is_simple_use): Add an optional
1997 stmt_vec_info * parameter before the optional gimple **.
1998 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
1999 (process_use, vect_get_vec_def_for_operand_1): Update callers.
2000 (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
2001 * tree-vect-loop.c (vectorizable_reduction): Likewise.
2002 (vectorizable_live_operation): Likewise.
2003 * tree-vect-patterns.c (type_conversion_p): Likewise.
2004 (vect_look_through_possible_promotion): Likewise.
2005 (vect_recog_rotate_pattern): Likewise.
2006 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
2008 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2010 * tree-vectorizer.h (stmt_vec_info): Temporarily change from
2011 a typedef to a wrapper class.
2012 (NULL_STMT_VEC_INFO): New macro.
2013 (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
2014 (stmt_vec_info::operator*): New function.
2015 (stmt_vec_info::operator gimple *): Likewise.
2016 (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
2017 (add_stmt_costs): Likewise.
2018 * tree-vect-loop-manip.c (iv_phi_p): Likewise.
2019 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
2020 (vect_get_known_peeling_cost): Likewise.
2021 (vect_estimate_min_profitable_iters): Likewise.
2022 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
2023 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
2024 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
2025 (vectorizable_store, free_stmt_vec_infos): Likewise.
2026 (new_stmt_vec_info): Change return type of xcalloc to
2029 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2031 * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
2032 * tree-vectorizer.c (vec_info::lookup_single_use): New function.
2033 * tree-vect-loop.c (vectorizable_reduction): Use it instead of
2034 a single_imm_use-based sequence.
2035 * tree-vect-stmts.c (supportable_widening_operation): Likewise.
2037 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2039 * tree-vectorizer.h (vec_info::lookup_def): Declare.
2040 * tree-vectorizer.c (vec_info::lookup_def): New function.
2041 * tree-vect-patterns.c (vect_get_internal_def): Use it.
2042 (vect_widened_op_tree): Likewise.
2043 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
2044 * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
2045 (vectorizable_reduction): Likewise.
2046 (vect_valid_reduction_input_p): Take a stmt_vec_info instead
2048 (vect_is_slp_reduction): Update calls accordingly. Use
2049 vec_info::lookup_def.
2050 (vect_is_simple_reduction): Likewise
2051 * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
2053 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2055 * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
2056 * tree-vectorizer.c (vec_info::lookup_stmt): New function.
2057 * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
2059 (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
2060 (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
2061 (vect_update_vf_for_slp, vect_analyze_loop_operations)
2062 (vect_is_slp_reduction, vectorizable_induction)
2063 (vect_transform_loop_stmt, vect_transform_loop): Likewise.
2064 * tree-vect-patterns.c (vect_init_pattern_stmt):
2065 (vect_determine_min_output_precision_1, vect_determine_precisions)
2066 (vect_pattern_recog): Likewise.
2067 * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
2068 * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
2069 * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
2070 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
2071 (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
2072 (vect_detect_hybrid_slp): Likewise. Change the walk_stmt_info
2073 info field from a loop to a loop_vec_info.
2075 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2077 * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
2078 (vec_info::add_stmt): Declare.
2079 * tree-vectorizer.c (vec_info::add_stmt): New function.
2080 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
2081 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
2082 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
2083 (vectorizable_induction): Likewise.
2084 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
2085 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
2086 (vectorizable_simd_clone_call, vectorizable_store): Likewise.
2087 (vectorizable_load): Likewise.
2088 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
2089 (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
2090 (vect_recog_gather_scatter_pattern): Likewise.
2091 (append_pattern_def_seq): Likewise. Remove a check that is
2092 performed by add_stmt itself.
2094 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2096 * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
2097 which make_ssa_name was called with new_stmt before new_stmt
2100 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2102 * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
2104 (vect_is_slp_reduction): ...here...
2105 (vect_is_simple_reduction): ...and here. Remove repetition of tests
2106 that are already known to be false.
2108 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2110 * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
2111 * tree-vect-slp.c (vect_free_slp_tree): Likewise. Don't update
2112 STMT_VINFO_NUM_SLP_USES when it's true.
2113 (vect_free_slp_instance): Add a final_p parameter and pass it to
2115 (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
2116 (vect_analyze_slp_instance): Likewise.
2117 (vect_slp_analyze_operations): Likewise.
2118 (vect_slp_analyze_bb_1): Likewise.
2119 * tree-vectorizer.c (vec_info): Likewise.
2120 * tree-vect-loop.c (vect_transform_loop): Likewise.
2122 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2124 * tree-vect-loop.c (vectorizable_reduction): Assert that the
2125 function is not called for second and subsequent members of
2128 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2130 * tree-vect-loop.c (get_initial_def_for_reduction): Move special
2131 cases for nested loops from here to ...
2132 (vect_create_epilog_for_reduction): ...here. Only call
2133 vect_is_simple_use for inner-loop reductions.
2135 2018-07-31 Martin Liska <mliska@suse.cz>
2137 PR gcov-profile/85338
2138 PR gcov-profile/85350
2139 PR gcov-profile/85372
2140 * profile.c (struct location_triplet): New.
2141 (struct location_triplet_hash): Likewise.
2142 (output_location): Do not output a BB that
2143 is already recorded for a line.
2144 (branch_prob): Use streamed_locations.
2146 2018-07-31 Martin Liska <mliska@suse.cz>
2148 PR gcov-profile/85370
2149 * coverage.c (coverage_begin_function): Do not mark target
2150 clones as artificial functions.
2152 2018-07-31 Martin Liska <mliska@suse.cz>
2154 PR gcov-profile/83813
2155 PR gcov-profile/84758
2156 PR gcov-profile/85217
2157 PR gcov-profile/85332
2158 * profile.c (branch_prob): Do not record GOTO expressions
2159 for GIMPLE statements which locations are already streamed.
2161 2018-07-31 Olivier Hainque <hainque@adacore.com>
2163 * gcc.c (handle_spec_function): Accept a soft_matched_part
2164 argument, as do_spec_1. Pass it down to ...
2165 (eval_spec_function): Accept a soft_matched_part argument,
2166 and pass it down to ...
2167 (do_spec_2): Accept a soft_matched_part argument, and pass
2168 it down to do_spec_1.
2169 (do_spec_1): Pass soft_matched_part to handle_spec_function.
2170 (handle_braces): Update call to handle_spec_function.
2171 (driver::set_up_specs): Update calls to do_spec_2.
2172 (compare_debug_dump_opt_spec_function): Likewise.
2173 (compare_debug_self_opt_spec_function): Likewise.
2175 2018-07-31 Olivier Hainque <hainque@adacore.com>
2177 * common.opt (nolibc): New option.
2178 * doc/invoke.texi (Link Options): Document it.
2179 * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
2180 * config/alpha/linux.h: Likewise.
2181 * config/arc/elf.h: Likewise.
2182 * config/arm/uclinux-elf.h: Likewise.
2183 * config/arm/unknown-elf.h: Likewise.
2184 * config/avr/avrlibc.h: Likewise.
2185 * config/bfin/bfin.h: Likewise.
2186 * config/bfin/linux.h: Likewise.
2187 * config/bfin/uclinux.h: Likewise.
2188 * config/darwin.h: Likewise.
2189 * config/darwin10.h: Likewise.
2190 * config/darwin12.h: Likewise.
2191 * config/gnu-user.h: Likewise.
2192 * config/lm32/uclinux-elf.h: Likewise.
2193 * config/pa/pa-hpux11.h: Likewise.
2194 * config/pa/pa64-hpux.h: Likewise.
2195 * config/sparc/sparc.h: Likewise.
2197 2018-07-31 Olivier Hainque <hainque@adacore.com>
2199 * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
2200 undefined variables.
2202 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
2205 * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
2208 2018-07-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
2210 * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
2211 terminated string literal.
2213 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
2215 PR rtl-optimization/85160
2216 * combine.c (is_just_move): New function.
2217 (try_combine): Allow combining two instructions into two if neither of
2218 the original instructions was a move.
2220 2018-07-30 Alexander Monakov <amonakov@ispras.ru>
2223 * doc/extend.texi (Global Register Variables): Discourage use of type
2225 (Local Register Variables): Likewise.
2227 2018-07-30 Richard Sandiford <richard.sandiford@arm.com>
2229 PR tree-optimization/86506
2230 * hwint.h (ceil_log2): Resync with hwint.c implementation.
2232 2018-07-30 Ilya Leoshkevich <iii@linux.ibm.com>
2235 * lra-constraints.c (spill_hard_reg_in_range): When selecting the
2236 hard_regno, make sure no insn between `from` and `to` clobbers it.
2238 2018-07-30 Cesar Philippidis <cesar@codesourcery.com>
2239 Tom de Vries <tdevries@suse.de>
2241 * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
2242 (PTX_DEFAULT_RUNTIME_DIM): ... this.
2243 (nvptx_goacc_validate_dims): Set default worker and gang dims to
2244 PTX_DEFAULT_RUNTIME_DIM.
2245 (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
2247 2018-07-29 John David Anglin <danglin@gcc.gnu.org>
2249 * config/pa/pa.c (pa_output_addr_vec): Align address table.
2250 * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
2251 * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
2253 2018-07-27 Michael Meissner <meissner@linux.ibm.com>
2255 * config/rs6000/constraints.md (wG constraint): Delete, no longer
2257 * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
2258 predicate to reflect toc fusion has been deleted.
2259 (toc_fusion_mem_raw): Delete, no longer used.
2260 (toc_fusion_mem_wrapped): Likewise.
2261 * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
2263 * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
2264 Delete, no longer used.
2265 * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
2266 meant to be used for toc fusion.
2267 (rs6000_debug_print_mode): Delete toc fusion debugging.
2268 (rs6000_debug_reg_global): Likewise.
2269 (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
2270 fusion and secondary reload support that were never used.
2271 (rs6000_option_override_internal): Delete TOC fusion, that was only
2272 partially defined, and it did not work unless you also used the
2274 (rs6000_legitimate_address_p): Delete TOC fusion support.
2275 (rs6000_opt_masks): Likewise.
2276 (fusion_wrap_memory_address): Delete function, no longer used.
2277 (fusion_split_address); Delete TOC fusion support.
2278 * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
2279 longer used with toc fusion being deleted.
2280 (TARGET_TOC_FUSION_FP): Likewise.
2281 * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
2283 (toc fusion spliter): Delete TOC fusion support.
2284 (toc_fusionload_<mode>): Likewise.
2285 (toc_fusionload_di): Likewise.
2286 (fusion_gpr_load_<mode>): Delete generator function, this insn no
2287 longer needs to be named. Rename predicate to delete TOC fusion.
2288 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
2289 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
2290 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
2291 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
2292 (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
2294 2018-07-27 Kelvin Nilsen <kelvin@gcc.gnu.org>
2296 * doc/extend.texi (Basic PowerPC Built-in Functions Available on
2297 ISA 2.05): Replace __uint128_t with __uint128 and __int128_t with
2298 __int128 in built-in function prototypes.
2299 (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
2300 (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
2302 2018-07-27 Martin Sebor <msebor@redhat.com>
2304 PR tree-optimization/86696
2305 * tree-ssa-strlen.c (get_min_string_length): Handle all integer
2306 types, including enums.
2307 (handle_char_store): Be prepared for the above function to fail.
2309 2018-07-26 Qing Zhao <qing.zhao@oracle.com>
2311 * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
2312 when optimization level is lower than 2 or optimize for size.
2314 2018-07-26 Martin Sebor <msebor@redhat.com>
2316 PR tree-optimization/86043
2317 PR tree-optimization/86042
2318 * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
2319 (get_string_cst_length): Rename...
2320 (get_min_string_length): ...to this. Add argument.
2321 (handle_char_store): Extend to handle multi-character stores by
2323 * tree.c (initializer_zerop): Use new argument. Handle MEM_REF.
2324 * tree.h (initializer_zerop): Add argument.
2326 2018-07-26 Jakub Jelinek <jakub@redhat.com>
2329 * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
2330 declare target to variables if they have always,{to,from,tofrom} map
2333 2018-07-26 Martin Liska <mliska@suse.cz>
2336 * lto-wrapper.c: Add linker_output as prefix
2337 for ltrans_output_file.
2339 2018-07-26 Segher Boessenkool <segher@kernel.crashing.org>
2341 PR rtl-optimization/85805
2342 * combine.c (reg_nonzero_bits_for_combine): Only use the last set
2343 value for hard registers if that was written in the same mode.
2345 2018-07-26 Martin Liska <mliska@suse.cz>
2347 PR gcov-profile/86536
2348 * gcov.c (format_gcov): Use printf format %.*f directly
2349 and do not handle special values.
2351 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
2353 * common/config/arc/arc-common.c (arc_option_optimization_table):
2354 Update default optimizations for size.
2356 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
2358 * config/arc/arc.md (movsf_insn): Add short instruction selection.
2359 * config/arc/constraints.md (CfZ): New constraint.
2360 * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
2361 (subsf3_fpu): Likewise.
2362 (cmpsf_fpu): Likewise.
2363 (cmpsf_fpu_uneq): Likewise.
2365 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
2367 * config/arc/arc.c (compact_memory_operand_p): Check for uncached
2369 (arc_is_uncached_mem_p): uncached applies to both the variable and
2372 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
2374 * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
2377 2018-07-25 David Malcolm <dmalcolm@redhat.com>
2379 * optinfo-emit-json.cc (class optrecord_json_writer): Convert
2380 field "m_scopes" from vec to auto_vec.
2382 2018-07-25 Martin Liska <mliska@suse.cz>
2384 * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
2387 2018-07-25 Richard Biener <rguenther@suse.de>
2390 * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
2391 special wrt context_die late.
2392 (gen_subprogram_die): Re-use DIEs in local scope.
2394 2018-07-25 Richard Sandiford <richard.sandiford@arm.com>
2396 PR tree-optimization/86644
2397 * hwint.c (ceil_log2): Fix comment. Return 0 for 0.
2399 2018-07-25 Martin Liska <mliska@suse.cz>
2402 * dumpfile.c: And excluded values with TDF_ALL_VALUES.
2403 * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
2405 2018-07-25 Martin Liska <mliska@suse.cz>
2408 * params.def: Explain ASan abbreviation and provide
2409 a documentation link.
2411 2018-07-24 Martin Sebor <msebor@redhat.com>
2413 PR tree-optimization/86622
2414 PR tree-optimization/86532
2415 * builtins.h (string_length): Declare.
2416 * builtins.c (c_strlen): Correct handling of non-constant offsets.
2417 (check_access): Be prepared for non-constant length ranges.
2418 (string_length): Make extern.
2419 * expr.c (string_constant): Only handle the minor non-constant
2420 array index. Use string_constant to compute the length of
2421 a generic string constant.
2423 2018-07-24 Richard Sandiford <richard.sandiford@arm.com>
2425 PR tree-optimization/86618
2426 * tree-vect-stmts.c (vectorizable_call): Don't take the address
2427 of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
2429 2018-07-24 David Malcolm <dmalcolm@redhat.com>
2431 PR tree-optimization/86636
2432 * json.cc (json::object::set): Fix comment. Add assertions.
2433 (json::array::append): Move here from json.h. Add comment and an
2435 (json::string::string): Likewise.
2436 * json.h (json::array::append): Move to json.cc.
2437 (json::string::string): Likewise.
2438 * optinfo-emit-json.cc
2439 (optrecord_json_writer::impl_location_to_json): Assert that we
2440 aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
2441 wrapper around it. Expand the location once, rather than three
2443 (optrecord_json_writer::inlining_chain_to_json): Fix the check for
2444 UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
2446 (optrecord_json_writer::optinfo_to_json): Likewise, in four
2447 places. Fix some overlong lines.
2449 2018-07-24 Matthew Malcomson <matthew.malcomson@arm.com>
2451 * config/aarch64/aarch64-simd.md
2452 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
2453 (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
2454 (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
2455 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
2456 (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
2457 (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
2458 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
2459 (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
2460 (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
2462 2018-07-24 Jakub Jelinek <jakub@redhat.com>
2465 * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
2466 and size > HOST_BITS_PER_WIDE_INT. For size > HOST_BITS_PER_WIDE_INT
2467 and abs_d == d, do the power of two handling if profitable.
2469 2018-07-24 Richard Biener <rguenther@suse.de>
2471 * match.pd: Add BIT_FIELD_REF canonicalizations.
2473 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
2476 * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
2478 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
2480 * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
2481 terminated STRING_CST object.
2483 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
2485 hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
2487 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
2489 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
2490 * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
2491 * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
2492 the elements into a register.
2493 (rs6000_split_v4si_init_di_reg): Delete.
2494 (rs6000_split_v4si_init): Delete.
2495 * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
2496 (vsx_init_v4si): Rewrite as a define_expand.
2498 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
2500 * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
2501 zero_extend argument from memory): New.
2503 2018-07-22 Martin Sebor <msebor@redhat.com>
2506 * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
2507 diagnosing calls with unknown arguments unless -Walloca-larger-than
2508 is restricted to less than PTRDIFF_MAX bytes.
2510 2018-07-22 Gerald Pfeifer <gerald@pfeifer.com>
2512 * doc/gcov.texi (Invoking Gcov): Editorial changes.
2514 2018-07-20 David Malcolm <dmalcolm@redhat.com>
2516 * pretty-print.c (text_info::set_location): Remove redundant
2517 "line_table" parameter from call to rich_location::set_range.
2519 2018-07-20 Martin Sebor <msebor@redhat.com>
2522 * builtins.c (expand_builtin_alloca): Adjust.
2523 * calls.c (alloc_max_size): Simplify.
2524 * cgraphunit.c (cgraph_node::expand): Adjust.
2525 * common.opt (larger_than_size, warn_frame_larger_than): Remove
2527 (frame_larger_than_size): Same.
2528 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
2529 to take a HOST_WIDE_INT argument and accept a byte-size suffix.
2531 * doc/invoke.texi (GCC Command Options): Document option arguments.
2532 Explain byte-size arguments and suffixes.
2533 (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
2534 (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
2535 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
2536 * doc/options.texi (UInteger): Expand.
2537 (Host_Wide_Int, ByteSize): Document new properties.
2538 * final.c (final_start_function_1): Include sizes in an error message.
2539 * function.c (frame_offset_overflow): Same.
2540 * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
2541 (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
2542 Diagnose unbounded alloca calls only for limits of less than
2544 (alloca_call_type): Adjust. Diagnose possibly out-of-bounds alloca
2545 calls and VLA size only for limits of less than PTRDIFF_MAX. Same
2547 (pass_walloca::execute): Adjust. Diagnose alloca calls in loops
2548 only for limits of less than PTRDIFF_MAX.
2549 * langhooks-def.h (lhd_handle_option): Change function argument
2551 * langhooks.c (lhd_handle_option): Same.
2552 * langhooks.h (handle_option): Same.
2553 * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
2555 (var_type, var_type_struct): Same.
2556 (var_set): Handle ByteSize flag.
2557 * optc-gen.awk: Add comments to output to ease debugging. Make
2558 use of HOST_WIDE_INT where appropriate.
2559 * opts-gen-save.awk: Use %lx to format unsigned long.
2560 * opth-gen.awk: Change function argument to HOST_WIDE_INT.
2561 * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
2562 arguments. Parse bytes-size suffixes.
2563 (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
2564 (enum_value_to_arg): Same.
2565 (decode_cmdline_option): Handle cl_host_wide_int. Adjust.
2566 (handle_option): Adjust.
2567 (generate_option): Change function argument to HOST_WIDE_INT.
2568 (cmdline_handle_error): Adjust.
2569 (read_cmdline_option): Change function argument to HOST_WIDE_INT.
2570 (set_option): Change function argument to HOST_WIDE_INT.
2571 (option_enabled): Handle cl_host_wide_int.
2572 (get_option_state): Handle CLVC_SIZE.
2573 (control_warning_option): Same.
2574 * opts.c (common_handle_option): Change function argument to
2575 HOST_WIDE_INT. Remove handling of OPT_Walloca_larger_than_ and
2576 OPT_Wvla_larger_than_.
2577 * opts.h (enum cl_var_type): Add an enumerator.
2578 * stor-layout.c (layout_decl): Print a more meaningful warning.
2579 * toplev.c (output_stack_usage): Adjust.
2581 2018-07-20 Qing Zhao <qing.zhao@oracle.com>
2583 * builtins.c (expand_builtin_memcmp): Delete the last parameter for
2584 call to inline_expand_builtin_string_cmp.
2585 (expand_builtin_strcmp): Likewise.
2586 (expand_builtin_strncmp): Likewise.
2587 (inline_string_cmp): Delete the last parameter, change char_type_node
2588 to unsigned_char_type_node for strcmp/strncmp, add conversions to the
2590 (inline_expand_builtin_string_cmp): Delete the last parameter, give up
2591 the inlining expansion on target where the type of the call has same or
2592 narrower precision than unsigned char.
2594 2018-07-20 David Malcolm <dmalcolm@redhat.com>
2596 * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
2597 (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
2598 * common.opt (fsave-optimization-record): New option.
2599 * coretypes.h (struct kv_pair): Move here from dumpfile.c.
2600 * doc/invoke.texi (-fsave-optimization-record): New option.
2601 * dumpfile.c: Include "optinfo-emit-json.h".
2602 (struct kv_pair): Move to coretypes.h.
2603 (optgroup_options): Make non-static.
2604 (dump_context::end_scope): Call
2605 optimization_records_maybe_pop_dump_scope.
2606 * dumpfile.h (optgroup_options): New decl.
2607 * json.cc: New file.
2609 * optinfo-emit-json.cc: New file.
2610 * optinfo-emit-json.h: New file.
2611 * optinfo.cc: Include "optinfo-emit-json.h".
2612 (optinfo::emit): Call optimization_records_maybe_record_optinfo.
2613 (optinfo_enabled_p): Check optimization_records_enabled_p.
2614 (optinfo_wants_inlining_info_p): Likewise.
2615 * optinfo.h: Update comment.
2616 * profile-count.c (profile_quality_as_string): New function.
2617 * profile-count.h (profile_quality_as_string): New decl.
2618 (profile_count::quality): New accessor.
2619 * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
2620 and optinfo_emit_json_cc_tests.
2621 * selftest.h (selftest::json_cc_tests): New decl.
2622 (selftest::optinfo_emit_json_cc_tests): New decl.
2623 * toplev.c: Include "optinfo-emit-json.h".
2624 (compile_file): Call optimization_records_finish.
2625 (do_compile): Call optimization_records_start.
2626 * tree-ssa-live.c: Include optinfo.h.
2627 (remove_unused_scope_block_p): Retain inlining information if
2628 optinfo_wants_inlining_info_p returns true.
2630 2018-07-20 Richard Biener <rguenther@suse.de>
2633 * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
2634 to cover -flto-partition=none.
2636 2018-07-20 Martin Liska <mliska@suse.cz>
2638 * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
2639 (get_decl_source_range): Remove unused function.
2641 2018-07-20 Richard Biener <rguenther@suse.de>
2643 * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
2644 (struct vn_phi_s): Likewise.
2645 (struct vn_reference_s): Likewise.
2646 * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
2647 for searching the slot of an entry known to be in the hash itself.
2648 (vn_phi_hasher::equal): Likewise.
2649 (vn_reference_hasher::equal): Likewise.
2650 (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
2652 (optimistic_info, current_info): Remove, keeping only valid_info.
2653 (vn_reference_lookup_1): Remove fallback lookup.
2654 (vn_reference_lookup_2): Likewise.
2655 (vn_nary_op_lookup_1): Likewise.
2656 (vn_phi_lookup): Likewise.
2657 (vn_nary_build_or_lookup_1): Make sure to not chain the built
2659 (vn_reference_insert): Adjust, chain the inserted hash element
2660 at last_inserted_ref.
2661 (vn_reference_insert_pieces): Likewise.
2662 (visit_reference_op_call): Likewise.
2663 (vn_nary_op_insert_into): Chain the inserted hash element at
2665 (vn_nary_op_insert_pieces): Adjust.
2666 (vn_nary_op_insert): Likewise.
2667 (vn_nary_op_insert_stmt): Likewise.
2668 (vn_phi_insert): Adjust, chain the inserted hash element at
2670 (process_scc): Remove clearing and copying the optimistic
2671 table. Instead remove elements inserted during an optimistic
2672 iteration from the single table we maintain.
2673 (init_scc_vn): Adjust.
2674 (free_scc_vn): Likewise.
2675 (sccvn_dom_walker::record_cond): Likewise.
2676 (sccvn_dom_walker::after_dom_children): Likewise.
2678 2018-07-19 Martin Sebor <msebor@redhat.com>
2680 PR tree-optimization/84047
2681 PR tree-optimization/83776
2682 * tree-vrp.c (vrp_prop::check_mem_ref): New function.
2683 (check_array_bounds): Call it.
2685 2018-07-19 Martin Sebor <msebor@redhat.com>
2687 * align.h (align_flags): Use member initialization.
2689 2018-07-19 David Malcolm <dmalcolm@redhat.com>
2691 * Makefile.in (OBJS): Add optinfo.o.
2692 * coretypes.h (class symtab_node): New forward decl.
2693 (struct cgraph_node): New forward decl.
2694 (class varpool_node): New forward decl.
2695 * dump-context.h: New file.
2696 * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
2698 (refresh_dumps_are_enabled): Use optinfo_enabled_p.
2699 (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
2700 (set_alt_dump_file): Likewise.
2701 (dump_context::~dump_context): New dtor.
2702 (dump_gimple_stmt): Move implementation to...
2703 (dump_context::dump_gimple_stmt): ...this new member function.
2704 Add the stmt to any pending optinfo, creating one if need be.
2705 (dump_gimple_stmt_loc): Move implementation to...
2706 (dump_context::dump_gimple_stmt_loc): ...this new member function.
2707 Start a new optinfo and add the stmt to it.
2708 (dump_gimple_expr): Move implementation to...
2709 (dump_context::dump_gimple_expr): ...this new member function.
2710 Add the stmt to any pending optinfo, creating one if need be.
2711 (dump_gimple_expr_loc): Move implementation to...
2712 (dump_context::dump_gimple_expr_loc): ...this new member function.
2713 Start a new optinfo and add the stmt to it.
2714 (dump_generic_expr): Move implementation to...
2715 (dump_context::dump_generic_expr): ...this new member function.
2716 Add the tree to any pending optinfo, creating one if need be.
2717 (dump_generic_expr_loc): Move implementation to...
2718 (dump_context::dump_generic_expr_loc): ...this new member
2719 function. Add the tree to any pending optinfo, creating one if
2721 (dump_printf): Move implementation to...
2722 (dump_context::dump_printf_va): ...this new member function. Add
2723 the text to any pending optinfo, creating one if need be.
2724 (dump_printf_loc): Move implementation to...
2725 (dump_context::dump_printf_loc_va): ...this new member function.
2726 Start a new optinfo and add the stmt to it.
2727 (dump_dec): Move implementation to...
2728 (dump_context::dump_dec): ...this new member function. Add the
2729 value to any pending optinfo, creating one if need be.
2730 (dump_context::dump_symtab_node): New member function.
2731 (dump_context::get_scope_depth): New member function.
2732 (dump_context::begin_scope): New member function.
2733 (dump_context::end_scope): New member function.
2734 (dump_context::ensure_pending_optinfo): New member function.
2735 (dump_context::begin_next_optinfo): New member function.
2736 (dump_context::end_any_optinfo): New member function.
2737 (dump_context::s_current): New global.
2738 (dump_context::s_default): New global.
2739 (dump_scope_depth): Delete global.
2740 (dumpfile_ensure_any_optinfo_are_flushed): New function.
2741 (dump_symtab_node): New function.
2742 (get_dump_scope_depth): Reimplement in terms of dump_context.
2743 (dump_begin_scope): Likewise.
2744 (dump_end_scope): Likewise.
2745 (selftest::temp_dump_context::temp_dump_context): New ctor.
2746 (selftest::temp_dump_context::~temp_dump_context): New dtor.
2747 (selftest::verify_item): New function.
2748 (ASSERT_IS_TEXT): New macro.
2749 (ASSERT_IS_TREE): New macro.
2750 (ASSERT_IS_GIMPLE): New macro.
2751 (selftest::test_capture_of_dump_calls): New test.
2752 (selftest::dumpfile_c_tests): Call it.
2753 * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
2754 (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
2755 (dump_gimple_stmt, dump_dec): Gather these related decls and add a
2756 descriptive comment.
2757 (dump_function, print_combine_total_stats, enable_rtl_dump_file)
2758 (dump_node, dump_bb): Move these unrelated decls.
2759 (class dump_manager): Add leading comment.
2760 * optinfo.cc: New file.
2761 * optinfo.h: New file.
2763 2018-07-19 Michael Collison <michael.collison@arm.com>
2764 Richard Henderson <rth@redhat.com>
2766 * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
2767 (subti): Handle op1 zero.
2768 (subvti4, usub4ti4): New.
2769 (*sub<GPI>3_compare1_imm): New.
2770 (sub<GPI>3_carryinCV): New.
2771 (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
2772 (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
2774 2018-07-19 Michael Collison <michael.collison@arm.com>
2775 Richard Henderson <rth@redhat.com>
2777 * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
2778 (addti3): Create simpler code if low part is already known to be 0.
2779 (addvti4, uaddvti4): New.
2780 (*add<GPI>3_compareC_cconly_imm): New.
2781 (*add<GPI>3_compareC_cconly): New.
2782 (*add<GPI>3_compareC_imm): New.
2783 (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
2784 handle constants within this pattern..
2785 (*add<GPI>3_compareV_cconly_imm): New.
2786 (*add<GPI>3_compareV_cconly): New.
2787 (*add<GPI>3_compareV_imm): New.
2788 (add<GPI>3_compareV): New.
2789 (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
2790 (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
2791 (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
2792 ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
2793 with 'comparison' operator.
2794 (*add<GPI>3_compareV_cconly_imm): Ditto.
2795 (*add<GPI>3_compareV_cconly): Ditto.
2796 (*add<GPI>3_compareV_imm): Ditto.
2797 (add<GPI>3_compareV): Ditto.
2798 (add<mode>3_carryinC): Ditto.
2799 (*add<mode>3_carryinC_zero): Ditto.
2800 (*add<mode>3_carryinC): Ditto.
2801 (add<mode>3_carryinV): Ditto.
2802 (*add<mode>3_carryinV_zero): Ditto.
2803 (*add<mode>3_carryinV): Ditto.
2805 2018-07-19 Michael Collison <michael.collison@arm.com>
2806 Richard Henderson <rth@redhat.com>
2808 * config/aarch64/aarch64-modes.def (CC_V): New.
2809 * config/aarch64/aarch64-protos.h
2810 (aarch64_addti_scratch_regs): Declare
2811 (aarch64_subvti_scratch_regs): Declare.
2812 (aarch64_expand_subvti): Declare.
2813 (aarch64_gen_unlikely_cbranch): Declare
2814 * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
2815 for signed overflow using CC_Vmode.
2816 (aarch64_get_condition_code_1): Handle CC_Vmode.
2817 (aarch64_gen_unlikely_cbranch): New function.
2818 (aarch64_addti_scratch_regs): New function.
2819 (aarch64_subvti_scratch_regs): New function.
2820 (aarch64_expand_subvti): New function.
2822 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
2824 * config/aarch64/aarch64-option-extensions.def: New entry for profile
2826 * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
2827 * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
2830 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
2833 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
2834 address check not strict.
2836 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
2838 * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
2840 (store_pair_lanes<mode>): Likewise.
2841 * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
2842 enum value 'ADDR_QUERY_LDP_STP_N'.
2843 * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
2844 (aarch64_print_address_internal): Add declaration.
2845 (aarch64_print_ldpstp_address): Remove.
2846 (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
2847 (aarch64_print_operand): Change printing of 'y'.
2848 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
2849 new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
2850 'true' rather than '1'.
2851 * config/aarch64/constraints.md (Uml): Likewise.
2852 (Uml): Rename to Umn.
2855 2018-07-19 Richard Biener <rguenther@suse.de>
2857 * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
2859 * tree-ssa-sccvn.c: Remove alloc-pool.h use.
2860 (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
2861 (vn_reference_hasher): Likewise.
2862 (struct vn_tables_s): Remove obstack and alloc-pool members.
2863 (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
2864 (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
2865 (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
2866 (vn_reference_insert_pieces): Likewise.
2867 (alloc_vn_nary_op_noinit): Adjust.
2868 (vn_nary_op_insert_stmt): Allocate phiargs in-place.
2869 (vn_phi_eq): Adjust.
2870 (shared_lookup_phiargs): Remove.
2871 (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
2872 (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
2873 (visit_reference_op_call): Likewise.
2874 (copy_nary, copy_phi, copy_reference): Remove.
2875 (process_scc): Rewind the obstack when iterating. Do not
2876 copy the elements to valid_info but just move them from one
2877 hashtable to the other.
2878 (allocate_vn_table): Adjust.
2879 (free_vn_table): Likewise.
2880 (init_scc_vn): Likewise.
2881 (free_scc_vn): Likewise.
2883 2018-07-19 H.J. Lu <hongjiu.lu@intel.com>
2886 * config/i386/i386.c (rest_of_insert_endbranch): Lookup
2887 indirect_return as function type attribute.
2888 (ix86_attribute_table): Change indirect_return to function
2890 * doc/extend.texi: Update indirect_return attribute.
2892 2018-07-19 Aldy Hernandez <aldyh@redhat.com>
2894 * wide-int.h (widest2_int): New.
2895 * gimple-fold.c (arith_overflowed_p): Use it.
2896 * tree.h (widest2_int_cst): New.
2897 * tree-vrp.c (wide_int_binop_overflow): Rename from
2898 vrp_int_const_binop.
2899 Rewrite to work on trees.
2900 (extract_range_from_multiplicative_op_1): Abstract code to...
2901 (wide_int_range_min_max): ...here.
2902 (wide_int_range_cross_product): ...and here.
2903 (extract_range_from_binary_expr_1): Abstract overflow code to...
2904 (wide_int_range_mult_wrapping): ...here.
2905 * tree-vrp.h (wide_int_range_cross_product): New.
2906 (wide_int_range_mult_wrapping): New.
2908 2018-07-19 Andrew Senkevich <andrew.senkevich@intel.com>
2909 Julia Koval <julia.koval@intel.com>
2911 * config/i386/x86-tune-costs.h (skylake_memcpy,
2912 skylake_memset): Replace rep_prefix with unrolling for size 512.
2914 2018-07-18 Kugan Vivekanandarajah <kuganv@linaro.org>
2917 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle comparision with EQ_EXPR
2920 2018-07-18 Kelvin Nilsen <kelvin@gcc.gnu.org>
2922 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
2923 this subsection to "PowerPC AltiVec/VSX Built-in Functions".
2924 (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
2925 previously known as "PowerPC AltiVec Built-in Functions". Move
2926 some material to new subsubsections "PowerPC AltiVec Built-in
2927 Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
2929 (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
2930 (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
2931 (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
2932 (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
2934 2018-07-18 Richard Biener <rguenther@suse.de>
2936 PR tree-optimization/86557
2937 * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
2940 2018-07-18 Ilya Leoshkevich <iii@linux.ibm.com>
2942 * config/s390/s390.c (s390_function_profiler): Generate CFI.
2944 2018-07-17 Jeff Law <law@redhat.com>
2946 * config/arm/arm.c (get_label_padding): Update for recent
2947 changes to label_to_alignment.
2949 PR tree-optimization/86010
2950 * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
2952 * config/mips/mips.c (vr4130_align_insns): Update for recent
2953 changes to label_to_alignment.
2955 * config/frv/frv.c (frv_label_align): Update for recent changes
2956 to label_to_alignment.
2958 * config/nios2/nios2.c (nios2_label_align): Update for recent
2959 changes which dropped ALIGN_LABELS_LOG.
2961 2018-07-17 Andreas Schwab <schwab@linux-m68k.org>
2963 * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
2964 (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
2966 2018-07-17 Claudiu Zissulescu <claziss@synopsys.com>
2968 * config/arc/arc.c (arc_label_align): Use align_labels instead of
2969 deprecated align_labels_log.
2971 2018-07-17 Richard Biener <rguenther@suse.de>
2974 * dwarf2out.c (init_sections_and_labels): Always generate
2975 a debug_line_str_section for early LTO debug.
2976 (dwarf2out_finish): Reset debug_line_str_hash output early.
2977 Bump counter for extra dwarf5 .debug_loc labels to not conflict
2979 (dwarf2out_early_finish): Output debug_line_str.
2981 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
2983 * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
2984 index register on z196 or later.
2986 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
2988 * config/s390/s390.c (s390_default_align): Set default function
2990 (s390_override_options_after_change): Call s390_default align.
2991 (s390_option_override_internal): Call s390_default_align.
2992 (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
2994 2018-07-17 Jakub Jelinek <jakub@redhat.com>
2997 * omp-low.c (create_task_copyfn): Copy over also fields corresponding
2998 to _looptemp_ clauses, other than the first two.
3000 2018-07-17 Martin Liska <mliska@suse.cz>
3002 * opts.c: Do not enable OPT_falign_* for -Os.
3004 2018-07-17 Martin Liska <mliska@suse.cz>
3006 * align.h (MAX_CODE_ALIGN): New.
3007 (MAX_CODE_ALIGN_VALUE): New.
3008 * common/config/i386/i386-common.c (ix86_handle_option):
3009 (MAX_CODE_ALIGN): Moved to align.h.
3010 * final.c (MAX_CODE_ALIGN): Likewise.
3011 * opts.c (parse_and_check_align_values):
3012 (MAX_CODE_ALIGN): Likewise.
3013 (MAX_CODE_ALIGN_VALUE): Likewise.
3015 2018-07-17 Martin Liska <mliska@suse.cz>
3017 * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
3018 in order to fulfil coding style.
3019 * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
3020 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
3021 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
3022 * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
3023 * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
3024 * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
3025 * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
3026 * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
3027 (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
3029 2018-07-17 Martin Liska <mliska@suse.cz>
3031 * align.h: New file.
3032 * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions directly.
3033 * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
3034 align_flags of label_to_alignment.
3035 * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into align_flags
3037 * config/m68k/m68k.c: Do not use removed align_labels_value and
3039 * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
3040 (LOOP_ALIGN): Likewise.
3041 (LABEL_ALIGN): Likewise.
3042 * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
3043 Remove not used macro.
3044 (rs6000_loop_align): Change return type to align_flags.
3045 (rs6000_loop_align_max_skip): Remove.
3046 * config/rs6000/rs6000-protos.h (rs6000_loop_align):
3047 Change return type to align_flags.
3048 * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
3049 Remove not used macro.
3050 (rs6000_loop_align): Change return type to align_flags.
3051 (rs6000_loop_align_max_skip): Remove.
3052 * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
3053 * config/rx/rx-protos.h (rx_align_for_label): Make it
3055 * config/rx/rx.c (rx_align_for_label): Change return type
3057 (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
3059 into align_flags class.
3060 (LABEL_ALIGN): Likewise.
3061 (LOOP_ALIGN): Likewise.
3062 * config/s390/s390.c (s390_label_align): Use align_flags
3064 (s390_asm_output_function_label): Likewise.
3065 * config/sh/sh.c (sh_override_options_after_change):
3066 Use align_flags class directly without macros.
3067 (find_barrier): Likewise.
3068 (barrier_align): Likewise.
3069 (sh_loop_align): Likewise.
3070 * config/spu/spu.c (spu_option_override):
3071 Use align_flags_tuple::get_value instead of removed macros.
3072 (spu_sched_init): Likewise.
3073 * config/spu/spu.h (GTY): Likewise.
3074 * config/visium/visium.c (visium_option_override):
3075 Set "8" as default secondary alignment.
3076 * config/visium/visium.h (SUBALIGN_LOG): Define to 3
3077 in order to guarantee secondary alignment of 8.
3078 * coretypes.h: Include align.h header file.
3079 * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
3080 TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
3081 and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
3082 * doc/tm.texi.in: Likewise.
3083 * final.c (struct label_alignment): Remove not used structure.
3084 (LABEL_ALIGN): Change type to align_flags.
3085 (LOOP_ALIGN): Likewise.
3086 (JUMP_ALIGN): Likewise.
3087 (default_loop_align_max_skip): Remove.
3088 (default_label_align_max_skip): Likewise.
3089 (default_jump_align_max_skip): Likewise.
3090 (default_label_align_after_barrier_max_skip):
3091 (LABEL_TO_ALIGNMENT): Change to access label_align vector.
3092 (LABEL_TO_MAX_SKIP): Remove.
3093 (label_to_alignment): Return align_flags type instead of integer.
3094 (label_to_max_skip): Remove.
3095 (align_fuzz): Use align_flags type.
3096 (compute_alignments): Use align_flags type and use align_flags::max
3097 to combine multiple alignments.
3098 (grow_label_align): Grow vec instead of C array.
3099 (update_alignments): Assign just LABEL_TO_ALIGNMENT.
3100 (shorten_branches): Use align_flags type and use align_flags::max
3101 to combine multiple alignments.
3102 (final_scan_insn_1): Remove usage of secondary alignment that comes
3103 from label alignment, but instead use proper secondary alignment
3104 which is computed in grow_label_align.
3105 * flags.h (struct align_flags_tuple): Move to align.h.
3106 (struct align_flags): Likewise.
3107 (state_align_loops): Rename to align_loops.
3108 (state_align_jumps): Rename to align_jumps.
3109 (state_align_labels): Rename to align_labels.
3110 (state_align_functions): Rename to align_functions.
3111 (align_loops_log): Remove.
3112 (align_jumps_log): Remove.
3113 (align_labels_log): Remove.
3114 (align_functions_log): Remove.
3115 (align_loops_max_skip): Remove.
3116 (align_jumps_max_skip): Remove.
3117 (align_labels_max_skip): Remove.
3118 (align_functions_max_skip): Remove.
3119 (align_loops_value): Remove.
3120 (align_jumps_value): Remove.
3121 (align_labels_value): Remove.
3122 (align_functions_value): Remove.
3123 * output.h (label_to_alignment): Change return type to align_flags.
3124 (label_to_max_skip): Remove.
3125 * target.def: Remove loop_align_max_skip, label_align_max_skip,
3126 jump_align_max_skip macros.
3127 * targhooks.h (default_loop_align_max_skip): Remove.
3128 (default_label_align_max_skip): Likewise.
3129 (default_jump_align_max_skip): Likewise.
3130 (default_label_align_after_barrier_max_skip): Remove.
3131 * toplev.c (read_log_maxskip): Use ::normalize function.
3132 (parse_N_M): Remove not used argument and also call ::normalize.
3133 (parse_alignment_opts): Do not pass unused arguments.
3134 * varasm.c (assemble_start_function): Use directly align_functions
3135 instead of removed macros.
3136 * system.h: Do not poison removed macros.
3138 2018-07-17 Jakub Jelinek <jakub@redhat.com>
3141 * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
3142 and cond temporaries don't have reference type if iterator has
3143 pointer type. For init use &for_pre_body instead of pre_p if
3144 for_pre_body is non-empty.
3146 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
3148 * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
3149 double-double modes to SFmode directly directly.
3150 (trunc<mode>sf2_fprs): Delete.
3152 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
3154 * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
3155 for conversions between IFmode and the decimal floating point modes.
3156 (init_float128_ieee): Use the correct names for conversions between
3157 KFmode and the decimal floating point modes.
3159 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
3161 * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
3162 for the conversions between TDmode and IFmode.
3163 (init_float128_ieee): Use more correct names for the conversions
3164 between TDmode and KFmode.
3166 2018-07-16 Jakub Jelinek <jakub@redhat.com>
3168 PR tree-optimization/86526
3169 * builtins.c (expand_builtin_memcmp): Formatting fixes.
3170 (inline_expand_builtin_string_cmp): Likewise.
3171 (inline_string_cmp): Likewise. Use c_readstr instead of
3172 builtin_memcpy_read_str. Add unit_mode temporary.
3174 2018-07-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
3177 * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
3178 * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
3180 2018-07-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
3182 * doc/extend.texi (PowerPC AltiVec Built-in Functions):
3183 Alphabetize prototypes of built-in functions, separating out
3184 built-in functions that are listed in this section but should be
3185 described elsewhere.
3187 2018-07-16 Uros Bizjak <ubizjak@gmail.com>
3190 * expmed.c (emit_store_flag): Do not emit setcc followed by a
3191 conditional move when trapping comparison was split to a
3192 non-trapping one (and vice versa).
3194 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
3196 * config/s390/s390.c (s390_function_profiler): Generate nops
3197 instead of profiler call sequences.
3198 * config/s390/s390.opt: Add the new option.
3200 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
3202 * config/s390/s390.c (s390_function_profiler): Generate
3203 __mcount_loc section.
3204 * config/s390/s390.opt: Add the new option.
3206 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
3208 * common.opt: Add the new warning.
3209 * config/s390/s390.c (s390_function_profiler): Emit "brasl
3210 %r0,__fentry__" when -mfentry is specified.
3211 (s390_option_override_internal): Disallow -mfentry for 31-bit
3213 * config/s390/s390.opt: Add the new option.
3215 2018-07-16 Richard Biener <rguenther@suse.de>
3218 * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
3219 for function-local FUNCTION_DECL and RESULT_DECL immediately.
3221 2018-07-16 Martin Liska <mliska@suse.cz>
3224 * ipa-pure-const.c (malloc_candidate_p): Revert ::get
3227 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
3229 * config/arc/arcHS.md: Update ARCHS scheduling rules.
3231 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
3233 * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
3235 * config/arc/arc-cpus.def (hs4x): New cpu.
3237 * config/arc/arc-tables.opt: Regenerate.
3238 * config/arc/arc.c (arc_sched_issue_rate): New function.
3239 (TARGET_SCHED_ISSUE_RATE): Define.
3240 (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
3241 * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
3243 (attr tune): Add ARCHS4x tune values.
3244 (attr tune_dspmpy): Define.
3245 (*tst): Correct instruction type.
3246 * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
3247 * config/arc/arcHS4x.md: New file.
3248 * config/arc/fpu.md: Update instruction type attributes.
3249 * config/arc/t-multilib: Regenerate.
3251 2018-07-16 Tom de Vries <tdevries@suse.de>
3254 * var-tracking.c (vt_initialize): Fix pre_dec handling.
3256 2018-07-16 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3258 * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
3261 2018-07-16 Eric Botcazou <ebotcazou@adacore.com>
3263 PR tree-optimization/86514
3264 * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
3265 conversion to a boolean type from a type with greater precision.
3267 2018-07-16 Tom de Vries <tdevries@suse.de>
3269 * var-tracking.c (vt_initialize): Print adjusted insn slim if
3270 dump_flags request TDF_SLIM.
3272 2018-07-16 Aldy Hernandez <aldyh@redhat.com>
3274 * fold-const.c (int_const_binop_1): Abstract...
3275 (wide_int_binop): ...wide int code here.
3276 (poly_int_binop): ...poly int code here.
3277 Abstract the rest of int_const_binop_1 into int_const_binop.
3278 * fold-const.h (wide_int_binop): New.
3279 * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
3280 Remove useless PLUS/MINUS_EXPR case.
3281 (zero_nonzero_bits_from_vr): Move wide int code...
3282 (zero_nonzero_bits_from_bounds): ...here.
3283 (extract_range_from_binary_expr_1): Move mask optimization code...
3284 (range_easy_mask_min_max): ...here.
3285 * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
3286 (range_easy_mask_min_max): New.
3288 2018-07-15 Jeff Law <law@redhat.com>
3291 * config/sh/sh.c (output_mi_thunk): Remove dead conditional
3294 2018-07-14 Jim Wilson <jimw@sifive.com>
3296 * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
3298 2018-07-14 Paul Koning <ni1d@arrl.net>
3300 * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
3302 2018-07-13 Jan Hubicka <hubicka@ucw.cz>
3304 * lto-streamer-out.c (copy_function_or_variable): Dump info about
3307 2018-07-13 Bill Schmidt <wschmidt@linux.ibm.com>
3308 Steve Munroe <munroesj52@gmail.com>
3310 * config/rs6000/emmintrin.h (_mm_and_si128): New function.
3311 (_mm_andnot_si128): Likewise.
3312 (_mm_or_si128): Likewise.
3313 (_mm_xor_si128): Likewise.
3315 2018-07-13 Qing Zhao <qing.zhao@oracle.com>
3318 * builtins.c (expand_builtin_memcmp): Inline the calls first
3319 when result_eq is false.
3320 (expand_builtin_strcmp): Inline the calls first.
3321 (expand_builtin_strncmp): Likewise.
3322 (inline_string_cmp): New routine. Expand a string compare
3323 call by using a sequence of char comparison.
3324 (inline_expand_builtin_string_cmp): New routine. Inline expansion
3325 a call to str(n)cmp/memcmp.
3326 * doc/invoke.texi (--param builtin-string-cmp-inline-length):
3328 * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
3330 2018-07-13 Richard Earnshaw <rearnsha@arm.com>
3332 * config/arm/driver-arm.c: Include arm-native.h.
3333 (host_detect_local_cpu): Use auto-generated data tables.
3334 (vendors, arm_cpu_table): Delete. Move part information to ...
3335 * config/arm/arm-cpus.in: ... here.
3336 * config/arm/parsecpu.awk (gen_native): New function.
3337 (vendor, part): New CPU fields.
3338 (END): Add support for building the native CPU detection tables.
3339 * config/arm/t-arm (arm-native.h): Add build rule.
3340 (driver-arm.o): Add dependency on arm-native.h.
3342 2018-07-13 Richard Biener <rguenther@suse.de>
3345 * match.pd (addr1 - addr2): Allow either of the operand to
3348 2018-07-13 Tom de Vries <tdevries@suse.de>
3350 * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
3353 2018-07-13 Jackson Woodruff <jackson.woodruff@arm.com>
3355 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
3356 arrays instead of numbered variables.
3358 2018-07-13 Eric Botcazou <ebotcazou@adacore.com>
3360 * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
3361 * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
3363 2018-07-13 Richard Biener <rguenther@suse.de>
3366 * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
3367 instead of get_context_die.
3369 2018-07-13 Kugan Vivekanandarajah <kuganv@linaro.org>
3370 Richard Biener <rguenther@suse.de>
3373 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
3374 that the loop latch destination where phi is defined.
3376 2018-07-12 Kito Cheng <kito.cheng@gmail.com>
3378 * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
3379 (riscv_expand_epilogue): Add assertion to check interrupt mode.
3380 (riscv_set_current_function): Extract getting interrupt type to new
3382 (riscv_get_interrupt_type): New function.
3383 (riscv_merge_decl_attributes): New function, checking interrupt type is
3385 (TARGET_MERGE_DECL_ATTRIBUTES): Define.
3387 2018-07-12 Paul Koning <ni1d@arrl.net>
3389 * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
3392 2018-07-12 Paul Koning <ni1d@arrl.net>
3394 * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
3395 zero reference, add doloop_end instead.
3396 * doc/md.texi (decrement_and_branch_until_zero): Remove.
3397 (Looping patterns): Remove decrement_and_branch_until_zero. Add
3398 detail for doloop_end.
3400 2018-07-12 Martin Sebor <msebor@redhat.com>
3403 * attribs.c (decl_attributes): Reject conflicting attributes before
3404 calling attribute handlers.
3406 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
3408 * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
3410 (gcc::dump_manager::get_dump_file_name): likewise.
3411 (dump_begin): Likewise.
3412 * dumpfile.h (dump_begin): Update prototype.
3413 (gcc::dump_manager::get_dump_file_name,
3414 gcc::dump_manager::get_dump_file_name): Update prototype.
3416 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
3418 * internal-fn.h (vectorizable_internal_fn_p): New function.
3419 * tree-vect-slp.c (compatible_calls_p): Likewise.
3420 (vect_build_slp_tree_1): Remove nops argument. Handle calls
3421 to internal functions.
3422 (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
3424 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
3426 * fold-const.h (inverse_conditions_p): Declare.
3427 * fold-const.c (inverse_conditions_p): New function.
3428 * match.pd: Use inverse_conditions_p. Add folds of view_converts
3429 that test the inverse condition of a conditional internal function.
3430 * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
3431 * internal-fn.c (internal_fn_mask_index): Handle conditional
3433 (vectorized_internal_fn_supported_p): New function.
3434 * tree-if-conv.c: Include internal-fn.h and fold-const.h.
3435 (any_pred_load_store): Replace with...
3436 (need_to_predicate): ...this new variable.
3437 (redundant_ssa_names): New variable.
3438 (ifcvt_can_use_mask_load_store): Move initial checks to...
3439 (ifcvt_can_predicate): ...this new function. Handle tree codes
3440 for which a conditional internal function exists.
3441 (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
3442 instead of ifcvt_can_use_mask_load_store. Update after variable
3444 (predicate_load_or_store): New function, split out from
3445 predicate_mem_writes.
3446 (check_redundant_cond_expr): New function.
3447 (value_available_p): Likewise.
3448 (predicate_rhs_code): Likewise.
3449 (predicate_mem_writes): Rename to...
3450 (predicate_statements): ...this. Use predicate_load_or_store
3451 and predicate_rhs_code.
3452 (combine_blocks, tree_if_conversion): Update after above name changes.
3453 (ifcvt_local_dce): Handle redundant_ssa_names.
3454 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
3455 general conditional functions.
3456 * tree-vect-stmts.c (vectorizable_call): Likewise.
3458 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
3459 Alan Hayward <alan.hayward@arm.com>
3460 David Sherwood <david.sherwood@arm.com>
3462 * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
3463 * internal-fn.c (can_interpret_as_conditional_op_p): New function.
3464 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
3465 plus and minus and convert them into IFN_COND_FMA-based sequences.
3466 (convert_mult_to_fma): Handle conditional plus and minus.
3468 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
3470 * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
3471 * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
3472 (cond_fnms_optab): New optabs.
3473 * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
3475 (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
3476 * internal-fn.h (get_conditional_internal_fn): Declare.
3477 (get_unconditional_internal_fn): Likewise.
3478 * internal-fn.c (cond_ternary_direct): New macro.
3479 (expand_cond_ternary_optab_fn): Likewise.
3480 (direct_cond_ternary_optab_supported_p): Likewise.
3481 (FOR_EACH_COND_FN_PAIR): Likewise.
3482 (get_conditional_internal_fn): New function.
3483 (get_unconditional_internal_fn): Likewise.
3484 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
3485 (gimple_match_op::gimple_match_op): Add a new overload for 5
3487 (gimple_match_op::set_op): Likewise.
3488 (gimple_resimplify5): Declare.
3489 * genmatch.c (decision_tree::gen): Generate simplifications for
3491 * gimple-match-head.c (gimple_simplify): Define an overload for
3492 5 operands. Handle calls with 5 arguments in the top-level overload.
3493 (convert_conditional_op): Handle conversions from unconditional
3494 internal functions to conditional ones.
3495 (gimple_resimplify5): New function.
3496 (build_call_internal): Pass a fifth operand.
3497 (maybe_push_res_to_seq): Likewise.
3498 (try_conditional_simplification): Try converting conditional
3499 internal functions to unconditional internal functions.
3500 Handle 3-operand unconditional forms.
3501 * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
3502 Define ternary equivalents of the current rules for binary conditional
3504 * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
3506 * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
3507 (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
3508 (optab): Handle them.
3509 (SVE_COND_FP_TERNARY): New int iterator.
3510 (sve_fmla_op, sve_fmad_op): New int attributes.
3511 * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
3512 (*cond_<optab><mode>_2, *cond_<optab><mode_4)
3513 (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
3515 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
3517 * target.def (preferred_else_value): New target hook.
3518 * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
3519 * doc/tm.texi: Regenerate.
3520 * targhooks.h (default_preferred_else_value): Declare.
3521 * targhooks.c (default_preferred_else_value): New function.
3522 * internal-fn.h (conditional_internal_fn_code): Declare.
3523 * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
3524 (get_conditional_internal_fn): Use it.
3525 (conditional_internal_fn_code): New function.
3526 * gimple-match.h (gimple_match_cond): New struct.
3527 (gimple_match_op): Add a cond member function.
3528 (gimple_match_op::gimple_match_op): Update all forms to take a
3530 * genmatch.c (expr::gen_transform): Use the same condition as res_op
3531 for the suboperation, but don't specify a particular else_value.
3532 * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
3533 (visit_nary_op, visit_reference_op_load): Pass
3534 gimple_match_cond::UNCOND to the gimple_match_op constructor.
3535 * gimple-match-head.c: Include tree-eh.h
3536 (convert_conditional_op): New function.
3537 (maybe_resimplify_conditional_op): Likewise.
3538 (gimple_resimplify1): Call maybe_resimplify_conditional_op.
3539 (gimple_resimplify2): Likewise.
3540 (gimple_resimplify3): Likewise.
3541 (gimple_resimplify4): Likewise.
3542 (maybe_push_res_to_seq): Return null for conditional operations.
3543 (try_conditional_simplification): New function.
3544 (gimple_simplify): Call it. Pass conditions to the gimple_match_op
3546 * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
3548 * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
3550 (TARGET_PREFERRED_ELSE_VALUE): Redefine.
3552 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
3554 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
3556 (hash_tree): Do not hash DECL_FCONTEXT
3557 * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
3558 Do not stream DECL_FCONTEXT.
3559 * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
3560 * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
3562 2018-07-12 Richard Biener <rguenther@suse.de>
3565 * dwarf2out.c (gen_block_die): Only output blocks when they have
3566 at least one !DECL_IGNORED_P variable.
3568 2018-07-12 Richard Biener <rguenther@suse.de>
3571 * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
3572 Remove -mieee-fp handling.
3574 2018-07-12 Richard Biener <rguenther@suse.de>
3576 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
3577 left-over from last patch.
3579 2018-07-12 Jakub Jelinek <jakub@redhat.com>
3581 PR tree-optimization/86492
3582 * gimple-ssa-store-merging.c
3583 (imm_store_chain_info::coalesce_immediate_stores): Call
3584 check_no_overlap even for the merge_overlapping case. Formatting fix.
3586 2018-07-12 Richard Biener <rguenther@suse.de>
3589 * fold-const.c (fold_binary_op_with_conditional_arg): Do not
3590 move possibly trapping operations into the conditional.
3592 2018-07-12 Richard Biener <rguenther@suse.de>
3594 * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
3595 (vn_lookup_simplify_result): Remove recursion limit applied
3597 (vn_nary_build_or_lookup_1): Adjust.
3598 (try_to_simplify): Likewise.
3599 * gimple-match-head.c (gimple_resimplify1): Instead apply one
3601 (gimple_resimplify2): Likewise.
3602 (gimple_resimplify3): Likewise.
3603 (gimple_resimplify4): Likewise.
3605 2018-07-11 Jakub Jelinek <jakub@redhat.com>
3607 * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
3608 Use __mmask64 type instead of __mmask8 for __M argument.
3609 * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
3610 _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
3612 (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
3613 __mmask16 for __M argument.
3614 (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
3615 _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
3616 to __mmask16 instead of __mmask8.
3617 * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
3618 _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
3619 _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
3620 _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
3621 instead of __mmask16 for __U argument.
3622 * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
3623 __mmask16 instead of __mmask8 for __U argument.
3624 (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
3626 (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
3628 (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
3630 (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
3632 (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
3634 (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
3636 (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
3638 (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
3640 (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
3642 (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
3644 (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
3645 _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
3646 return type as well as __M argument type and all casts from __mmask8
3648 (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
3649 _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
3650 return type as well as __M argument type and all casts from __mmask8
3652 (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
3653 _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
3654 return type as well as __M argument type and all casts from __mmask8
3656 (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
3657 _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
3658 return type as well as __M argument type and all casts from __mmask8
3660 * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
3661 _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
3664 2018-07-11 Grazvydas Ignotas <notasas@gmail.com>
3666 * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
3667 _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
3670 2018-07-11 Paul Koning <ni1d@arrl.net>
3672 * doc/md.texi (define_subst): Document how multiple occurrences of
3673 the same argument in the replacement pattern are handled.
3675 2018-07-11 Paul Koning <ni1d@arrl.net>
3677 * doc/extend.texi (Common Variable Attributes): Move "mode" into
3679 (Common Type Attributes): Add "mode" attribute.
3681 2018-07-11 Jan Hubicka <hubicka@ucw.cz>
3683 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
3684 stream DECL_ORIGINAL_TYPE.
3685 (DFS::DFS_write_tree_body): Drop hack handling local external decls.
3686 (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
3687 * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
3688 Do not walk original type.
3689 * tree-streamer-out.c (streamer_write_chain): Drop hack handling
3691 (write_ts_decl_non_common_tree_pointers): Do not stream
3693 * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
3694 (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
3696 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
3698 * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
3699 thread twice from the same starting edge.
3701 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
3703 * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
3704 * gimple.c (gimple_call_nonnull_result_p): ...here...
3705 (gimple_call_nonnull_arg): ...and here.
3706 * gimple.h (gimple_call_nonnull_result_p): New.
3707 (gimple_call_nonnull_arg): New.
3709 2018-07-11 Richard Earnshaw <rearnsha@arm.com>
3711 * config/arm/arm-cpus.in: Move information from fpu field of each
3712 cpu definition to the isa field.
3713 * config/arm/parsecpu.awk (fpu): Delete match rule.
3714 (gen_comm_data): Don't add bits from the CPU's FPU entry.
3716 2018-07-11 Richard Biener <rguenther@suse.de>
3719 * dwarf2out.c (init_sections_and_labels): Use
3720 output_asm_line_debug_info consistently.
3721 (dwarf2out_early_finish): Likewise.
3722 (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
3725 2018-07-11 Richard Biener <rguenther@suse.de>
3727 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
3728 Rework father_bb setting in a way to avoid propagating constants
3729 multiple times on a loop body.
3731 2018-07-10 Mark Wielaard <mark@klomp.org>
3734 * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
3736 2018-07-10 Richard Biener <rguenther@suse.de>
3738 * hash-map.h (hash_map::iterator::operator*): Return
3739 references to key and value.
3741 2018-07-10 Jakub Jelinek <jakub@redhat.com>
3744 * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
3745 to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
3746 BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
3747 (gimplify_omp_for): For composite loops, move outer
3748 OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
3749 OMP_FOR/OMP_SIMD if there are any non-trivial wrappers. For class
3750 iterators add any needed clauses. Allow OMP_FOR_ORIG_DECLS to contain
3751 TREE_LIST for both the original class iterator and the "last" helper
3752 var. Gimplify OMP_FOR_PRE_BODY before the outermost composite
3753 loop, remember has_decl_expr from outer composite loops for the
3754 innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
3756 2018-07-09 Martin Sebor <msebor@redhat.com>
3760 * builtins.c (c_strlen): Avoid out-of-bounds warnings when
3761 accessing implicitly initialized array elements.
3762 * expr.c (string_constant): Handle string initializers of
3763 character arrays within aggregates.
3764 * gimple-fold.c (fold_array_ctor_reference): Add argument.
3765 Store element offset. As a special case, handle zero size.
3766 (fold_nonarray_ctor_reference): Same.
3767 (fold_ctor_reference): Add argument. Store subobject offset.
3768 * gimple-fold.h (fold_ctor_reference): Add argument.
3770 2018-07-09 Paul Koning <ni1d@arrl.net>
3772 * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
3773 (pdp11_insn_cost): New function.
3774 (pdp11_md_asm_adjust): New function.
3775 (TARGET_INVALID_WITHIN_DOLOOP): Define.
3776 (pdp11_rtx_costs): Update to match machine better.
3777 (output_addr_const_pdp11): Correct format mismatch warnings.
3778 * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
3779 * config/pdp11/pdp11.md: General change to add base_cost and/or
3780 length attributes for use by new pdp11_insn_cost function.
3781 (MIN_BRANCH): Correct definition.
3783 (doloop_end): Use standard pattern name for looping pattern.
3784 (doloop_end_nocc): New.
3785 (movsf): Add another constraint alternative.
3786 (zero_extendqihi2): Add constraint alternatives for not in place
3788 (zero_extendhisi2): Remove.
3789 (shift patterns): Add CC handling variants.
3793 (define_peephole2): New peephole to recognize mov that sets CC for
3796 2018-07-09 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3798 * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
3799 SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
3802 2018-07-09 Jakub Jelinek <jakub@redhat.com>
3805 * real.c (real_nextafter): Return true if result is denormal.
3807 2018-07-09 Martin Liska <mliska@suse.cz>
3809 * common.opt: Add back wrongly removed attribute.
3811 2018-07-09 Richard Biener <rguenther@suse.de>
3814 * dwarf2out.c (gen_block_die): For an early generated DIE
3815 always output high/low PC attributes.
3817 2018-07-09 Tom de Vries <tdevries@suse.de>
3819 * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
3820 * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
3821 onto VAR_DECL with abstract origin.
3823 2018-07-07 Jim Wilson <jimw@sifive.com>
3825 * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
3827 2018-07-07 Tom de Vries <tdevries@suse.de>
3829 * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
3831 2018-07-07 Aldy Hernandez <aldyh@redhat.com>
3833 * tree-vrp.c (vrp_int_const_binop): Change overflow type to
3835 (combine_bound): Use wide-int overflow calculation instead of
3837 * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
3839 * fold-const.c (int_const_binop_2): Same.
3840 (extract_muldiv_1): Same.
3841 (fold_div_compare): Same.
3842 (fold_abs_const): Same.
3844 * poly-int.h (add): Same.
3848 * predict.c (predict_iv_comparison): Same.
3849 * profile-count.c (slow_safe_scale_64bit): Same.
3850 * simplify-rtx.c (simplify_const_binary_operation): Same.
3851 * tree-chrec.c (tree_fold_binomial): Same.
3852 * tree-data-ref.c (split_constant_offset_1): Same.
3853 * tree-if-conv.c (idx_within_array_bound): Same.
3854 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
3855 * tree-ssa-phiopt.c (minmax_replacement): Same.
3856 * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
3857 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
3858 * vr-values.c (vr_values::adjust_range_with_scev): Same.
3859 * wide-int.cc (wi::add_large): Same.
3860 (wi::mul_internal): Same.
3861 (wi::sub_large): Same.
3862 (wi::divmod_internal): Same.
3863 * wide-int.h: Change overflow type to overflow_type for neg, add,
3864 mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
3865 mod_trunc, mod_ceil, mod_round, add_large, sub_large,
3866 mul_internal, divmod_internal.
3867 (overflow_type): New enum.
3868 (accumulate_overflow): New.
3870 2018-07-06 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
3872 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
3873 (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
3875 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
3877 * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
3878 argument is checked for zero before entering loop, avoid checking again.
3880 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
3882 * gimplify.h (generic_expr_could_trap_p): Set as global function.
3883 * gimplify.h (generic_expr_could_trap_p): Likwise.
3884 * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
3886 2018-07-06 Jakub Jelinek <jakub@redhat.com>
3888 PR tree-optimization/86401
3889 * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
3890 ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
3891 (fold_bit_and_mask): ... here. New helper function for match.pd.
3892 * fold-const.h (fold_bit_and_mask): Declare.
3893 * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
3895 2018-07-06 Peter Bergner <bergner@linux.ibm.com>
3898 * target.def (translate_mode_attribute): New hook.
3899 * targhooks.h (default_translate_mode_attribute): Declare.
3900 * targhooks.c (default_translate_mode_attribute): New function.
3901 * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
3902 * doc/tm.texi: Regenerate.
3903 * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
3904 (rs6000_translate_mode_attribute): New function.
3906 2018-07-06 Paul Koning <ni1d@arrl.net>
3908 * doc/md.texi (define_split): Document DONE and FAIL.
3909 (define_peephole2): Ditto.
3911 2018-07-05 Jeff Law <law@redhat.com>
3913 PR tree-optimization/86010
3914 * tree-ssa-dse.c (compute_trims): More aggressively trim at
3915 both the head and tail of mem* and str* calls.
3917 2018-07-05 Jim Wilson <jimw@sifive.com>
3919 * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
3921 2018-07-05 Indu Bhagat <indu.bhagat@oracle.com>
3923 * config/aarch64/aarch64-simd.md: correct flags text for
3924 MIN_EXPR replacement.
3926 2018-07-05 James Clarke <jrtc27@jrtc27.com>
3928 * configure: Regenerated.
3930 2018-07-05 Carl Love <cel@us.ibm.com>
3932 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
3933 float argument to VSX_BUILTIN_DOUBLEH_V4SF.
3934 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
3935 VSX_BUILTIN_DOUBLEL_V4SF.
3937 2018-07-05 Martin Sebor <msebor@redhat.com>
3940 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
3941 than its domain to compute its the upper bound of a char array.
3943 2018-07-05 Nathan Sidwell <nathan@acm.org>
3945 Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
3946 * doc/cpp.texi: Update comment.
3947 * doc/tm.texi: Rebuilt.
3948 * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
3949 (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
3950 * doc/extend.texi (Backwards Compatibility): Clarify it is system
3951 headers affected by extern "C".
3952 * system.h: Poison NO_IMPLICIT_EXTERN_C.
3953 * config/alpha/alpha.h, config/arm/uclinux-elf.h,
3954 config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
3955 config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
3956 config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
3957 config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
3958 config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
3959 config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
3960 config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
3961 config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
3962 config/sparc/openbsd64.h, config/sparc/sp-elf.h,
3963 config/sparc/sp64-elf.h, config/spu/spu.h,
3964 config/stormy16/stormy16.h, config/v850/v850.h,
3965 config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
3966 define NO_IMPLICIT_EXTERN_C.
3967 * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
3969 2018-07-05 Tamar Christina <tamar.christina@arm.com>
3972 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
3973 instead of GET_MODE_SIZE when comparing Units.
3975 2018-07-05 Tamar Christina <tamar.christina@arm.com>
3978 * rtlanal.c (set_noop_p): Constrain on mode change,
3979 include hard-reg-set.h
3981 2018-07-05 Tamar Christina <tamar.christina@arm.com>
3983 * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
3985 2018-07-05 Jakub Jelinek <jakub@redhat.com>
3988 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
3991 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
3993 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
3995 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
3998 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
4000 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
4002 2018-07-04 Eric Botcazou <ebotcazou@adacore.com>
4005 * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
4007 2018-07-04 Aldy Hernandez <aldyh@redhat.com>
4009 * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
4010 neg_*_op* variables.
4012 2018-07-04 Martin Liska <mliska@suse.cz>
4014 * tree-switch-conversion.c: Define
4015 max_ratio_for_speed and max_ratio_for_size constants.
4017 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
4018 Martin Liska <mliska@suse.cz>
4023 * common.opt: Rename align options with 'str_' prefix.
4024 * common/config/i386/i386-common.c (set_malign_value): New
4026 (ix86_handle_option): Use it to set -falign-* options/
4027 * config/aarch64/aarch64-protos.h (struct tune_params): Change
4028 type from int to string.
4029 * config/aarch64/aarch64.c: Update default values from int
4031 * config/alpha/alpha.c (alpha_override_options_after_change):
4033 * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
4034 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
4035 max skip conditionally.
4036 * config/i386/freebsd.h (SUBALIGN_LOG): New.
4037 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
4038 max skip conditionally.
4039 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
4040 max skip conditionally.
4041 * config/i386/gnu-user.h (SUBALIGN_LOG): New.
4042 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
4043 max skip conditionally.
4044 * config/i386/i386.c (struct ptt): Change type from int to
4046 (ix86_default_align): Set default values.
4047 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
4048 max skip conditionally.
4049 * config/i386/iamcu.h (SUBALIGN_LOG): New.
4050 (ASM_OUTPUT_MAX_SKIP_ALIGN):
4051 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
4052 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
4053 max skip conditionally.
4054 * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
4055 (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
4056 * config/i386/x86-64.h (SUBALIGN_LOG): New.
4057 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
4058 max skip conditionally.
4059 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
4060 * config/ia64/ia64.c (ia64_option_override): Set default values
4061 for alignment options.
4062 * config/m68k/m68k.c: Handle new str_align_* options.
4063 * config/mips/mips.c (mips_set_compression_mode): Change
4065 (mips_option_override): Set default values for options.
4066 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
4068 * config/rs6000/rs6000.c (rs6000_option_override_internal):
4070 * config/rx/rx.c (rx_option_override): Likewise.
4071 * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
4072 (LABEL_ALIGN): Use align_labels_log.
4073 (LOOP_ALIGN): Use align_loops_align.
4074 * config/s390/s390.c (s390_asm_output_function_label): Use new
4076 * config/sh/sh.c (sh_override_options_after_change):
4077 Change type of constants.
4078 * config/spu/spu.c (spu_sched_init): Likewise.
4079 * config/sparc/sparc.c (sparc_option_override): Set default
4081 * config/visium/visium.c (visium_option_override): Likewise.
4082 * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
4083 emit p2align format with last argument if it's not needed.
4084 * doc/invoke.texi: Document extended format of -falign-*.
4085 * final.c: Use align_labels alignment.
4086 * flags.h (struct target_flag_state): Change type to use
4088 (struct align_flags_tuple): New.
4089 (struct align_flags): Likewise.
4090 (align_loops_log): Redefine macro to use new types.
4091 (align_loops_max_skip): Redefine macro to use new types.
4092 (align_jumps_log): Redefine macro to use new types.
4093 (align_jumps_max_skip): Redefine macro to use new types.
4094 (align_labels_log): Redefine macro to use new types.
4095 (align_labels_max_skip): Redefine macro to use new types.
4096 (align_functions_log): Redefine macro to use new types.
4097 (align_loops): Redefine macro to use new types.
4098 (align_jumps): Redefine macro to use new types.
4099 (align_labels): Redefine macro to use new types.
4100 (align_functions): Redefine macro to use new types.
4101 (align_functions_max_skip): Redefine macro to use new types.
4102 (align_loops_value): New macro.
4103 (align_jumps_value): New macro.
4104 (align_labels_value): New macro.
4105 (align_functions_value): New macro.
4106 * function.c (invoke_set_current_function_hook): Propagate
4107 alignment values from flags to global variables default in
4109 * ipa-icf.c (sem_function::equals_wpa): Use
4110 cl_optimization_option_eq instead of memcmp.
4111 * lto-streamer.h (cl_optimization_stream_out): Support streaming
4113 (cl_optimization_stream_in): Likewise.
4114 * optc-save-gen.awk: Support strings in cl_optimization.
4115 * opth-gen.awk: Likewise.
4116 * opts.c (finish_options): Remove error checking of invalid
4118 (MAX_CODE_ALIGN): Remove.
4119 (MAX_CODE_ALIGN_VALUE): Likewise.
4120 (parse_and_check_align_values): New function.
4121 (check_alignment_argument): Likewise.
4122 (common_handle_option): Use check_alignment_argument.
4123 * opts.h (parse_and_check_align_values): Declare.
4124 * toplev.c (init_alignments): Remove.
4125 (read_log_maxskip): New.
4126 (parse_N_M): Likewise.
4127 (parse_alignment_opts): Likewise.
4128 (backend_init_target): Remove usage of init_alignments.
4129 * toplev.h (parse_alignment_opts): Declare.
4130 * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
4132 * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
4133 * tree.c (cl_option_hasher::equal): New.
4134 * varasm.c: Use new global macros.
4136 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
4138 * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
4139 Use a simpler align directive also if MAXSKIP = ALIGN-1.
4140 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4141 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4142 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4143 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
4144 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
4145 is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
4146 define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
4147 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4148 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4149 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4150 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4152 2018-07-04 Martin Liska <mliska@suse.cz>
4153 Jonathan Wakely <jwakely@redhat.com>
4155 * coverage.c: Use correct type.
4156 * doc/invoke.texi: Language correction.
4158 2018-07-03 H.J. Lu <hongjiu.lu@intel.com>
4161 * config/i386/i386.c (rest_of_insert_endbranch): Also generate
4162 ENDBRANCH for non-tail call which may return via indirect branch.
4163 * doc/extend.texi: Document indirect_return attribute.
4165 2018-07-03 Martin Sebor <msebor@redhat.com>
4167 PR tree-optimization/86274
4168 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
4170 (format_floating): Correct handling of infinities and NaNs.
4172 2018-07-03 Martin Sebor <msebor@redhat.com>
4174 * print-tree.c (print_real_cst): New function.
4175 (print_node_brief): Call it.
4176 (print_node): Ditto.
4178 2018-07-03 Jeff Law <law@redhat.com>
4180 * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
4181 into a single pattern.
4183 * config/h8300/h8300.md (ors code_iterator): New.
4184 (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
4185 a single pattern and single splitter.
4186 (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
4187 (iorqi3_1, xorqi3_1): Likewise.
4188 (iorqi3, xorqi3 expanders): Similarly.
4190 * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
4191 (movmd_internal) into a single pattern using the P mode iterator.
4192 (movmd splitters): Similarly.
4193 (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
4194 (movsd splitters): Similarly.
4196 * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
4197 ADDB, ADDW and ADDL into a single ADD attribute which selects the
4198 right table based on the size of the operand.
4199 * config/h8300/h8300.md (length_table): Corresponding changes. All
4200 references to "addb", "addw" and "addl" changed to "add".
4201 (btst patterns): Merge two variants into a single pattern.
4202 (tstqi, tsthi): Likewise.
4203 (addhi3_incdec, addsi3_incdec): Likewise.
4204 (subhi3_h8300hs, subsi3_h8300hs): Likewise.
4205 (mulhi3, mulsi3): Likewise.
4206 (udivhi3, udivsi3): Likewise.
4207 (divhi3, divsi3): Likewise.
4208 (andorqi3, andorhi3, andorsi3): Likewise.
4210 2018-07-03 Uros Bizjak <ubizjak@gmail.com>
4213 * config/i386/sse.md (uavg<mode>3_ceil): New expander.
4214 (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
4216 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
4218 PR tree-optimization/85694
4219 * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
4220 (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
4222 * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
4223 (<u>avg<mode>3_ceil): New patterns.
4225 2018-07-03 David Malcolm <dmalcolm@redhat.com>
4227 * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
4228 scan-tree-dump directive.
4229 * gcc.dg/vect/slp-perm-2.c: Likewise.
4230 * gcc.dg/vect/slp-perm-3.c: Likewise.
4231 * gcc.dg/vect/slp-perm-5.c: Likewise.
4232 * gcc.dg/vect/slp-perm-6.c: Likewise.
4233 * gcc.dg/vect/slp-perm-7.c: Likewise.
4234 * gcc.dg/vect/slp-perm-8.c: Likewise.
4236 2018-07-03 Marek Polacek <polacek@redhat.com>
4239 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
4241 2018-07-03 Richard Biener <rguenther@suse.de>
4244 * tree-ssa-structalias.c (find_func_clobbers): Properly
4245 handle indirect calls.
4247 2018-07-03 Jeff Law <law@redhat.com>
4249 * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
4250 (shifts): New code iterator.
4251 (movqi, movhi, movsi, movsf expanders): Consolidate into a single
4252 expander. Fix HImode handling on H8/SX.
4253 (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
4254 (subqi3, subhi3, subsi3 expanders): Likewise.
4255 (andqi3, andhi3, andsi3 expanders): Likewise.
4256 (iorqi3, iorhi3, iorsi3 expanders): Likewise.
4257 (xorqi3, xorhi3, xorsi3 expanders): Likewise.
4258 (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
4259 (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
4260 (zero_extendqihi2, zero_extendqisi2): Likewise.
4261 (extendqihi2, extendqisi2): Likewise.
4262 (rotlqi3, rotlhi3, rotlsi3): Likewise.
4263 (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
4264 (rotlqi3_1, rotlhi3_1): Likewise.
4265 (logicalhi3_sn, logicalsi3_sn): Likewise.
4266 (logicalhi3, logicalsi3): Likewise.
4268 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
4270 * tree-vect-patterns.c (vect_recog_rotate_pattern)
4271 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
4272 (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
4273 (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
4274 type to append_pattern_def_seq instead of creating a stmt_vec_info
4276 (build_mask_conversion): Likewise. Remove vinfo argument.
4277 (vect_add_conversion_to_patterm): Likewise, renaming to...
4278 (vect_add_conversion_to_pattern): ...this.
4279 (vect_recog_mask_conversion_pattern): Update call to
4280 build_mask_conversion. Pass the vector type to
4281 append_pattern_def_seq here too.
4282 (vect_recog_gather_scatter_pattern): Update call to
4283 vect_add_conversion_to_pattern.
4285 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
4287 * tree-vect-patterns.c (new_pattern_def_seq): Delete.
4288 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
4289 (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
4290 (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
4291 STMT_VINFO_PATTERN_DEF_SEQ to null here.
4292 (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
4293 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
4294 append_pattern_def_seq instead of new_pattern_def_seq.
4295 (vect_recog_divmod_pattern): Do both of the above.
4296 (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
4299 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
4301 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
4302 (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
4303 (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
4304 (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
4305 (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
4306 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
4307 (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
4308 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
4309 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
4310 (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
4311 parameter with a single stmt_vec_info.
4312 (vect_recog_func_ptr): Likewise.
4313 (vect_recog_gather_scatter_pattern): Likewise, folding in...
4314 (vect_try_gather_scatter_pattern): ...this.
4315 (vect_pattern_recog_1): Remove stmts_to_replace and just pass
4316 the stmt_vec_info of the statement to be matched. Don't clear
4317 STMT_VINFO_RELATED_STMT.
4318 (vect_pattern_recog): Update call accordingly.
4320 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
4322 PR tree-optimization/85694
4323 * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
4324 (uavgM3_ceil): Document new optabs.
4325 * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
4326 * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
4328 * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
4329 (savg_ceil_optab): New optabs.
4330 * tree-vect-patterns.c (vect_recog_average_pattern): New function.
4331 (vect_vect_recog_func_ptrs): Add it.
4332 * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
4333 constant directly from the associated lhs.
4335 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
4337 * tree-vect-patterns.c (vect_split_statement): New function.
4338 (vect_convert_input): Use it to try to split an existing cast.
4340 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
4342 * poly-int.h (print_hex): New function.
4343 * dumpfile.h (dump_dec, dump_hex): Declare.
4344 * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
4345 * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
4346 min_input_precision, operation_precision and operation_sign.
4347 * tree-vect-patterns.c (vect_get_range_info): New function.
4348 (vect_same_loop_or_bb_p, vect_single_imm_use)
4349 (vect_operation_fits_smaller_type): Delete.
4350 (vect_look_through_possible_promotion): Add an optional
4351 single_use_p parameter.
4352 (vect_recog_over_widening_pattern): Rewrite to use new
4353 stmt_vec_info infomration. Handle one operation at a time.
4354 (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
4355 (vect_truncatable_operation_p, vect_set_operation_type)
4356 (vect_set_min_input_precision): New functions.
4357 (vect_determine_min_output_precision_1): Likewise.
4358 (vect_determine_min_output_precision): Likewise.
4359 (vect_determine_precisions_from_range): Likewise.
4360 (vect_determine_precisions_from_users): Likewise.
4361 (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
4362 (vect_vect_recog_func_ptrs): Put over_widening first.
4364 (vect_pattern_recog): Call vect_determine_precisions.
4366 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
4368 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
4369 statements that have been replaced by further pattern statements.
4370 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
4372 2018-07-03 Richard Biener <rguenther@suse.de>
4374 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
4375 always set *dt. Dump vectype in vectype overload.
4376 * dumpfile.h (dump_gimple_expr): New function.
4377 (dump_gimple_expr_loc): Likewise.
4378 * dumpfile.c (dump_gimple_expr): New function.
4379 (dump_gimple_expr_loc): Likewise.
4381 2018-07-02 Jeff Law <law@redhat.com>
4383 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
4384 the H8/300, H8/300H and H8/S variants into a single pattern.
4385 (movhi_h8300, movqi_h8300hs): Similarly.
4386 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
4387 (QHI mode iterator): New.
4389 * config/h8300/h8300.md: Remove trailing whitespace.
4391 2018-07-02 Jim Wilson <jimw@sifive.com>
4393 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
4394 instead of emit_insn for interrupt returns.
4395 * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
4396 (riscv_sret, riscv_uret): Likewise.
4398 2018-07-02 David Malcolm <dmalcolm@redhat.com>
4400 * pretty-print.c (selftest::test_pp_format): Move save and restore
4401 of quotes to class auto_fix_quotes, and add an instance.
4402 * selftest.c: Include "intl.h".
4403 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
4404 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
4405 * selftest.h (selftest::auto_fix_quotes): New class.
4407 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
4409 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
4410 (aarch64_sve_prepare_conditional_op): Remove.
4411 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
4412 Allow aarch64_simd_reg_or_zero as select operand; remove
4413 the aarch64_sve_prepare_conditional_op call.
4414 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
4415 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
4416 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
4417 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
4418 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
4419 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
4420 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
4421 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
4422 and a splitters to match all of the *_any patterns.
4423 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
4425 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
4426 (SVE_COND_FP_BINARY_REV): Remove.
4427 (sve_int_op_rev, sve_fp_op_rev): New.
4428 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
4429 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
4430 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
4431 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
4432 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
4433 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
4434 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
4435 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
4436 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
4438 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
4439 Remove match_dup 1 from the inner unspec.
4440 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
4442 * config/aarch64/aarch64.md (movprfx): New attr.
4443 (length): Default movprfx to 8.
4444 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
4445 (*madd<SVE_I>, *msub<SVE_I): Likewise.
4446 (*<su>mul<SVE_I>3_highpart): Likewise.
4447 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
4448 (*v<ASHIFT><SVE_I>3): Likewise.
4449 (*<su><MAXMIN><SVE_I>3): Likewise.
4450 (*<su><MAXMIN><SVE_F>3): Likewise.
4451 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
4452 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
4453 (*div<SVE_F>4): Likewise.
4455 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
4457 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
4460 2018-07-02 Richard Biener <rguenther@suse.de>
4462 PR tree-optimization/86363
4463 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
4464 memset argument refers to a non-variable address.
4466 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
4468 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
4469 {PLUS,MINUS}_EXPR code to...
4470 (adjust_symbolic_bound): ...here,
4471 (combine_bound): ...here,
4472 (set_value_range_with_overflow): ...and here.
4474 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
4476 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
4478 (extract_range_from_abs_expr): ...here.
4480 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
4482 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
4483 -fno-omit-frame-pointer when not optimizing.
4485 2018-07-02 Martin Liska <mliska@suse.cz>
4488 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
4489 (propagate_nothrow): Likewise.
4491 2018-07-02 Martin Liska <mliska@suse.cz>
4494 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
4496 2018-07-02 David Malcolm <dmalcolm@redhat.com>
4498 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
4499 function in r262149, changing "loc" param from source_location to
4500 const dump_location_t &.
4501 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
4502 declaration, as above.
4504 2018-07-01 Paul Koning <ni1d@arrl.net>
4506 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
4507 -munit-asm, -mgnu-asm, -mdec-asm.
4508 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
4509 (pdp11_output_labelref): New.
4510 (pdp11_output_def): New.
4511 (pdp11_output_addr_vec_elt): New.
4512 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
4513 %# and %@ format codes.
4514 (pdp11_option_override): New.
4515 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
4516 (pdp11_output_ident): New.
4517 (pdp11_asm_named_section): New.
4518 (pdp11_asm_init_sections): New.
4519 (pdp11_file_start): New.
4520 (pdp11_file_end): New.
4521 (output_ascii): Use .ascii/.asciz for -mdec-asm.
4522 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
4523 %o, like %c but octal.
4524 (pdp11_option_override): New.
4525 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
4527 (DATA_SECTION_ASM_OP): Ditto.
4528 (READONLY_DATA_SECTION_ASM_OP): New.
4529 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
4530 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
4531 (ASM_OUTPUT_LABELREF): Ditto.
4532 (ASM_OUTPUT_DEF): Ditto.
4533 (ASM_OUTPUT_EXTERNAL): New.
4534 (ASM_OUTPUT_SOURCE_FILENAME): New.
4535 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
4536 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
4537 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
4538 %# and %@ format codes.
4539 * config/pdp11/pdp11.opt (mgnu-asm): New.
4540 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
4541 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
4542 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
4544 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
4546 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
4547 dereferencing path[] beyond its length.
4549 (debug_all_paths): New.
4550 (rewire_first_differing_edge): New.
4551 (adjust_paths_after_duplication): New.
4552 (duplicate_thread_path): Call adjust_paths_after_duplication.
4554 (thread_through_all_blocks): Add new argument to
4555 duplicate_thread_path.
4557 2018-06-30 Jim Wilson <jimw@sifive.com>
4559 * config/riscv/predicates.md (p2m1_shift_operand): New.
4560 (high_mask_shift_operand): New.
4561 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
4562 pattern using p2m1_shift_operand.
4563 (lshsi3_zero_extend_3+2): New combiner pattern using
4564 high_mask_shift_operand.
4566 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
4568 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
4570 (vect_recog_rotate_pattern): ...here.
4571 (vect_convert_input): Try to insert casts of invariants in the
4573 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
4574 preheader to be empty.
4576 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
4578 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
4579 vector type. If given, install it in the new statement's
4581 (vect_element_precision): New function.
4582 (vect_unpromoted_value): New struct.
4583 (vect_unpromoted_value::vect_unpromoted_value): New function.
4584 (vect_unpromoted_value::set_op): Likewise.
4585 (vect_look_through_possible_promotion): Likewise.
4586 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
4587 (vect_widened_op_tree, vect_convert_input): Likewise.
4588 (vect_convert_inputs, vect_convert_output): Likewise.
4589 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
4590 to handle the optional cast of the multiplication result and
4591 vect_widened_op_tree to detect the widened multiplication itself.
4592 Do not require the input and output of promotion casts to have
4593 the same sign, but base the signedness of the operation on the
4594 input rather than the result. If the pattern includes two
4595 promotions, check that those promotions have the same sign.
4596 Do not restrict the MULT_EXPR handling to a double-width result;
4597 handle quadruple-width results and wider. Use vect_convert_inputs
4598 to convert the inputs to the common type.
4599 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
4600 to handle the optional cast of the ABS result. Also allow a sign
4601 change or a sign extension between the ABS and MINUS.
4602 Use vect_widened_op_tree to detect the widened subtraction and use
4603 vect_convert_inputs to convert the inputs to the common type.
4604 (vect_handle_widen_op_by_const): Delete.
4605 (vect_recog_widen_op_pattern): New function.
4606 (vect_recog_widen_mult_pattern): Use it.
4607 (vect_recog_widen_shift_pattern): Likewise.
4608 (vect_recog_widen_sum_pattern): Use
4609 vect_look_through_possible_promotion to handle the promoted
4612 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
4614 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
4615 the containing gimple_seq *.
4616 * gimple-iterator.h (gsi_for_stmt): Declare it.
4617 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
4618 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
4619 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
4620 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
4621 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
4623 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
4625 (vect_mark_pattern_stmts): ...here. Handle cases in which the
4626 statement being replaced is part of an existing pattern
4627 definition sequence, inserting the new pattern statements before
4629 (vect_pattern_recog_1): Don't return a bool. If the statement
4630 is already part of a pattern, instead apply pattern matching
4631 to the pattern definition statements. Don't clear the
4632 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
4633 (vect_pattern_recog): Don't break after the first match;
4634 continue processing the pattern definition statements instead.
4635 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
4637 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
4639 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
4640 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
4641 (vect_recog_widen_sum_pattern): Use it.
4643 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
4645 * tree-vect-loop.c (vectorizable_reduction): Assert that the
4646 phi is not a pattern statement and has not been replaced by
4647 a pattern statement.
4648 * tree-vect-patterns.c (type_conversion_p): Don't check
4649 STMT_VINFO_IN_PATTERN_P.
4650 (vect_recog_vector_vector_shift_pattern): Likewise.
4651 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
4652 the pattern statement rather than the original statement; check
4653 directly for a WIDEN_MULT_EXPR here.
4654 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
4655 vect_is_simple_use to return the pattern statement rather
4656 than the original statement; use is_pattern_stmt_p to check
4657 for such a pattern statement.
4658 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
4659 to return the pattern statement rather than the original statement;
4660 don't do the same transformation here.
4661 (vect_is_simple_use): If the defining statement has been replaced
4662 by a pattern statement, return the pattern statement instead.
4663 Remove the corresponding (local) transformation from the vectype
4666 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
4668 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
4669 end and default to null.
4670 * tree-vect-loop.c (vect_create_epilog_for_reduction)
4671 (vectorizable_reduction): Update calls accordingly, dropping the
4672 gimple ** argument if the passed-back statement isn't needed.
4673 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
4674 (vect_recog_rotate_pattern): Likewise.
4675 (vect_recog_mask_conversion_pattern): Likewise.
4676 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
4677 (vect_mask_constant_operand_p): Likewise.
4678 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
4679 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
4680 (get_group_load_store_type, get_load_store_type): Likewise.
4681 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
4682 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
4683 (vectorizable_conversion, vectorizable_assignment): Likewise.
4684 (vectorizable_shift, vectorizable_operation): Likewise.
4685 (vectorizable_store, vect_is_simple_cond): Likewise.
4686 (vectorizable_condition, vectorizable_comparison): Likewise.
4687 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
4688 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
4689 and move it to the end. Cope with null def_stmt_outs.
4691 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
4693 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
4695 2018-06-29 Jeff Law <law@redhat.com>
4697 * config/v850/v850.c (v850_legitimate_address_p): Handle large
4698 displacements for TARGET_V850E2V3 and newer.
4699 (TARGET_LRA_P): Remove. Defaults to LRA now.
4700 * config/v850/v850.md (sign23byte_load): Remove.
4701 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
4702 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
4704 2018-06-29 Martin Liska <mliska@suse.cz>
4707 * coverage.c (coverage_init): Mangle full path name.
4708 * doc/invoke.texi: Document the change.
4709 * gcov-io.c (mangle_path): New.
4710 * gcov-io.h (mangle_path): Likewise.
4711 * gcov.c (mangle_name): Use mangle_path for path mangling.
4713 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4715 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
4716 if starting source register is not even.
4718 2018-06-29 Martin Liska <mliska@suse.cz>
4720 PR tree-optimization/86263
4721 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
4722 Make edge redirection.
4724 2018-06-29 David Malcolm <dmalcolm@redhat.com>
4726 * dumpfile.c (dump_loc): Add indentation based on scope depth.
4727 (dump_scope_depth): New variable.
4728 (get_dump_scope_depth): New function.
4729 (dump_begin_scope): New function.
4730 (dump_end_scope): New function.
4731 * dumpfile.h (get_dump_scope_depth): New declaration.
4732 (dump_begin_scope): New declaration.
4733 (dump_end_scope): New declaration.
4734 (class auto_dump_scope): New class.
4735 (AUTO_DUMP_SCOPE): New macro.
4736 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
4739 2018-06-29 Richard Biener <rguenther@suse.de>
4741 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
4742 compute_all_dependences succeeds.
4743 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
4744 exceed --param loop-max-datarefs-for-datadeps.
4746 2018-06-29 Jakub Jelinek <jakub@redhat.com>
4748 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
4750 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
4753 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
4754 alternative 0 in preferred_for_speed attribute.
4756 2018-06-28 Paul Koning <ni1d@arrl.net>
4758 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
4759 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
4760 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
4761 * config/pdp11/pdp11.md: Correct "length" attribute calculation
4762 for shift insn patterns.
4764 2018-06-28 David Malcolm <dmalcolm@redhat.com>
4766 * cgraph.c (cgraph_node::get_body): Replace assignments to
4767 "dump_file" with calls to set_dump_file.
4768 * dumpfile.c (alt_dump_file): Make static, and group with...
4769 (alt_flags): ...this definition.
4770 (dumps_are_enabled): New variable.
4771 (refresh_dumps_are_enabled): New function.
4772 (set_dump_file): New function.
4773 (set_alt_dump_file): New function.
4774 (gcc::dump_manager::dump_start): Replace assignments to
4775 "dump_file" and "alt_dump_file" with calls to set_dump_file and
4777 (gcc::dump_manager::dump_finish): Likewise.
4778 * dumpfile.h (alt_dump_file): Delete decl.
4779 (dumps_are_enabled): New variable decl.
4780 (set_dump_file): New function decl.
4781 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
4783 * tree-nested.c (lower_nested_functions): Replace assignments to
4784 "dump_file" with calls to set_dump_file.
4786 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
4788 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
4789 goto_locus of each outgoing edge of each basic block.
4791 2018-06-28 Richard Biener <rguenther@suse.de>
4793 * dwarf2out.c (decl_scope_table): Remove.
4794 (push_decl_scope): Likewise.
4795 (pop_decl_scope): Likewise.
4796 (gen_type_die_for_member): Do not call push/pop_decl_scope.
4797 (gen_struct_or_union_type_die): Likewise.
4798 (gen_tagged_type_die): Likewise.
4799 (dwarf2out_init): Do not initialize decl_scope_table.
4800 (dwarf2out_c_finalize): Do not free it.
4802 2018-06-28 Richard Biener <rguenther@suse.de>
4804 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
4805 deciding whether to not re-use a DIE.
4807 2018-06-28 Richard Biener <rguenther@suse.de>
4809 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
4810 DW_AT_abstract_origin attribute.
4812 2018-06-28 Martin Liska <mliska@suse.cz>
4814 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
4815 Use newly introduced constants.
4816 * tree-switch-conversion.h (struct jump_table_cluster):
4817 Define max_ratio_for_size and max_ratio_for_speed.
4819 2018-06-28 Martin Liska <mliska@suse.cz>
4821 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
4822 Add new checking assert to catch invalid state.
4823 (jump_table_cluster::can_be_handled): Handle single case
4825 (jump_table_cluster::is_beneficial): Bail out for such case.
4826 (bit_test_cluster::find_bit_tests):
4827 Add new checking assert to catch invalid state.
4828 (bit_test_cluster::can_be_handled): Handle single case
4830 (bit_test_cluster::is_beneficial): Bail out for such case.
4831 (switch_decision_tree::analyze_switch_statement):
4834 2018-06-28 Martin Liska <mliska@suse.cz>
4836 * common.opt: Introduce -completion option.
4837 * gcc.c (driver_handle_option): Handle it.
4838 (driver::main): Print completions if completion
4840 * opt-suggestions.c (option_proposer::get_completions):
4842 (option_proposer::suggest_completion): Likewise.
4843 (option_proposer::find_param_completions): Likewise.
4844 (verify_autocompletions): Likewise.
4845 (test_completion_valid_options): Likewise.
4846 (test_completion_valid_params): Likewise.
4847 (in_completion_p): Likewise.
4848 (empty_completion_p): Likewise.
4849 (test_completion_partial_match): Likewise.
4850 (test_completion_garbage): Likewise.
4851 (opt_proposer_c_tests): Likewise.
4852 * opt-suggestions.h: Declare new functions.
4853 * opts.c (common_handle_option): Handle OPT__completion_.
4854 * selftest-run-tests.c (selftest::run_tests): Add
4855 opt_proposer_c_tests.
4856 * selftest.c (assert_str_startswith): New.
4857 * selftest.h (assert_str_startswith): Likewise.
4858 (opt_proposer_c_tests): New.
4859 (ASSERT_STR_STARTSWITH): Likewise.
4861 2018-06-28 Martin Liska <mliska@suse.cz>
4863 * Makefile.in: Add opt-suggestions.o.
4864 * gcc-main.c: Include opt-suggestions.h.
4865 * gcc.c (driver::driver): Likewise.
4866 (driver::~driver): Remove m_option_suggestions.
4867 (driver::build_option_suggestions): Moved to option_proposer.
4868 (driver::suggest_option): Likewise.
4869 (driver::handle_unrecognized_options): Use option_proposer.
4870 * gcc.h (class driver): Add new memver m_option_proposer.
4871 * opt-suggestions.c: New file.
4872 * opt-suggestions.h: New file.
4874 2018-06-28 Martin Liska <mliska@suse.cz>
4876 * vec.h (class auto_string_vec): New (moved from auto_argvec).
4877 (auto_string_vec::~auto_string_vec): Likewise.
4879 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
4881 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
4882 prevent_decl_creation_for_types fields up and add reset_location field.
4883 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
4884 statement if id->reset_location is true.
4885 (copy_edges_for_bb): Do not set goto_locus on the new edges if
4886 id->reset_location is true.
4887 (copy_phis_for_bb): Force input_location on the arguments if
4888 id->reset_location is true.
4889 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
4890 is set on the function to be inlined.
4892 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
4894 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
4896 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
4898 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
4899 registers for Pmode.
4900 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
4901 hard registers for the clobbered pseudo.
4903 2018-06-27 Paul Koning <ni1d@arrl.net>
4905 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
4906 mutually exclusive options.
4907 * config/pdp11/constraints.md (h): New constraint.
4908 (O): Update definition to match shift code generation.
4909 (D): New constraint.
4910 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
4912 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
4914 (output_jump): Change arguments.
4915 (pdp11_fixed_cc_regs): New function.
4916 (pdp11_cc_mode): Ditto.
4917 (pdp11_expand_shift): Ditto.
4918 (pdp11_assemble_shift): Ditto.
4919 (pdp11_small_shift): Ditto.
4920 (pdp11_branch_cost): Remove.
4921 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
4923 (pdp11_register_move_cost): Update for CC registers.
4924 (pdp11_rtx_costs): Add case for LSHIFTRT.
4925 (pdp11_output_jump): Add CCNZ mode conditional branches.
4926 (notice_update_cc_on_set): Remove.
4927 (pdp11_cc_mode): New function.
4928 (simple_memory_operand): Correct pre/post decrement case.
4929 (no_side_effect_operand): New function.
4930 (pdp11_regno_reg_class): Add CC_REGS class.
4931 (pdp11_fixed_cc_regs): New function.
4932 (pdp11_small_shift): New function.
4933 (pdp11_expand_shift): New function to expand shift insns.
4934 (pdp11_assemble_shift): New function to output shifts.
4935 (pdp11_branch_cost): Remove.
4936 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
4937 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
4938 (WCHAR_TYPE): Ditto.
4939 (PTRDIFF_TYPE): Ditto.
4940 (ADJUST_INSN_LENGTH): New macro.
4941 (FIXED_REGISTERS): Add CC registers.
4942 (CALL_USED_REGISTERS): Ditto.
4944 (REG_CLASS_NAMES): Ditto.
4945 (REG_CLASS_CONTENTS): Ditto.
4946 (SELECT_CC_MODE): Use new function.
4947 (TARGET_FLAGS_REGNUM): New macro.
4948 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
4949 (cc0_reg_rtx): Remove.
4950 (CC_STATUS_MDEP): Remove.
4951 (CC_STATUS_MDEFP_INIT): Remove.
4952 (CC_IN_FPU): Remove.
4953 (NOTICE_UPDATE_CC): Remove.
4954 (REGISTER_NAMES): Add CC registers.
4955 (BRANCH_COST): Change to constant 1.
4956 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
4958 * config/pdp11/pdp11.opt (mbcopy): Remove.
4959 (mbcopy-builtin): Remove.
4960 (mbranch-cheap): Remove.
4961 (mbranch-expensive): Remove.
4962 * config/pdp11/predicates.md (expand_shift_operand): Update to
4963 match shift code generation.
4964 (ccnz_operator): New predicate.
4965 * doc/invoke.texi (PDP-11 Options): Remove deleted options
4966 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
4967 Remove non-existent option -mabshi, -mno-abshi. Document mutually
4969 * doc/md.texi (PDP-11): Document new D and h constraints. Update
4970 description of O constraint.
4972 2018-06-27 Jeff Law <law@redhat.com>
4973 Austin Law <austinklaw@gmail.com>
4975 * config/v850/v850.md (addsi3_set_flags): New pattern.
4976 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
4977 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
4978 (zero_extendhisi2_v850_set_flags): Likewise.
4979 (zero_extendqisi2_v850_set_flags): Likewise.
4980 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
4981 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
4982 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
4984 * config/v850/v850-protos.h (notice_update_cc): Remove.
4985 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
4986 (v850_print_operand): Handle 'D' and "d".
4987 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
4988 Add handling of arithmetic/logical operations compared against zero.
4989 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
4990 Do not look at v850_compare_op, instead get mode from last argument.
4991 (v850_gen_compare): Remove
4992 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
4993 after reload for prologue insns.
4994 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
4996 (construct_save_jarl): Likewise.
4997 (TARGET_FLAGS_REGNUM): Define.
4998 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
4999 (NOTICE_UPDATE_CC): Remove.
5000 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
5001 than cc0. Conditionalize on reload_completed.
5002 (cmpsi_insn, setfcc_insn): Likewise.
5003 (tst1 splitter): Turn into define_and_split which sets the flags
5005 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
5006 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
5007 (cstoresf4, cstoredf4): Clobber the flags.
5008 (cmpsi, cmpsf, cmpdf): Remove expanders.
5009 (setf_insn): Remove pattern.
5010 (addsi3): Turn into define_and_split which clobbers the flags after
5011 reload and a suitable pattern (addsi3_clobber_flags) for use after
5013 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
5014 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
5015 (ashrsi3, ashrsi3_v850e2): Likewise.
5016 (bins): Clobber the flags.
5017 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
5018 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
5019 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
5020 (call_value_internal_short, call_value_internal_long): Likewise.
5021 (callt_save_interrupt, callt_return_interrupt): Likewise.
5022 (save_interrupt, return_interrupt): Likewise.
5023 (callt_save_all_interrupt, save_all_interrupt): Likewise.
5024 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
5025 (restore_all_interrupt, _restore_all_interrupt): Likewise.
5026 (All FP comparisons): Only allow after reload has completed.
5028 (divh, divhu): Tweak output template.
5029 (branch_z_normal, branch_z_invert): Remove
5030 (branch_nz_normal, branch_nz_invert): Likewise.
5031 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
5033 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
5034 * config/v850/v850.c (notice_update_cc): Remove.
5035 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
5036 (CC_NO_CARRY): Likewise.
5037 (NOTICE_UPDATE_CC): Define to nothing.
5038 * config/v850/v850.md: Remove block comment on cc0 handling
5039 Remove "cc" attribute from all patterns. Remove cc_status handling
5040 from all patterns. Minor formatting fixes.
5042 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5044 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
5045 (cortex-a76.cortex-a55): Likewise.
5046 * config/aarch64/aarch64-tune.md: Regenerate.
5047 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
5048 cortex-a76.cortex-a55.
5050 2018-06-27 Jeff Law <law@redhat.com>
5052 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
5053 (MULTILIB_DIRNAMES): Similarly.
5055 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
5057 * gimple.h (gimple_return_retbnd): Delete.
5058 (gimple_return_set_retbnd): Likewise.
5059 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
5060 gimple_return_set_retbnd.
5061 * gimple-pretty-print.c (dump_gimple_return): Remove call to
5062 gimple_return_retbnd and adjust.
5063 * tree-inline.h (struct copy_body_data): Remove retbnd field.
5064 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
5065 Explicitly return NULL in a couple more cases. Move assertion
5066 on debug statements and remove unreachable code.
5067 (reset_debug_binding): Do not test id->retbnd.
5068 (expand_call_inline): Do not set it.
5070 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
5072 * configure.ac: Add --disable-gcov option.
5073 * configure: Regenerate.
5074 * Makefile.in: Honour @enable_gcov@.
5075 * doc/install.texi: Document --disable-gcov.
5077 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5079 * config/arm/arm-cpus.in (cortex-a76): New entry.
5080 (cortex-a76.cortex-a55): Likewise.
5081 * config/arm/arm-tables.opt: Regenerate.
5082 * config/arm/arm-tune.md: Likewise.
5083 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
5084 * doc/invoke.texi (ARM Options): Document cortex-a76 and
5085 cortex-a76.cortex-a55.
5087 2018-06-27 Tamar Christina <tamar.christina@arm.com>
5090 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
5092 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
5094 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
5096 (EPILOGUE_USES): Likewise.
5098 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
5100 * tree-inline.c (remap_location): New function extracted from...
5101 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
5102 (copy_phis_for_bb): ...here. Call remap_location.
5103 (copy_cfg_body): Adjust call to copy_edges_for_bb.
5105 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
5107 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
5108 unaligned vsx for 16B memset.
5110 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
5113 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
5114 ieee128_float_type_node to long_double_type_node unless
5115 TARGET_LONG_DOUBLE_128 is set.
5117 2018-06-26 David Malcolm <dmalcolm@redhat.com>
5119 * cfgloop.c (get_loop_location): Convert return type from
5120 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
5121 by implicit construction from rtx_insn *, and using
5122 dump_user_location_t::from_function_decl for the fallback case.
5123 * cfgloop.h (get_loop_location): Convert return type from
5124 location_t to dump_user_location_t.
5125 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
5126 dump_printf_loc to pass in a dump_location_t rather than a
5127 location_t, via the gimple stmt.
5128 * coverage.c (get_coverage_counts): Update calls to
5129 dump_printf_loc to pass in dump_location_t rather than a
5131 * doc/optinfo.texi (Dump types): Convert example of
5132 dump_printf_loc from taking "locus" to taking "insn". Update
5133 description of the "_loc" calls to cover dump_location_t.
5134 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
5136 (dump_user_location_t::dump_user_location_t): New constructors,
5137 from gimple *stmt and rtx_insn *.
5138 (dump_user_location_t::from_function_decl): New function.
5139 (dump_loc): Make static.
5140 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
5141 const dump_location_t &.
5142 (dump_generic_expr_loc): Delete.
5143 (dump_printf_loc): Convert param "loc" from location_t to
5144 const dump_location_t &.
5145 (selftest::test_impl_location): New function.
5146 (selftest::dumpfile_c_tests): New function.
5147 * dumpfile.h: Include "profile-count.h".
5148 (class dump_user_location_t): New class.
5149 (struct dump_impl_location_t): New struct.
5150 (class dump_location_t): New class.
5151 (dump_printf_loc): Convert 2nd param from source_location to
5152 const dump_location_t &.
5153 (dump_generic_expr_loc): Delete.
5154 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
5155 const dump_location_t &.
5156 * gimple-fold.c (fold_gimple_assign): Update call to
5157 dump_printf_loc to pass in a dump_location_t rather than a
5158 location_t, via the gimple stmt.
5159 (gimple_fold_call): Likewise.
5160 * gimple-loop-interchange.cc
5161 (loop_cand::analyze_iloop_reduction_var): Update for change to
5162 check_reduction_path.
5163 (tree_loop_interchange::interchange): Update for change to
5165 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
5166 change in return-type of find_loop_location.
5167 (graphite_regenerate_ast_isl): Likewise.
5168 * graphite-optimize-isl.c (optimize_isl): Likewise.
5169 * graphite.c (graphite_transform_loops): Likewise.
5170 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
5171 pass in a dump_location_t rather than a location_t, via the
5173 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
5174 * ipa.c (walk_polymorphic_call_targets): Likewise.
5175 * loop-unroll.c (report_unroll): Convert "locus" param from
5176 location_t to dump_location_t.
5177 (decide_unrolling): Update for change to get_loop_location's
5179 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
5180 location_t to dump_user_location_t.
5181 (grid_find_single_omp_among_assignments_1): Updates calls to
5182 dump_printf_loc to pass in a dump_location_t rather than a
5183 location_t, via the gimple stmt.
5184 (grid_parallel_clauses_gridifiable): Convert "tloc" from
5185 location_t to dump_location_t. Updates calls to dump_printf_loc
5186 to pass in a dump_location_t rather than a location_t, via the
5188 (grid_inner_loop_gridifiable_p): Likewise.
5189 (grid_dist_follows_simple_pattern): Likewise.
5190 (grid_gfor_follows_tiling_pattern): Likewise.
5191 (grid_target_follows_gridifiable_pattern): Likewise.
5192 (grid_attempt_target_gridification): Convert initialization
5193 of local "grid" from memset to zero-initialization; FIXME: does
5194 this require C++11? Update call to dump_printf_loc to pass in a
5195 optinfo_location rather than a location_t, via the gimple stmt.
5196 * profile.c (read_profile_edge_counts): Updates call to
5197 dump_printf_loc to pass in a dump_location_t rather than a
5199 (compute_branch_probabilities): Likewise.
5200 * selftest-run-tests.c (selftest::run_tests): Call
5202 * selftest.h (dumpfile_c_tests): New decl.
5203 * tree-loop-distribution.c (pass_loop_distribution::execute):
5204 Update for change in return type of find_loop_location.
5205 * tree-parloops.c (parallelize_loops): Likewise.
5206 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
5207 "locus" from location_t to dump_user_location_t.
5208 (canonicalize_loop_induction_variables): Likewise.
5209 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
5210 for change in return type of find_loop_location.
5211 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
5212 to dump_printf_loc to pass in a dump_location_t rather than a
5213 location_t, via the stmt.
5214 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
5216 * tree-vect-loop-manip.c (find_loop_location): Convert return
5217 type from source_location to dump_user_location_t.
5218 (vect_do_peeling): Update for above change.
5219 (vect_loop_versioning): Update for change in type of
5221 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
5222 from location_t to dump_user_location_t.
5223 (vect_estimate_min_profitable_iters): Update for change in type
5225 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
5226 location_t to dump_location_t.
5227 (vect_slp_bb): Update for change in type of vect_location.
5228 * tree-vectorizer.c (vect_location): Convert from source_location
5229 to dump_user_location_t.
5230 (try_vectorize_loop_1): Update for change in vect_location's type.
5231 (vectorize_loops): Likewise.
5232 (increase_alignment): Likewise.
5233 * tree-vectorizer.h (vect_location): Convert from source_location
5234 to dump_user_location_t.
5235 (find_loop_location): Convert return type from source_location to
5236 dump_user_location_t.
5237 (check_reduction_path): Convert 1st param from location_t to
5238 dump_user_location_t.
5239 * value-prof.c (check_counter): Update call to dump_printf_loc to
5240 pass in a dump_user_location_t rather than a location_t; update
5241 call to error_at for change in type of "locus".
5242 (check_ic_target): Update call to dump_printf_loc to
5243 pass in a dump_user_location_t rather than a location_t, via the
5246 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
5248 * config/s390/s390.h (enum processor_flags): Do not use
5249 default tune parameter when -march was specified.
5251 2018-06-26 Jakub Jelinek <jakub@redhat.com>
5254 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
5255 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
5258 2018-06-26 Richard Biener <rguenther@suse.de>
5260 PR tree-optimization/86287
5262 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
5263 (vect_analyze_loop): Initialize n_stmts.
5265 2018-06-26 Richard Biener <rguenther@suse.de>
5268 * fold-const.c (fold_convertible_p): Pointer extension
5271 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
5274 * dwarf2out.c (loc_list_has_views): Adjust comments.
5275 (dw_loc_list): Split single cross-partition range with
5278 2018-06-25 Jeff Law <law@redhat.com>
5280 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
5281 on -mbig-switch by default.
5283 * config/v850/predicates.md (const_float_1_operand): Fix match_code
5285 (const_float_0_operand): Remove unused predicate.
5286 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
5287 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
5288 (recipsf2): New expander. Original pattern now called
5290 (recipdf2, recipdf2_insn): Similarly.
5291 (rsqrtsf2, rsqrtsf2_insn): Similarly
5292 (rsqrtdf2, rsqrtdf2_insn): Similarly
5294 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
5296 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
5297 Simplify logic for FreeBSD (twice).
5299 2018-06-25 Martin Sebor <msebor@redhat.com>
5301 PR tree-optimization/86204
5302 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
5303 a strnlen result if it's less than the length of the string.
5305 2018-06-25 Martin Sebor <msebor@redhat.com>
5307 PR tree-optimization/85700
5308 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
5309 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
5310 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
5312 2018-06-25 Martin Sebor <msebor@redhat.com>
5314 * doc/extend.texi (Zero-length arrays): Update and clarify.
5316 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
5318 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
5319 added IEEE/IBM long double multilib support on PowerPC little
5320 endian Linux systems.
5321 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
5322 (MULTILIB_DEFAULTS): Likewise.
5323 * config/rs6000/rs6000.c (rs6000_option_override_internal):
5325 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
5326 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
5327 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
5329 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
5332 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
5333 (REORDER_45): Likewise.
5335 2018-06-25 Jeff Law <law@redhat.com>
5337 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
5338 dividend to 32 bits. Adjust length.
5339 (udivmodhi4): Cleanup output template. Fix length.
5341 2018-06-25 Carl Love <cel@us.ibm.com>
5343 * config/rs6000/vsx.md: Change word selector to prefered location.
5345 2018-06-25 Richard Biener <rguenther@suse.de>
5347 PR tree-optimization/86304
5348 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
5349 epilogue-if-converted loops as well.
5351 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
5353 * lto-section-out.c (lto_begin_section): Do not print section
5354 name for noaddr and unnumbered dumps.
5356 2018-06-25 Richard Biener <rguenther@suse.de>
5358 * tree-vectorizer.h (struct vec_info_shared): New structure
5359 with parts split out from struct vec_info and loop_nest from
5360 struct _loop_vec_info.
5361 (struct vec_info): Adjust accordingly.
5362 (struct _loop_vec_info): Likewise.
5363 (LOOP_VINFO_LOOP_NEST): Adjust.
5364 (LOOP_VINFO_DATAREFS): Likewise.
5365 (LOOP_VINFO_DDRS): Likewise.
5366 (struct _bb_vec_info): Likewise.
5367 (BB_VINFO_DATAREFS): Likewise.
5368 (BB_VINFO_DDRS): Likewise.
5369 (struct _stmt_vec_info): Add dr_aux member.
5370 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
5371 (DR_MISALIGNMENT_UNINITIALIZED): New.
5372 (set_dr_misalignment): Adjust.
5373 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
5374 (vect_analyze_loop): Adjust prototype.
5375 (vect_analyze_loop_form): Likewise.
5376 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
5377 Compute dependences lazily.
5378 (vect_record_base_alignments): Use shared datarefs/ddrs.
5379 (vect_verify_datarefs_alignment): Likewise.
5380 (vect_analyze_data_refs_alignment): Likewise.
5381 (vect_analyze_data_ref_accesses): Likewise.
5382 (vect_analyze_data_refs): Likewise.
5383 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
5384 constructor parameter for shared part.
5385 (vect_analyze_loop_form): Pass in shared part and adjust.
5386 (vect_analyze_loop_2): Pass in storage for the number of
5387 stmts. Move loop nest finding to the caller. Compute
5389 (vect_analyze_loop): Pass in shared part.
5390 (vect_transform_loop): Verify shared datarefs are unchanged.
5391 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
5392 constructor parameter for shared part.
5393 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
5394 (vect_slp_bb): Verify shared datarefs are unchanged before
5396 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
5398 (new_stmt_vec_info): Initialize DR_AUX misalignment to
5399 DR_MISALIGNMENT_UNINITIALIZED.
5400 * tree-vectorizer.c (vec_info::vec_info): Add constructor
5401 parameter for shared part.
5402 (vec_info::~vec_info): Adjust.
5403 (vec_info_shared::vec_info_shared): New.
5404 (vec_info_shared::~vec_info_shared): Likewise.
5405 (vec_info_shared::save_datarefs): Likewise.
5406 (vec_info_shared::check_datarefs): Likewise.
5407 (try_vectorize_loop_1): Construct shared part live for analyses
5408 of a single loop for multiple vector sizes.
5409 * tree-parloops.c (gather_scalar_reductions): Adjust.
5411 2018-06-25 Richard Biener <rguenther@suse.de>
5413 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
5414 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
5415 (vect_analyze_data_refs): Remove similar code from here and
5416 simplify accordingly.
5418 2018-06-25 Richard Biener <rguenther@suse.de>
5420 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
5421 for reverse storage order accesses rather than asserting
5422 they cannot happen here.
5424 2018-06-25 Tom de Vries <tdevries@suse.de>
5427 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
5428 Use data16 instead of .byte for insn prefix.
5430 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
5433 * parser.c (make_char_string_pack): Pass this literal chars
5434 through cpp_interpret_string.
5435 (cp_parser_userdef_numeric_literal): Check the result of
5436 make_char_string_pack.
5438 2018-06-24 Maya Rashish <coypu@sdf.org>
5440 * ginclude/stddef.h: Simplify conditions around avoiding
5441 re-definition of __size_t.
5443 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
5445 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
5446 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
5448 2018-06-22 Maya Rashish <coypu@sdf.org>
5450 * doc/invoke.texi (mno-fancy-math-387): Update for changes
5451 made to OpenBSD and NetBSD through the years.
5453 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
5455 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
5456 behavior of vec_pack (vector double, vector double) to match
5457 behavior of vec_float2 (vector double, vector double).
5459 2018-06-22 Olivier Hainque <hainque@adacore.com>
5461 * gimplify.c (gimplify_function_tree): Prevent creation
5462 of a trampoline for the address of the current function
5463 passed to entry/exit instrumentation hooks.
5465 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
5468 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
5471 2018-06-22 Martin Liska <mliska@suse.cz>
5473 PR tree-optimization/86263
5474 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
5475 Bail out if is_enabled is false.
5476 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
5478 (jump_table_cluster::is_enabled): New function.
5480 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
5482 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
5483 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
5484 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
5485 (lto_input_ts_binfo_tree_pointers): Likewise.
5486 * tree-streamer-out.c (streamer_write_tree_bitfields,
5487 write_ts_binfo_tree_pointers): Likewise.
5488 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
5490 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
5492 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
5494 2018-06-22 Martin Liska <mliska@suse.cz>
5496 * symbol-summary.h (get): Make it pure and inline move
5497 functionality from ::get function.
5498 (get): Remove and inline into ::get and ::get_create.
5499 (get_create): Move code from ::get function.
5501 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5504 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
5505 -x assembler-with-cpp.
5507 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5509 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
5510 _FILE_OFFSET_BITS=64 for C++.
5512 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
5514 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
5515 conversion insn that shows up when pr85657-3.c is compiled using
5516 IEEE 128-bit long double.
5517 (neg<mode>2_internal): Use the correct mode to check whether the
5518 mode is IBM extended.
5519 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
5520 multiply and divide external functions from being created more
5523 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
5525 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
5527 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
5528 the edge can be forwarded.
5529 (cfg_layout_merge_blocks): Likewise.
5531 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
5533 * except.c (finish_eh_generation): Commit edge insertions only after
5534 the EH edges have been redirected from post-landing to landing pads.
5536 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
5538 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
5539 create_tmp_var_for to create the FRAME decl.
5540 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
5542 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
5544 * tree-inline.c (copy_edges_for_bb): Minor tweak.
5545 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
5546 debug statement when resetting its value.
5547 (expand_call_inline): Copy the locus of the call onto the assignment
5548 of the return value, if any. Use local variable in more cases.
5550 2018-06-21 Martin Liska <mliska@suse.cz>
5552 * ipa-pure-const.c (propagate_nothrow): Use
5553 funct_state_summaries->get.
5554 (dump_malloc_lattice): Likewise.
5555 (propagate_malloc): Likewise.
5557 2018-06-21 Richard Biener <rguenther@suse.de>
5559 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
5560 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
5561 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
5563 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
5564 BLOCK_ABSTRACT_ORIGIN unconditionally.
5566 2018-06-21 David Malcolm <dmalcolm@redhat.com>
5568 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
5570 * ipa-reference.c (ipa_reference_c_finalize): Delete
5571 ipa_ref_opt_sum_summaries and set it to NULL.
5573 2018-06-21 Tom de Vries <tdevries@suse.de>
5575 PR tree-optimization/85859
5576 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
5577 test with comment from bb_no_side_effects_p.
5579 2018-06-21 Richard Biener <rguenther@suse.de>
5581 PR tree-optimization/86232
5582 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
5583 max for constant niter.
5585 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
5587 * config/aarch64/aarch64-simd.md
5588 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
5590 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
5592 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
5593 Make opernads of the unspec commutative.
5595 2018-06-21 Richard Biener <rguenther@suse.de>
5597 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
5598 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
5599 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
5600 (vect_analyze_data_ref_dependence): Re-order checks to deal with
5602 (vect_record_base_alignments): Do not record base alignment
5603 for gathers or scatters.
5604 (vect_compute_data_ref_alignment): Drop return value that is always
5605 true. Bail out early for gathers or scatters.
5606 (vect_enhance_data_refs_alignment): Bail out early for gathers
5608 (vect_find_same_alignment_drs): Likewise.
5609 (vect_analyze_data_refs_alignment): Remove dead code.
5610 (vect_slp_analyze_and_verify_node_alignment): Likewise.
5611 (vect_analyze_data_refs): For possible gathers or scatters do
5612 not create an alternate DR, just check their possible validity
5613 and mark them. Adjust DECL_NONALIASED handling to not rely
5615 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
5616 update inits of gathers or scatters.
5617 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
5618 Also copy gather/scatter flag to pattern vinfo.
5620 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
5622 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
5623 behavior of vec_packsu (vector unsigned long long, vector unsigned
5624 long long) to match behavior of vec_packs with same signature.
5626 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
5627 Thomas Schwinge <thomas@codesourcery.com>
5628 Cesar Philippidis <cesar@codesourcery.com>
5630 * gimplify.c (gimplify_scan_omp_clauses): Add support for
5631 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
5632 (gimplify_adjust_omp_clauses): Likewise.
5633 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
5634 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
5635 (gimplify_omp_target_update): Update handling of acc update and
5637 * omp-low.c (install_var_field): Remove unused parameter
5638 base_pointers_restrict.
5639 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
5640 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
5642 (omp_target_base_pointers_restrict_p): Delete.
5643 (scan_omp_target): Update call to scan_sharing_clauses.
5644 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
5646 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
5647 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
5648 (convert_local_omp_clauses): Likewise.
5649 * tree-pretty-print.c (dump_omp_clause): Likewise.
5650 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
5652 (omp_clause_code_name): Likewise.
5654 2018-06-20 Jakub Jelinek <jakub@redhat.com>
5657 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
5660 PR tree-optimization/86231
5661 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
5662 anti-range don't overwrite *vr0min before using it to compute *vr0max.
5664 2018-06-20 Tom de Vries <tdevries@suse.de>
5666 PR tree-optimization/86097
5667 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
5668 iv type if signedness of iv type is not the same as that of *nit.
5670 2018-06-20 Jakub Jelinek <jakub@redhat.com>
5672 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
5673 EDGE_EH edges, verify they are all EDGE_EH.
5675 2018-06-20 Maya Rashish <coypu@sdf.org>
5677 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
5679 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5681 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
5682 * config/aarch64/aarch64.c (xgene1_tunings): Add
5683 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
5684 (aarch64_mode_valid_for_sched_fusion_p):
5685 Allow 16-byte modes.
5686 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
5687 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
5689 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
5691 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
5692 * config/aarch64/iterators.md (VQ2): New mode iterator.
5694 2018-06-20 Martin Liska <mliska@suse.cz>
5696 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
5697 Change default ratio from 10 to 8.
5699 2018-06-20 Martin Liska <mliska@suse.cz>
5701 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
5703 (bit_test_cluster::find_bit_tests): Likewise.
5704 (switch_decision_tree::analyze_switch_statement): Find clusters.
5705 * tree-switch-conversion.h (struct jump_table_cluster): Document
5708 2018-06-20 Martin Liska <mliska@suse.cz>
5710 * tree-switch-conversion.c (switch_conversion::collect):
5711 Record m_uniq property.
5712 (switch_conversion::expand): Bail out for special conditions.
5713 (group_cluster::~group_cluster): New.
5714 (group_cluster::group_cluster): Likewise.
5715 (group_cluster::dump): Likewise.
5716 (jump_table_cluster::emit): New.
5717 (switch_decision_tree::fix_phi_operands_for_edges): New.
5718 (struct case_node): Remove struct.
5719 (jump_table_cluster::can_be_handled): New.
5720 (case_values_threshold): Moved to header.
5721 (reset_out_edges_aux): Likewise.
5722 (jump_table_cluster::is_beneficial): New.
5723 (bit_test_cluster::can_be_handled): Likewise.
5724 (add_case_node): Remove.
5725 (bit_test_cluster::is_beneficial): New.
5726 (case_bit_test::cmp): New.
5727 (bit_test_cluster::emit): New.
5728 (expand_switch_as_decision_tree_p): Remove.
5729 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
5730 (fix_phi_operands_for_edge): Likewise.
5731 (switch_decision_tree::analyze_switch_statement): New.
5732 (compute_cases_per_edge): Move ...
5733 (switch_decision_tree::compute_cases_per_edge): ... here.
5734 (try_switch_expansion): Likewise.
5735 (switch_decision_tree::try_switch_expansion): Likewise.
5736 (record_phi_operand_mapping): Likewise.
5737 (switch_decision_tree::record_phi_operand_mapping): Likewise.
5738 (emit_case_decision_tree): Likewise.
5739 (switch_decision_tree::emit): Likewise.
5740 (balance_case_nodes): Likewise.
5741 (switch_decision_tree::balance_case_nodes): Likewise.
5742 (dump_case_nodes): Likewise.
5743 (switch_decision_tree::dump_case_nodes): Likewise.
5744 (emit_jump): Likewise.
5745 (switch_decision_tree::emit_jump): Likewise.
5746 (emit_cmp_and_jump_insns): Likewise.
5747 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
5748 (emit_case_nodes): Likewise.
5749 (switch_decision_tree::emit_case_nodes): Likewise.
5750 (conditional_probability): Remove.
5751 * tree-switch-conversion.h (enum cluster_type): New.
5753 (struct cluster): Likewise.
5754 (cluster::cluster): Likewise.
5755 (struct simple_cluster): Likewise.
5756 (simple_cluster::simple_cluster): Likewise.
5757 (struct group_cluster): Likewise.
5758 (struct jump_table_cluster): Likewise.
5759 (struct bit_test_cluster): Likewise.
5760 (struct min_cluster_item): Likewise.
5761 (struct case_tree_node): Likewise.
5762 (case_tree_node::case_tree_node): Likewise.
5763 (jump_table_cluster::case_values_threshold): Likewise.
5764 (struct case_bit_test): Likewise.
5765 (struct switch_decision_tree): Likewise.
5766 (struct switch_conversion): Likewise.
5767 (switch_decision_tree::reset_out_edges_aux): Likewise.
5769 2018-06-20 Martin Liska <mliska@suse.cz>
5771 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
5772 (hoist_edge_and_branch_if_true): Likewise.
5773 (expand_switch_using_bit_tests_p): Likewise.
5774 (struct case_bit_test): Likewise.
5775 (case_bit_test_cmp): Likewise.
5776 (emit_case_bit_tests): Likewise.
5777 (switch_conversion::switch_conversion): New class.
5778 (struct switch_conv_info): Remove old struct.
5779 (collect_switch_conv_info): More to ...
5780 (switch_conversion::collect): ... this.
5781 (check_range): Likewise.
5782 (switch_conversion::check_range): Likewise.
5783 (check_all_empty_except_final): Likewise.
5784 (switch_conversion::check_all_empty_except_final): Likewise.
5785 (check_final_bb): Likewise.
5786 (switch_conversion::check_final_bb): Likewise.
5787 (create_temp_arrays): Likewise.
5788 (switch_conversion::create_temp_arrays): Likewise.
5789 (free_temp_arrays): Likewise.
5790 (gather_default_values): Likewise.
5791 (switch_conversion::gather_default_values): Likewise.
5792 (build_constructors): Likewise.
5793 (switch_conversion::build_constructors): Likewise.
5794 (constructor_contains_same_values_p): Likewise.
5795 (switch_conversion::contains_same_values_p): Likewise.
5796 (array_value_type): Likewise.
5797 (switch_conversion::array_value_type): Likewise.
5798 (build_one_array): Likewise.
5799 (switch_conversion::build_one_array): Likewise.
5800 (build_arrays): Likewise.
5801 (switch_conversion::build_arrays): Likewise.
5802 (gen_def_assigns): Likewise.
5803 (switch_conversion::gen_def_assigns): Likewise.
5804 (prune_bbs): Likewise.
5805 (switch_conversion::prune_bbs): Likewise.
5806 (fix_phi_nodes): Likewise.
5807 (switch_conversion::fix_phi_nodes): Likewise.
5808 (gen_inbound_check): Likewise.
5809 (switch_conversion::gen_inbound_check): Likewise.
5810 (process_switch): Use the newly created class.
5811 (switch_conversion::expand): New.
5812 (switch_conversion::~switch_conversion): New.
5813 * tree-switch-conversion.h: New file.
5815 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
5817 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
5818 tree-vect-patterns.c.
5819 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
5820 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
5821 (vect_recog_sad_pattern): Likewise.
5822 (vect_recog_widen_sum_pattern): Likewise.
5823 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
5824 (vect_recog_widen_shift_pattern): Remove the type_in argument.
5825 (vect_recog_rotate_pattern): Likewise.
5826 (vect_recog_mult_pattern): Likewise.
5827 (vect_recog_vector_vector_shift_pattern): Likewise.
5828 (vect_recog_divmod_pattern): Likewise.
5829 (vect_recog_mixed_size_cond_pattern): Likewise.
5830 (vect_recog_bool_pattern): Likewise.
5831 (vect_recog_mask_conversion_pattern): Likewise.
5832 (vect_try_gather_scatter_pattern): Likewise.
5833 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
5834 (vect_recog_over_widening_pattern): Likewise.
5835 (vect_recog_gather_scatter_pattern): Likewise.
5836 (vect_recog_func_ptr): Move from tree-vectorizer.h
5837 (vect_vect_recog_func_ptrs): Move further down the file.
5838 (vect_recog_func): Likewise. Remove the third argument.
5839 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
5840 (vect_pattern_recog_1): Expect the pattern function to do any
5841 necessary target tests. Also expect it to provide a vector type.
5842 Remove the type_in handling.
5844 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
5846 * tree-vect-patterns.c (vect_pattern_detected): New function.
5847 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
5848 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
5849 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
5850 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
5851 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
5852 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
5853 (vect_recog_mask_conversion_pattern)
5854 (vect_try_gather_scatter_pattern): Likewise.
5856 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
5858 * tree-vect-patterns.c (vect_get_internal_def): New function.
5859 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
5860 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
5861 (search_type_for_mask_1): Use it.
5863 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
5865 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
5866 redundant WIDEN_SUM_EXPR handling.
5867 (vect_recog_sad_pattern): Likewise.
5869 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
5871 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
5872 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
5873 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
5874 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
5875 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
5877 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
5879 * tree-vect-stmts.c (vectorizable_call): Make sure that we
5880 use the stmt_vec_info of the original bb statement for the
5881 new zero assignment, even if the call is part of a pattern.
5883 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
5885 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
5886 that the sequence is attached to the original statement rather
5887 than the pattern statement.
5888 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
5889 PATTERN_DEF_SEQ from the original statement rather than
5890 the main pattern statement.
5891 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
5892 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
5893 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
5895 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
5897 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
5898 definition statements before the early exit for statements that aren't
5900 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
5902 (vect_transform_loop): ...here. Process pattern definition
5903 statements without first checking whether the main pattern
5904 statement is live or relevant.
5906 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
5908 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
5909 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
5911 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
5913 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
5914 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
5915 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
5916 (expand_block_compare): Change select_block_compare_mode call.
5917 (expand_strncmp_align_check): Use new functions, fix comment.
5918 (emit_final_str_compare_gpr): New function.
5919 (expand_strn_compare): Refactor and clean up code.
5920 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
5922 2018-06-19 Tony Reix <tony.reix@atos.com>
5923 Damien Bergamini <damien.bergamini@atos.com>
5924 David Edelsohn <dje.gcc@gmail.com>
5926 * collect2.c (static_obj): New variable.
5927 (static_libs): New variable.
5928 (is_in_list): Uncomment declaration.
5929 (main): Track AIX libraries linked statically.
5930 (is_in_list): Uncomment definition.
5931 (scan_prog_file): Don't add AIX shared libraries initializer
5932 to constructor list if linking statically.
5934 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
5936 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
5938 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
5940 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
5942 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
5945 2018-06-19 Martin Liska <mliska@suse.cz>
5947 * config/i386/i386.c (ix86_can_inline_p): Do not use
5948 ipa_fn_summaries::get_create.
5949 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
5951 (devirtualization_time_bonus): Likewise.
5952 (ipcp_propagate_stage): Likewise.
5953 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
5954 (edge_set_predicate): Likewise.
5955 (evaluate_conditions_for_known_args): Likewise.
5956 (evaluate_properties_for_edge): Likewise.
5957 (ipa_call_summary::reset): Tranform to ...
5958 (ipa_call_summary::~ipa_call_summary): ... this.
5959 (ipa_fn_summary::reset): Transform to ...
5960 (ipa_fn_summary::~ipa_fn_summary): ... this.
5961 (ipa_fn_summary_t::remove): Rename to ...
5962 (ipa_fn_summary_t::remove_callees): ... this.
5963 (ipa_fn_summary_t::duplicate): Use placement new
5964 instead of memory copy.
5965 (ipa_call_summary_t::duplicate): Likewise.
5966 (ipa_call_summary_t::remove): Remove.
5967 (dump_ipa_call_summary): Change get_create to get.
5968 (ipa_dump_fn_summary): Dump only when summary exists.
5969 (analyze_function_body): Use symbol_summary::get instead
5971 (compute_fn_summary): Likewise.
5972 (estimate_edge_devirt_benefit): Likewise.
5973 (estimate_edge_size_and_time): Likewise.
5974 (inline_update_callee_summaries): Likewise.
5975 (remap_edge_change_prob): Likewise.
5976 (remap_edge_summaries): Likewise.
5977 (ipa_merge_fn_summary_after_inlining): Likewise.
5978 (write_ipa_call_summary): Likewise.
5979 (ipa_fn_summary_write): Likewise.
5980 (ipa_free_fn_summary): Likewise.
5981 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
5982 (struct ipa_call_summary): Likewise.
5983 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
5985 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
5986 (estimate_size_after_inlining): Likewise.
5987 (estimate_growth): Likewise.
5988 (growth_likely_positive): Likewise.
5989 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
5990 (inline_call): Likewise.
5991 * ipa-inline.c (caller_growth_limits): Likewise.
5992 (can_inline_edge_p): Likewise.
5993 (can_inline_edge_by_limits_p): Likewise.
5994 (compute_uninlined_call_time): Likewise.
5995 (compute_inlined_call_time): Likewise.
5996 (want_inline_small_function_p): Likewise.
5997 (edge_badness): Likewise.
5998 (update_caller_keys): Likewise.
5999 (update_callee_keys): Likewise.
6000 (inline_small_functions): Likewise.
6001 (inline_to_all_callers_1): Likewise.
6002 (dump_overall_stats): Likewise.
6003 (early_inline_small_functions): Likewise.
6004 (early_inliner): Likewise.
6005 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
6006 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
6007 * ipa-pure-const.c (malloc_candidate_p): Likewise.
6008 * ipa-split.c (execute_split_functions): Likewise.
6009 * symbol-summary.h: Likewise.
6010 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
6012 2018-06-19 Richard Biener <rguenther@suse.de>
6014 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
6015 (vectorize_loops): ... here. Fix dbgcnt handling.
6016 (try_vectorize_loop): Wrap try_vectorize_loop_1.
6018 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
6021 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
6022 ieee128 argument takes up only one (vector) register, not two (floating
6025 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
6027 * gimplify.c (gimplify_init_constructor): Really never clear for an
6028 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
6030 2018-06-19 Richard Biener <rguenther@suse.de>
6032 PR tree-optimization/86179
6033 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
6034 after failed recognition.
6036 2018-06-18 Martin Sebor <msebor@redhat.com>
6039 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
6040 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
6041 Handle integer subtraction.
6042 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
6043 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
6045 2018-06-18 David Malcolm <dmalcolm@redhat.com>
6047 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
6048 param from rtx to rtx_insn *.
6049 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
6051 (frv_ifcvt_modify_insn): Likwise.
6052 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
6053 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
6054 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
6055 as_a <rtx_insn *> cast to local "unprotected_region" once
6056 it's been established that it's not NULL or pc_rtx.
6057 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
6058 param "sethi" from rtx to rtx_insn *.
6059 (nds32_group_float_insns): Likewise for param "insn".
6060 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
6062 (vax_output_int_subtract): Likewise.
6063 * config/vax/vax.c (vax_output_int_add): Likewise for param
6065 (vax_output_int_subtract): Likewise.
6066 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
6067 (emit_pattern_after): Likewise for param "after".
6068 (emit_insn_after): Likewise.
6069 (emit_jump_insn_after): Likewise.
6070 (emit_call_insn_after): Likewise.
6071 (emit_debug_insn_after): Likewise.
6072 (emit_pattern_before): Likewise for param "before".
6073 (emit_insn_before): Likewise.
6074 (emit_jump_insn_before): Likewise.
6075 * final.c (get_insn_template): Likewise for param "insn", removing
6077 * output.h (get_insn_template): Likewise for 2nd param.
6078 * rtl.h (emit_insn_before): Likewise.
6079 (emit_jump_insn_before): Likewise.
6080 (emit_debug_insn_before_noloc): Likewise.
6081 (emit_insn_after): Likewise.
6082 (emit_jump_insn_after): Likewise.
6083 (emit_call_insn_after): Likewise.
6084 (emit_debug_insn_after): Likewise.
6085 (set_insn_deleted): Likewise for param.
6087 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
6090 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
6091 floating point modes, so that IFmode is numerically greater than
6092 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
6093 to declare the ordering. This prevents IFmode from being
6094 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
6095 machine. Include rs6000-modes.h to share the fractional values
6096 between genmodes* and the rest of the compiler.
6100 * config/rs6000/rs6000-modes.h: New file.
6101 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
6102 meaning of rs6000_long_double_size so that 126..128 selects an
6103 appropriate 128-bit floating point type.
6104 (rs6000_option_override_internal): Likewise.
6105 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
6106 (TARGET_LONG_DOUBLE_128): Change the meaning of
6107 rs6000_long_double_size so that 126..128 selects an appropriate
6108 128-bit floating point type.
6109 (LONG_DOUBLE_TYPE_SIZE): Update comment.
6110 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
6111 source and destination to match the standard usage.
6112 (truncifkf2): Likewise.
6113 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
6114 ISA 2.07 to use an explicit clobber, instead of passing in a
6116 (copysign<mode>3_soft): Likewise.
6118 2018-06-18 David Malcolm <dmalcolm@redhat.com>
6120 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
6121 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
6122 (vect_slp_analyze_instance_dependence): Likewise.
6123 (vect_enhance_data_refs_alignment): Likewise.
6124 (vect_analyze_data_refs_alignment): Likewise.
6125 (vect_slp_analyze_and_verify_instance_alignment
6126 (vect_analyze_data_ref_accesses): Likewise.
6127 (vect_prune_runtime_alias_test_list): Likewise.
6128 (vect_analyze_data_refs): Likewise.
6129 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
6130 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
6131 (vect_analyze_scalar_cycles_1): Likewise.
6132 (vect_get_loop_niters): Likewise.
6133 (vect_analyze_loop_form_1): Likewise.
6134 (vect_update_vf_for_slp): Likewise.
6135 (vect_analyze_loop_operations): Likewise.
6136 (vect_analyze_loop): Likewise.
6137 (vectorizable_induction): Likewise.
6138 (vect_transform_loop): Likewise.
6139 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
6140 * tree-vect-slp.c (vect_analyze_slp): Likewise.
6141 (vect_make_slp_decision): Likewise.
6142 (vect_detect_hybrid_slp): Likewise.
6143 (vect_slp_analyze_operations): Likewise.
6144 (vect_slp_bb): Likewise.
6145 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
6146 (vectorizable_bswap): Likewise.
6147 (vectorizable_call): Likewise.
6148 (vectorizable_simd_clone_call): Likewise.
6149 (vectorizable_conversion): Likewise.
6150 (vectorizable_assignment): Likewise.
6151 (vectorizable_shift): Likewise.
6152 (vectorizable_operation): Likewise.
6153 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
6155 2018-06-18 Martin Sebor <msebor@redhat.com>
6157 PR tree-optimization/81384
6158 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
6159 * builtins.c (expand_builtin_strnlen): New function.
6160 (expand_builtin): Call it.
6161 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
6162 * builtins.def (BUILT_IN_STRNLEN): New.
6163 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
6164 Warn for bounds in excess of maximum object size.
6165 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
6166 single-value ranges. Handle strnlen.
6167 (handle_builtin_strlen): Handle strnlen.
6168 (strlen_check_and_optimize_stmt): Same.
6169 * doc/extend.texi (Other Builtins): Document strnlen.
6171 2018-06-18 Maya Rashish <coypu@sdf.org>
6173 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
6174 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
6175 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
6177 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
6179 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
6181 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6183 * tree.c (escaped_string::escape): Replace cast to char * by
6184 const_cast<char *> (unescaped).
6186 2018-06-18 Nick Clifton <nickc@redhat.com>
6189 * tree.c (escaped_string): New class. Converts an unescaped
6190 string into its escaped equivalent.
6191 (warn_deprecated_use): Use the new class to convert the
6192 deprecation message, if present.
6193 (test_escaped_strings): New self test.
6194 (test_c_tests): Add test_escaped_strings.
6195 * doc/extend.texi (deprecated): Add a note that the
6196 deprecation message is affected by the -fmessage-length
6197 option, and that control characters will be escaped.
6198 (#pragma GCC error): Document this pragma.
6199 (#pragma GCC warning): Likewise.
6200 * doc/invoke.texi (-fmessage-length): Document this option's
6201 effect on the #warning and #error preprocessor directives and
6202 the deprecated attribute.
6204 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
6206 * tree.c (decl_value_expr_lookup): Revert latest change.
6207 (decl_value_expr_insert): Likewise.
6209 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
6211 * gimplify.c (nonlocal_vlas): Delete.
6212 (nonlocal_vla_vars): Likewise.
6213 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
6215 (gimplify_body): Do not create and destroy nonlocal_vlas.
6216 * tree-nested.c: Include diagnostic.h.
6217 (use_pointer_in_frame): Tweak.
6218 (lookup_field_for_decl): Add assertion and declare the transformation.
6219 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
6220 internal error when the reference is in a wrong context. Do not
6221 create a debug decl by default.
6222 (note_nonlocal_block_vlas): Delete.
6223 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
6224 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
6225 create a debug decl by default.
6226 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
6227 call is in a wrong context.
6228 (fixup_vla_decls): New function.
6229 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
6230 debug variables were created.
6231 * tree.c (decl_value_expr_lookup): Add checking assertion.
6232 (decl_value_expr_insert): Likewise.
6234 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
6237 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
6238 * tree-scalar-evolution.c (interpret_expr): Likewise.
6239 (expression_expensive_p): Likewise.
6240 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
6241 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
6242 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
6243 (ssa_defined_by_minus_one_stmt_p): New.
6245 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
6248 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
6249 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
6250 * dojump.c (do_jump): Likewise.
6251 * expr.c (expand_expr_real_2): Check operand type's sign.
6252 * fold-const.c (const_unop): Handle ABSU_EXPR.
6253 (fold_abs_const): Likewise.
6254 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
6255 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
6256 (strip_sign_op_1): Likesise.
6257 * match.pd: Add new pattern to generate ABSU_EXPR.
6258 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
6259 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
6260 * tree-eh.c (operation_could_trap_helper_p): Likewise.
6261 * tree-inline.c (estimate_operator_cost): Likewise.
6262 * tree-pretty-print.c (dump_generic_node): Likewise.
6263 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
6264 * tree.def (ABSU_EXPR): New.
6266 2018-06-16 Jakub Jelinek <jakub@redhat.com>
6269 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
6270 documented as preserved for backward compatibility only.
6271 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
6273 PR rtl-optimization/86108
6274 * bb-reorder.c (create_forwarder_block): Renamed to ...
6275 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
6276 jump from new landing pad to the second part.
6277 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
6280 2018-06-15 Jakub Jelinek <jakub@redhat.com>
6283 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
6284 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
6285 Only call store_expr for halves if the mode is the same.
6288 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
6289 Fix up comment formatting.
6291 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
6293 * typed-splay-tree.h (typed_splay_tree::remove): New function.
6294 (typed_splay_tree::closure,
6295 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
6296 (typed_splay_tree::typed_splay_tree,
6297 typed_splay_tree::operator =): Declared private.
6298 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
6299 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
6300 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
6301 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
6302 typed_splay_tree::splay_tree_splay,
6303 typed_splay_tree::splay_tree_foreach_helper,
6304 typed_splay_tree::splay_tree_insert,
6305 typed_splay_tree::splay_tree_remove,
6306 typed_splay_tree::splay_tree_lookup,
6307 typed_splay_tree::splay_tree_predecessor,
6308 typed_splay_tree::splay_tree_successor,
6309 typed_splay_tree::splay_tree_min,
6310 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
6311 (typed_splay_tree::root, typed_splay_tree::comp,
6312 typed_splay_tree::delete_key,
6313 typed_splay_tree::delete_value): New data members.
6314 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
6315 typed_splay_tree::remove.
6317 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
6319 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
6320 -mginv and -mno-ginv to the assembler.
6321 * config/mips/mips.opt (-mcrc): New option.
6323 * doc/invoke.text (-mcrc): Document.
6326 2018-06-15 Nick Clifton <nickc@redhat.com>
6329 * tree.c (escaped_string): New class. Converts an unescaped
6330 string into its escaped equivalent.
6331 (warn_deprecated_use): Use the new class to convert the
6332 deprecation message, if present.
6333 (test_escaped_strings): New self test.
6334 (test_c_tests): Add test_escaped_strings.
6335 * doc/extend.texi (deprecated): Add a note that the
6336 deprecation message is affected by the -fmessage-length
6337 option, and that control characters will be escaped.
6338 (#pragma GCC error): Document this pragma.
6339 (#pragma GCC warning): Likewise.
6340 * doc/invoke.texi (-fmessage-length): Document this option's
6341 effect on the #warning and #error preprocessor directives and
6342 the deprecated attribute.
6344 2018-06-15 Richard Biener <rguenther@suse.de>
6346 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
6347 here, also noting vector size used.
6348 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
6349 size used in MSG_OPTIMIZED_LOCATIONS dump.
6350 (pass_slp_vectorize::execute): Adjust.
6352 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
6355 * config/arc/arc.c (arc_return_address_register): Fix
6358 2018-06-15 Richard Biener <rguenther@suse.de>
6361 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
6362 leave useless conversion stripping to force_gimple_operand_gsi.
6363 (gimplify_build2): Likewise.
6364 (gimplify_build1): Likewise.
6366 2018-06-15 Richard Biener <rguenther@suse.de>
6369 * tree-cfg.c (move_stmt_op): unshare invariant addresses
6370 before adjusting their block.
6372 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
6374 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
6375 multilibs for *-*-rtems*.
6376 * config/riscv/t-rtems: New file.
6378 2018-06-14 Jakub Jelinek <jakub@redhat.com>
6381 * match.pd ((A +- CST1) +- CST2): Punt if last resort
6382 unsigned_type_for returns NULL.
6385 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
6386 subregs of multi-word pseudos unless the float mode has word size.
6388 2018-06-14 Richard Biener <rguenther@suse.de>
6391 * tree-vect-generic.c (build_word_mode_vector_type): Remove
6392 duplicate and harmful type_hash_canon.
6393 * tree.c (type_hash_canon): Assert we didn't find ourselves.
6395 2018-06-14 Richard Biener <rguenther@suse.de>
6398 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
6401 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
6403 * config/rtems.h (STDINT_LONG32): Define.
6405 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
6406 Prachi Godbole <prachi.godbole@imgtec.com>
6408 * config/mips/mips-cpus.def: Define P6600.
6409 * config/mips/mips-tables.opt: Regenerate.
6410 * config/mips/mips.c (mips_ucbranch_type): New enum.
6411 (mips_rtx_cost_data): Add support for P6600.
6412 (mips_issue_rate): Likewise.
6413 (mips_multipass_dfa_lookahead): Likewise.
6414 (mips_avoid_hazard): Likewise.
6415 (mips_reorg_process_insns): Likewise.
6416 (mips_classify_branch_p6600): New function.
6417 * config/mips/mips.h (TUNE_P6600): New define.
6418 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
6419 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
6420 * config/mips/mips.md: Include p6600.md.
6421 (processor): Add p6600.
6422 * config/mips/p6600.md: New file.
6423 * doc/invoke.texi: Add p6600 to supported architectures.
6425 2018-06-13 Martin Sebor <msebor@redhat.com>
6427 PR tree-optimization/86114
6428 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
6430 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
6432 2018-06-13 Richard Biener <rguenther@suse.de>
6434 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
6435 Properly set vector type of the intermediate stmt.
6436 * tree-vect-stmts.c (vectorizable_operation): The destination
6437 var always has vectype_out type.
6439 2018-06-13 Jeff Law <law@redhat.com>
6441 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
6442 integer 0 for argument to print_rtl_with_bb.
6443 (rl78_reorg): Likewise.
6445 2018-06-13 David Malcolm <dmalcolm@redhat.com>
6447 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
6448 from rtx to rtx_insn *.
6449 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
6451 (add_sched_insns_for_speculation): Likewise for local "target",
6452 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
6453 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
6454 from rtx_insn ** to rtx_code_label **.
6455 (reorg_emit_nops): Likewise.
6456 (c6x_reorg): Likewise for local "call_labels".
6457 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
6459 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
6460 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
6461 the loops over LABEL_REFS.
6462 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
6464 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
6465 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
6466 (split_branches): Strengthen local "olabel" from rtx to
6467 rtx_insn *, adding a safe_as_a cast.
6468 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
6470 (add_insn_after): Likewise for first two params.
6471 (add_insn_before): Likewise.
6472 (remove_insn): Likewise for param.
6473 (emit_pattern_before_noloc): Likewise for second and third params.
6474 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
6475 (emit_call_insn_before_noloc): Likewise.
6476 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
6478 (emit_barrier_before): Likewise.
6479 (emit_label_before): Strengthen "label" param from "rtx" to
6480 "rtx_code_label *". Strengthen "before" param from "rtx" to
6482 (emit_insn_after_1): Strengthen "after" param from "rtx" to
6484 (emit_pattern_after_noloc): Likewise.
6485 (emit_insn_after_noloc): Likewise.
6486 (emit_jump_insn_after_noloc): Likewise.
6487 (emit_call_insn_after_noloc): Likewise.
6488 (emit_debug_insn_after_noloc): Likewise.
6489 (emit_barrier_after): Likewise.
6490 (emit_label_after): Likewise for both params.
6491 (emit_pattern_after_setloc): Likewise for "after" param. Convert
6492 "loc" param from "int" to "location_t".
6493 (emit_insn_after_setloc): Likewise.
6494 (emit_jump_insn_after_setloc): Likewise.
6495 (emit_call_insn_after_setloc): Likewise.
6496 (emit_debug_insn_after_setloc): Likewise.
6497 (emit_pattern_before_setloc): Likewise for "before" param. Convert
6498 "loc" param from "int" to "location_t".
6499 (emit_pattern_before): Convert NULL_RTX to NULL.
6500 (emit_insn_before_setloc): Convert "loc" param from "int" to
6502 (emit_jump_insn_before_setloc): Likewise.
6503 (emit_call_insn_before_setloc): Likewise.
6504 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
6505 rtx_insn *. Convert "loc" param from "int" to "location_t".
6506 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
6507 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
6508 Convert 3rd param from "int" to "location_t".
6509 (emit_barrier_before, emit_barrier_after, next_real_insn):
6510 Strengthen param from rtx to rtx_insn *.
6511 (emit_label_before): Strengthen 1st param from "rtx" to
6512 "rtx_code_label *". Strengthen 2nd param from "rtx" to
6514 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
6515 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
6516 Strengthen 2nd param from "rtx" to "rtx_insn *".
6517 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
6518 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
6519 Likewise. Convert 3rd param from "int" to "location_t".
6520 (emit_label_after): Strengthen 1st param from "rtx" to
6522 (next_real_insn, remove_insn): Strengthen param from "rtx" to
6524 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
6525 from "rtx" to "rtx_insn *".
6527 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
6529 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
6530 bodies streamed in with -Q.
6531 * dumpfile.c (dump_files): Add lto-stream-out dump file.
6532 * dumpfile.h (tree_dump_index): Add lto_stream_out.
6533 * gimple-streamer-out.c: Include gimple-pretty-print.h
6534 (output_bb): Dump stmts streamed.
6535 * lto-section-out.c: Include print-tree.h
6536 (lto_begin_section): Dump sections created.
6537 (lto_output_decl_index): Dump decl encoded.
6538 * lto-streamer-out.c: Include print-tree.h
6539 (create_output_block): Dump output block created.
6540 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
6541 (output_function): Dump function output.
6542 (output_constructor): Dump constructor streamed.
6543 (write_global_stream): Output indexes encoded.
6544 (produce_asm_for_decls): Dump streams encoded.
6545 * lto-streamer.c (streamer_dump_file): New global var.
6546 * lto-streamer.h (streamer_dump_file): Declare.
6547 * passes.c (ipa_write_summaries): Initialize streamer dump.
6548 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
6551 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
6554 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
6555 offsets for register save directives. Emit a second batch of save
6556 directives, if need be, when the function accesses prior frames.
6558 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
6560 * config/arc/fpu.md (fmasf4): Force operand to register.
6561 (fnmasf4): Likewise.
6563 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
6565 * config/arc/arc-protos.h (arc_pad_return): Remove.
6566 * config/arc/arc.c (machine_function): Remove force_short_suffix
6568 (arc_print_operand): Adjust printing of '&'.
6569 (arc_verify_short): Remove conditional printing of short suffix.
6570 (arc_final_prescan_insn): Remove reference to size_reason.
6571 (pad_return): New function.
6572 (arc_reorg): Call pad_return.
6573 (arc_pad_return): Remove.
6574 (arc_init_machine_status): Remove reference to force_short_suffix.
6575 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
6576 (attr length): When attribute iscompact is true force to 2
6577 regardless; in the case of maybe check if we want to force the
6578 instruction to have 4 bytes length.
6579 (nopv): Change it to generate 4 byte long nop as well.
6580 (blockage): New pattern.
6581 (simple_return): Remove call to arc_pad_return.
6582 (p_return_i): Likewise.
6584 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
6586 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
6588 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
6590 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
6593 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
6595 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
6596 for ARC700 and ARCv2.
6598 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
6601 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
6602 operands[2] instead of operands[1].
6605 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
6607 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
6608 case, check whether the outer register overlaps an unallocatable
6609 register, not just whether it fits the required class.
6611 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
6613 * poly-int.h (can_div_trunc_p): Add new overload in which all values
6615 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
6616 (memrefs_conflict_p): Likewise.
6617 (init_alias_analysis): Likewise.
6618 * cfgexpand.c (expand_debug_expr): Likewise.
6619 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
6620 * cse.c (fold_rtx): Likewise.
6621 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
6622 * expr.c (emit_block_move_hints): Likewise.
6623 (clear_storage_hints, push_block, emit_push_insn): Likewise.
6624 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
6625 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
6626 (emit_group_store): Likewise.
6627 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
6628 to read the PRE/POST_MODIFY increment.
6629 * calls.c (store_one_arg): Use strip_offset.
6630 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
6632 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
6634 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
6635 (simplify_binary_operation_1): Extend CONST_INT handling to
6637 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
6638 than a HOST_WIDE_INT.
6639 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
6641 (adjust_mems, add_stores): Update accodingly.
6642 (vt_canonicalize_addr): Track polynomial offsets.
6643 (emit_note_insn_var_location): Likewise.
6644 (vt_add_function_parameter): Likewise.
6645 (vt_initialize): Likewise.
6647 2018-06-12 Jeff Law <law@redhat.com>
6649 * config.gcc (alpha*-*-freebsd*): Remove.
6650 * config/alpha/freebsd.h: Remove.
6652 2018-06-12 David Malcolm <dmalcolm@redhat.com>
6655 * spellcheck-tree.c (levenshtein_distance): Rename to...
6656 (get_edit_distance): ...this, and update for underlying renaming.
6657 * spellcheck-tree.h (levenshtein_distance): Rename to...
6658 (get_edit_distance): ...this.
6659 * spellcheck.c (levenshtein_distance): Rename to...
6660 (get_edit_distance): ...this. Convert from Levenshtein distance
6661 to Damerau-Levenshtein distance by supporting transpositions of
6662 adjacent characters. Rename "v1" to "v_next" and "v0" to
6664 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
6665 (selftest::test_edit_distance_unit_test_oneway): ...this, and
6666 update for underlying renaming.
6667 (selftest::levenshtein_distance_unit_test): Rename to...
6668 (selftest::test_get_edit_distance_unit): ...this, and update for
6669 underlying renaming.
6670 (selftest::test_find_closest_string): Add example from PR 69968
6671 where transposition helps
6672 (selftest::test_metric_conditions): Update for renaming.
6673 (selftest::test_metric_conditions): Likewise.
6674 (selftest::spellcheck_c_tests): Likewise.
6675 * spellcheck.h (levenshtein_distance): Rename both overloads to...
6676 (get_edit_distance): ...this.
6677 (best_match::consider): Update for renaming.
6679 2018-06-12 Martin Sebor <msebor@redhat.com>
6681 PR tree-optimization/85259
6682 * builtins.c (compute_objsize): Handle constant offsets.
6683 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
6684 true iff a warning has been issued.
6685 * gimple.h (gimple_nonartificial_location): New function.
6686 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
6687 gimple_nonartificial_location and handle -Wno-system-headers.
6688 (handle_builtin_stxncpy): Same.
6690 2018-06-12 Martin Sebor <msebor@redhat.com>
6693 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
6695 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
6697 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
6698 BUILTIN_VEC_XST entries for pointer to double and long long.
6700 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
6703 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
6705 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
6708 2018-06-12 Martin Liska <mliska@suse.cz>
6710 * doc/options.texi: Document IntegerRange.
6712 2018-06-12 Martin Liska <mliska@suse.cz>
6714 * config/i386/i386.opt: Make MPX-related options as Deprecated.
6715 * opt-functions.awk: Handle Deprecated flag.
6716 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
6718 (read_cmdline_option): Report warning for a deprecated option.
6719 * opts.h (struct cl_option): Add new field cl_deprecated.
6720 (CL_ERR_DEPRECATED): New.
6722 2018-06-12 Martin Liska <mliska@suse.cz>
6724 * doc/options.texi: Document Deprecated option flag.
6726 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
6728 * config/arc/arc-arch.h (arc_extras): New enum.
6729 (arc_cpu_t):Add field extra.
6730 (arc_cpu_types): Consider the extras.
6731 * config/arc/arc-cpus.def: Add extras info.
6732 * config/arc/arc-opts.h (processor_type): Consider extra field.
6733 * config/arc/arc.c (arc_override_options): Handle extra field.
6735 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
6737 * config/arc/arc-arch.h: Update ARC_OPTX macro.
6738 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
6740 * config/arc/arc.c (arc_init): Update pic warning.
6741 (irq_range): Update irq range parsing warnings.
6742 (arc_override_options): Update various warning messages.
6743 (arc_handle_aux_attribute): Likewise.
6745 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
6747 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
6749 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
6751 * doc/sourcebuild.texi: Document usage of line number 0 in verify
6752 compiler messages directives.
6754 2018-06-12 Matthew Fortune <mfortune@gmail.com>
6756 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
6757 * config/mips/mips-tables.opt: Regenerate.
6758 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
6760 * doc/invoke.texi: Document -march=i6500.
6762 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
6764 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
6765 (i6400_gpmul): Add cpu_unit.
6766 (i6400_gpdiv): Likewise.
6767 (i6400_msa_add_d): Update reservations.
6768 (i6400_msa_int_add) Likewise.
6769 (i6400_msa_short_logic3) Likewise.
6770 (i6400_msa_short_logic2) Likewise.
6771 (i6400_msa_short_logic) Likewise.
6772 (i6400_msa_move) Likewise.
6773 (i6400_msa_cmp) Likewise.
6774 (i6400_msa_short_float2) Likewise.
6775 (i6400_msa_div_d) Likewise.
6776 (i6400_msa_long_logic1) Likewise.
6777 (i6400_msa_long_logic2) Likewise.
6778 (i6400_msa_mult) Likewise.
6779 (i6400_msa_long_float2) Likewise.
6780 (i6400_msa_long_float4) Likewise.
6781 (i6400_msa_long_float5) Likewise.
6782 (i6400_msa_long_float8) Likewise.
6783 (i6400_fpu_fadd): Include frint type.
6784 (i6400_fpu_store): New define_insn_reservation.
6785 (i6400_fpu_load): Likewise.
6786 (i6400_fpu_move): Likewise.
6787 (i6400_fpu_fcmp): Likewise.
6788 (i6400_fpu_fmadd): Likewise.
6789 (i6400_int_mult): Include imul3nc type and update reservation.
6790 (i6400_int_div): Include idiv3 type and update reservation.
6791 (i6400_int_load): Update to check type not move_type.
6792 (i6400_int_store): Likewise.
6793 (i6400_int_prefetch): Set zero latency.
6795 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
6797 * gcc.c: Document new %@{...} sequence.
6798 (LINK_COMMAND_SPEC): Use it for the -L switches.
6799 (cpp_unique_options): Use it for the -I switches.
6800 (at_file_argbuf): New global variable.
6801 (in_at_file): Likewise.
6802 (alloc_args): Create at_file_argbuf.
6803 (clear_args): Truncate at_file_argbuf.
6804 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
6805 (open_at_file): New function.
6806 (close_at_file): Likewise.
6807 (create_at_file): Delete.
6808 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
6811 (validate_switches_from_spec): Deal with %@{...} sequence.
6812 (validate_switches): Likewise.
6813 (driver::finalize): Call clear_args.
6815 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
6817 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
6819 2018-06-11 Martin Sebor <msebor@redhat.com>
6821 * doc/invoke.texi (-Wall): List -Wc++17-compat.
6822 (Wno-class-memaccess): Add @opindex.
6823 (Wno-templates, Wno-multiple-inheritance): Same.
6824 (Wno-virtual-inheritance, Wno-namespaces): Same.
6825 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
6826 (Wno-format-overflow, Wno-format-truncation): Same.
6827 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
6828 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
6829 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
6830 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
6831 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
6832 (Wno-misspelled-isr): Same.
6834 2018-06-11 Martin Sebor <msebor@redhat.com>
6836 * PR tree-optimization/86083
6837 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
6839 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
6841 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
6843 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
6846 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
6847 on the correct operand.
6848 (*movdi_internal64): Ditto.
6850 2018-06-11 Martin Liska <mliska@suse.cz>
6852 PR tree-optimization/86089
6853 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
6855 2018-06-11 Julia Koval <julia.koval@intel.com>
6857 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
6858 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
6859 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
6861 2018-06-11 Olivier Hainque <hainque@adacore.com>
6863 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
6864 for Ada with strict dwarf2.
6866 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
6869 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
6872 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
6874 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
6876 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
6878 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
6881 2018-06-08 Martin Liska <mliska@suse.cz>
6883 * tree-cfg.h (debug_function): Fix argument type to match
6886 2018-06-08 Martin Liska <mliska@suse.cz>
6888 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
6889 Remove usage of MPX-related (and removed) fields.
6890 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
6892 2018-06-08 David Malcolm <dmalcolm@redhat.com>
6894 * cfg.c (debug): Use TDF_NONE rather than 0.
6895 * cfghooks.c (debug): Likewise.
6896 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
6897 (struct dump_option_value_info): Convert to...
6898 (struct kv_pair): ...this template type.
6899 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
6901 (optinfo_verbosity_options): Likewise.
6902 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
6904 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
6905 than int for "optgroup_flags" param.
6906 (dump_generic_expr_loc): Use dump_flags_t rather than int for
6908 (dump_dec): Likewise.
6909 (dump_finish): Use TDF_NONE rather than 0.
6910 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
6911 rather than int for "optgroup_flags" param. Use TDF_NONE rather
6912 than 0. Update for change to option_ptr.
6913 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
6914 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
6915 0. Update for changes to optinfo_verbosity_options and
6917 (opt_info_switch_p): Convert optgroup_flags from int to
6919 (dump_basic_block): Use dump_flags_t rather than int
6920 for "dump_kind" param.
6921 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
6922 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
6923 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
6924 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
6925 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
6926 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
6927 TDF_NONE): Convert from macros to...
6928 (enum dump_flag): ...this new enum.
6929 (dump_flags_t): Update to use enum.
6930 (operator|, operator&, operator~, operator|=, operator&=):
6931 Implement for dump_flags_t.
6932 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
6933 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
6934 Convert from macros to...
6935 (enum optgroup_flag): ...this new enum.
6936 (optgroup_flags_t): New typedef.
6937 (operator|, operator|=): Implement for optgroup_flags_t.
6938 (struct dump_file_info): Convert field "alt_flags" to
6939 dump_flags_t. Convert field "optgroup_flags" to
6941 (dump_basic_block): Use dump_flags_t rather than int for param.
6942 (dump_generic_expr_loc): Likewise.
6943 (dump_dec): Likewise.
6944 (dump_register): Convert param "optgroup_flags" to
6946 (opt_info_enable_passes): Likewise.
6947 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
6949 * gimple-pretty-print.c (debug): Likewise.
6950 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
6951 (merged_store_group::apply_stores): Likewise.
6952 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
6953 * gimple.c (verify_gimple_pp): Likewise.
6954 * graphite-poly.c (print_pbb_body): Likewise.
6955 * passes.c (pass_manager::register_one_dump_file): Convert
6956 local "optgroup_flags" to optgroup_flags_t.
6957 * print-tree.c (print_node): Use TDF_NONE rather than 0.
6959 (debug_body): Likewise.
6960 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
6961 to optgroup_flags_t.
6962 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
6964 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
6965 (convert_mult_to_fma): Likewise.
6966 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
6967 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
6968 * tree-vect-data-refs.c (dump_lower_bound): Convert param
6969 "dump_kind" to dump_flags_t.
6971 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
6973 * config/rs6000/rs6000.c (min, max): Delete.
6975 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
6977 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
6978 -mabi=spe and -mabi=no-spe.
6980 2018-06-08 Martin Liska <mliska@suse.cz>
6982 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
6983 where we expect an existing summary.
6985 2018-06-08 Martin Liska <mliska@suse.cz>
6987 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
6988 * ipa-inline.h (estimate_edge_growth): Likewise.
6990 2018-06-08 Martin Liska <mliska@suse.cz>
6992 * cgraph.c (function_version_hasher::hash): Use
6993 cgraph_node::get_uid ().
6994 (function_version_hasher::equal):
6995 * cgraph.h (cgraph_node::get_uid): New method.
6996 * ipa-inline.c (update_caller_keys): Use
6997 cgraph_node::get_uid ().
6998 (update_callee_keys): Likewise.
6999 * ipa-utils.c (searchc): Likewise.
7000 (ipa_reduced_postorder): Likewise.
7001 * lto-cgraph.c (input_node): Likewise.
7002 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
7003 * symbol-summary.h (symtab_insertion): Likewise.
7004 (symtab_removal): Likewise.
7005 (symtab_duplication): Likewise.
7006 * tree-pretty-print.c (dump_function_header): Likewise.
7007 * tree-sra.c (convert_callers_for_node): Likewise.
7009 2018-06-08 Martin Liska <mliska@suse.cz>
7011 * cgraph.c (symbol_table::create_edge): Always assign a new
7013 (symbol_table::free_edge): Do not recycle numbers.
7014 * cgraph.h (cgraph_edge::get): New method.
7015 * symbol-summary.h (symtab_removal): Use it.
7016 (symtab_duplication): Likewise.
7017 (call_summary::hashable_uid): Remove.
7019 2018-06-08 Martin Liska <mliska@suse.cz>
7021 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
7022 (initialize_growth_caches): Remove.
7023 (free_growth_caches): Likewise.
7024 (do_estimate_edge_time): Use edge_growth_cache.
7025 (do_estimate_edge_size): Likewise.
7026 (do_estimate_edge_hints): Likewise.
7027 * ipa-inline.c (reset_edge_caches): Likewise.
7028 (recursive_inlining): Likewise.
7029 (inline_small_functions): Likewise.
7030 * ipa-inline.h (initialize_growth_caches): Remove.
7031 (estimate_edge_size): Likewise.
7032 (estimate_edge_time): Likewise.
7033 (estimate_edge_hints): Likewise.
7034 (reset_edge_growth_cache): Likewise.
7035 * symbol-summary.h (call_summary::remove): New method.
7037 2018-06-08 Martin Liska <mliska@suse.cz>
7039 * ipa-cp.c (class edge_clone_summary): New summary.
7040 (grow_edge_clone_vectors): Remove.
7041 (ipcp_edge_duplication_hook): Remove.
7042 (class edge_clone_summary_t): New call_summary class.
7043 (ipcp_edge_removal_hook): Remove.
7044 (edge_clone_summary_t::duplicate): New function.
7045 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
7046 (create_specialized_node): Likewise.
7047 (ipcp_driver): Initialize edge_clone_summaries and do not
7050 2018-06-08 Martin Liska <mliska@suse.cz>
7052 * symbol-summary.h (get): New function.
7053 (call_summary::m_initialize_when_cloning): New class member.
7055 2018-06-08 Martin Liska <mliska@suse.cz>
7057 * cgraph.c (cgraph_node::remove): Do not recycle uid.
7058 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
7059 (symbol_table::allocate_cgraph_symbol): Do not set uid.
7060 * passes.c (uid_hash_t): Record removed_nodes by their uids.
7061 (remove_cgraph_node_from_order): Use the removed_nodes set.
7062 (do_per_function_toporder): Likwise.
7063 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
7064 instead of summary_uid.
7065 (symtab_removal): Likewise.
7066 (symtab_duplication): Likewise.
7068 2018-06-08 Martin Liska <mliska@suse.cz>
7070 * ipa-cp.c (ipcp_store_bits_results): Use
7071 ipcp_transformation_sum.
7072 (ipcp_store_vr_results): Likewise.
7073 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
7075 (ipcp_transformation_initialize): ... this.
7076 (ipa_set_node_agg_value_chain):
7077 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
7078 (write_ipcp_transformation_info): Likewise.
7079 (read_ipcp_transformation_info): Likewise.
7080 (ipcp_update_bits): Likewise.
7081 (ipcp_update_vr): Likewise.
7082 (ipcp_transform_function): Likewise.
7083 * ipa-prop.h: Rename ipcp_transformation_summary to
7084 ipcp_transformation.
7085 (class ipcp_transformation_t): New function summary.
7086 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
7087 (ipa_get_agg_replacements_for_node): Likewise.
7089 2018-06-08 Martin Liska <mliska@suse.cz>
7091 * ipa-pure-const.c (struct funct_state_d): Do it class instead
7093 (class funct_state_summary_t): New function_summary class.
7094 (has_function_state): Remove.
7095 (get_function_state): Likewise.
7096 (set_function_state): Likewise.
7097 (add_new_function): Likewise.
7098 (funct_state_summary_t::insert): New function.
7099 (duplicate_node_data): Remove.
7100 (remove_node_data): Remove.
7101 (funct_state_summary_t::duplicate): New function.
7102 (register_hooks): Create new funct_state_summaries.
7103 (pure_const_generate_summary): Use it.
7104 (pure_const_write_summary): Likewise.
7105 (pure_const_read_summary): Likewise.
7106 (propagate_pure_const): Likewise.
7107 (propagate_nothrow): Likewise.
7108 (dump_malloc_lattice): Likewise.
7109 (propagate_malloc): Likewise.
7110 (execute): Do not register hooks, just remove summary
7112 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
7115 2018-06-08 Martin Liska <mliska@suse.cz>
7117 * ipa-reference.c (remove_node_data): Remove.
7118 (duplicate_node_data): Likewise.
7119 (class ipa_ref_var_info_summary_t): New class.
7120 (class ipa_ref_opt_summary_t): Likewise.
7121 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
7122 (get_reference_optimization_summary): Use
7123 ipa_ref_opt_sum_summaries.
7124 (set_reference_vars_info): Remove.
7125 (set_reference_optimization_summary): Likewise.
7126 (ipa_init): Create summaries.
7127 (init_function_info): Use function summary.
7128 (ipa_ref_opt_summary_t::duplicate): New function.
7129 (ipa_ref_opt_summary_t::remove): New function.
7130 (get_read_write_all_from_node): Fix GNU coding style.
7131 (propagate): Use function summary.
7132 (write_node_summary_p): Fix GNU coding style.
7133 (stream_out_bitmap): Likewise.
7134 (ipa_reference_read_optimization_summary): Use function summary.
7135 (ipa_reference_c_finalize): Do not release hooks.
7137 2018-06-08 Martin Liska <mliska@suse.cz>
7139 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
7140 (analyze_function_body): Extract multiple calls of get_create.
7141 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
7142 * ipa-inline.c (recursive_inlining): Use ::get method.
7143 * ipa-inline.h (estimate_edge_growth): Likewise.
7145 2018-06-08 Martin Liska <mliska@suse.cz>
7147 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
7149 (hsa_function_summary::hsa_function_summary): Use the new enum
7151 (hsa_gpu_implementation_p): Use hsa_summaries::get.
7152 * hsa-gen.c (hsa_get_host_function): Likewise.
7153 (get_brig_function_name): Likewise.
7154 * ipa-hsa.c (process_hsa_functions): Likewise.
7155 (ipa_hsa_write_summary): Likewise.
7156 * symbol-summary.h (symtab_duplication): Use ::get function/
7157 (get): New function.
7159 2018-06-08 Martin Liska <mliska@suse.cz>
7161 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
7163 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
7164 (hsa_register_kernel): Likewise.
7165 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
7166 * hsa-gen.c (hsa_get_host_function): Likewise.
7167 (get_brig_function_name): Likewise.
7168 (generate_hsa): Likewise.
7169 (pass_gen_hsail::execute): Likewise.
7170 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
7171 (devirtualization_time_bonus): Likewise.
7172 (ipcp_propagate_stage): Likewise.
7173 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
7174 (edge_set_predicate): Likewise.
7175 (evaluate_conditions_for_known_args): Likewise.
7176 (evaluate_properties_for_edge): Likewise.
7177 (ipa_fn_summary::reset): Likewise.
7178 (ipa_fn_summary_t::duplicate): Likewise.
7179 (dump_ipa_call_summary): Likewise.
7180 (ipa_dump_fn_summary): Likewise.
7181 (analyze_function_body): Likewise.
7182 (compute_fn_summary): Likewise.
7183 (estimate_edge_devirt_benefit): Likewise.
7184 (estimate_edge_size_and_time): Likewise.
7185 (estimate_calls_size_and_time): Likewise.
7186 (estimate_node_size_and_time): Likewise.
7187 (inline_update_callee_summaries): Likewise.
7188 (remap_edge_change_prob): Likewise.
7189 (remap_edge_summaries): Likewise.
7190 (ipa_merge_fn_summary_after_inlining): Likewise.
7191 (ipa_update_overall_fn_summary): Likewise.
7192 (read_ipa_call_summary): Likewise.
7193 (inline_read_section): Likewise.
7194 (write_ipa_call_summary): Likewise.
7195 (ipa_fn_summary_write): Likewise.
7196 (ipa_free_fn_summary): Likewise.
7197 * ipa-hsa.c (process_hsa_functions): Likewise.
7198 (ipa_hsa_write_summary): Likewise.
7199 (ipa_hsa_read_section): Likewise.
7200 * ipa-icf.c (sem_function::merge): Likewise.
7201 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
7202 (do_estimate_edge_time): Likewise.
7203 (estimate_size_after_inlining): Likewise.
7204 (estimate_growth): Likewise.
7205 (growth_likely_positive): Likewise.
7206 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
7207 (inline_call): Likewise.
7208 * ipa-inline.c (caller_growth_limits): Likewise.
7209 (can_inline_edge_p): Likewise.
7210 (can_inline_edge_by_limits_p): Likewise.
7211 (compute_uninlined_call_time): Likewise.
7212 (compute_inlined_call_time): Likewise.
7213 (want_inline_small_function_p): Likewise.
7214 (edge_badness): Likewise.
7215 (update_caller_keys): Likewise.
7216 (update_callee_keys): Likewise.
7217 (recursive_inlining): Likewise.
7218 (inline_small_functions): Likewise.
7219 (inline_to_all_callers_1): Likewise.
7220 (dump_overall_stats): Likewise.
7221 (early_inline_small_functions): Likewise.
7222 (early_inliner): Likewise.
7223 * ipa-inline.h (estimate_edge_growth): Likewise.
7224 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
7225 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
7226 * ipa-prop.h (IPA_NODE_REF): Likewise.
7227 (IPA_EDGE_REF): Likewise.
7228 * ipa-pure-const.c (malloc_candidate_p): Likewise.
7229 (propagate_malloc): Likewise.
7230 * ipa-split.c (execute_split_functions): Likewise.
7231 * symbol-summary.h: Rename get to get_create.
7233 (get_create): Likewise.
7234 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
7236 2018-06-08 Martin Liska <mliska@suse.cz>
7238 * symbol-summary.h (release): Move definition out of class
7240 (symtab_removal): Likewise.
7241 (symtab_duplication): Likewise.
7243 2018-06-08 Martin Liska <mliska@suse.cz>
7245 * symbol-summary.h (function_summary): Move constructor
7246 implementation out of class declaration.
7247 (release): Likewise.
7248 (symtab_insertion): Likewise.
7249 (symtab_removal): Likewise.
7250 (symtab_duplication): Likewise.
7253 2018-06-08 Martin Liska <mliska@suse.cz>
7255 * Makefile.in: Remove support for MPX (macros, related functions,
7256 fields in cgraph_node, ...).
7257 * builtin-types.def (BT_BND): Likewise.
7258 (BT_FN_BND_CONST_PTR): Likewise.
7259 (BT_FN_CONST_PTR_BND): Likewise.
7260 (BT_FN_VOID_PTR_BND): Likewise.
7261 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
7262 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
7263 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
7264 (expand_builtin_mempcpy_with_bounds): Likewise.
7265 (expand_builtin_memset_with_bounds): Likewise.
7266 (expand_builtin_memset_args): Likewise.
7267 (std_expand_builtin_va_start): Likewise.
7268 (expand_builtin): Likewise.
7269 (expand_builtin_with_bounds): Likewise.
7270 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
7271 (DEF_LIB_BUILTIN_CHKP): Likewise.
7272 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
7273 (DEF_CHKP_BUILTIN): Likewise.
7274 (BUILT_IN_MEMCPY): Likewise.
7275 (BUILT_IN_MEMMOVE): Likewise.
7276 (BUILT_IN_MEMPCPY): Likewise.
7277 (BUILT_IN_MEMSET): Likewise.
7278 (BUILT_IN_STPCPY): Likewise.
7279 (BUILT_IN_STRCAT): Likewise.
7280 (BUILT_IN_STRCHR): Likewise.
7281 (BUILT_IN_STRCPY): Likewise.
7282 (BUILT_IN_STRLEN): Likewise.
7283 (BUILT_IN_MEMCPY_CHK): Likewise.
7284 (BUILT_IN_MEMMOVE_CHK): Likewise.
7285 (BUILT_IN_MEMPCPY_CHK): Likewise.
7286 (BUILT_IN_MEMSET_CHK): Likewise.
7287 (BUILT_IN_STPCPY_CHK): Likewise.
7288 (BUILT_IN_STRCAT_CHK): Likewise.
7289 (BUILT_IN_STRCPY_CHK): Likewise.
7290 * calls.c (store_bounds): Likewise.
7291 (emit_call_1): Likewise.
7292 (special_function_p): Likewise.
7293 (maybe_warn_nonstring_arg): Likewise.
7294 (initialize_argument_information): Likewise.
7295 (finalize_must_preallocate): Likewise.
7296 (compute_argument_addresses): Likewise.
7297 (expand_call): Likewise.
7298 * cfgexpand.c (expand_call_stmt): Likewise.
7299 (expand_return): Likewise.
7300 (expand_gimple_stmt_1): Likewise.
7301 (pass_expand::execute): Likewise.
7302 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
7303 (cgraph_node::remove): Likewise.
7304 (cgraph_node::dump): Likewise.
7305 (cgraph_node::verify_node): Likewise.
7306 * cgraph.h (chkp_function_instrumented_p): Likewise.
7307 (symtab_node::get_alias_target): Likewise.
7308 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
7309 (cgraph_local_p): Likewise.
7310 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
7311 (cgraph_edge::rebuild_references): Likewise.
7312 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
7313 (walk_polymorphic_call_targets): Likewise.
7314 (cgraph_node::expand_thunk): Likewise.
7315 (symbol_table::output_weakrefs): Likewise.
7316 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
7317 (ix86_handle_option): Likewise.
7318 * config/i386/constraints.md: Likewise.
7319 * config/i386/i386-builtin-types.def (BND): Likewise.
7323 * config/i386/i386-builtin.def (BDESC_END): Likewise.
7324 (BDESC_FIRST): Likewise.
7326 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
7327 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
7328 * config/i386/i386.c (enum reg_class): Likewise.
7329 (ix86_target_string): Likewise.
7330 (ix86_option_override_internal): Likewise.
7331 (ix86_conditional_register_usage): Likewise.
7332 (ix86_valid_target_attribute_inner_p): Likewise.
7333 (ix86_set_indirect_branch_type): Likewise.
7334 (ix86_set_current_function): Likewise.
7335 (ix86_function_arg_regno_p): Likewise.
7336 (init_cumulative_args): Likewise.
7337 (ix86_function_arg_advance): Likewise.
7338 (ix86_function_arg): Likewise.
7339 (ix86_pass_by_reference): Likewise.
7340 (ix86_function_value_regno_p): Likewise.
7341 (ix86_function_value_1): Likewise.
7342 (ix86_function_value_bounds): Likewise.
7343 (ix86_return_in_memory): Likewise.
7344 (ix86_setup_incoming_vararg_bounds): Likewise.
7345 (ix86_va_start): Likewise.
7346 (indirect_thunk_need_prefix): Likewise.
7347 (print_reg): Likewise.
7348 (ix86_print_operand): Likewise.
7349 (ix86_expand_call): Likewise.
7350 (ix86_output_function_return): Likewise.
7351 (reg_encoded_number): Likewise.
7352 (BDESC_VERIFYS): Likewise.
7353 (ix86_init_mpx_builtins): Likewise.
7354 (ix86_init_builtins): Likewise.
7355 (ix86_emit_cmove): Likewise.
7356 (ix86_emit_move_max): Likewise.
7357 (ix86_expand_builtin): Likewise.
7358 (ix86_builtin_mpx_function): Likewise.
7359 (ix86_get_arg_address_for_bt): Likewise.
7360 (ix86_load_bounds): Likewise.
7361 (ix86_store_bounds): Likewise.
7362 (ix86_load_returned_bounds): Likewise.
7363 (ix86_store_returned_bounds): Likewise.
7364 (ix86_class_likely_spilled_p): Likewise.
7365 (ix86_hard_regno_mode_ok): Likewise.
7366 (x86_order_regs_for_local_alloc): Likewise.
7367 (ix86_mitigate_rop): Likewise.
7368 (ix86_bnd_prefixed_insn_p): Likewise.
7369 (ix86_mpx_bound_mode): Likewise.
7370 (ix86_make_bounds_constant): Likewise.
7371 (ix86_initialize_bounds): Likewise.
7372 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
7373 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
7374 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
7375 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
7376 (TARGET_CHKP_BOUND_MODE): Likewise.
7377 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
7378 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
7379 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
7380 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
7381 * config/i386/i386.h (TARGET_MPX): Likewise.
7382 (TARGET_MPX_P): Likewise.
7383 (VALID_BND_REG_MODE): Likewise.
7384 (FIRST_BND_REG): Likewise.
7385 (LAST_BND_REG): Likewise.
7386 (enum reg_class): Likewise.
7387 (BND_REG_P): Likewise.
7388 (BND_REGNO_P): Likewise.
7389 (BNDmode): Likewise.
7390 (ADJUST_INSN_LENGTH): Likewise.
7391 * config/i386/i386.md: Likewise.
7392 * config/i386/i386.opt: Likewise.
7393 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
7394 (defined): Likewise.
7395 (LINK_MPX): Likewise.
7396 (MPX_SPEC): Likewise.
7397 (LIBMPX_SPEC): Likewise.
7398 (LIBMPXWRAPPERS_SPEC): Likewise.
7399 (CHKP_SPEC): Likewise.
7400 * config/i386/predicates.md: Likewise.
7401 * dbxout.c (dbxout_type): Likewise.
7402 * doc/extend.texi: Likewise.
7403 * doc/invoke.texi: Likewise.
7404 * doc/md.texi: Likewise.
7405 * doc/tm.texi: Likewise.
7406 * doc/tm.texi.in: Likewise.
7407 * dwarf2out.c (is_base_type): Likewise.
7408 (gen_formal_types_die): Likewise.
7409 (gen_subprogram_die): Likewise.
7410 (gen_type_die_with_usage): Likewise.
7411 (gen_decl_die): Likewise.
7412 (dwarf2out_late_global_decl): Likewise.
7413 * expr.c (expand_assignment): Likewise.
7414 (emit_storent_insn): Likewise.
7415 (store_expr_with_bounds): Likewise.
7416 (store_expr): Likewise.
7417 (expand_expr_real_1): Likewise.
7418 * expr.h (store_expr_with_bounds): Likewise.
7419 * function.c (use_register_for_decl): Likewise.
7420 (struct bounds_parm_data): Likewise.
7421 (assign_parms_augmented_arg_list): Likewise.
7422 (assign_parm_find_entry_rtl): Likewise.
7423 (assign_parm_is_stack_parm): Likewise.
7424 (assign_parm_load_bounds): Likewise.
7425 (assign_bounds): Likewise.
7426 (assign_parms): Likewise.
7427 (expand_function_start): Likewise.
7428 * gcc.c (CHKP_SPEC): Likewise.
7429 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
7430 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
7431 (wrestrict_dom_walker::check_call): Likewise.
7432 * gimple.c (gimple_build_call_from_tree): Likewise.
7433 * gimple.h (enum gf_mask): Likewise.
7434 (gimple_call_with_bounds_p): Likewise.
7435 (gimple_call_set_with_bounds): Likewise.
7436 * gimplify.c (gimplify_init_constructor): Likewise.
7437 * ipa-cp.c (initialize_node_lattices): Likewise.
7438 (propagate_constants_across_call): Likewise.
7439 (find_more_scalar_values_for_callers_subset): Likewise.
7440 * ipa-hsa.c (process_hsa_functions): Likewise.
7441 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
7442 * ipa-icf.c (sem_function::merge): Likewise.
7443 * ipa-inline.c (early_inliner): Likewise.
7444 * ipa-pure-const.c (warn_function_noreturn): Likewise.
7445 (warn_function_cold): Likewise.
7446 (propagate_pure_const): Likewise.
7447 * ipa-ref.h (enum GTY): Likewise.
7448 * ipa-split.c (find_retbnd): Likewise.
7449 (consider_split): Likewise.
7450 (split_function): Likewise.
7451 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
7452 * ipa.c (walk_polymorphic_call_targets): Likewise.
7453 (symbol_table::remove_unreachable_nodes): Likewise.
7454 (process_references): Likewise.
7455 (cgraph_build_static_cdtor_1): Likewise.
7456 * lto-cgraph.c (lto_output_node): Likewise.
7457 (output_refs): Likewise.
7458 (compute_ltrans_boundary): Likewise.
7459 (input_overwrite_node): Likewise.
7460 (input_node): Likewise.
7461 (input_cgraph_1): Likewise.
7462 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
7463 * passes.c (pass_manager::execute_early_local_passes): Likewise.
7464 (class pass_chkp_instrumentation_passes): Likewise.
7465 (make_pass_chkp_instrumentation_passes): Likewise.
7466 * passes.def: Likewise.
7467 * rtl.h (struct GTY): Likewise.
7468 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
7469 * stor-layout.c (layout_type): Likewise.
7470 * symtab.c: Likewise.
7471 * target.def: Likewise.
7472 * targhooks.c (default_chkp_bound_type): Likewise.
7473 (default_chkp_bound_mode): Likewise.
7474 (default_builtin_chkp_function): Likewise.
7475 (default_chkp_function_value_bounds): Likewise.
7476 (default_chkp_make_bounds_constant): Likewise.
7477 (default_chkp_initialize_bounds): Likewise.
7478 * targhooks.h (default_chkp_bound_type): Likewise.
7479 (default_chkp_bound_mode): Likewise.
7480 (default_builtin_chkp_function): Likewise.
7481 (default_chkp_function_value_bounds): Likewise.
7482 (default_chkp_make_bounds_constant): Likewise.
7483 (default_chkp_initialize_bounds): Likewise.
7484 * toplev.c (compile_file): Likewise.
7485 (process_options): Likewise.
7486 * tree-core.h (DEF_BUILTIN): Likewise.
7487 (DEF_BUILTIN_CHKP): Likewise.
7488 * tree-inline.c (declare_return_variable): Likewise.
7489 (remap_gimple_stmt): Likewise.
7490 (copy_bb): Likewise.
7491 (initialize_inlined_parameters): Likewise.
7492 (expand_call_inline): Likewise.
7493 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
7494 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
7495 (make_pass_ipa_chkp_produce_thunks): Likewise.
7496 (make_pass_chkp): Likewise.
7497 (make_pass_chkp_opt): Likewise.
7498 (make_pass_chkp_instrumentation_passes): Likewise.
7499 * tree-pretty-print.c (dump_generic_node): Likewise.
7500 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
7501 * tree-ssa-dce.c (propagate_necessity): Likewise.
7502 (eliminate_unnecessary_stmts): Likewise.
7503 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
7504 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
7505 * tree-ssa-sccvn.h: Likewise.
7506 * tree-ssa-strlen.c (get_string_length): Likewise.
7507 (valid_builtin_call): Likewise.
7508 (adjust_last_stmt): Likewise.
7509 (handle_builtin_strchr): Likewise.
7510 (handle_builtin_strcpy): Likewise.
7511 (handle_builtin_stxncpy): Likewise.
7512 (handle_builtin_memcpy): Likewise.
7513 (handle_builtin_strcat): Likewise.
7514 (strlen_check_and_optimize_stmt): Likewise.
7515 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
7516 * tree-streamer-in.c: Likewise.
7517 * tree-streamer.c (record_common_node): Likewise.
7518 * tree.c (tree_code_size): Likewise.
7519 (wide_int_to_tree_1): Likewise.
7520 (type_contains_placeholder_1): Likewise.
7521 (build_common_tree_nodes): Likewise.
7522 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
7523 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
7524 (POINTER_BOUNDS_P): Likewise.
7525 (BOUNDED_TYPE_P): Likewise.
7526 (BOUNDED_P): Likewise.
7527 (CALL_WITH_BOUNDS_P): Likewise.
7528 (pointer_bounds_type_node): Likewise.
7529 * value-prof.c (gimple_ic): Likewise.
7530 * var-tracking.c (vt_add_function_parameters): Likewise.
7531 * varasm.c (make_decl_rtl): Likewise.
7532 (assemble_start_function): Likewise.
7533 (output_constant): Likewise.
7534 (maybe_assemble_visibility): Likewise.
7535 * varpool.c (ctor_for_folding): Likewise.
7536 * chkp-builtins.def: Remove.
7537 * ipa-chkp.c: Remove.
7538 * ipa-chkp.h: Remove.
7539 * rtl-chkp.c: Remove.
7540 * rtl-chkp.h: Remove.
7541 * tree-chkp-opt.c: Remove.
7542 * tree-chkp.c: Remove.
7543 * tree-chkp.h: Remove.
7545 2018-06-07 Carl Love <cel@us.ibm.com>
7547 * config/rs6000/vsx.md (vextract_fp_from_shorth,
7548 vextract_fp_from_shortl): Add BE support.
7550 2018-06-07 Paul Koning <ni1d@arrl.net>
7552 * compare-elim.c (try_merge_compare): Don't merge compare if
7553 address contains a side effect.
7554 (try_eliminate_compare): Likewise.
7556 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
7558 * config.gcc: Support "tremont".
7559 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
7560 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
7562 * config/i386/i386.c (m_TREMONT): Define.
7563 (processor_target_table): Add "tremont".
7564 (PTA_TREMONT): Define.
7565 (ix86_lea_outperforms): Add TARGET_TREMONT.
7566 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
7567 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
7568 and M_INTEL_GOLDMONT_PLUS.
7569 (fold_builtin_cpu): Add "tremont".
7570 (ix86_add_stmt_cost): Add TARGET_TREMONT.
7571 (ix86_option_override_internal): Add "tremont".
7572 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
7573 (processor_type): Add PROCESSOR_TREMONT.
7574 * config/i386/x86-tune.def: Add m_TREMONT.
7575 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
7577 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
7579 * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
7580 symbol defined for msp430i* devices to be lower case.
7582 2018-06-07 Richard Biener <rguenther@suse.de>
7584 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
7585 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
7586 Properly wrap signed arithmetic if overflow wraps.
7588 2018-06-07 Jakub Jelinek <jakub@redhat.com>
7590 PR tree-optimization/69615
7591 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
7592 of a cast from a same precision integral SSA_NAME in a bb dominated
7593 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
7594 cast to utype if rhs2 has already a compatible type.
7596 2018-06-07 Richard Biener <rguenther@suse.de>
7598 PR tree-optimization/85935
7599 * graphite-scop-detection.c (find_params_in_bb): Analyze
7600 condition operands with respect to the correct loop. Assert
7601 the analysis doesn't fail.
7603 2018-06-04 Carl Love <cel@us.ibm.com>
7605 * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
7606 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
7607 as it is slightly cheaper.
7608 (first_match_or_eos_index_<mode>):
7609 Calculate index using natural element order.
7610 (first_match_index_<mode>):
7611 Calculate index using natural element order.
7612 (first_match_or_eos_index_<mode>):
7613 Calculate index using natural order.
7614 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
7615 for BE and LE modes.
7616 * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
7617 P9V_BUILTIN_VCLZLSBB_V16QI.
7618 * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
7621 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
7623 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
7624 indentation and line wrap for many prototypes. Add missing
7625 @smallexample directives around block of prototypes for vec_xl and
7628 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
7630 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
7631 track if we pass or return IEEE 128-bit floating point.
7632 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
7633 C++ mangling that is compatible with GCC 8.1.
7634 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
7635 (init_cumulative_args): Note if we pass or return IEEE 128-bit
7636 floating point types.
7637 (rs6000_function_arg_advance_1): Likewise.
7638 (rs6000_mangle_type): Optionally generate mangled names that match
7639 what GCC 8.1 generated for IEEE 128-bit floating point types.
7640 (rs6000_globalize_decl_name): If we have an external function that
7641 passes or returns IEEE 128-bit types, generate a weak reference
7642 from the mangled name used in GCC 8.1 to the current mangled
7644 (rs6000_init_builtins): Make __ibm128 use the long double type if
7645 long double is IBM extended double. Make __float128 use the long
7646 double type if long double is IEEE 128-bit.
7649 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
7650 macro for __ibm128 built-in functions.
7651 (PACK_IF): Add __ibm128 pack/unpack functions.
7652 (UNPACK_IF): Likewise.
7653 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
7654 enable long double built-in functions if long double is IEEE
7655 128-bit floating point.
7656 (rs6000_invalid_builtin): Update long double built-in function
7658 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
7659 functions, adjust the built-in function to use the long double
7660 built-in function if __ibm128 and long double are the same type.
7661 * doc/extend.texi (PowerPC builtins): Update documention for
7662 __builtin_{,un}pack_longdouble. Add documentation for
7663 __builtin_{,un}pack_ibm128.
7665 2018-06-06 Jim Wilson <jimw@sifive.com>
7667 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
7668 (struct machine_function): New field interrupt_mode.
7669 (riscv_handle_type_attribute): New function. Add forward declaration.
7670 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
7671 (riscv_expand_epilogue): Check interrupt_mode field.
7672 (riscv_set_current_function): Check interrupt attribute args and
7673 set interrupt_mode field.
7674 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
7675 (riscv_sret, riscv_uret): New.
7676 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
7677 new arguments to interrupt attribute.
7679 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
7682 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
7683 Don't handle -mcpu=power8 if -mpower9-vector is also used.
7685 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
7687 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
7688 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
7689 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
7690 several redundant entries.
7692 2018-06-06 David Malcolm <dmalcolm@redhat.com>
7694 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
7695 type from "rtx" to "rtx_insn *".
7696 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
7697 for local "call_insn", removing cast.
7698 (ix86_expand_call): Likewise, introducing a "call_insn" local.
7700 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
7702 PR tree-optimization/86066
7703 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
7704 for BIT_INSERT_EXPR stores.
7706 2018-06-06 Richard Biener <rguenther@suse.de>
7708 PR tree-optimization/86062
7709 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
7710 component refs ontop
7711 of to be offsetted base.
7713 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
7715 * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
7716 to be static and remove check on interrupt attribute name.
7718 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
7720 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
7721 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
7723 2018-06-05 Steve Ellcey <sellcey@cavium.com>
7726 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
7728 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
7729 Remove second argument, change how error is called.
7730 (aarch64_layout_arg): Remove second argument from
7731 aarch64_err_no_fpadvsimd call.
7732 (aarch64_init_cumulative_args): Ditto.
7733 (aarch64_gimplify_va_arg_expr): Ditto.
7734 * config/aarch64/aarch64.md (mov<mode>): Ditto.
7736 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
7738 * config/i386/i386.md (simple_return_indirect_internal): New expander.
7739 (*simple_return_indirect_internal<mode>): Rename from
7740 simple_return_indirect_internal. Use W mode iterator.
7741 (rstorssp): New expander.
7742 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
7743 (clrssbsy): New expander.
7744 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
7746 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
7748 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
7750 (cmse_check_pointed_object): Likewise.
7752 2018-06-05 Martin Liska <mliska@suse.cz>
7754 PR gcov-profile/47618
7755 * doc/invoke.texi: Document how -fprofile-dir format
7758 2018-06-05 Richard Biener <rguenther@suse.de>
7760 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
7761 removal pretend DOM info isn't available so we do not update
7762 it and only remove edges, not dominated blocks. Actually free
7763 DOM info in case we removed something. Remove unreachable blocks.
7764 (mfb_keep_latches): Work with either DOM info or marked backedges.
7765 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
7766 first. Mark backedges if DOM info isn't available.
7767 (Re-)compute DOM info after cleanup_control_flow_pre.
7769 2018-06-05 Richard Biener <rguenther@suse.de>
7771 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
7772 (locus_discrim_hasher::hash): Adjust.
7773 (locus_discrim_hasher::equal): Likewise.
7774 (next_discriminator_for_locus): Work on line directly.
7775 (same_line_p): Pass in expanded locus1 as well.
7776 (assign_discriminators): Avoid redundant location expansions.
7778 2018-06-05 Richard Biener <rguenther@suse.de>
7780 PR tree-optimization/86046
7781 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
7782 if required after clearing TREE_ADDRESSABLE.
7784 2018-06-05 Richard Biener <rguenther@suse.de>
7786 PR tree-optimization/86047
7787 * tree-ssa-loop.c (for_each_index): Glob handling of all
7788 decls and constants and really handle all of them.
7790 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7793 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
7794 qualifier_void_pointer and qualifier_const_void_pointer.
7795 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
7796 (arm_init_builtins): Handle the above.
7797 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
7798 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
7799 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
7802 2018-06-05 Martin Liska <mliska@suse.cz>
7804 * auto-profile.c (read_autofdo_file): Do not use
7805 gcov_ctr_summary struct.
7806 (afdo_callsite_hot_enough_for_early_inline): Likewise.
7807 * coverage.c (struct counts_entry): Likewise.
7808 (read_counts_file): Read just single summary entry.
7809 (get_coverage_counts): Use gcov_summary struct.
7810 * coverage.h (get_coverage_counts): Likewise.
7811 * gcov-dump.c (dump_working_sets): Likewise.
7812 (tag_summary): Dump just single summary.
7813 * gcov-io.c (gcov_write_summary): Write just histogram
7815 (gcov_read_summary): Read just single summary.
7816 (compute_working_sets): Use gcov_summary struct.
7817 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
7818 of GCOV_COUNTERS_SUMMABLE.
7819 (GCOV_COUNTERS_SUMMABLE): Remove.
7820 (GCOV_FIRST_VALUE_COUNTER): Replace with
7821 GCOV_COUNTER_V_INTERVAL.
7822 (struct gcov_ctr_summary): Remove.
7823 (struct gcov_summary): Directly use fields of former
7825 (compute_working_sets): Use gcov_summary struct.
7826 * gcov.c (read_count_file): Do not use ctrs fields.
7827 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
7829 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
7831 * profile.c: Likewise.
7832 * profile.h: Likewise.
7834 2018-06-05 Martin Liska <mliska@suse.cz>
7836 PR gcov-profile/84846
7837 * gcov.c (output_lines): Print working directory only
7838 in intermediate format.
7840 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
7842 * config/s390/s390-builtin-types.def: Add void function type.
7843 * config/s390/s390-builtins.def: Use the function type for the
7846 2018-06-04 Jim Wilson <jimw@sifive.com>
7848 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
7850 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
7851 and maybe_eh_return. Change regno to unsigned int. Use new args to
7852 handle EH_RETURN_DATA_REGNO registers properly.
7853 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
7854 (riscv_expand_epilogue): Update comment. Change argument name and
7855 type. Update code to use new name and type. Pass new args to
7856 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
7858 * config/riscv/riscv.md (NORMAL_RETURN): New.
7859 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
7860 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
7861 (eh_return): Call gen_eh_return_internal and emit barrier.
7862 (eh_return_internal): Call riscv_expand_epilogue.
7864 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
7866 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
7867 bit_insertion field and declare can_be_merged_into method.
7868 (merged_store_group::can_be_merged_into): New method.
7869 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
7870 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
7871 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
7873 2018-06-04 Richard Biener <rguenther@suse.de>
7875 PR tree-optimization/85955
7876 * builtins.c (fold_builtin_sincos): Convert pointers to
7877 destination to appropriate type before dereferencing.
7879 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
7881 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
7883 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
7885 * expr.c (expand_expr_real_1): Force the operand into memory if
7886 its TYPE_MODE is BLKmode and if there is no integer mode for
7887 the number of bits being extracted.
7889 2018-06-04 Jakub Jelinek <jakub@redhat.com>
7893 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
7894 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
7896 2018-06-04 Richard Biener <rguenther@suse.de>
7898 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
7899 (cleanup_tree_cfg_noloop): ... single caller. Do
7900 start_recording_case_labels later.
7902 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
7904 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
7905 to _IMMINTRIN_H_INCLUDED.
7906 * config/i386/pconfigintrin.h: Ditto.
7907 * config/i386/waitpkgintrin.h: Ditto.
7908 * config/i386/immintrin.h: Add includes for sgxintrin.h,
7909 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
7910 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
7911 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
7912 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
7913 waitpkgintrin.h and cldemoteintrin.h.
7915 2018-06-04 Richard Biener <rguenther@suse.de>
7917 PR tree-optimization/86038
7918 * tracer.c (find_best_successor): Check probability for
7919 being initialized, bail out if not.
7921 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
7924 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
7925 of bits to ignore when comparing architectures.
7927 2018-06-04 Jakub Jelinek <jakub@redhat.com>
7929 PR tree-optimization/69615
7930 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
7931 maximum or minimum of the type, try to merge it also as if
7932 range1 is + [-, x - 1] or + [x + 1, -].
7935 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
7937 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
7939 PR tree-optimization/86034
7940 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
7941 the unsigned bitfield type in a bit insertion sequence if it does not
7942 have a larger precision than the bitfield size.
7943 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
7945 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
7947 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
7949 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
7951 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
7952 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
7953 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
7954 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
7956 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
7958 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
7959 Disable -fdelete-null-pointer-checks for ELF toolchain.
7961 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
7962 Kito Cheng <kito.cheng@gmail.com>
7964 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
7965 (nds32le-*-*, nds32be-*-*): Integrate checking process.
7966 (nds32*-*-*): Add glibc and uclibc conditions.
7967 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
7968 (TARGET_EXCEPT_UNWIND_INFO): Define.
7969 * config/nds32/elf.h: New file.
7970 * config/nds32/linux.h: New file.
7971 * config/nds32/nds32-elf.opt: New file.
7972 * config/nds32/nds32-linux.opt: New file.
7973 * config/nds32/nds32-fp-as-gp.c
7974 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
7975 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
7977 (nds32_asm_file_end): Ditto.
7978 (nds32_print_operand): Ditto.
7979 (nds32_insert_attributes): Ditto.
7980 (nds32_init_libfuncs): New function.
7981 (TARGET_HAVE_TLS): Define.
7982 (TARGET_INIT_LIBFUNCS): Define.
7983 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
7985 (TARGET_ELF): Apply different mcmodel setting.
7986 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
7987 been migrated into elf.h and linux.h files.
7988 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
7989 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
7990 (mcmodel): The content has been migrated into nds32-elf.opt and
7991 nds32-linux.opt files.
7992 * config/nds32/t-elf: New file.
7993 * config/nds32/t-linux: New file.
7995 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
7996 Shiva Chen <shiva0217@gmail.com>
7998 * config/nds32/constants.md (unspec_volatile_element): Add
7999 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
8000 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
8002 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
8003 (make_pass_nds32_fp_as_gp): Declare.
8004 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
8006 (nds32_asm_function_end_prologue): Remove unused asm output.
8007 (nds32_asm_function_begin_epilogue): Remove unused asm output.
8008 (nds32_asm_file_start): Output necessary fp_as_gp information.
8009 (nds32_option_override): Adjust register usage.
8010 (nds32_expand_prologue): Consider fp_as_gp situation.
8011 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
8012 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
8015 (simple_return): Ditto.
8016 (omit_fp_begin): Output special directive for fp_as_gp.
8017 (omit_fp_end): Output special directive for fp_as_gp.
8018 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
8019 mforbid-fp-as-gp): New options.
8021 2018-06-01 Mark Wielaard <mark@klomp.org>
8023 * dwarf2out.c (dwarf2out_finish): Remove generation of
8024 DW_AT_loclists_base.
8026 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
8028 * gimple-ssa-store-merging.c: Include gimple-fold.h.
8029 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
8030 (struct merged_store_group): Add bit_insertion field.
8031 (dump_char_array): Use standard hexadecimal format.
8032 (merged_store_group::merged_store_group): Set bit_insertion to false.
8033 (merged_store_group::apply_stores): Use optimal buffer size. Deal
8034 with BIT_INSERT_EXPR stores. Move up code updating the mask and
8035 also print the mask in the dump file.
8036 (pass_store_merging::gate): Minor tweak.
8037 (imm_store_chain_info::coalesce_immediate): Fix wrong association
8038 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
8039 stores with INTEGER_CST stores.
8040 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
8041 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
8042 and use it throughout. Generate bit insertion sequences if need be.
8043 (pass_store_merging::process_store): Remove redundant condition.
8044 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
8046 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
8048 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
8049 the 128-bit floating point types. Fix function comment.
8051 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8053 * config/aarch64/aarch64-simd.md
8054 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
8056 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
8059 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
8061 PR tree-optimization/85989
8062 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
8064 (backprop::intersect_uses): Check it when deciding whether this
8065 is a backedge reference.
8066 (backprop::process_block): Add each phi to m_visited_phis
8067 after visiting it, then clear it at the end.
8069 2018-06-01 Richard Biener <rguenther@suse.de>
8071 * tree-vectorizer.h (vect_dr_stmt): New function.
8072 (vect_get_load_cost): Adjust.
8073 (vect_get_store_cost): Likewise.
8074 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
8075 Use vect_dr_stmt instead of DR_SMTT.
8076 (vect_record_base_alignments): Likewise.
8077 (vect_calculate_target_alignment): Likewise.
8078 (vect_compute_data_ref_alignment): Likewise and make static.
8079 (vect_update_misalignment_for_peel): Likewise.
8080 (vect_verify_datarefs_alignment): Likewise.
8081 (vector_alignment_reachable_p): Likewise.
8082 (vect_get_data_access_cost): Likewise. Pass down
8083 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
8084 (vect_get_peeling_costs_all_drs): Likewise.
8085 (vect_peeling_hash_get_lowest_cost): Likewise.
8086 (vect_enhance_data_refs_alignment): Likewise.
8087 (vect_find_same_alignment_drs): Likewise.
8088 (vect_analyze_data_refs_alignment): Likewise.
8089 (vect_analyze_group_access_1): Likewise.
8090 (vect_analyze_group_access): Likewise.
8091 (vect_analyze_data_ref_access): Likewise.
8092 (vect_analyze_data_ref_accesses): Likewise.
8093 (vect_vfa_segment_size): Likewise.
8094 (vect_small_gap_p): Likewise.
8095 (vectorizable_with_step_bound_p): Likewise.
8096 (vect_prune_runtime_alias_test_list): Likewise.
8097 (vect_analyze_data_refs): Likewise.
8098 (vect_supportable_dr_alignment): Likewise.
8099 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
8100 (vect_gen_prolog_loop_niters): Likewise.
8101 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
8102 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
8104 (vect_recog_mask_conversion_pattern): Likewise.
8105 (vect_try_gather_scatter_pattern): Likewise.
8106 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
8107 to vect_get_store_cost.
8108 (vect_get_store_cost): Get stmt_info instead of DR.
8109 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
8110 (vect_get_load_cost): Get stmt_info instead of DR.
8112 2018-06-01 Richard Biener <rguenther@suse.de>
8115 * gimple-fold.c (var_decl_component_p): Also allow offsetted
8116 vars wrapped in MEM_REFs.
8118 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
8120 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
8121 Fix subreg tests so that we only return a choice between
8122 GENERAL_REGS and FP_REGS if the original classes included both.
8124 2018-06-01 Richard Biener <rguenther@suse.de>
8127 * tree-ssa-structalias.c (get_function_part_constraint):
8128 Handle NULL fi->decl.
8129 (find_func_aliases_for_call): Properly handle indirect
8130 fi from direct call.
8131 (find_func_clobbers): Likewise.
8132 (ipa_pta_execute): Likewise.
8133 (create_variable_info_for): For functions that are ifunc_resolver
8134 resolve to a varinfo that contains the result of the resolver call.
8135 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
8138 2018-05-31 Michael Collison <michael.collison@arm.com>
8140 * config/aarch64/aarch64.md:
8141 (*fix_to_zero_extenddfdi2): New pattern.
8142 * gcc.target/aarch64/fix_extend1.c: New testcase.
8144 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
8148 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
8149 and BUILT_IN_STRNCMP_EQ.
8150 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
8151 BUILT_IN_STRNCMP_EQ.
8152 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
8153 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
8154 (gimple_fold_builtin): Likewise.
8155 * tree-ssa-strlen.c (compute_string_length): New function.
8156 (determine_min_obsize): New function.
8157 (handle_builtin_string_cmp): New function to handle calls to
8158 string compare functions.
8159 (strlen_optimize_stmt): Add handling to builtin string compare
8161 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
8162 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
8163 * tree.c (build_common_builtin_nodes): Add new defines of
8164 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
8166 2018-05-31 Jakub Jelinek <jakub@redhat.com>
8169 * bb-reorder.c (pass_partition_blocks::gate): Return false for
8170 functions with naked attribute.
8172 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
8174 * config/i386/sse.md (avx_vec_concat<mode>):
8175 Substitute concat_tg_mode mode attribute with xtg_mode.
8176 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
8177 (concat_tg_mode): Remove mode attribute.
8179 2018-05-31 Martin Sebor <msebor@redhat.com>
8182 * calls.c (alloc_max_size): Correct a logic error/typo.
8183 Treat excessive arguments as infinite. Warn for invalid arguments.
8184 * doc/invoke.texi (-Walloc-size-larger-than): Update.
8186 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
8189 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
8190 and movx for Haswell.
8192 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
8193 Cesar Philippidis <cesar@codesourcery.com>
8196 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
8197 when emitting error on private/firstprivate reductions.
8198 * omp-low.c (lower_omp_target): Avoid reference-type processing
8199 on pointers for firstprivate clause.
8201 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
8203 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
8206 * config/aarch64/aarch64-simd.md
8207 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
8208 (aarch64_ld1_x3_<mode>): Likewise
8209 (aarch64_st1x2<VALLDIF:mode>): Likewise
8210 (aarch64_st1_x2_<mode>): Likewise
8211 (aarch64_st1x3<VALLDIF:mode>): Likewise
8212 (aarch64_st1_x3_<mode>): Likewise
8213 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
8214 (vld1_s8_x3): Likewise.
8215 (vld1_u16_x3): Likewise.
8216 (vld1_s16_x3): Likewise.
8217 (vld1_u32_x3): Likewise.
8218 (vld1_s32_x3): Likewise.
8219 (vld1_u64_x3): Likewise.
8220 (vld1_s64_x3): Likewise.
8221 (vld1_f16_x3): Likewise.
8222 (vld1_f32_x3): Likewise.
8223 (vld1_f64_x3): Likewise.
8224 (vld1_p8_x3): Likewise.
8225 (vld1_p16_x3): Likewise.
8226 (vld1_p64_x3): Likewise.
8227 (vld1q_u8_x3): Likewise.
8228 (vld1q_s8_x3): Likewise.
8229 (vld1q_u16_x3): Likewise.
8230 (vld1q_s16_x3): Likewise.
8231 (vld1q_u32_x3): Likewise.
8232 (vld1q_s32_x3): Likewise.
8233 (vld1q_u64_x3): Likewise.
8234 (vld1q_s64_x3): Likewise.
8235 (vld1q_f16_x3): Likewise.
8236 (vld1q_f32_x3): Likewise.
8237 (vld1q_f64_x3): Likewise.
8238 (vld1q_p8_x3): Likewise.
8239 (vld1q_p16_x3): Likewise.
8240 (vld1q_p64_x3): Likewise.
8241 (vst1_s64_x2): Likewise.
8242 (vst1_u64_x2): Likewise.
8243 (vst1_f64_x2): Likewise.
8244 (vst1_s8_x2): Likewise.
8245 (vst1_p8_x2): Likewise.
8246 (vst1_s16_x2): Likewise.
8247 (vst1_p16_x2): Likewise.
8248 (vst1_s32_x2): Likewise.
8249 (vst1_u8_x2): Likewise.
8250 (vst1_u16_x2): Likewise.
8251 (vst1_u32_x2): Likewise.
8252 (vst1_f16_x2): Likewise.
8253 (vst1_f32_x2): Likewise.
8254 (vst1_p64_x2): Likewise.
8255 (vst1q_s8_x2): Likewise.
8256 (vst1q_p8_x2): Likewise.
8257 (vst1q_s16_x2): Likewise.
8258 (vst1q_p16_x2): Likewise.
8259 (vst1q_s32_x2): Likewise.
8260 (vst1q_s64_x2): Likewise.
8261 (vst1q_u8_x2): Likewise.
8262 (vst1q_u16_x2): Likewise.
8263 (vst1q_u32_x2): Likewise.
8264 (vst1q_u64_x2): Likewise.
8265 (vst1q_f16_x2): Likewise.
8266 (vst1q_f32_x2): Likewise.
8267 (vst1q_f64_x2): Likewise.
8268 (vst1q_p64_x2): Likewise.
8269 (vst1_s64_x3): Likewise.
8270 (vst1_u64_x3): Likewise.
8271 (vst1_f64_x3): Likewise.
8272 (vst1_s8_x3): Likewise.
8273 (vst1_p8_x3): Likewise.
8274 (vst1_s16_x3): Likewise.
8275 (vst1_p16_x3): Likewise.
8276 (vst1_s32_x3): Likewise.
8277 (vst1_u8_x3): Likewise.
8278 (vst1_u16_x3): Likewise.
8279 (vst1_u32_x3): Likewise.
8280 (vst1_f16_x3): Likewise.
8281 (vst1_f32_x3): Likewise.
8282 (vst1_p64_x3): Likewise.
8283 (vst1q_s8_x3): Likewise.
8284 (vst1q_p8_x3): Likewise.
8285 (vst1q_s16_x3): Likewise.
8286 (vst1q_p16_x3): Likewise.
8287 (vst1q_s32_x3): Likewise.
8288 (vst1q_s64_x3): Likewise.
8289 (vst1q_u8_x3): Likewise.
8290 (vst1q_u16_x3): Likewise.
8291 (vst1q_u32_x3): Likewise.
8292 (vst1q_u64_x3): Likewise.
8293 (vst1q_f16_x3): Likewise.
8294 (vst1q_f32_x3): Likewise.
8295 (vst1q_f64_x3): Likewise.
8296 (vst1q_p64_x3): Likewise.
8298 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
8300 * config/msp430/msp430.c (msp430_output_labelref): Prepend
8301 user_label_prefix to name.
8303 * tree-core.h: Update comment about the format of NAME string
8304 passed to handler in attribute_spec.
8306 * config/msp430/msp430.md: Remove erroneous subreg expression from
8307 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
8308 zero_extend{q,h}isi2.
8310 2018-05-30 Borislav Petkov <bp@suse.de>
8312 * doc/extend.texi: Document some architecture specific
8313 constraints and sort entries.
8315 2018-05-30 Martin Sebor <msebor@redhat.com>
8318 * builtins.c (expand_builtin_stpcpy_1): New function.
8319 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
8320 only if the former succeeds.
8322 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
8324 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
8327 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
8329 * doc/invoke.texi (-flinker-output): Document
8331 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
8333 * passes.c (ipa_write_summaries): Only modify statements if body
8335 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
8336 incrementally linking.
8337 (ipa_passes): Likewise.
8338 * lto-cgraph.c (lto_output_node): When incrementally linking do not
8339 pass down resolution info.
8340 * common.opt (flag_incremental_link): Update info.
8341 * gcc.c (plugin specs): Turn flinker-output=* to
8342 -plugin-opt=-linker-output-known
8343 * toplev.c (compile_file): Also cut compilation when doing incremental
8345 * flag-types. (enum lto_partition_model): Add
8346 LTO_LINKER_OUTPUT_NOLTOREL.
8347 (invoke.texi): Add -flinker-output docs.
8348 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
8349 link same way as WPA; do not stream in dead initializers.
8351 * dwarf2out.c (dwarf2out_die_ref_for_decl,
8352 darf2out_register_external_decl): Support incremental link.
8354 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
8356 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
8358 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
8360 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
8361 it down to simple_object_copy_lto_debug_sections.
8362 (run_gcc): Determine incremental LTO link time and configure
8363 lto1 into non-wpa mode, disable renaming of debug sections.
8365 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
8367 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
8368 descriptions of various incorrectly documented functions.
8370 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
8373 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
8374 address check not strict.
8376 2018-05-30 Richard Biener <rguenther@suse.de>
8378 PR tree-optimization/85964
8379 * tracer.c (better_p): Drop initialized count check, we only
8380 call the function with initialized counts now.
8381 (find_best_successor): Do find a best edge if one
8382 has uninitialized count.
8383 (find_best_predecessor): Likewise. Do BB frequency check only
8384 if count is initialized.
8386 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
8388 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
8389 (aarch64_ldrstr_offset_compare): New.
8390 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
8391 load/store orderings.
8392 (aarch64_gen_adjusted_ldpstp): Likewise.
8394 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
8396 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
8397 Check for subset of GENERAL_REGS and FP_REGS.
8398 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
8401 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
8403 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
8404 and wi::to_poly_offset. Add the current offset and then check
8405 whether the sum fits, rather than using an unchecked addition of
8406 a checked term. Check for a shwi rather than a uhwi.
8407 * expr.c (get_bit_range): Use tree_to_poly_uint64.
8408 (store_constructor): Use poly_int_tree_p.
8409 (expand_expr_real_1): Likewise.
8410 * function.c (assign_temp): Likewise.
8411 * fold-const.c (const_binop): Use poly_int_tree_p and
8413 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
8415 * ipa-icf-gimple.c (func_checker::compare_operand): Use
8416 to_poly_offset for MEM offsets.
8417 * ipa-icf.c (sem_variable::equals): Likewise.
8418 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
8419 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
8420 wi::to_poly_offset for BIT_FIELD_REF offsets.
8421 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
8423 * var-tracking.c (emit_note_insn_var_location): Use
8424 tree_to_poly_uint64.
8426 2018-05-29 Jim Wilson <jimw@sifive.com>
8428 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
8430 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
8433 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
8434 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
8436 (sse4_1_round<mode>2): Use nonimmediate_operand
8437 for operand 1 predicate.
8439 2018-05-29 Martin Sebor <msebor@redhat.com>
8440 Richard Biener <rguenther@suse.de>
8443 * calls.c (get_size_range): Call determine_value_range instead
8444 of get_value_range..
8445 * tree-vrp.h (determine_value_range): Declared new function.
8446 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
8448 2018-05-29 Richard Biener <rguenther@suse.de>
8450 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
8451 sure to use non-pattern stmts for get_earlier_stmt arguments.
8452 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
8453 called on pattern stmts.
8454 (get_later_stmt): Likewise.
8456 2018-05-29 Martin Liska <mliska@suse.cz>
8458 PR gcov-profile/85759
8459 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
8462 2018-05-29 Jakub Jelinek <jakub@redhat.com>
8464 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
8466 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
8470 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
8471 VEC_PACK_FLOAT_EXPR): New tree codes.
8472 * tree-pretty-print.c (op_code_prio): Handle
8473 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
8474 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
8475 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
8476 * tree-inline.c (estimate_operator_cost): Likewise.
8477 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
8478 * fold-const.c (const_binop): Likewise.
8479 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
8480 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
8481 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
8482 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
8483 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
8484 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
8485 * expr.c (expand_expr_real_2): Likewise.
8486 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
8487 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
8488 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
8490 * optabs.c (expand_widen_pattern_expr): For
8491 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
8492 sign from result type rather than operand's type.
8493 (expand_binop_directly): For vec_packu_float_optab and
8494 vec_packs_float_optab allow result type to be different from operand's
8496 * optabs-tree.c (optab_for_tree_code): Handle
8497 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
8498 VEC_PACK_FLOAT_EXPR. Formatting fixes.
8499 * tree-vect-generic.c (expand_vector_operations_1): Handle
8500 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
8501 VEC_PACK_FLOAT_EXPR.
8502 * tree-vect-stmts.c (supportable_widening_operation): Handle
8504 (supportable_narrowing_operation): Handle FLOAT_EXPR.
8505 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
8506 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
8507 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
8508 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
8510 (vec_pack<floatprefix>_float_<mode>): New expander.
8511 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
8513 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
8514 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
8515 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
8516 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
8517 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
8519 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
8520 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
8521 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
8522 VEC_PACK_FLOAT_EXPR): Document.
8524 2018-05-29 Richard Biener <rguenther@suse.de>
8526 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
8528 (stmt_vec_info_vec): Make pointer.
8529 (init_stmt_vec_info_vec): Remove.
8530 (free_stmt_vec_info_vec): Likewise.
8531 (set_stmt_vec_info_vec): New function.
8532 (free_stmt_vec_infos): Likewise.
8533 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
8534 (set_vinfo_for_stmt): Likewise.
8535 (get_earlier_stmt): Likewise.
8536 (get_later_stmt): Likewise.
8537 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
8538 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
8539 (vec_info::~vec_info): Free stmt_vec_infos.
8540 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
8541 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
8542 (pass_slp_vectorize::execute): Likewise.
8543 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
8544 (free_stmt_vec_info_vec): Likewise.
8545 (set_stmt_vec_info_vec): New function.
8546 (free_stmt_vec_infos): Likewise.
8547 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
8548 the global stmt_vec_info_vec.
8549 * tree-parloops.c (gather_scalar_reductions): Use
8550 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
8553 2018-05-29 Richard Biener <rguenther@suse.de>
8555 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
8557 2018-05-29 Martin Liska <mliska@suse.cz>
8558 David Malcolm <dmalcolm@redhat.com>
8560 * vec.c (test_reverse): New.
8561 (vec_c_tests): Add new test.
8562 * vec.h (vl_ptr>::reverse): New function.
8564 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
8566 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
8568 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
8571 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
8573 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
8575 2018-05-28 Richard Biener <rguenther@suse.de>
8577 PR tree-optimization/85933
8578 * tree-vect-data-refs.c (vect_record_base_alignments): Only
8579 look at stmts marked as vectorizable.
8581 2018-05-28 Richard Biener <rguenther@suse.de>
8583 PR tree-optimization/85934
8584 * tree-vect-generic.c (expand_vector_operations_1): Hoist
8585 vector boolean check before scalar optimization.
8587 2018-05-28 Jakub Jelinek <jakub@redhat.com>
8589 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
8592 2018-05-28 Mark Wielaard <mark@klomp.org>
8594 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
8595 if it is an expression containing a minus sign.
8597 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
8599 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
8601 2018-05-27 Paul Koning <ni1d@arrl.net>
8603 * config/pdp11/pdp11.md (truncsihi2): Remove.
8605 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
8606 Chung-Ju Wu <jasonwucj@gmail.com>
8608 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
8610 (unaligned_store_dw): Ditto.
8611 * config/nds32/nds32-memory-manipulation.c
8612 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
8613 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
8614 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
8615 (emit_setmem_word_loop): Rename to ...
8616 (emit_setmem_doubleword_loop): ... this.
8617 (nds32_gen_dup_4_byte_to_word_value): New function.
8618 (nds32_gen_dup_8_byte_to_double_word_value): New function.
8619 (nds32_expand_setmem_loop): Refine implementation.
8620 (nds32_expand_setmem_loop_v3m): Ditto.
8621 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
8624 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
8626 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
8628 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
8630 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
8631 (nds32_init_machine_status): Initialize machine->attr_naked_p and
8632 machine->attr_no_prologue_p.
8633 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
8634 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
8635 (nds32_expand_epilogue): Consider attr_naked_p.
8636 (nds32_expand_epilogue_v3pop): Likewise.
8637 (nds32_can_use_return_insn): Likewise.
8638 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
8639 attr_no_prologue_p fields.
8640 * config/nds32/nds32.opt (mret-in-naked-func): New option.
8642 2018-05-27 Jakub Jelinek <jakub@redhat.com>
8645 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
8647 * config/i386/sse.md
8648 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
8650 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
8652 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
8654 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
8656 (*<floatsuffix>floatv2div2sf2): Rename to ...
8657 (*float<floatunssuffix>v2div2sf2): ... this.
8658 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
8659 (float<floatunssuffix>v2div2sf2_mask): ... this.
8660 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
8661 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
8662 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
8664 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
8666 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
8668 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
8670 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
8672 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
8674 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
8675 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
8676 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
8677 gen_ufix_truncv8dfv8si2.
8678 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
8679 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
8680 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
8681 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
8682 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
8683 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
8684 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
8685 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
8687 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
8691 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
8693 2018-05-25 Jim Wilson <jimw@sifive.com>
8695 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
8696 * config/riscv/riscv.c (struct machine_function): Add
8697 interrupt_handler_p and attribute_checked_p fields.
8698 (riscv_attribute_table): Add interrupt.
8699 (riscv_interrupt_type_p): New.
8700 (riscv_save_reg_p): Save extra regs for interrupt handler.
8701 (riscv_use_save_libcall): Return false for interrupt handler.
8702 (riscv_first_stack_step): Add forward declaration.
8703 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
8704 for interrupt handler with large frame. Use it for saved reg list.
8705 (riscv_expand_prologue): Move flag_stack_usage_info support to
8706 eliminate duplication.
8707 (riscv_expand_epilogue): Generate mret for interrupt handler.
8708 (riscv_epilogue_uses): New.
8709 (riscv_can_use_return_insn): Return false for interrupt handler.
8710 (riscv_function_ok_for_sibcall): Likewise.
8711 (riscv_set_current_function): Add interrupt handler support.
8712 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
8713 * config/riscv/riscv.md (UNSPECV_MRET): New.
8715 (riscv_frflags, riscv_fsflags): Use tab after opcode.
8717 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
8719 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
8721 PR tree-optimization/85712
8722 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
8723 this candidate has already been replaced in-situ by a copy.
8725 2018-05-25 Jason Merrill <jason@redhat.com>
8727 PR c++/80485 - inline function non-zero address.
8728 * symtab.c (nonzero_address): Check DECL_COMDAT.
8730 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
8733 * config/alpha/alpha.md (ashlsi3): New insn pattern.
8734 (*ashlsi_se): Rename from *ashldi_se. Define as sign
8735 extension of SImode operation. Use const123_operand predicate.
8736 (*saddsi_1): Remove.
8737 (*saddl_se_1): Ditto.
8739 (*ssubl_se_1): Ditto.
8740 * config/alpha/predicates.md (const123_operand): New predicate.
8741 * config/alpha/constraints.md (P): Use IN_RANGE.
8743 2018-05-25 Richard Biener <rguenther@suse.de>
8745 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
8747 (ref_maybe_used_by_stmt_p): Likewise.
8748 (stmt_may_clobber_ref_p): Likewise.
8749 (stmt_may_clobber_ref_p_1): Likewise.
8750 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
8752 (ref_maybe_used_by_stmt_p): Likewise.
8753 (stmt_may_clobber_ref_p): Likewise.
8754 (stmt_may_clobber_ref_p_1): Likewise.
8755 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
8756 the alias oracle to disambiguate DRs with stmts DR analysis
8758 (vect_analyze_data_refs): Do not give up on not analyzable
8759 DRs for BB vectorization. Remove code truncating the dataref
8762 2018-05-25 Jakub Jelinek <jakub@redhat.com>
8765 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
8766 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
8767 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
8769 2018-05-25 Richard Biener <rguenther@suse.de>
8771 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
8772 function, combining stmt data ref gathering and fatal analysis
8774 (vect_analyze_data_refs): Remove now redudnant code and simplify.
8775 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
8776 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
8777 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
8778 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
8780 2018-05-25 Bin Cheng <bin.cheng@arm.com>
8782 PR tree-optimization/85720
8783 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
8784 SCC if all partitions are builtins.
8785 (version_loop_by_alias_check): New parameter. Generate cancelable
8786 runtime alias check if all partitions are builtins.
8787 (distribute_loop): Update call to above function.
8789 2018-05-25 Bin Cheng <bin.cheng@arm.com>
8791 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
8792 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
8793 (parm_default_def_partition_arg): Ditto.
8794 (set_parm_default_def_partition): Ditto.
8795 (get_parm_default_def_partitions): Ditto and make it static.
8796 (get_undefined_value_partitions): Ditto and make it static.
8797 (remove_ssa_form): Refactor call to init_var_map here.
8798 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
8799 computation for loop region.
8800 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
8801 (register_default_def): Delete.
8802 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
8803 (parm_default_def_partition_arg): Ditto.
8804 (set_parm_default_def_partition): Ditto.
8805 (get_parm_default_def_partitions): Ditto and make it static.
8806 (get_undefined_value_partitions): Ditto and make it static.
8807 (coalesce_with_default, coalesce_with_default): Update comment.
8808 (create_coalesce_list_for_region): New func factored out from
8809 create_outofssa_var_map.
8810 (populate_coalesce_list_for_outofssa): New func factored out from
8811 create_outofssa_var_map and coalesce_ssa_name.
8812 (create_outofssa_var_map): Delete.
8813 (coalesce_ssa_name): Refactor to support live range computation.
8814 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
8815 (get_parm_default_def_partitions): Delete.
8816 (get_undefined_value_partitions): Ditto.
8817 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
8818 computation for loop region.
8819 (new_tree_live_info, loe_visit_block): Ditto.
8820 (live_worklist, set_var_live_on_entry): Ditto.
8821 (calculate_live_on_exit, verify_live_on_entry): Ditto.
8822 * tree-ssa-live.h (struct _var_map): New fields.
8823 (init_var_map): Change decl.
8824 (region_contains_p): New.
8826 2018-05-25 Bin Cheng <bin.cheng@arm.com>
8828 * tree-ssa-live.h (live_merge_and_clear): Delete.
8830 2018-05-25 Richard Biener <rguenther@suse.de>
8833 * tree-dump.c (dequeue_and_dump): Remove access to removed
8834 operand 2 of a SWITCH_EXPR.
8836 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
8838 * doc/sourcebuild.texi (vect_double_cond_arith): Include
8839 multiplication and division.
8840 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
8841 (cond_udiv@var{m}, cond_umod@var{m}): Document.
8842 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
8843 (cond_udiv_optab, cond_umod_optab): New optabs.
8844 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
8845 (IFN_COND_RDIV): New internal functions.
8846 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
8847 TRUNC_MOD_EXPR and RDIV_EXPR.
8848 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
8849 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
8851 (SVE_INT_BINARY): Include mult.
8852 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
8853 (optab, sve_int_op): Handle mult.
8854 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
8856 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
8857 for SVE_INT_BINARY_SD.
8859 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
8861 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
8862 (optab, sve_int_op): Handle div and udiv.
8863 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
8864 for SVE_INT_BINARY_SD.
8865 (*<optab><mode>3): New insn for the same.
8867 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
8869 * tree-vect-patterns.c: Include predict.h.
8870 (vect_recog_divmod_pattern): Restrict check for division support
8871 to when optimizing for size.
8873 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
8875 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
8876 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
8877 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
8878 (gimple_match_op::set_op): Likewise.
8879 (gimple_resimplify4): Declare.
8880 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
8881 (expr::gen_transform): Likewise.
8882 (decision_tree::gen): Generate a simplification routine for 4 operands.
8883 * gimple-match-head.c (gimple_simplify): Add an overload for
8884 4 operands. In the top-level function, handle up to 4 call
8885 arguments and call gimple_resimplify4.
8886 (gimple_resimplify4): New function.
8887 (build_call_internal): Pass a fourth operand.
8888 (maybe_push_to_seq): Likewise.
8889 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
8890 Fold VEC_COND_EXPRs of an operation and a default value into
8891 an IFN_COND_* function if possible.
8892 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
8894 (SVE_COND_FP_BINARY): Include them.
8895 (optab, sve_fp_op): Handle them.
8896 (SVE_INT_BINARY_REV): New code iterator.
8897 (SVE_COND_FP_BINARY_REV): New int iterator.
8898 (commutative): New int attribute.
8899 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
8901 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
8903 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
8904 (*cond_<optab><mode>): New patterns for reversed operands.
8906 2018-05-25 Richard Biener <rguenther@suse.de>
8908 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
8909 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
8910 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
8911 (STMT_VINFO_GROUPED_ACCESS): Adjust.
8912 * tree-vect-data-refs.c (everywhere): Adjust users.
8913 * tree-vect-loop.c (everywhere): Likewise.
8914 * tree-vect-slp.c (everywhere): Likewise.
8915 * tree-vect-stmts.c (everywhere): Likewise.
8916 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
8918 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8920 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
8922 (gcc_cv_as_section_exclude): ... this.
8923 Try Solaris as #exclude syntax.
8924 * configure: Regenerate.
8925 * config.in: Regenerate.
8926 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
8928 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
8929 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
8931 * varasm.c (default_elf_asm_named_section): Don't check if
8932 HAVE_GAS_SECTION_EXCLUDE is defined.
8934 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
8936 * doc/md.texi: Update the documentation of the cond_* optabs
8937 to mention the new final operand. Fix GET_MODE_NUNITS call.
8938 Describe the scalar case too.
8939 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
8940 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
8942 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
8943 (get_conditional_internal_fn): Update comment.
8944 * tree-vect-loop.c (vectorizable_reduction): Pass the original
8945 accumulator value as a final argument to conditional functions.
8946 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
8947 a define_expand and add an "else" operand. Assert for now that
8948 the else operand is equal to operand 2. Use SVE_INT_BINARY and
8949 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
8950 (*cond_<optab><mode>): New patterns.
8951 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
8952 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
8953 (UNSPEC_COND_EOR): Delete.
8954 (optab): Remove associated mappings.
8955 (SVE_INT_BINARY): New code iterator.
8956 (sve_int_op): Remove int attribute and add "minus" to the code
8958 (SVE_COND_INT_OP): Delete.
8959 (SVE_COND_FP_OP): Rename to...
8960 (SVE_COND_FP_BINARY): ...this.
8962 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
8964 * optabs.c (can_reuse_operands_p): New function.
8965 (maybe_legitimize_operands): Try to reuse the results for
8968 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
8970 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
8971 Add {q} suffix to insn mnemonic.
8973 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
8975 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
8976 (msp430_warn_func_return): New.
8978 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
8980 * fold-const.c (tree_nonzero_bits): New function.
8981 * fold-const.h (tree_nonzero_bits): Likewise.
8982 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
8983 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
8985 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
8989 * varasm.c (assemble_alias): Check ifunc_resolver only on
8992 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
8995 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
8996 when memory input operand is handled.
8998 2018-05-24 Luis Machado <luis.machado@linaro.org>
9000 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
9002 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
9004 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
9006 * match.pd: Delay FMA folds until after vectorization.
9008 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
9011 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
9012 address check not strict.
9014 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
9016 * gimple-match.h (gimple_match_op): New class.
9017 (mprts_hook): Replace parameters with a gimple_match_op *.
9018 (maybe_build_generic_op): Likewise.
9019 (gimple_simplified_result_is_gimple_val): Replace parameters with
9020 a const gimple_match_op *.
9021 (gimple_simplify): Replace code_helper * and tree * parameters with
9022 a gimple_match_op * parameter.
9023 (gimple_resimplify1): Replace code_helper *, tree and tree *
9024 parameters with a gimple_match_op * parameter.
9025 (gimple_resimplify2): Likewise.
9026 (gimple_resimplify3): Likewise.
9027 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
9028 parameters with a gimple_match_op * parameter.
9029 * gimple-match-head.c (gimple_simplify): Change prototypes of
9030 auto-generated functions to take a gimple_match_op * instead of
9031 separate code_helper * and tree * parameters. Make the same
9032 change in the top-level overload and update calls to the
9033 gimple_resimplify routines. Update calls to the auto-generated
9034 functions and to maybe_push_res_to_seq in the publicly-facing
9035 operation-specific gimple_simplify overloads.
9036 (gimple_match_op::MAX_NUM_OPS): Define.
9037 (gimple_resimplify1): Replace rcode and ops with a single res_op
9038 parameter. Update call to gimple_simplify.
9039 (gimple_resimplify2): Likewise.
9040 (gimple_resimplify3): Likewise.
9041 (mprts_hook): Replace parameters with a gimple_match_op *.
9042 (maybe_build_generic_op): Likewise.
9043 (build_call_internal): Replace type, nargs and ops with
9044 a gimple_match_op *.
9045 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
9046 with a single gimple_match_op *. Update calls to mprts_hook,
9047 build_call_internal and gimple_simplified_result_is_gimple_val.
9048 Factor out code that is common to the tree_code and combined_fn cases.
9049 * genmatch.c (expr::gen_transform): Replace tem_code and
9050 tem_ops with a gimple_match_op called tem_op. Update calls
9051 to the gimple_resimplify functions and maybe_push_res_to_seq.
9052 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
9053 res_ops. Update call to the gimple_resimplify functions.
9054 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
9055 (decision_tree::gen): Make the functions take a gimple_match_op *
9056 called res_op instead of separate res_code and res_ops parameters.
9057 Update call accordingly.
9058 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
9059 and ops with a single res_op parameter. Update calls to
9060 maybe_build_generic_op and maybe_push_res_to_seq.
9061 (fold_stmt_1): Update calls to gimple_simplify and
9062 replace_stmt_with_simplification.
9063 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
9064 and gimple_simplified_result_is_gimple_val.
9065 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
9067 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
9068 with a gimple_match_op *.
9069 (vn_nary_build_or_lookup): Likewise. Update call to
9070 vn_nary_build_or_lookup_1.
9071 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
9072 gimple_match_op *. Update calls to the gimple_resimplify routines
9073 and to gimple_simplified_result_is_gimple_val.
9074 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
9075 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
9076 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
9077 (visit_nary_op): Likewise.
9078 (visit_reference_op_load): Likewise.
9080 2018-05-23 Luis Machado <luis.machado@linaro.org>
9082 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
9083 modifier for printing the step amount.
9085 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
9088 * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
9091 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
9093 * doc/sourcebuild.texi (Endianness): New subsubsection.
9095 2018-05-23 Luis Machado <luis.machado@linaro.org>
9097 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
9098 <prefetch_dynamic_strides>: New const bool field.
9099 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
9100 prefetch_dynamic_strides.
9101 (exynosm1_prefetch_tune): Likewise.
9102 (thunderxt88_prefetch_tune): Likewise.
9103 (thunderx_prefetch_tune): Likewise.
9104 (thunderx2t99_prefetch_tune): Likewise.
9105 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
9107 (aarch64_override_options_internal): Update to set
9108 PARAM_PREFETCH_DYNAMIC_STRIDES.
9109 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
9110 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
9111 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
9112 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
9113 prefetch-dynamic-strides setting.
9115 2018-05-23 Luis Machado <luis.machado@linaro.org>
9117 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
9118 <minimum_stride>: New const int field.
9119 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
9120 minimum_stride field defaulting to -1.
9121 (exynosm1_prefetch_tune): Likewise.
9122 (thunderxt88_prefetch_tune): Likewise.
9123 (thunderx_prefetch_tune): Likewise.
9124 (thunderx2t99_prefetch_tune): Likewise.
9125 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
9126 <default_opt_level>: Set to 3.
9127 (aarch64_override_options_internal): Update to set
9128 PARAM_PREFETCH_MINIMUM_STRIDE.
9129 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
9130 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
9131 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
9132 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
9133 stride is constant and is below the minimum stride threshold.
9135 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9137 * config/arm/arm-cpus.in (mode26): Delete.
9138 (armv4): Delete mode26 reference.
9139 * config/arm/arm.c (arm_configure_build_target): Delete use of
9142 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
9144 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
9146 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
9147 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
9149 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
9150 (floatunsdidf2): Ditto.
9152 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
9154 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
9155 (fixuns_trunc<mode>si2_avx512f): Ditto.
9156 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
9157 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
9158 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
9160 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
9162 PR rtl-optimization/79985
9163 * df-scan.c (df_insn_refs_collect): Remove special case for
9164 global registers and asm statements.
9166 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
9168 * extend.texi (Global Register Variables): Rewrite the bullet list.
9169 Note that the register is available for allocation. Note that access
9170 via inline asm must use constraints. Add note about async-signal
9171 handlers. Remove paragraph about automagic register selection.
9173 2018-05-23 Richard Biener <rguenther@suse.de>
9175 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
9176 of fixed offset from memset VN.
9178 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
9180 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
9182 (alloc_cand_and_find_basis): Initialize first_interp field.
9183 (slsr_process_mul): Modify first_interp field.
9184 (slsr_process_add): Likewise.
9185 (slsr_process_cast): Modify first_interp field for each new
9187 (slsr_process_copy): Likewise.
9188 (dump_candidate): Dump first_interp field.
9189 (replace_mult_candidate): Process all interpretations, not just
9191 (replace_rhs_if_not_dup): Likewise.
9192 (replace_one_candidate): Likewise.
9194 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
9196 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
9198 (aarch64_needs_frame_chain): New function.
9199 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
9201 2018-05-23 Sudakshina Das <sudi.das@arm.com>
9204 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
9205 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
9206 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
9207 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
9208 as true for strict-align.
9209 (aarch64_can_inline_p): Perform checks even when callee has no
9210 attributes to check for strict alignment.
9211 * doc/extend.texi (AArch64 Function Attributes): Document
9213 * doc/invoke.texi: (AArch64 Options): Likewise.
9215 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
9217 PR tree-optimization/85853
9218 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
9219 the handling of the root of the node to...
9220 (vect_slp_analyze_node_operations_1): ...this new function,
9221 and run the whole thing with the child nodes' def types
9222 set according to their SLP node's def type.
9224 2018-05-23 Richard Biener <rguenther@suse.de>
9227 * tree-data-ref.c (create_runtime_alias_checks): Defer
9228 and ignore overflow warnings.
9230 2018-05-23 Yury Gribov <tetra2005@gmail.com>
9232 PR tree-optimization/85822
9233 * tree-vrp.c (is_masked_range_test): Fix handling of negative
9236 2018-05-23 Richard Biener <rguenther@suse.de>
9238 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
9239 memset constants via native_interpret_expr.
9241 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
9244 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
9246 (cgraph_node::create_alias): Likewise.
9247 (cgraph_node::get_availability): Check ifunc_resolver instead
9248 of looking up ifunc attribute.
9249 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
9250 * varasm.c (do_assemble_alias): Likewise.
9251 (assemble_alias): Likewise.
9252 (default_binds_local_p_3): Likewise.
9253 * cgraph.h (cgraph_node): Add ifunc_resolver.
9254 (cgraph_node::only_called_directly_or_aliased_p): Return false
9256 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
9258 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
9259 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
9260 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
9261 instead of looking up ifunc attribute.
9263 2018-05-22 Luis Machado <luis.machado@linaro.org>
9265 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
9267 2018-05-22 Martin Sebor <msebor@redhat.com>
9270 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
9271 only when expasion succeeds.
9272 (expand_builtin_strcmp): Same.
9273 (expand_builtin_strncmp): Same.
9275 2018-05-22 Martin Sebor <msebor@redhat.com>
9277 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
9279 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
9280 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9282 * config/aarch64/aarch64-ldpstp.md: Replace uses of
9283 aarch64_mem_pair_operand with memory_operand and delete operand swapping
9285 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
9286 Add check for legitimate_address.
9287 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
9288 (aarch64_swap_ldrstr_operands): New.
9289 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
9292 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
9293 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9295 * config/aarch64/aarch64.md: New patterns to generate stp
9297 (store_pair_sw, store_pair_dw): New patterns to generate stp for
9298 single words and double words.
9299 (load_pair_sw, load_pair_dw): Likewise.
9300 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
9302 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
9304 * config/aarch64/aarch64-ldpstp.md: Modify peephole
9305 for different mode ldpstp and add peephole for merged zero stores.
9307 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
9309 (aarch64_gen_store_pair): Rename calls to match new patterns.
9310 (aarch64_gen_load_pair): Rename calls to match new patterns.
9311 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
9312 (load_pair<DREG:mode><DREG2:mode>): ... This.
9313 (store_pair<mode>): Rename to...
9314 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
9315 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
9317 (V_INT_EQUIV): Handle SImode.
9318 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
9321 2018-05-22 Martin Sebor <msebor@redhat.com>
9324 * calls.c (maybe_warn_nonstring_arg): Use string length to set
9325 or ajust the presumed bound on an operation to avoid unnecessary
9328 2018-05-22 Martin Sebor <msebor@redhat.com>
9330 PR tree-optimization/85826
9331 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
9332 assuming that a DECL necesarily has a constant size.
9334 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
9337 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
9339 2018-05-22 Richard Biener <rguenther@suse.de>
9341 PR tree-optimization/85834
9342 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
9343 non-constant and non-zero memset arguments.
9345 2018-05-22 Martin Liska <mliska@suse.cz>
9348 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
9350 2018-05-22 Richard Biener <rguenther@suse.de>
9352 PR tree-optimization/85863
9353 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
9354 comparisons when vectype is specified.
9355 (vectorizable_condition): Do not specify vectype for
9356 vect_is_simple_cond when SLP vectorizing.
9358 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
9361 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
9362 define __ibm128 as long double.
9363 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
9364 as a distinct type when IEEE 128-bit support is enabled.
9365 (init_float128_ieee): Fix up conversions between IFmode and IEEE
9366 128-bit types to use the correct functions.
9367 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
9368 convert between 128-bit floating point types that have different
9369 modes but the same representation, instead of using gen_lowpart to
9371 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
9373 (IFKF_reg): New attributes to give the register constraints for
9375 (extend<mode>tf2_internal): New insns to mark an explicit
9376 conversion between 128-bit floating point types that have a
9377 different mode but share the same representation.
9379 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
9381 PR tree-optimization/85814
9382 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
9383 a null return from get_strinfo when unsharing the next
9384 strinfo in the chain.
9386 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
9389 * varasm.c (weak_finish): Clean up weak_decls.
9391 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9393 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
9394 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
9395 UNSPEC_UADALP values.
9396 * config/aarch64/iterators.md (ABAL): New int iterator.
9399 (sur): Add mappings for the above.
9400 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
9402 (aarch64_<sur>abal<mode>_4): Likewise.
9403 (aarch64_<sur>adalp<mode>_3): Likewise.
9404 (<sur>sadv16qi): New define_expand.
9406 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
9408 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
9409 (*movdf_internal): Ditto.
9410 (*rcpsf2_sse): Ditto.
9411 (*rsqrtsf2_sse): Ditto.
9412 (*sqrt<mode>2_sse): Ditto.
9414 2018-05-21 Tamar Christina <tamar.christina@arm.com>
9416 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
9418 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
9419 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
9420 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
9421 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
9423 * config/aarch64/arm_neon.h: Likewise.
9424 * config/aarch64/iterators.md (VQ_I): New.
9426 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
9428 * config.gcc: Add arc/t-multilib-linux to tmake_file for
9430 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
9433 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
9435 * config/nds32/constraints.md (S): New constraint.
9436 * config/nds32/nds32.md (call_internal): Use constraint S.
9437 (call_value_internal): Likewise.
9438 (sibcall_internal): Likewise.
9439 (sibcall_value_internal): Likewise.
9441 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
9442 Chung-Ju Wu <jasonwucj@gmail.com>
9444 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
9447 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
9448 Chung-Ju Wu <jasonwucj@gmail.com>
9450 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
9451 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
9452 (nds32_rtx_costs_impl): Simplify.
9453 (nds32_address_cost_impl): Simplify.
9454 (nds32_init_rtx_costs): New function.
9455 (nds32_rtx_costs_speed_prefer): Likewise.
9456 (nds32_rtx_costs_size_prefer): Likewise.
9457 (nds32_address_cost_speed_prefer): Likewise.
9458 (nds32_address_cost_speed_fwprop): Likewise.
9459 (nds32_address_cost_size_prefer): Likewise.
9460 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
9461 * config/nds32/nds32.c (nds32_option_override): Use
9462 nds32_init_rtx_costs function.
9464 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
9466 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
9467 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
9468 (TARGET_PIPELINE_N8): Likewise.
9469 (TARGET_PIPELINE_N10): Likewise.
9470 (TARGET_PIPELINE_N13): Likewise.
9471 (TARGET_PIPELINE_GRAYWOLF): Likewise.
9473 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
9475 * config/nds32/nds32-fpu.md: Update copyright year.
9477 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
9479 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
9481 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
9483 * config/nds32/nds32.c
9484 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
9485 * config/nds32/nds32.opt (minline-asm-r15): New option.
9487 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
9489 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
9491 * config/nds32/nds32.md (abssi2): New pattern.
9493 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
9495 * config/i386/i386.md (rex64namesuffix): New mode attribute.
9496 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
9497 Merge insn pattern from sse_cvtsi2ss<round_name> and
9498 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
9499 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
9500 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
9501 using SWI48 mode iterator.
9502 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
9503 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
9504 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
9505 pattern from sse_cvttss2si<round_saeonly_name>
9506 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
9507 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
9508 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
9509 using SWI48 mode iterator.
9510 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
9511 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
9512 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
9513 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
9514 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
9515 using SWI48 mode iterator.
9516 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
9517 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
9518 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
9519 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
9520 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
9521 SWI48 mode iterator.
9522 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
9523 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
9524 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
9525 pattern from sse_cvttsd2si<round_saeonly_name>
9526 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
9528 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
9530 * config/nds32/nds32-md-auxiliary.c
9531 (nds32_valid_smw_lwm_base_p): Refine.
9532 (nds32_output_smw_single_word): Refine.
9533 (nds32_output_smw_double_word): New.
9534 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
9536 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
9538 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
9539 (nds32_output_stack_pop): Refine.
9540 (nds32_expand_unaligned_load): Refine.
9541 (nds32_expand_unaligned_store): Refine.
9543 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
9544 Chung-Ju Wu <jasonwucj@gmail.com>
9546 * config/nds32/constants.md: Add TP_REGNUM constant.
9547 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
9548 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
9550 * config/nds32/nds32-doubleword.md: Consider flag_pic.
9551 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
9552 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
9553 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
9554 and PIC code generation.
9555 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
9557 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
9559 * config/nds32/nds32.md: Support TLS and PIC.
9560 * config/nds32/nds32.c: Support TLS and PIC.
9561 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
9562 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
9565 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
9567 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
9568 mode with E_ prefix.
9570 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
9571 Chung-Ju Wu <jasonwucj@gmail.com>
9573 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
9574 * config/nds32/nds32-md-auxiliary.c
9575 (symbolic_reference_mentioned_p): New.
9576 (nds32_legitimize_ict_address): New.
9577 (nds32_expand_ict_move): New.
9578 (nds32_indirect_call_referenced_p): New.
9579 (nds32_symbol_binds_local_p): Delete.
9580 (nds32_long_call_p): Modify.
9581 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
9582 * config/nds32/nds32-protos.h
9583 (symbolic_reference_mentioned_p): Declare.
9584 (nds32_legitimize_ict_address): Declare.
9585 (nds32_expand_ict_move): Declare.
9586 (nds32_indirect_call_referenced_p): Declare.
9587 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
9588 (nds32_relax_group): Use nds32_ict_const_p as condition.
9589 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
9590 (nds32_asm_file_start): Output ict_model directive in asm code.
9591 (nds32_legitimate_address_p): Consider indirect call.
9592 (nds32_print_operand): Consider indirect call.
9593 (nds32_print_operand_address): Consider indirect call.
9594 (nds32_insert_attributes): Handle "indirect_call" attribute.
9595 (TARGET_LEGITIMATE_ADDRESS_P): Define.
9596 (TARGET_LEGITIMATE_CONSTANT_P): Define.
9597 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
9598 (TARGET_DELEGITIMIZE_ADDRESS): Define.
9599 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
9600 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
9601 (TARGET_ICT_MODEL_SMALL): Define.
9602 (TARGET_ICT_MODEL_LARGE): Define.
9603 * config/nds32/nds32.md (movsi): Consider ict model.
9604 (call, call_value): Consider ict model.
9605 (sibcall, sibcall_value): Consider ict model.
9606 * config/nds32/nds32.opt (mict-model): New option.
9607 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
9610 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
9611 Monk Chiang <sh.chiang04@gmail.com>
9612 Jim Wilson <jimw@sifive.com>
9614 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
9615 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
9616 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
9617 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
9618 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
9619 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
9620 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
9621 compute save_libcall_adjustment properly.
9622 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
9623 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
9624 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
9625 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
9626 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
9627 (ABI_SPEC): Handle mabi=ilp32e.
9628 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
9629 (RVE): Add RVE mask.
9630 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
9631 <-march>: Add rv32e as an example.
9633 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
9636 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
9637 (intra_create_variable_infos): Handle C++ constructors.
9639 2018-05-18 Martin Liska <mliska@suse.cz>
9641 * passes.def: Remove a redundant pass.
9643 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
9646 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
9648 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9650 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
9652 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
9654 (armv2, armv2a, armv3, armv3m): Delete architectures.
9655 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
9656 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
9657 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
9659 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
9660 (*mulsidi3adddi): Likewise.
9661 (mulsidi3): Likewise.
9662 (*mulsidi3_nov6): Likewise.
9663 (umulsidi3): Likewise.
9664 (umulsidi3_nov6): Likewise.
9665 (umaddsidi4): Likewise.
9666 (*umulsidi3adddi): Likewise.
9667 (smulsi3_highpart): Likewise.
9668 (*smulsi3_highpart_nov6): Likewise.
9669 (umulsi3_highpart): Likewise.
9670 (*umulsi3_highpart_nov6): Likewise.
9671 * config/arm/arm.h (arm_arch3m): Delete.
9672 * config/arm/arm.c (arm_arch3m): Delete.
9673 (arm_option_override_internal): Update armv3-related comment.
9674 (arm_configure_build_target): Delete use of isa_bit_mode32.
9675 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
9676 (arm_rtx_costs_internal): Delete check of arm_arch3m.
9677 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
9679 (mulusa3): Likewise.
9680 * config/arm/arm-protos.h (arm_arch3m): Delete.
9681 * config/arm/arm-tables.opt: Regenerate.
9682 * config/arm/arm-tune.md: Likewise.
9683 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
9684 deleted architectures.
9686 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9688 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
9689 (armv5t, armv5te): New features.
9690 (ARMv5, ARMv5e): Delete fgroups.
9691 (ARMv5t, ARMv5te): Adjust for above changes.
9693 (armv5, armv5e): Delete arches.
9694 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
9696 (*call_reg_arm): Likewise.
9697 (*call_value_reg_armv5): Likewise.
9698 (*call_value_reg_arm): Likewise.
9699 (*call_symbol): Likewise.
9700 (*call_value_symbol): Likewise.
9701 (*sibcall_insn): Likewise.
9702 (*sibcall_value_insn): Likewise.
9704 (prefetch): Likewise.
9705 (define_split and define_peephole2 dependent on arm_arch5):
9707 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
9709 (TARGET_ARM_QBIT): Likewise.
9710 (TARGET_DSP_MULTIPLY): Likewise.
9711 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
9712 (arm_arch5, arm_arch5e): Delete.
9713 (arm_arch5t, arm_arch5te): Declare.
9714 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
9715 (arm_arch5t): Declare.
9716 (arm_option_reconfigure_globals): Update for the above.
9717 (arm_options_perform_arch_sanity_checks): Update comment, replace
9718 use of arm_arch5 with arm_arch5t.
9719 (use_return_insn): Likewise.
9720 (arm_emit_call_insn): Likewise.
9721 (output_return_instruction): Likewise.
9722 (arm_final_prescan_insn): Likewise.
9723 (arm_coproc_builtin_available): Likewise.
9724 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
9725 arm_arch5e with arm_arch5t and arm_arch5te.
9726 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
9727 (arm_arch5t, arm_arch5te): Declare.
9728 * config/arm/arm-tables.opt: Regenerate.
9729 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
9730 * config/arm/t-multilib: Likewise.
9731 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
9732 instead of arm_arch5.
9733 (*call_reg_thumb1): Likewise.
9734 (*call_value_reg_thumb1_v5): Likewise.
9735 (*call_value_reg_thumb1): Likewise.
9736 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
9738 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
9740 2018-05-18 Martin Liska <mliska@suse.cz>
9742 PR gcov-profile/84846
9743 * doc/gcov.texi: Document -t option of gcov tool.
9745 2018-05-18 Martin Liska <mliska@suse.cz>
9747 PR gcov-profile/84846
9748 * gcov.c (print_usage): Add new -t option.
9749 (process_args): Handle the option.
9750 (generate_results): Use stdout as output when requested by
9753 2018-05-18 Martin Liska <mliska@suse.cz>
9755 PR gcov-profile/84846
9756 * coverage.c (coverage_init): Write PWD to .gcno file.
9757 * doc/gcov.texi: Document how working directory is printed.
9758 * gcov-dump.c (dump_gcov_file): Print PWD.
9759 * gcov.c (output_intermediate_file): Likewise.
9760 (read_graph_file): Read PWD string.
9761 (output_lines): Print PWD.
9763 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9766 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
9767 for retval and return false if all args to phi are zero.
9769 2018-05-18 Richard Biener <rguenther@suse.de>
9771 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
9773 (evrp_dom_walker::before_dom_children): Call it.
9775 2018-05-18 Richard Biener <rguenther@suse.de>
9777 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
9778 results when processing array refs with variable index.
9780 2018-05-18 Toon Moene <toon@moene.org>
9782 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
9783 directly after that of -floop-interchange. Indicate that both
9784 options are enabled by default when specifying -O3.
9786 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9788 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
9789 iterator. Delete separate integer-mode vec_set<mode> expander.
9790 (aarch64_simd_vec_setv2di): Delete.
9791 (vec_setv2di): Delete.
9792 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
9793 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
9794 the "w, r" alternative.
9796 2018-05-18 Martin Liska <mliska@suse.cz>
9798 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
9799 * tree-pass.h (make_pass_lower_switch_O0): New function.
9800 * tree-switch-conversion.c (node_has_low_bound): Remove.
9801 (node_has_high_bound): Likewise.
9802 (node_is_bounded): Likewise.
9803 (class pass_lower_switch): Make it a template type and create
9805 (pass_lower_switch::execute): Add template argument.
9806 (make_pass_lower_switch): New function.
9807 (make_pass_lower_switch_O0): New function.
9808 (do_jump_if_equal): Remove.
9809 (emit_case_nodes): Simplify to just handle all 3 cases and leave
9810 all the hard work to tree optimization passes.
9812 2018-05-18 Martin Liska <mliska@suse.cz>
9814 * dbgcnt.c (limit_low): Renamed from limit.
9815 (limit_high): New variable.
9816 (dbg_cnt_is_enabled): Check for upper limit.
9817 (dbg_cnt): Adjust dumping.
9818 (dbg_cnt_set_limit_by_index): Add new argument for high
9820 (dbg_cnt_set_limit_by_name): Likewise.
9821 (dbg_cnt_process_single_pair): Parse new format.
9822 (dbg_cnt_process_opt): Use strtok.
9823 (dbg_cnt_list_all_counters): Remove 'value' and add
9825 * doc/invoke.texi: Document changes.
9827 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
9829 * doc/sourcebuild.texi (scalar_all_fma): Document.
9830 * tree.def (FMA_EXPR): Delete.
9831 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
9832 * internal-fn.c (ternary_direct): New macro.
9833 (expand_ternary_optab_fn): Likewise.
9834 (direct_ternary_optab_supported_p): Likewise.
9835 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
9836 * builtins.c (fold_builtin_fma): Delete.
9837 (fold_builtin_3): Don't call it.
9838 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
9839 * expr.c (expand_expr_real_2): Likewise.
9840 * fold-const.c (operand_equal_p): Likewise.
9841 (fold_ternary_loc): Likewise.
9842 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
9843 * gimple.c (DEFTREECODE): Likewise.
9844 * gimplify.c (gimplify_expr): Likewise.
9845 * optabs-tree.c (optab_for_tree_code): Likewise.
9846 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
9847 * tree-eh.c (operation_could_trap_p): Likewise.
9848 (stmt_could_throw_1_p): Likewise.
9849 * tree-inline.c (estimate_operator_cost): Likewise.
9850 * tree-pretty-print.c (dump_generic_node): Likewise.
9851 (op_code_prio): Likewise.
9852 * tree-ssa-loop-im.c (stmt_cost): Likewise.
9853 * tree-ssa-operands.c (get_expr_operands): Likewise.
9854 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
9855 * fold-const-call.h (fold_fma): Delete.
9856 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
9857 CFN_FNMA and CFN_FNMS.
9859 * genmatch.c (combined_fn): New enum.
9860 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
9861 (commutative_op): New function.
9862 (commutate): Use it. Handle more than 2 operands.
9863 (dt_operand::gen_gimple_expr): Use commutative_op.
9864 (parser::parse_expr): Allow :c to be used with non-binary
9865 operators if the commutative operand is known.
9866 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
9867 CFN_FMS, CFN_FNMA and CFN_FNMS.
9868 (backprop::process_assign_use): Remove FMA_EXPR handling.
9869 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
9870 (gen_hsa_fma): New function.
9871 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
9872 IFN_FNMA and IFN_FNMS.
9873 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
9874 * gimple-fold.h (follow_all_ssa_edges): Declare.
9875 * gimple-fold.c (follow_all_ssa_edges): New function.
9876 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
9877 gimple_build interface and use follow_all_ssa_edges to fold the result.
9878 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
9879 instead of checking for optabs directly.
9880 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
9881 rather than FMA_EXPRs.
9882 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
9883 call to IFN_FMA instead of an FMA_EXPR.
9885 2018-05-17 Jim Wilson <jimw@sifive.com>
9887 * expr.c (do_tablejump): When converting index to Pmode, if we have a
9888 sign extended promoted subreg, and the range does not have the sign bit
9889 set, then do a sign extend.
9891 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
9892 test, check for sign extended subreg and/or constant operands, and
9893 do a sign extend in that case.
9895 2018-05-17 Steve Ellcey <sellcey@cavium.com>
9897 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
9898 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
9900 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
9901 Change logics_shift_reg to logics_shift_imm.
9902 (thunderx2t99_fp_loadpair_basic): Delete.
9903 (thunderx2t99_fp_storepair_basic): Delete.
9904 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
9905 (thunderx2t99_asimd_polynomial): Delete.
9906 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
9907 and neon_fp_mul_d_scalar_q.
9908 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
9909 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
9910 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
9911 (thunderx2t99_asimd_lut): Add missing tbl types.
9912 (thunderx2t99_asimd_ext): Delete.
9913 (thunderx2t99_asimd_load1_1_mult): Delete.
9914 (thunderx2t99_asimd_load1_2_mult): Delete.
9915 (thunderx2t99_asimd_load1_ldp): New.
9916 (thunderx2t99_asimd_load1): New.
9917 (thunderx2t99_asimd_load2): Add missing *load2* types.
9918 (thunderx2t99_asimd_load3): New.
9919 (thunderx2t99_asimd_load4): New.
9920 (thunderx2t99_asimd_store1_1_mult): Delete.
9921 (thunderx2t99_asimd_store1_2_mult): Delete.
9922 (thunderx2t99_asimd_store2_mult): Delete.
9923 (thunderx2t99_asimd_store2_onelane): Delete.
9924 (thunderx2t99_asimd_store_stp): New.
9925 (thunderx2t99_asimd_store1): New.
9926 (thunderx2t99_asimd_store2): New.
9927 (thunderx2t99_asimd_store3): New.
9928 (thunderx2t99_asimd_store4): New.
9930 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
9932 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
9933 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
9935 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
9936 Segher Boessenkool <segher@kernel.crashing.org>
9939 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
9942 2018-05-17 Richard Biener <rguenther@suse.de>
9944 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
9945 for pruning loop and prune defs feeding only already visited PHIs.
9947 2018-05-17 Richard Biener <rguenther@suse.de>
9949 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
9951 2018-05-17 Bin Cheng <bin.cheng@arm.com>
9952 Richard Biener <rguenther@suse.de>
9954 PR tree-optimization/85793
9955 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
9956 for VMAT_ELEMENTWISE.
9958 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
9960 * internal-fn.h (lookup_internal_fn): Declare
9961 * internal-fn.c (lookup_internal_fn): New function.
9962 * gimple.c (gimple_build_call_from_tree): Handle calls to
9964 * gimple-pretty-print.c (dump_gimple_call): Print "." before
9965 internal function names.
9966 * tree-pretty-print.c (dump_generic_node): Likewise.
9967 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
9969 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
9971 * gimple-fold.h (gimple_build): Make the function forms take
9972 combined_fn rather than built_in_function.
9973 (gimple_simplify): Likewise.
9974 * gimple-match-head.c (gimple_simplify): Likewise.
9975 * gimple-fold.c (gimple_build): Likewise.
9976 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
9977 rather than gimple_build_call_internal.
9978 (get_initial_defs_for_reduction): Likewise.
9979 (vect_create_epilog_for_reduction): Likewise.
9980 (vectorizable_live_operation): Likewise.
9982 2018-05-17 Martin Liska <mliska@suse.cz>
9984 * gimple-ssa-sprintf.c (format_directive): Do not use
9985 space in between 'G_' and '('.
9987 2018-05-17 Jakub Jelinek <jakub@redhat.com>
9990 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
9991 even if the mask is not all ones.
9994 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
9996 (ix86_gimple_fold_builtin): Likewise.
9999 * config/i386/i386.c: Include tree-vector-builder.h.
10000 (ix86_vector_shift_count): New function.
10001 (ix86_fold_builtin): Fold shift builtins by scalar count.
10002 (ix86_gimple_fold_builtin): Likewise.
10004 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
10005 _mm512_setzero): New intrinsics.
10007 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
10008 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10010 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
10011 code generation for cases where splatting a value is not useful.
10012 * simplify-rtx.c (simplify_ternary_operation): Simplify
10013 vec_merge across a vec_duplicate and a paradoxical subreg forming
10014 a vector mode to a vec_concat.
10016 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
10018 * config.gcc: Support "goldmont-plus".
10019 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
10021 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
10022 PROCESSOR_GOLDMONT_PLUS.
10023 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
10024 (processor_target_table): Add "goldmont-plus".
10025 (PTA_GOLDMONT_PLUS): Define.
10026 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
10027 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
10028 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
10029 (fold_builtin_cpu): Add "goldmont-plus".
10030 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
10031 (ix86_option_override_internal): Add "goldmont-plus".
10032 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
10033 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
10034 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
10035 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
10037 2018-05-17 Richard Biener <rguenther@suse.de>
10039 PR tree-optimization/85757
10040 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
10041 remove defs that only feed that PHI from further processing.
10043 2018-05-16 Jim Wilson <jimw@sifive.com>
10045 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
10047 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
10049 2018-05-16 Mark Wielaard <mark@klomp.org>
10051 * dwarf2out.c (count_index_strings): New function.
10052 (output_indirect_strings): Call count_index_strings and generate
10053 header for dwarf_version >= 5.
10055 2018-05-16 Mark Wielaard <mark@klomp.org>
10057 * dwarf2out.c (dwarf_FORM): New function.
10058 (set_indirect_string): Use dwarf_FORM.
10059 (reset_indirect_string): Likewise.
10060 (size_of_die): Likewise.
10061 (value_format): Likewise.
10062 (output_die): Likewise.
10063 (add_skeleton_AT_string): Likewise.
10064 (output_macinfo_op): Likewise.
10065 (index_string): Likewise.
10066 (output_index_string_offset): Likewise.
10067 (output_index_string): Likewise.
10068 (count_index_strings): Likewise.
10070 2018-05-16 Carl Love <cel@us.ibm.com>
10072 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
10073 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
10075 2018-05-16 Martin Jambor <mjambor@suse.cz>
10077 * ipa-prop.c (ipa_free_all_edge_args): Remove.
10078 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
10080 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
10082 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
10083 (fnma<mode>4): Likewise.
10084 (fms<mode>4): Likewise.
10085 (fnms<mode>4): Likewise.
10086 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
10087 (aarch64_fnma<mode>4): Likewise.
10088 (aarch64_fms<mode>4): Likewise.
10089 (aarch64_fnms<mode>4): Likewise.
10090 (aarch64_fnmadd<mode>4): Likewise.
10092 2018-05-16 Jason Merrill <jason@redhat.com>
10094 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
10096 2018-05-16 Richard Biener <rguenther@suse.de>
10098 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
10099 (dump_stmt_cost): Declare.
10100 (add_stmt_cost): Dump cost we add.
10101 (add_stmt_costs): New function.
10102 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
10103 No longer exported.
10104 (vect_analyze_stmt): Adjust prototype.
10105 (vectorizable_condition): Likewise.
10106 (vectorizable_live_operation): Likewise.
10107 (vectorizable_reduction): Likewise.
10108 (vectorizable_induction): Likewise.
10109 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
10110 cost vector to pass to vectorizable_ and record afterwards.
10111 (vect_model_reduction_cost): Take cost vector argument and adjust.
10112 (vect_model_induction_cost): Likewise.
10113 (vectorizable_reduction): Likewise.
10114 (vectorizable_induction): Likewise.
10115 (vectorizable_live_operation): Likewise.
10116 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
10117 SLP_TREE_NUMBER_OF_VEC_STMTS.
10118 (vect_analyze_slp_cost_1): Remove.
10119 (vect_analyze_slp_cost): Likewise.
10120 (vect_slp_analyze_node_operations): Take visited args and
10121 a target cost vector. Avoid processing already visited stmt sets.
10122 (vect_slp_analyze_operations): Use a local cost vector to gather
10123 costs and register those of non-discarded instances.
10124 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
10125 (vect_schedule_slp_instance): Remove copying of
10126 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
10128 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
10129 adding cost. Record cost entry location.
10130 (vect_prologue_cost_for_slp_op): Function to compute cost of
10131 a constant or invariant generated for SLP vect in the prologue,
10132 split out from vect_analyze_slp_cost_1.
10133 (vect_model_simple_cost): Make static. Adjust for SLP costing.
10134 (vect_model_promotion_demotion_cost): Likewise.
10135 (vect_model_store_cost): Likewise, make static.
10136 (vect_model_load_cost): Likewise.
10137 (vectorizable_bswap): Add cost vector arg and adjust.
10138 (vectorizable_call): Likewise.
10139 (vectorizable_simd_clone_call): Likewise.
10140 (vectorizable_conversion): Likewise.
10141 (vectorizable_assignment): Likewise.
10142 (vectorizable_shift): Likewise.
10143 (vectorizable_operation): Likewise.
10144 (vectorizable_store): Likewise.
10145 (vectorizable_load): Likewise.
10146 (vectorizable_condition): Likewise.
10147 (vectorizable_comparison): Likewise.
10148 (can_vectorize_live_stmts): Likewise.
10149 (vect_analyze_stmt): Likewise.
10150 (vect_transform_stmt): Adjust calls to vectorizable_*.
10151 * tree-vectorizer.c: Include gimple-pretty-print.h.
10152 (dump_stmt_cost): New function.
10154 2018-05-16 Richard Biener <rguenther@suse.de>
10156 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
10157 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
10158 * tree-ssa-dse.c: Include tree-ssa-loop.h.
10159 (check_name): New callback.
10160 (dse_classify_store): Track cycles via a visited bitmap of PHI
10161 defs and simplify handling of in-loop and across loop dead stores
10162 and properly fail for loop-variant refs. Handle byte-tracking with
10163 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
10166 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
10168 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
10169 (vect_get_mask_type_for_stmt): Likewise.
10170 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
10172 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
10173 to determine the statement's vector type and the vector type that
10174 should be used for calculating nunits. Deal with cases in which
10175 the type has to be deferred.
10176 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
10177 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
10178 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
10179 (vect_determine_vf_for_stmt): New functions, split out from...
10180 (vect_determine_vectorization_factor): ...here.
10181 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
10182 (vect_get_mask_type_for_stmt): New functions, split out from
10183 vect_determine_vectorization_factor.
10185 2018-05-16 Richard Biener <rguenther@suse.de>
10187 * tree-cfg.c (verify_gimple_assign_ternary): Properly
10188 verify the [VEC_]COND_EXPR embedded comparison.
10190 2018-05-15 Martin Sebor <msebor@redhat.com>
10192 PR tree-optimization/85753
10193 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
10194 RECORD_TYPE in addition to ARRAY_TYPE.
10196 2018-05-15 Martin Sebor <msebor@redhat.com>
10198 PR middle-end/85643
10199 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
10201 2018-05-15 Richard Biener <rguenther@suse.de>
10203 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
10204 add by_clobber_p one. Change algorithm to collect all defs
10205 representing uses we need to walk and try reducing them to
10206 a single one before failing.
10207 (dse_dom_walker::dse_optimize_stmt): Adjust.
10209 2018-05-13 Mark Wielaard <mark@klomp.org>
10211 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
10212 (size_of_loc_descr): Likewise.
10213 (output_loc_operands): Likewise.
10214 (output_loc_operands_raw): Likewise.
10215 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
10216 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
10217 (hash_loc_operands): Likewise.
10218 (compare_loc_operands): Likewise.
10220 2018-05-14 Mark Wielaard <mark@klomp.org>
10222 * dwarf2out.c (count_index_addrs): New function.
10223 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
10225 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10227 PR tree-optimization/83648
10228 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
10229 return value as malloc candidate.
10231 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10234 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
10235 param as true in call to suggest_attribute.
10237 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
10239 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
10240 -mreadonly-in-sdata.
10242 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10244 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
10246 (aarch64_crypto_aesd_fused): Likewise.
10248 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
10250 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
10251 (movsi_aarch64): Likewise.
10252 (load_pairsi): Likewise.
10253 (load_pairdi): Likewise.
10254 (store_pairsi): Likewise.
10255 (store_pairdi): Likewise.
10256 (load_pairsf): Likewise.
10257 (load_pairdf): Likewise.
10258 (store_pairsf): Likewise.
10259 (store_pairdf): Likewise.
10260 (zero_extend): Likewise.
10261 (trunc): Swap alternatives.
10262 (fcvt_target): Add '?' to prefer w over r.
10264 2018-05-14 Jakub Jelinek <jakub@redhat.com>
10267 * config/i386/i386.md: Disallow non-commutative arithmetics in
10268 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
10269 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
10270 in the peephole2 before it.
10272 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
10274 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
10275 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
10276 (ix86_handle_option): Handle -mcldemote.
10277 * config.gcc: New header.
10278 * config/i386/cldemoteintrin.h: New file.
10279 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
10280 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
10282 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
10283 OPTION_MASK_ISA_CLDEMOTE.
10284 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
10285 (ix86_valid_target_attribute_inner_p): Ditto.
10286 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
10287 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
10288 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
10289 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
10290 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
10292 * config/i386/i386.opt: Add -mcldemote.
10293 * config/i386/x86intrin.h: New header.
10294 * doc/invoke.texi: Add -mcldemote.
10296 2018-05-14 Richard Biener <rguenther@suse.de>
10298 * doc/match-and-simplify.texi: Adjust :s documentation.
10300 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
10302 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
10303 intended memcpy size.
10304 (REORDER_45): Likewise.
10306 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
10308 * sort.cc: New file.
10309 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
10310 * vec.c (qsort_chk): Use gcc_qsort.
10311 * Makefile.in (OBJS-libcommon): Add sort.o.
10312 (build/sort.o): New target. Use it...
10313 (BUILD_RTL): ... here, and...
10314 (build/gencfn-macros): ... here, and...
10315 (build/genmatch): ... here.
10317 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
10318 Chung-Ju Wu <jasonwucj@gmail.com>
10320 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
10321 * config/nds32/nds32-graywolf.md: New file.
10322 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
10323 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
10325 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
10326 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
10327 * config/nds32/nds32.md (pipeline_model): Add graywolf.
10328 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
10329 * config/nds32/pipelines.md: Include n15 settings.
10331 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
10332 Chung-Ju Wu <jasonwucj@gmail.com>
10334 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
10335 * config/nds32/nds32-n13.md: New file.
10336 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
10337 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
10339 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
10340 * config/nds32/nds32.md (pipeline_model): Add n13.
10341 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
10342 * config/nds32/pipelines.md: Include n13 settings.
10344 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
10345 Chung-Ju Wu <jasonwucj@gmail.com>
10347 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
10348 * config/nds32/nds32-n10.md: New file.
10349 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
10350 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
10352 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
10353 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
10354 * config/nds32/nds32.md (pipeline_model): Add n10.
10355 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
10356 * config/nds32/pipelines.md: Include n10 settings.
10358 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
10359 Kito Cheng <kito.cheng@gmail.com>
10360 Chung-Ju Wu <jasonwucj@gmail.com>
10362 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
10363 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
10364 Add enum values for DSP extension instructions.
10365 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
10367 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
10368 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
10369 New code iterators.
10370 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
10371 * config/nds32/nds32-dspext.md: New file for DSP implementation.
10372 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
10373 * config/nds32/nds32-intrinsic.md: Likewise.
10374 * config/nds32/nds32_intrinsic.h: Likewise.
10375 * config/nds32/nds32-md-auxiliary.c: Likewise.
10376 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
10377 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
10378 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
10379 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
10380 * config/nds32/nds32-protos.h: New declarations for DSP extension.
10381 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
10382 TYPE_DMAC in switch statement.
10383 * config/nds32/nds32.c: New checking and implementation for DSP
10384 extension instructions.
10385 * config/nds32/nds32.h: Likewise.
10386 * config/nds32/nds32.md: Likewise.
10387 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
10388 * config/nds32/predicates.md: Implement new predicates for DSP
10391 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
10393 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
10394 Reformat alternatives and attributes so it is easier to identify
10395 which constraints/attributes go with which instruction.
10396 (mov<mode>_hardfloat32, FMOVE64): Likewise.
10397 (mov<mode>_softfloat32, FMOVE64): Likewise.
10398 (mov<mode>_hardfloat64, FMOVE64): Likewise.
10399 (mov<mode>_softfloat64, FMOVE64): Likewise.
10401 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
10403 * doc/extend.texi (PowerPC Built-in Functions): Rename this
10405 (Basic PowerPC Built-in Functions): The new name of the
10406 subsection previously known as "PowerPC Built-in Functions".
10407 (Basic PowerPC Built-in Functions Available on all Configurations):
10409 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
10410 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
10411 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
10412 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
10414 2018-05-11 Martin Jambor <mjambor@suse.cz>
10417 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
10420 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
10423 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
10425 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
10427 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
10428 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
10429 (ix86_handle_option): Handle -mwaitpkg.
10430 * config.gcc: New header.
10431 * config/i386/cpuid.h (bit_WAITPKG): New bit.
10432 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
10433 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
10435 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
10436 OPTION_MASK_ISA_WAITPKG.
10437 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
10438 (ix86_option_override_internal): Add PTA_WAITPKG.
10439 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
10440 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
10441 IX86_BUILTIN_TPAUSE.
10442 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
10443 __builtin_ia32_umwait and __builtin_ia32_tpause.
10444 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
10445 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
10446 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
10447 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
10448 UNSPECV_TPAUSE): New.
10449 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
10450 * config/i386/i386.opt: Add -mwaitpkg.
10451 * config/i386/waitpkgintrin.h: New file.
10452 * config/i386/x86intrin.h: New header.
10453 * doc/invoke.texi: Add -mwaitpkg.
10455 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
10458 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
10460 (cortex-m0): Use armv6s-m isa.
10461 (cortex-m0plus): Likewise.
10462 (cortex-m1): Likewise.
10463 (cortex-m0.small-multiply): Likewise.
10464 (cortex-m0plus.small-multiply): Likewise.
10465 (cortex-m1.small-multiply): Likewise.
10467 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
10468 Jakub Jelinek <jakub@redhat.com>
10470 PR tree-optimization/85692
10471 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
10472 source permute as well.
10474 2018-05-11 Martin Liska <mliska@suse.cz>
10477 * doc/extend.texi: Document LLVM style format for no_sanitize
10480 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
10482 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
10483 mode_supports_vsx_dform_quad to mode_supports_dq_form.
10484 (mode_supports_vsx_dform_quad): Likewise.
10485 (mode_supports_vmx_dform): Move these functions to be next to the
10486 other mode_supports functions.
10487 (mode_supports_dq_form): Likewise.
10488 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
10489 mode_supports_dq_form.
10490 (reg_offset_addressing_ok_p): Likewise.
10491 (offsettable_ok_by_alignment): Likewise.
10492 (rs6000_legitimate_offset_address_p): Likewise.
10493 (legitimate_lo_sum_address_p): Likewise.
10494 (rs6000_legitimize_address): Likewise.
10495 (rs6000_legitimize_reload_address): Likewise.
10496 (rs6000_secondary_reload_inner): Likewise.
10497 (rs6000_preferred_reload_class): Likewise.
10498 (rs6000_output_move_128bit): Likewise.
10500 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
10502 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
10503 Generate SImode target register for null target.
10504 <case IX86_BUILTIN_XGETBV>: Ditto.
10505 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
10506 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
10508 2018-05-10 Carl Love <cel@us.ibm.com>
10510 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
10511 dcbtt and dcbtstt if operands[2] is 0.
10513 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
10516 * config/i386/sse.md (usadv64qi): New expander.
10518 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
10520 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
10521 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
10522 -maltivec=be support.
10523 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
10524 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
10525 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
10526 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
10527 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
10528 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
10529 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
10530 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
10531 altivec_vsumsws): Adjust.
10532 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
10533 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
10534 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
10536 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
10537 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
10538 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
10539 (altivec_lve<VI_char>x): Delete expand.
10540 (*altivec_lve<VI_char>x_internal): Rename to...
10541 (altivec_lve<VI_char>x): ... this.
10542 (altivec_lvxl_<mode>): Delete expand.
10543 (*altivec_lvxl_<mode>_internal): Rename to ...
10544 (altivec_lvxl_<mode>): ... this.
10545 (altivec_stvxl_<mode>): Delete expand.
10546 (*altivec_stvxl_<mode>_internal): Rename to ...
10547 (altivec_stvxl_<mode>): ... this.
10548 (altivec_stve<VI_char>x): Delete expand.
10549 (*altivec_stve<VI_char>x_internal): Rename to ...
10550 (altivec_stve<VI_char>x): ... this.
10551 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
10552 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
10553 reduc_plus_scal_<mode>): Adjust.
10554 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
10556 (rs6000_cpu_cpp_builtins): Adjust.
10557 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
10558 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
10559 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
10560 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
10561 -maltivec=be support.
10562 (rs6000_split_vec_extract_var): Adjust.
10563 (rs6000_split_v4si_init): Adjust.
10564 (swap_selector_for_mode): Delete.
10565 (altivec_expand_lvx_be, altivec_expand_stvx_be,
10566 altivec_expand_stvex_be): Delete.
10567 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
10568 -maltivec=be support.
10569 (rs6000_gimple_fold_builtin): Ditto.
10570 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
10572 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
10573 (TARGET_DIRECT_MOVE_64BIT): Adjust.
10574 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
10575 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
10576 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
10577 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
10578 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
10579 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
10580 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
10581 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
10582 anonymous split): Adjust.
10583 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
10584 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
10586 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
10588 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
10589 when --with-gxx-include-dir is also specified.
10590 * configure: Regenerate.
10592 2018-05-09 Jim Wilson <jimw@sifive.com>
10595 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
10596 * config/riscv/t-withmultilib: New.
10597 * config/riscv/withmultilib.h: New.
10598 * doc/install.texi: Document RISC-V --with-multilib-list support.
10600 2018-05-09 Richard Biener <rguenther@suse.de>
10602 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
10604 (vect_bb_vectorization_profitable_p): Adjust. Compute
10605 actual scalar cost using the cost vector and the add_stmt_cost
10608 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
10610 PR rtl-optimization/85645
10611 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
10612 in the REG_CFA_REGISTER note for LR, don't leave it empty.
10614 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
10616 PR rtl-optimization/85645
10617 * shrink-wrap.c (spread_components): Return a boolean saying if
10618 anything was changed.
10619 (try_shrink_wrapping_separate): Iterate spread_components until
10620 nothing changes anymore.
10622 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
10624 PR rtl-optimization/85645
10625 * regrename.c (build_def_use): Also kill the chains that include the
10626 destination of a REG_CFA_REGISTER note.
10628 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
10630 PR rtl-optimization/85645
10631 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
10632 insn that has a REG_CFA_REGISTER note.
10634 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
10636 * cfgexpand.c (expand_clobber): New function.
10637 (expand_gimple_stmt_1): Use it.
10638 * tree-vect-stmts.c (vect_clobber_variable): New function,
10640 (vectorizable_simd_clone_call): ...here.
10641 (vectorizable_store): Emit a clobber either side of an
10642 IFN_STORE_LANES sequence.
10643 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
10645 2018-05-09 Tom de Vries <tom@codesourcery.com>
10648 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
10649 (define_insn "trap_if_false"): Add exit after trap.
10651 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
10653 PR rtl-optimization/85638
10654 * bb-reorder.c: Include common/common-target.h.
10655 (create_forwarder_block): New function extracted from...
10656 (fix_up_crossing_landing_pad): ...here. Rename into...
10657 (dw2_fix_up_crossing_landing_pad): ...this.
10658 (sjlj_fix_up_crossing_landing_pad): New function.
10659 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
10660 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
10661 from both partitions and exit the loop after one iteration.
10663 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
10666 * doc/extend.texi (PowerPC Built-in Functions): Rename this
10668 (Basic PowerPC Built-in Functions): The new name of the
10669 subsection previously known as "PowerPC Built-in Functions".
10670 (Basic PowerPC Built-in Functions Available on all Configurations):
10672 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
10674 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
10676 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
10678 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
10681 2018-05-08 Jim Wilson <jimw@sifive.com>
10683 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
10684 (LD_EMUL_SUFFIX): New.
10685 (LINK_SPEC): Use it.
10687 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
10689 * doc/extend.texi (PowerPC Built-in Functions): Rename this
10691 (Basic PowerPC Built-in Functions): The new name of the
10692 subsection previously known as "PowerPC Built-in Functions".
10693 (Basic PowerPC Built-in Functions Available on all Configurations):
10695 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
10697 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
10699 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
10701 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
10704 2018-05-08 Jakub Jelinek <jakub@redhat.com>
10707 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
10708 after cmpelim optimization.
10710 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
10712 * config.gcc: Support "goldmont".
10713 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
10714 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
10715 PROCESSOR_GOLDMONT.
10716 * config/i386/i386.c (m_GOLDMONT): Define.
10717 (processor_target_table): Add "goldmont".
10718 (PTA_GOLDMONT): Define.
10719 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
10720 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
10721 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
10722 (fold_builtin_cpu): Add "goldmont".
10723 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
10724 (ix86_option_override_internal): Add "goldmont".
10725 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
10726 (processor_type): Add PROCESSOR_GOLDMONT.
10727 * config/i386/i386.md: Add CPU "glm".
10728 * config/i386/glm.md: New file.
10729 * config/i386/x86-tune.def: Add m_GOLDMONT.
10730 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
10732 2018-05-08 Jakub Jelinek <jakub@redhat.com>
10735 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
10737 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
10738 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
10739 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
10742 * config/i386/i386.c (ix86_fold_builtin): Handle
10743 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
10746 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
10747 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
10749 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
10752 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
10753 (check_arch): Likewise.
10754 (check_fpu): Return the result rather than printing it.
10755 (end arch): Fix operator precedence.
10756 (end cpu): Likewise.
10757 (END): Print the result from check_fpu.
10759 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
10760 Alan Hayward <alan.hayward@arm.com>
10761 David Sherwood <david.sherwood@arm.com>
10763 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
10764 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
10765 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
10766 (*fcmuo<mode>_and): New patterns.
10768 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
10770 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
10771 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
10772 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
10773 (cmp_op, sve_imm_con): New code attributes.
10774 (SVE_COND_INT_CMP, imm_con): Delete.
10775 (cmp_op): Remove above unspecs from int attribute.
10776 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
10778 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
10779 comparison-specific unspecs.
10780 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
10781 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
10782 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
10783 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
10784 (*vec_fcm<cmp_op><mode>): Rename to...
10785 (*fcm<cmp_op><mode>): ...this and adjust likewise.
10786 (*vec_fcmuo<mode>): Rename to...
10787 (*fcmuo<mode>): ...this and adjust likewise.
10788 (*pred_fcm<cmp_op><mode>): New pattern.
10789 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
10790 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
10792 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
10794 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
10795 (aarch64_emit_sve_predicated_cond): New function.
10796 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
10797 (aarch64_emit_unspec_cond_or): Replace with...
10798 (aarch64_emit_sve_or_conds): ...this new function. Use
10799 aarch64_emit_sve_ptrue_op for the individual comparisons and
10800 aarch64_emit_binop to OR them together.
10801 (aarch64_emit_inverted_unspec_cond): Replace with...
10802 (aarch64_emit_sve_inverted_cond): ...this new function. Use
10803 aarch64_emit_sve_ptrue_op for the comparison and
10804 aarch64_emit_unop to invert the result.
10805 (aarch64_expand_sve_vec_cmp_float): Update after the above
10806 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
10808 2018-05-07 Nathan Sidwell <nathan@acm.org>
10810 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
10811 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
10812 (Backwards Compatibility): Likewise.
10814 2018-05-07 Luis Machado <luis.machado@linaro.org>
10818 2018-05-07 Luis Machado <luis.machado@linaro.org>
10820 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
10821 <prefetch_dynamic_strides>: New const bool field.
10822 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
10823 prefetch_dynamic_strides.
10824 (exynosm1_prefetch_tune): Likewise.
10825 (thunderxt88_prefetch_tune): Likewise.
10826 (thunderx_prefetch_tune): Likewise.
10827 (thunderx2t99_prefetch_tune): Likewise.
10828 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
10830 (aarch64_override_options_internal): Update to set
10831 PARAM_PREFETCH_DYNAMIC_STRIDES.
10832 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
10833 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
10834 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
10835 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
10836 prefetch-dynamic-strides setting.
10838 2018-05-07 Luis Machado <luis.machado@linaro.org>
10840 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
10841 <minimum_stride>: New const int field.
10842 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
10843 minimum_stride field.
10844 (exynosm1_prefetch_tune): Likewise.
10845 (thunderxt88_prefetch_tune): Likewise.
10846 (thunderx_prefetch_tune): Likewise.
10847 (thunderx2t99_prefetch_tune): Likewise.
10848 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
10849 (aarch64_override_options_internal): Update to set
10850 PARAM_PREFETCH_MINIMUM_STRIDE.
10851 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
10852 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
10853 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
10854 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
10855 stride is constant and is below the minimum stride threshold.
10857 2018-05-07 Luis Machado <luis.machado@linaro.org>
10859 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
10862 2018-05-07 Luis Machado <luis.machado@linaro.org>
10864 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
10865 <prefetch_dynamic_strides>: New const bool field.
10866 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
10867 prefetch_dynamic_strides.
10868 (exynosm1_prefetch_tune): Likewise.
10869 (thunderxt88_prefetch_tune): Likewise.
10870 (thunderx_prefetch_tune): Likewise.
10871 (thunderx2t99_prefetch_tune): Likewise.
10872 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
10874 (aarch64_override_options_internal): Update to set
10875 PARAM_PREFETCH_DYNAMIC_STRIDES.
10876 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
10877 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
10878 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
10879 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
10880 prefetch-dynamic-strides setting.
10882 2018-05-07 Luis Machado <luis.machado@linaro.org>
10884 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
10885 <minimum_stride>: New const int field.
10886 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
10887 minimum_stride field.
10888 (exynosm1_prefetch_tune): Likewise.
10889 (thunderxt88_prefetch_tune): Likewise.
10890 (thunderx_prefetch_tune): Likewise.
10891 (thunderx2t99_prefetch_tune): Likewise.
10892 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
10893 (aarch64_override_options_internal): Update to set
10894 PARAM_PREFETCH_MINIMUM_STRIDE.
10895 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
10896 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
10897 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
10898 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
10899 stride is constant and is below the minimum stride threshold.
10901 2018-05-06 Jakub Jelinek <jakub@redhat.com>
10904 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
10905 the type is addressable. Don't force op into register if it has
10908 2018-05-05 Roland McGrath <mcgrathr@google.com>
10911 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
10912 any section for which we don't know a specific type it should have,
10913 regardless of name. Previously this was done only for the exact
10914 names ".init_array", ".fini_array", and ".preinit_array".
10915 (default_elf_asm_named_section): Add comment about
10916 relationship with default_section_type_flags and SECTION_NOTYPE.
10917 (get_section): Don't consider it a type conflict if one side has
10918 SECTION_NOTYPE and the other doesn't, as long as neither has the
10919 SECTION_BSS et al used in the default_section_type_flags logic.
10921 2018-05-05 Tom de Vries <tom@codesourcery.com>
10924 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
10925 (workaround_barsyncs): New function.
10926 (nvptx_reorg): Use workaround_barsyncs.
10927 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
10928 (define_expand "nvptx_membar_cta"): New define_expand.
10929 (define_insn "*nvptx_membar_cta"): New insn.
10931 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
10933 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
10934 To improve optimization opportunities.
10935 * builtin-types.def: The new needed builtin types for the above.
10937 2018-05-04 Richard Biener <rguenther@suse.de>
10939 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
10940 * gimple-ssa-store-merging.c
10941 (imm_store_chain_info::output_merged_store): Remove redundant create,
10942 release split_store vector contents on failure.
10943 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
10944 scalar stmt vector on cache hit.
10946 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
10948 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
10950 * config.gcc (powerpc-xilinx-eabi*): Remove.
10951 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
10953 (fusion_addis_mem_combo_load): Ditto.
10954 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
10956 (rs6000_cpu_cpp_builtins): Ditto.
10957 * config/rs6000/rs6000-linux.c
10958 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
10959 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
10960 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
10962 (rs6000_setup_reg_addr_masks): Ditto.
10963 (rs6000_init_hard_regno_mode_ok): Ditto.
10964 (rs6000_option_override_internal): Ditto.
10965 (legitimate_lo_sum_address_p): Ditto.
10966 (rs6000_legitimize_address): Ditto.
10967 (rs6000_legitimize_reload_address): Ditto.
10968 (rs6000_legitimate_address_p): Ditto.
10969 (abi_v4_pass_in_fpr): Ditto.
10970 (setup_incoming_varargs): Ditto.
10971 (rs6000_gimplify_va_arg): Ditto.
10972 (rs6000_split_multireg_move): Ditto.
10973 (rs6000_savres_strategy): Ditto.
10974 (rs6000_emit_prologue_components): Ditto.
10975 (rs6000_emit_epilogue_components): Ditto.
10976 (rs6000_emit_prologue): Ditto.
10977 (rs6000_emit_epilogue): Ditto.
10978 (rs6000_elf_file_end): Ditto.
10979 (rs6000_function_value): Ditto.
10980 (rs6000_libcall_value): Ditto.
10981 * config/rs6000/rs6000.h: Ditto.
10982 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
10983 (TARGET_MINMAX): ... this. New.
10984 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
10985 * config/rs6000/rs6000.md: Remove Xilinx FP support.
10986 (*movsi_internal1_single): Delete.
10987 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
10988 mfpu=, mxilinx-fpu): Delete.
10989 * config/rs6000/singlefp.h: Delete.
10990 * config/rs6000/sysv4.h: Remove Xilinx FP support.
10991 * config/rs6000/t-rs6000: Ditto.
10992 * config/rs6000/t-xilinx: Delete.
10993 * config/rs6000/titan.md: Adjust for fp_type removal.
10994 * config/rs6000/vsx.md: Remove Xilinx FP support.
10995 (VStype_simple): Delete.
10996 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
10997 * config/rs6000/xfpu.h: Delete.
10998 * config/rs6000/xfpu.md: Delete.
10999 * config/rs6000/xilinx.h: Delete.
11000 * config/rs6000/xilinx.opt: Delete.
11001 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
11002 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
11004 2018-05-04 Tom de Vries <tom@codesourcery.com>
11007 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
11010 2018-05-04 Richard Biener <rguenther@suse.de>
11012 PR middle-end/85627
11013 * tree-complex.c (update_complex_assignment): We are always in SSA form.
11014 (expand_complex_div_wide): Likewise.
11015 (expand_complex_operations_1): Likewise.
11016 (expand_complex_libcall): Preserve EH info of the original stmt.
11017 (tree_lower_complex): Handle removed blocks.
11018 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
11019 on complex multiplication and division libcall builtins.
11021 2018-05-04 Richard Biener <rguenther@suse.de>
11023 PR middle-end/85574
11024 * fold-const.c (negate_expr_p): Restrict negation of operand
11025 zero of a division to when we know that can happen without
11027 (fold_negate_expr_1): Likewise.
11029 2018-05-04 Jakub Jelinek <jakub@redhat.com>
11032 * real.h (real_nextafter): Declare.
11033 * real.c (real_nextafter): New function.
11034 * fold-const-call.c (fold_const_nextafter): New function.
11035 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
11036 CASE_CFN_NEXTTOWARD.
11037 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
11038 even when arg1_mode is different from arg0_mode.
11040 2018-05-03 Nathan Sidwell <nathan@acm.org>
11042 * doc/extend.texi (Deprecated Features): Remove
11043 -ffriend-injection.
11044 (Backwards Compatibility): Likewise.
11045 * doc/invoke.texi (C++ Language Options): Likewise.
11046 (C++ Dialect Options): Likewise.
11048 2018-05-03 Jakub Jelinek <jakub@redhat.com>
11051 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
11052 _mm512_mask_mullox_epi64): New intrinsics.
11054 2018-05-03 Tom de Vries <tom@codesourcery.com>
11057 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
11058 dump files): Add offload-tree.
11060 2018-05-03 Richard Biener <rguenther@suse.de>
11062 PR tree-optimization/85615
11063 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
11064 to loops not nested in BBs loop father to avoid creating multi-entry
11067 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11069 PR tree-optimization/70291
11070 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
11071 arguments. Change return type to tree. Emit libcall as a new
11072 statement rather than replacing existing one when inplace_p is true.
11073 (expand_complex_multiplication_components): New function.
11074 (expand_complex_multiplication): Expand floating-point complex
11075 multiplication using the above.
11076 (expand_complex_division): Rename inner_type parameter to type.
11077 Update expand_complex_libcall call-site.
11078 (expand_complex_operations_1): Update expand_complex_multiplication
11079 and expand_complex_division call-sites.
11081 2018-05-02 Jakub Jelinek <jakub@redhat.com>
11084 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
11085 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
11086 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
11087 the highest significant bit of the shift count mask is clear. In
11088 check whether and[sq]i3 is needed verify that all significant bits
11089 of the shift count other than the highest are set.
11091 2018-05-02 Tom de Vries <tom@codesourcery.com>
11094 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
11095 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
11096 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
11097 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
11098 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
11099 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
11100 __builtin_goacc_parlevel_size.
11102 2018-05-02 Richard Biener <rguenther@suse.de>
11104 PR tree-optimization/85597
11105 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
11106 do not use split vect_get_vec_defs call but call vect_get_slp_defs
11109 2018-05-02 Tom de Vries <tom@codesourcery.com>
11112 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
11113 dump files): Add ltrans-tree.
11115 2018-05-02 Tom de Vries <tom@codesourcery.com>
11118 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
11119 dump files): Add wpa-ipa.
11121 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
11123 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
11124 powerpc*-*-linux*paired* target.
11125 * config/rs6000/750cl.h: Delete.
11126 * config/rs6000/paired.h: Delete.
11127 * config/rs6000/paired.md: Delete.
11128 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
11130 * config/rs6000/rs6000-builtin.def: Remove paired float support.
11131 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
11132 comment. Remove paired float support.
11133 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
11134 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
11136 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
11137 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
11139 * config/rs6000/rs6000.c: Remove paired float support.
11140 (paired_expand_vector_init, paired_expand_vector_move,
11141 paired_emit_vector_compare, paired_emit_vector_cond_expr,
11142 (paired_expand_lv_builtin, paired_expand_stv_builtin,
11143 paired_expand_builtin, paired_expand_predicate_builtin,
11144 paired_init_builtins): Delete.
11145 * config/rs6000/rs6000.h: Remove paired float support.
11146 * config/rs6000/rs6000.md: Remove paired float support.
11147 (move_from_CR_ov_bit): Delete.
11148 * config/rs6000/rs6000.opt (mpaired): Delete.
11149 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
11150 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
11152 2018-05-02 Richard Biener <rguenther@suse.de>
11154 PR middle-end/85567
11155 * gimplify.c (gimplify_save_expr): When in SSA form allow
11156 SAVE_EXPRs to compute to SSA vars.
11158 2018-05-02 Jakub Jelinek <jakub@redhat.com>
11161 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
11162 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
11163 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
11164 clobber operands[2], instead use a new pseudo. Formatting fixes.
11166 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
11168 PR tree-optimization/85586
11169 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
11170 exit early for statements in the same group if the accesses are
11173 2018-05-02 Tom de Vries <tom@codesourcery.com>
11176 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
11179 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
11181 PR tree-optimization/85143
11182 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
11184 2018-05-01 Tom de Vries <tom@codesourcery.com>
11187 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
11188 not found" error message.
11190 2018-05-01 Tom de Vries <tom@codesourcery.com>
11193 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
11194 * vec.c (test_ordered_remove_if): New function.
11195 (vec_c_tests): Call test_ordered_remove_if.
11196 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
11197 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
11198 * tree-vect-patterns.c (vect_pattern_recog_1): Use
11199 VEC_ORDERED_REMOVE_IF.
11201 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11203 PR tree-optimization/82665
11204 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
11205 pointer subtraction where arguments come from a memchr call.
11207 2018-05-01 Jakub Jelinek <jakub@redhat.com>
11209 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
11210 --push-state --as-needed and --pop-state instead of --as-needed and
11211 --no-as-needed if ld supports it.
11212 * configure: Regenerated.
11215 * doc/install.texi2html: Replace _002d with - and _002a with * in
11216 generated html files using sed.
11218 2018-04-30 David Malcolm <dmalcolm@redhat.com>
11221 * gcc-rich-location.c (blank_line_before_p): New function.
11222 (use_new_line): New function.
11223 (gcc_rich_location::add_fixit_insert_formatted): New function.
11224 * gcc-rich-location.h
11225 (gcc_rich_location::add_fixit_insert_formatted): New function.
11227 2018-04-30 David Malcolm <dmalcolm@redhat.com>
11229 * selftest.c (assert_streq): Rename "expected" and "actual" to
11230 "val1" and "val2". Extend NULL-handling to cover both inputs
11231 symmetrically, while still requiring both to be non-NULL for a pass.
11232 * selftest.h (assert_streq): Rename "expected" and "actual" to
11234 (ASSERT_EQ): Likewise.
11235 (ASSERT_EQ_AT): Likewise.
11236 (ASSERT_KNOWN_EQ): Likewise.
11237 (ASSERT_KNOWN_EQ_AT): Likewise.
11238 (ASSERT_NE): Likewise.
11239 (ASSERT_MAYBE_NE): Likewise.
11240 (ASSERT_MAYBE_NE_AT): Likewise.
11241 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
11242 the assertion to pass.
11243 (ASSERT_STREQ_AT): Likewise.
11245 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
11247 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
11248 interaction with -pie.
11250 2018-04-30 David Malcolm <dmalcolm@redhat.com>
11252 * selftest.h: Fix alphabetization of per-source-file selftest
11255 2018-04-30 Jason Merrill <jason@redhat.com>
11257 PR c++/61982 - dead stores to destroyed objects.
11258 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
11261 2018-04-30 Jason Merrill <jason@redhat.com>
11263 * tree.c (build_clobber): New.
11264 * tree.h: Declare it.
11265 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
11267 2018-04-30 David Malcolm <dmalcolm@redhat.com>
11269 * diagnostic-show-locus.c (layout::layout): Update for
11270 location_get_source_line returning a char_span.
11271 (struct char_span): Move to input.h.
11272 (struct correction): Update for fields in char_span becoming
11274 (struct source_line): Update for location_get_source_line
11275 returning a char_span.
11276 (layout::print_line): Likewise.
11277 * edit-context.c (edited_file::print_content): Likewise.
11278 (edited_file::print_diff_hunk): Likewise.
11279 (edited_file::print_run_of_changed_lines): Likewise.
11280 (edited_file::get_num_lines): Likewise.
11281 (edited_line::edited_line): Likewise.
11282 * final.c (asm_show_source): Likewise.
11283 * input.c (location_get_source_line): Convert return type
11284 from const char * to char_span, losing the final "line_len"
11286 (dump_location_info): Update for the above.
11287 (get_substring_ranges_for_loc): Likewise. Use a char_span
11288 when handling the literal within the line.
11289 (test_reading_source_line): Update for location_get_source_line
11290 returning a char_span.
11291 * input.h (class char_span): Move here from
11292 diagnostic-show-locus.c, converting from a struct to a class.
11293 Make data members private.
11294 (char_span::operator bool): New.
11295 (char_span::length): New.
11296 (char_span::get_buffer): New.
11297 (char_span::operator[]): New.
11298 (char_span::subspan): Make const.
11299 (char_span::xstrdup): New.
11300 (location_get_source_line): Convert return type from const char *
11301 to char_span, losing the final "line_size" param.
11303 2018-04-30 Jan Hubicka <jh@suse.cz>
11305 * lto-wrapper.c (ltrans_priorities): New static var.
11306 (cmp_priority): New.
11307 (run_gcc): Read priorities and if doing parallel build order
11308 the Makefile by them.
11310 2018-04-30 David Malcolm <dmalcolm@redhat.com>
11312 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
11314 2018-04-30 Richard Biener <rguenther@suse.de>
11316 * tree-cfg.c (verify_address): Remove base argument, add
11317 flag whether to check TREE_ADDRESSABLE and do that.
11318 (verify_expr): Remove.
11319 (verify_types_in_gimple_reference): Add pieces from verify_expr.
11320 (verify_gimple_assign_single): Likewise.
11321 (verify_gimple_switch): Likewise.
11322 (verify_expr_location_1): Dereference tp once. Add (disabled)
11323 piece from verify_expr.
11324 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
11326 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
11328 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
11330 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
11332 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
11333 (small_data_pattern): Likewise.
11334 (arc_rewrite_small_data): Likewise.
11335 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
11336 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
11337 (get_symbol_alignment): New function.
11338 (legitimate_small_data_address_p): Likewise.
11339 (legitimate_scaled_address): Update, call
11340 legitimate_small_data_address_p.
11341 (output_sdata): New static variable.
11342 (arc_print_operand): Update how we handle small data operands.
11343 (arc_print_operand_address): Likewise.
11344 (arc_legitimate_address_p): Update, use
11345 legitimate_small_data_address_p.
11346 (arc_rewrite_small_data_p): Remove.
11347 (arc_rewrite_small_data_1): Likewise.
11348 (arc_rewrite_small_data): Likewise.
11349 (small_data_pattern): Likewise.
11350 (compact_sda_memory_operand): Update to use
11351 legitimate_small_data_address_p and get_symbol_alignment.
11352 (prepare_move_operands): Don't rewite sdata pattern.
11353 (prepare_extend_operands): Remove.
11354 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
11356 (zero_extendqisi2): Likewise.
11357 (zero_extendhisi2): Likewise.
11358 (extendqihi2): Likewise.
11359 (extendqisi2): Likewise.
11360 (extendhisi2): Likewise.
11361 (addsi3): Likewise.
11362 (subsi3): Likewise.
11363 (andsi3): Likewise.
11364 * config/arc/constraints.md (Usd): Change it to memory constraint.
11366 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
11368 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
11369 as source of std instructions.
11370 * config/arc/arc.md (movsi_insn): Update pattern predicate to
11371 allow 6-bit constants as source for store instructions.
11372 (movdi_insn): Update instruction pattern to allow 6-bit constants
11373 as source for store instructions.
11375 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
11377 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
11379 2018-04-30 Nathan Sidwell <nathan@acm.org>
11380 Sandra Loosemore <sandra@codesourcery.com>
11382 * dumpfile.c (dump_open): Allow '-' for stdout.
11383 * doc/invoke.texi (Developer Options): Document dump filename
11384 determination early. Document stdin/stdout selection.
11386 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
11388 Microblaze Target: PIC data text relative
11390 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
11391 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
11393 * config/microblaze/microblaze.h (microblaze_constant_address_p):
11394 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
11395 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
11396 New addressing mode for data-text relative position indepenedent code.
11397 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
11398 'ADDRESS_SYMBOLIC_TXT_REL'.
11399 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
11400 (microblaze_legitimate_pic_operand): Exclude function calls from
11401 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
11402 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
11404 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
11405 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
11406 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
11407 for 'address + offset'.
11408 (microblaze_expand_prologue): Add new function prologue call for
11410 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
11411 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
11412 table in case of TARGET_PIC_DATA_TEXT_REL.
11413 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
11414 * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
11415 Add new macros 'UNSPEC_TEXT',
11416 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
11417 + exclude function calls from 'UNSPEC_PLT' in case of data text
11419 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
11420 new target hook for generating address diff vector tables in case of
11422 * doc/tm.texi : Regenerate.
11423 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
11424 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
11425 of addr diff vector generation.
11426 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
11427 target hook definition.
11428 * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
11429 Add default function for generate_pic_addr_diff_vec -> flag_pic.
11430 * doc/invoke.texi (Add new pic option): Add new microblaze pic
11431 option for data text relative.
11433 2018-04-30 Richard Biener <rguenther@suse.de>
11435 * tree-chrec.h (evolution_function_is_constant_p): Remove
11437 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
11439 2018-04-30 Richard Biener <rguenther@suse.de>
11442 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
11444 2018-04-30 Richard Biener <rguenther@suse.de>
11446 PR tree-optimization/28364
11447 PR tree-optimization/85275
11448 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
11449 copying first exit test.
11451 2018-04-28 Mark Wielaard <mark@klomp.org>
11453 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
11454 dwarf_version >= 5.
11455 (dwarf_AT): Handle DW_AT_addr_base.
11456 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
11458 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
11461 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
11462 (*ashl<dwi>3_doubleword_mask_1): Ditto.
11463 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
11464 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
11466 2018-04-28 Richard Biener <rguenther@suse.de>
11468 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
11469 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
11470 to reflect use. Only add interesting stmts.
11472 2018-04-27 Martin Jambor <mjambor@suse.cz>
11475 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
11476 the jump function allows for passing through aggregate values.
11478 2018-04-27 David Malcolm <dmalcolm@redhat.com>
11480 * input.h (in_system_header_at): Convert from macro to inline
11482 (from_macro_expansion_at): Likewise.
11483 (from_macro_definition_at): Likewise.
11485 2018-04-27 Jeff Law <law@redhat.com>
11487 * config.gcc: Mark tile* targets as deprecated/obsolete.
11489 2018-04-27 Richard Biener <rguenther@suse.de>
11491 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
11494 2018-04-27 Richard Biener <rguenther@suse.de>
11496 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
11498 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
11500 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
11501 with Yd constraint. Set "preferred_for_speed" attribute from
11502 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
11503 with Yd constraint.
11504 (*movdi_internal): Ditto.
11505 (movti_interunit splitters): Remove
11506 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
11507 (movdi_interunit splitters): Ditto.
11508 * config/i386/constraints.md (Ye): Remove.
11509 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
11511 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11514 * config/aarch64/constraints.md (Usg): Limit to 31.
11515 (Usj): Limit to 63.
11517 2018-04-27 Jakub Jelinek <jakub@redhat.com>
11519 PR tree-optimization/85529
11520 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
11521 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
11522 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
11523 zero extension or masking of the MSB bit.
11524 (optimize_range_tests): Add FIRST_BB argument, pass it through
11525 to optimize_range_tests_var_bound.
11526 (maybe_optimize_range_tests, reassociate_bb): Adjust
11527 optimize_range_tests callers.
11529 2018-04-26 Richard Biener <rguenther@suse.de>
11530 Jakub Jelinek <jakub@redhat.com>
11532 * cgraph.h (symbol_table): Just declare debug method here.
11533 * symtab.c (symbol_table::debug): Define.
11535 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
11537 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
11539 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
11541 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
11542 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
11543 (*movdi_internal): Substitute Yi and Yj constraint with x
11544 and Ym and Yn constraint with y constraint. Update "isa"
11545 attribute and set "preferred_for_speed" attribute from
11546 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
11547 (*movsi_internal): Ditto.
11548 (*movdf_internal): Ditto.
11549 (*movsf_internal): Ditto.
11550 (*zero_extendsidi2): Ditto.
11551 * config/i386/sse.md (vec_set<mode>_0): Ditto.
11552 (sse2_loadld): Ditto.
11553 (*vec_extract<ssevecmodelower>_0): Ditto.
11554 (*vec_extractv4si_0_zext_sse4): Ditto.
11555 (vec_concatv2di): Ditto.
11556 (*vec_dup<mode>): Ditto.
11557 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
11558 * config/i386/constraints.md (Yi): Remove.
11563 2018-04-26 Nathan Sidwell <nathan@acm.org>
11565 * dumpfile.c (dump_open): New.
11566 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
11567 (dump_finish): Detect stdio/stderr by value not name.
11569 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
11571 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
11573 2018-04-26 Tom de Vries <tom@codesourcery.com>
11576 * config/nvptx/nvptx.c (verify_neutering_jumps)
11577 (verify_neutering_labels): New function
11578 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
11580 2018-04-26 Tom de Vries <tom@codesourcery.com>
11583 * config/nvptx/nvptx.c (needs_neutering_p): New function.
11584 (nvptx_single): Use needs_neutering_p to skip over insns that do not
11587 2018-04-26 Richard Biener <rguenther@suse.de>
11588 Tom de Vries <tom@codesourcery.com>
11591 * lto-streamer-out.c (output_function): Fixup loops if required to match
11592 discovery done in the reader.
11594 2018-04-26 Richard Biener <rguenther@suse.de>
11596 PR tree-optimization/85116
11597 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
11598 have a loop exit from the single latch predecessor. Remove
11599 case of header with just condition.
11600 (ch_base::copy_headers): Exclude infinite loops from any
11602 (pass_ch::execute): Record exits.
11604 2018-04-26 Richard Biener <rguenther@suse.de>
11606 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
11607 prologue cost vector and pass it to vect_get_load_cost.
11608 (vect_get_peeling_costs_all_drs): Likewise.
11609 (vect_peeling_hash_get_lowest_cost): Likewise.
11610 (vect_enhance_data_refs_alignment): Likewise.
11612 2018-04-26 Richard Biener <rguenther@suse.de>
11614 PR middle-end/85450
11615 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
11616 checking of integer<->pointer conversions.
11617 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
11618 sign-/zero-extending pointer types.
11619 (expand_omp_for_static_chunk): Likewise.
11621 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
11622 Jean Lee <xiaoyur347@gmail.com>
11624 * config/mips/mips.c (mips_asan_shadow_offset): New function.
11625 (TARGET_ASAN_SHADOW_OFFSET): Define.
11626 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
11627 true for -fsanitize=address.
11629 2018-04-25 Mark Wielaard <mark@klomp.org>
11631 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
11634 2018-04-25 Jakub Jelinek <jakub@redhat.com>
11636 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
11637 than "alu", remove explicit "memory" and "imm_disp" attributes.
11638 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
11640 PR middle-end/85414
11641 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
11642 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
11643 gen_lowpart_no_emit.
11645 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
11648 * config/i386/i386.c (ix86_expand_builtin): Change memory
11649 operand to XI, extend p0 to Pmode.
11650 * config/i386/i386.md: Change unspec volatile and operand
11651 1 mode to XI, change operand 0 mode to P.
11653 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
11655 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
11656 GET_MODE_MASK before any checking.
11657 (nds32_can_use_bset_p): Likewise.
11658 (nds32_can_use_btgl_p): Likewise.
11660 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
11662 * config/nds32/nds32-doubleword.md: New define_split pattern for
11663 illegal register number.
11665 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
11667 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
11669 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
11671 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
11673 2018-04-25 Richard Biener <rguenther@suse.de>
11675 * lto-streamer.h (LTO_major_version): Bump to 8.
11677 2018-04-25 Jakub Jelinek <jakub@redhat.com>
11679 * BASE-VER: Set to 9.0.0.
11681 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
11683 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
11684 in __abskf2 and __powikf2.
11686 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11689 * config/aarch64/constraints.md (Usg, Usj): New constraints.
11690 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
11691 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
11692 Use the above on operand 2. Reindent.
11693 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
11695 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
11698 * common/config/i386/i386-common.c (ix86_handle_option): Don't
11700 * config/i386/i386.opt (mcet): Removed.
11701 * doc/install.texi: Remove -mcet documentation.
11702 * doc/invoke.texi: Likewise.
11704 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
11707 * doc/install.texi: Remove -mcet from bootstrap-cet.
11709 2018-04-24 Jakub Jelinek <jakub@redhat.com>
11712 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
11713 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
11717 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
11718 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
11719 containing a CONST_VECTOR.
11721 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
11723 * doc/install.texi: Update newlib dependency for nvptx.
11725 2018-04-24 Jakub Jelinek <jakub@redhat.com>
11728 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
11729 instead of INTVAL when shifting x left.
11731 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
11733 PR tree-optimization/85478
11734 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
11735 vect_grouped_store_supported for single element vectors.
11737 2018-04-24 Richard Biener <rguenther@suse.de>
11740 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
11741 load cost increase to the case of non-constant step.
11743 2018-04-24 Jakub Jelinek <jakub@redhat.com>
11746 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
11747 destination if any_malformed_asm.
11749 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
11751 PR middle-end/85496
11752 * expr.c (store_field): In the bitfield case, if the value comes from
11753 a function call and is returned in registers by means of a PARALLEL,
11754 do not change the mode of the temporary unless BLKmode and VOIDmode.
11756 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
11758 PR rtl-optimization/85423
11759 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
11760 dependencies to debug insns when the previous insn is non-debug.
11762 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
11764 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
11765 enums into a single definition.
11766 (fls): Fix predicates and printing.
11769 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
11771 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
11772 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
11773 and short u6 immediate.
11774 (check_if_valid_sleep_operand): Remove.
11775 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
11777 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
11779 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
11780 flag_always_save_lp condition.
11781 * config/nds32/nds32.opt (malways-save-lp): New option.
11783 2018-04-22 Shiva Chen <shiva0217@gmail.com>
11785 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
11786 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
11787 * config/nds32/nds32.h
11788 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
11789 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
11791 2018-04-22 Shiva Chen <shiva0217@gmail.com>
11793 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
11794 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
11796 2018-04-22 Shiva Chen <shiva0217@gmail.com>
11797 Chung-Ju Wu <jasonwucj@gmail.com>
11799 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
11801 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
11802 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
11804 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
11806 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
11808 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
11810 * config/nds32/nds32-protos.h (nds32_data_alignment,
11811 nds32_local_alignment): Declare.
11812 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
11813 nds32_local_alignment): New functions.
11814 (TARGET_CONSTANT_ALIGNMENT): Define.
11815 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
11817 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
11819 * config/nds32/nds32.c
11820 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
11821 (TARGET_MODES_TIEABLE_P): Likewise.
11823 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
11825 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
11826 level Ofast and Og.
11828 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
11829 Chung-Ju Wu <jasonwucj@gmail.com>
11831 * config/nds32/constants.md (unspec_volatile_element): Add enum values
11832 for unaligned access.
11833 * config/nds32/nds32-intrinsic.c: Implementation of expanding
11835 * config/nds32/nds32-intrinsic.md: Likewise.
11836 * config/nds32/nds32_intrinsic.h: Likewise.
11837 * config/nds32/nds32.h (nds32_builtins): Likewise.
11838 * config/nds32/nds32.opt (munaligned-access): New option.
11839 * config/nds32/nds32.c (nds32_asm_file_start): Display
11840 flag_unaligned_access status.
11842 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
11844 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
11845 -mno-relax is present.
11846 * config/riscv/linux.h (LINK_SPEC): Ditto.
11848 2018-04-20 Martin Sebor <msebor@redhat.com>
11851 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
11853 (gimple_fold_builtin_stxcpy_chk): Same.
11854 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
11856 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
11859 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
11860 __powikf2 when long double is IEEE 128-bit.
11862 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
11864 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
11865 step to make sure stack always aligned.
11867 2018-04-20 Carl Love <cel@us.ibm.com>
11870 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
11871 size check for arg0.
11873 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
11874 Tom de Vries <tom@codesourcery.com>
11877 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
11878 Emit insns for calls too.
11879 (nvptx_find_par): Always look for worker-level predecessor insn.
11880 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
11882 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
11883 (nvptx_process_pars): Propagate frames for calls.
11885 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
11888 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
11890 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
11891 (ix86_handle_option): Don't handle OPT_mibt.
11892 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
11894 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
11896 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
11897 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
11898 (ix86_target_macros): Define __CET__ with flag_cf_protection
11899 for -fcf-protection.
11900 * config/i386/i386.c (isa2_opts): Remove -mibt.
11901 * config/i386/i386.h (TARGET_IBT): Removed.
11902 (TARGET_IBT_P): Likewise.
11903 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
11904 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
11905 * config/i386/i386.opt (mcet): Update help message.
11906 (mshstk): Likewise.
11908 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
11909 -mcet as an alias for -mshstk.
11911 2018-04-20 Richard Biener <rguenther@suse.de>
11913 PR middle-end/85475
11914 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
11915 complexity by forcing a single use of the multiply operand.
11917 2018-04-20 Martin Jambor <mjambor@suse.cz>
11920 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
11921 recursion dependency to only apply to non-clones.
11923 2018-04-20 Martin Jambor <mjambor@suse.cz>
11926 * ipa-cp.c (create_specialized_node): Check that clones of
11927 self-recursive edges exist during IPA-CP.
11929 2018-04-19 Toon Moene <toon@moene.org>
11931 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
11934 2018-04-19 Jakub Jelinek <jakub@redhat.com>
11936 PR tree-optimization/85467
11937 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
11938 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
11939 VECTOR_CST element to type.
11941 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
11944 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
11945 * config/i386/i386.md (builtin_setjmp_setup): Removed.
11946 (builtin_longjmp): Likewise.
11947 (save_stack_nonlocal): New pattern.
11948 (restore_stack_nonlocal): Likewise.
11950 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
11953 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
11954 Replace ASM_OUTPUT_LABEL with fprintf.
11956 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
11959 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
11960 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
11961 * config/i386/i386-c.c (ix86_target_macros_internal): Also
11962 define __IBT__ and __SHSTK__ for -fcf-protection.
11963 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
11965 (ix86_trampoline_init): Likewise.
11966 (x86_output_mi_thunk): Likewise.
11967 (ix86_notrack_prefixed_insn_p): Likewise.
11968 (ix86_option_override_internal): Don't disallow -fcf-protection.
11969 * config/i386/i386.md (rdssp<mode>): Also enable for
11971 (incssp<mode>): Likewise.
11972 (nop_endbr): Likewise.
11973 * config/i386/i386.opt (mcet): Change help message to built-in
11976 (mshstk): Likewise.
11977 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
11978 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
11979 enable CET built-in functions.
11981 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
11983 * common/config/i386/i386-common.c
11984 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
11985 OPTION_MASK_ISA_MOVDIRI_UNSET,
11986 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
11987 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
11988 * config.gcc (movdirintrin.h): New header.
11989 * config/i386/cpuid.h (bit_MOVDIRI,
11990 bit_MOVDIR64B): New bits.
11991 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
11993 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
11994 (VOID, PVOID, PCVOID)): New function types.
11995 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
11996 __builtin_ia32_directstoreu_u64,
11997 __builtin_ia32_movdir64b): New builtins.
11998 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
11999 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
12001 (ix86_valid_target_attribute_inner_p): Ditto.
12002 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
12003 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
12004 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
12005 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
12006 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
12007 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
12008 (movdiri<mode>, movdir64b_<mode>): New.
12009 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
12010 * config/i386/immintrin.h: Include movdirintrin.h.
12011 * config/i386/movdirintrin.h: New file.
12012 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
12014 2018-04-19 Richard Biener <rguenther@suse.de>
12016 PR middle-end/85455
12017 * cfg.c (clear_bb_flags): When loop state says we have
12018 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
12020 2018-04-19 Richard Biener <rguenther@suse.de>
12022 PR tree-optimization/84737
12023 * tree-vect-data-refs.c (vect_copy_ref_info): New function
12024 copying restrict info.
12025 (vect_setup_realignment): Use it.
12026 * tree-vectorizer.h (vect_copy_ref_info): Declare.
12027 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
12028 the first DR to all generated stores.
12029 (vectorizable_load): Likewise for loads.
12031 2018-04-19 Jakub Jelinek <jakub@redhat.com>
12033 PR tree-optimization/85446
12034 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
12035 the integral and pointer types to have the same precision.
12037 * doc/install.texi: Document --disable-cet being the default and
12040 2018-04-18 Martin Liska <mliska@suse.cz>
12042 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
12045 2018-04-18 Martin Liska <mliska@suse.cz>
12048 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
12051 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
12052 arguments if they are comparable.
12054 2018-04-18 Martin Liska <mliska@suse.cz>
12057 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
12060 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
12063 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
12066 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
12067 ENDBR after calling __morestack.
12069 2018-04-18 David Malcolm <dmalcolm@redhat.com>
12072 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
12073 by using gcc_base_ver to generate a gcc_driver_version, and use
12074 it when generating GCC_DRIVER_NAME.
12075 * configure: Regenerate.
12077 2018-04-18 Jakub Jelinek <jakub@redhat.com>
12080 * config.gcc: Obsolete powerpc*-*-*spe*.
12082 2018-04-17 Jakub Jelinek <jakub@redhat.com>
12085 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
12086 (stabstr_D): Change type of unum from unsigned int to
12087 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
12090 2018-04-17 Jim Wilson <jimw@sifive.com>
12093 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
12094 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
12095 Set arg_pointer_offset after using pretend_args_size.
12097 2018-04-17 Jakub Jelinek <jakub@redhat.com>
12099 PR rtl-optimization/85431
12100 * dse.c (record_store): Ignore zero width stores.
12103 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
12104 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
12105 __builtin_stack_restore rather than after it.
12106 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
12107 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
12108 argument instead of virtual_dynamic_stack_rtx.
12110 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
12112 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
12114 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
12115 Add note to error message to explain internal mapping of overloaded
12116 built-in function name to non-overloaded built-in function name.
12117 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
12120 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
12123 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
12124 where the inputs overlap with the output.
12126 2018-04-17 Jakub Jelinek <jakub@redhat.com>
12128 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
12129 (=v, v) alternative and explicit "memory" attribute.
12130 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
12131 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
12133 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
12134 "sselog1" type instead of "sselog".
12135 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
12136 "sselog". Remove explicit "memory" attribute.
12137 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
12138 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
12140 (vec_extract_hi_v32hi): Merge all alternatives into one, use
12141 "sselog1" type instead of "sselog". Remove explicit "memory"
12143 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
12144 use "sselog1" type instead of "sselog". Remove explicit "memory"
12146 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
12147 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
12149 (vec_extract_hi_v64qi): Merge all alternatives into one, use
12150 "sselog1" type instead of "sselog". Remove explicit "memory"
12152 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
12153 use "sselog1" type instead of "sselog". Remove explicit "memory"
12157 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
12159 PR middle-end/85414
12160 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
12163 2018-04-17 Martin Jambor <mjambor@suse.cz>
12166 * ipa-cp.c (create_specialized_node): Call
12167 expand_all_artificial_thunks if necessary.
12169 2018-04-17 Martin Liska <mliska@suse.cz>
12172 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
12173 in message, remote space in between '_G' and '('.
12175 2018-04-17 Jakub Jelinek <jakub@redhat.com>
12178 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
12179 avx512f_vmcmp<mode>3<round_saeonly_name>,
12180 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
12181 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
12182 avx512f_rndscale<mode><round_saeonly_name>,
12183 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
12184 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
12185 Use %<iptr>2 instead of %2 for -masm=intel.
12186 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
12187 avx512f_vcvttss2usi<round_saeonly_name>,
12188 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
12190 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
12191 avx512f_vcvttsd2usi<round_saeonly_name>,
12192 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
12193 Use %q1 instead of %1 for -masm=intel.
12194 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
12195 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
12196 of %3 for -masm=intel.
12197 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
12199 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
12201 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
12203 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
12204 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
12206 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
12208 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
12209 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
12210 %g1 and one with %0 and %1.
12211 (avx512er_vmrcp28<mode><round_saeonly_name>,
12212 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
12213 %1 for -masm=intel.
12214 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
12215 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
12216 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
12217 of %0 and %{%4%} for -masm=intel.
12218 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
12219 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
12220 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
12221 order of %0 and %{%5%}%{z%} for -masm=intel.
12223 2018-04-17 Jan Hubicka <jh@suse.cz>
12226 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
12228 2018-04-17 Martin Liska <mliska@suse.cz>
12231 * multiple_target.c (create_dispatcher_calls): Set apostrophes
12232 for target_clone error message. Make default implementation
12233 clone to be a local declaration.
12234 (separate_attrs): Add new argument and check for an empty
12236 (expand_target_clones): Handle it.
12237 (ipa_target_clone): Make redirection just for target_clones
12240 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
12241 Tom de Vries <tom@codesourcery.com>
12243 PR middle-end/84955
12244 * omp-expand.c (expand_oacc_for): Add dummy false branch for
12245 tiled basic blocks without omp continue statements.
12247 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
12250 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
12251 vec_extract expression as having side effects to make sure it gets
12254 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
12257 * config/i386/i386.c (get_builtin_code_for_version): Check
12260 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
12263 * config.gcc: Support "skylake".
12264 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
12266 * config/i386/i386.c (m_SKYLAKE): Define.
12267 (processor_target_table): Add "skylake".
12268 (ix86_option_override_internal): Add "skylake".
12269 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
12270 PROCESSOR_CANNONLAKE.
12271 (get_builtin_code_for_version): Fix priority for
12272 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
12273 PROCESSOR_SKYLAKE-AVX512.
12274 * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
12275 (processor_type): Add PROCESSOR_SKYLAKE.
12277 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
12278 Jason Merrill <jason@redhat.com>
12281 * convert.c (convert_to_integer_1): Use direct recursion for
12282 enumeral types and types with a precision less than the number
12283 of bits in their mode.
12285 2018-04-16 Julia Koval <julia.koval@intel.com>
12288 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
12289 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
12291 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
12294 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
12295 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
12296 and -mno-direct-move.
12298 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
12301 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
12302 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
12303 Ensure negative shifts result in {0}.
12305 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
12307 PR rtl-optimization/79916
12308 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
12309 regs (if any) to define how to gnerate SD moves when LRA is in
12312 2018-04-13 Jakub Jelinek <jakub@redhat.com>
12314 PR rtl-optimization/85393
12315 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
12316 * except.c (expand_dw2_landing_pad_for_region): Make static.
12317 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
12318 a label and unconditional jump to old_bb, rather than
12319 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
12322 PR rtl-optimization/85376
12323 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
12324 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
12325 instead of a specific value.
12327 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
12328 Bin Cheng <bin.cheng@arm.com>
12330 PR tree-optimization/82965
12331 PR tree-optimization/83991
12332 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
12333 by_profile_only parameter.
12334 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
12335 information if the loop was predicted to iterate too many times.
12336 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
12338 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
12341 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
12344 2018-04-13 Martin Liska <mliska@suse.cz>
12345 Jakub Jelinek <jakub@redhat.com>
12347 PR middle-end/81657
12348 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
12349 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
12350 * builtins.c (expand_builtin_memory_copy_args): Use
12351 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
12352 handle dest_addr == pc_rtx.
12354 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
12357 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
12358 asked to not generate direct moves.
12359 (fix_trunc<mode>si2_stfiwx): Similar.
12360 (fix_trunc<mode>si2_internal): Similar.
12362 2018-04-12 Jakub Jelinek <jakub@redhat.com>
12365 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
12366 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
12367 lookup if dest in some wider mode is known to be const0_rtx and
12368 if so, record permanent equivalence for it to be ZERO_EXTEND of
12369 the narrower mode destination.
12371 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
12373 * lto-streamer-out.c (output_function): Revert 259346.
12374 * omp-expand.c (expand_oacc_for): Likewise.
12376 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
12378 PR rtl-optimization/85354
12379 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
12380 * sel-sched.c (sel_global_init): ... here.
12382 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
12385 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
12386 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
12387 mode for PE-COFF targets.
12388 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
12389 (i386_pe_asm_lto_end): Likewise.
12390 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
12391 (TARGET_ASM_LTO_END): Likewise.
12392 * config/i386/winnt.c (saved_debug_info_level): New static variable.
12393 (i386_pe_asm_lto_start): New function.
12394 (i386_pe_asm_lto_end): Likewise.
12396 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
12397 Richard Biener <rguenther@suse.de>
12399 PR middle-end/84955
12400 * lto-streamer-out.c (output_function): Fix CFG loop state before
12402 * omp-expand.c (expand_oacc_for): Handle calls to internal
12403 functions like regular functions.
12405 2018-04-12 Richard Biener <rguenther@suse.de>
12408 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
12409 for the early LTO debug to properly generate references to it
12410 during DIE emission. Do not re-use that for the skeleton for
12412 (dwarf2out_early_finish): Likewise.
12414 2018-04-12 Jakub Jelinek <jakub@redhat.com>
12417 * config/i386/sse.md
12418 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
12419 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
12420 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
12421 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
12422 and output is a reg, avoid creating invalid lowpart subreg, but
12423 instead split into a 512-bit move. Don't split if not AVX512VL,
12424 input is xmm16+ reg and output is a mem.
12425 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
12426 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
12427 xmm16+ reg and output is a mem.
12429 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
12431 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
12432 also for flag_dwarf2_cfi_asm.
12434 2018-04-12 Jakub Jelinek <jakub@redhat.com>
12436 PR rtl-optimization/85342
12437 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
12438 a bool scalar var inside of the loop instead. Don't try to update
12439 recog_data.operand after failed apply_change_group.
12441 2018-04-12 Tom de Vries <tom@codesourcery.com>
12444 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
12445 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
12446 array with flexible array member as array without given dimension.
12447 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
12448 argument for undefined param to true.
12450 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
12453 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
12454 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
12455 from PowerPC section.
12456 * config/rs6000/sysv4.opt (mcall-): Improve help text.
12457 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
12458 help text that is too long.
12459 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
12460 help text that is too long.
12461 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
12462 help text that is too long.
12464 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
12466 * config/alpha/alpha.md (stack_probe_internal): Rename
12467 from "probe_stack". Update all callers.
12469 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
12471 PR rtl-optimization/84566
12472 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
12473 sched_macro_fuse_insns.
12475 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
12478 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
12479 (compute_block_dependences): ... from here.
12481 2018-04-11 Jakub Jelinek <jakub@redhat.com>
12483 PR tree-optimization/85331
12484 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
12485 from int to HOST_WIDE_INT.
12487 2018-04-11 Martin Jambor <mjambor@suse.cz>
12490 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
12491 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
12492 not the same as the source val.
12493 (cgraph_edge_brings_value_p): New parameter.
12494 (gather_edges_for_value): Pass destination value to
12495 cgraph_edge_brings_value_p.
12496 (perhaps_add_new_callers): Likewise.
12497 (get_info_about_necessary_edges): Likewise and exclude values brought
12498 only by self-recursive edges.
12499 (create_specialized_node): Redirect only clones of self-calling edges.
12500 (+self_recursive_pass_through_p): New function.
12501 (find_more_scalar_values_for_callers_subset): Use it.
12502 (find_aggregate_values_for_callers_subset): Likewise.
12503 (known_aggs_to_agg_replacement_list): Removed.
12504 (decide_whether_version_node): Re-calculate known constants for all
12505 remaining context clones.
12507 2018-04-11 Richard Biener <rguenther@suse.de>
12510 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
12511 from early DWARF output.
12512 (dwarf2out_early_finish): Output line info unconditionally into
12513 early DWARF and add reference to it.
12515 2018-04-11 Jakub Jelinek <jakub@redhat.com>
12518 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
12519 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
12520 other than V2DFmode using iptr mode attribute.
12521 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
12523 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
12525 PR rtl-optimization/84659
12526 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
12528 2018-04-11 Jakub Jelinek <jakub@redhat.com>
12531 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
12533 (output_loc_list): Pass address of a dummy size variable even in the
12534 locview handling loop.
12535 (index_location_lists): Add comment on why skip_loc_list_entry can't
12538 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
12541 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
12544 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
12547 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
12548 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
12549 and -mstring-compare-inline-limit.
12551 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
12554 * config/rs6000/rs6000.md (allocate_stack): Put the residual size
12555 for stack clash protection in a register whenever we need it to be in
12558 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
12560 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
12561 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
12563 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
12566 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
12568 (mlong-double-): Ditto.
12569 * config/rs6000/sysv4.opt (msdata=): Ditto.
12570 (mtls-size=): Ditto.
12572 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
12574 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
12575 erroneous entries for
12576 "vector int vec_ldl (int, long int *)", and
12577 "vector unsigned int vec_ldl (int, unsigned long int *)".
12578 Add comments and entries for
12579 "vector bool char vec_ldl (int, bool char *)",
12580 "vector bool short vec_ldl (int, bool short *)",
12581 "vector bool int vec_ldl (int, bool int *)",
12582 "vector bool long long vec_ldl (int, bool long long *)",
12583 "vector pixel vec_ldl (int, pixel *)",
12584 "vector long long vec_ldl (int, long long *)",
12585 "vector unsigned long long vec_ldl (int, unsigned long long *)".
12586 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
12587 type tree bool_long_long_type_node and correct definition of
12588 bool_V2DI_type_node to make reference to this new type tree.
12589 (rs6000_mangle_type): Replace erroneous reference to
12590 bool_long_type_node with bool_long_long_type_node.
12591 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
12592 comments to emphasize sign distinctions for char and int types and
12593 replace RS6000_BTI_bool_long constant with
12594 RS6000_BTI_bool_long_long constant. Also add comment to restrict
12595 use of RS6000_BTI_pixel.
12596 (bool_long_type_node): Remove this macro definition.
12597 (bool_long_long_type_node): New macro definition
12599 2018-04-10 Jakub Jelinek <jakub@redhat.com>
12601 PR rtl-optimization/85300
12602 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
12603 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
12604 simplify_unary_operation fails.
12606 2018-04-10 Martin Liska <mliska@suse.cz>
12608 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
12609 cgraph_edge and ipa_ref.
12611 2018-04-10 Jakub Jelinek <jakub@redhat.com>
12615 * config/i386/sse.md
12616 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
12617 computation of the VEC_MERGE selector from mask.
12618 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
12619 Fix decoding of the VEC_MERGE selector into mask.
12621 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
12623 PR tree-optimization/85286
12624 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
12626 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
12628 * final.c (final_1): Set insn_last_address as well as
12629 insn_current_address.
12631 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12634 * explow.c (emit_stack_probe): Call validize_mem on memory location
12635 before passing it to gen_probe_stack. Create address operand and
12636 legitimize it for the probe_stack_address case.
12638 2018-04-09 Jan Hubicka <jh@suse.cz>
12641 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
12642 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
12643 * tree.c (free_lang_data_in_type): Fix handling of binfos;
12645 (free_lang_data): Rebuild type inheritance graph.
12647 2018-04-09 Martin Sebor <msebor@redhat.com>
12649 * invoke.texi (-finline-small-functions): Mention other optimization
12651 (-findirect-inlining, -fpartial-inlining): Same.
12652 (-finline-functions-called-once): Same.
12653 (-freorder-blocks-and-partition): Same.
12655 2018-04-09 Jan Hubicka <jh@suse.cz>
12658 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
12659 jumps; choose last target that matches the criteria (i.e.
12660 no partition changes for non-crossing jumps).
12661 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
12662 support for redirecting crossing jumps to non-crossing.
12664 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
12666 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
12667 also for naked functions.
12669 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
12671 * config/arc/arc.md (add_shift): New pattern.
12672 (add_shift2): Likewise.
12673 (sub_shift): Likewise.
12674 (sub_shift_cmp0_noout): Likewise.
12675 (compare_si_ashiftsi): Likewise.
12676 (xbfu_cmp0_noout): New combine pattern.
12677 (xbfu_cmp0"): Likewise.
12678 (movsi_set_cc_insn): Place the predicable variant first.
12679 (commutative_binary_cmp0_noout): Remove clobber.
12680 (commutative_binary_cmp0): New pattern.
12681 (noncommutative_binary_cmp0): Likewise.
12682 (noncommutative_binary_cmp0_noout): Likewise.
12683 (noncommutative_binary_comparison_result_used): Removed.
12684 (rsub_cmp0): New pattern.
12685 (rsub_cmp0_noout): Likewise.
12686 (extzvsi): Changed, keep only meaningful variants.
12687 (SQH, SEZ): New iterators.
12688 (SQH_postfix): New mode attribute.
12689 (SEZ_prefix): New code attribute.
12690 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
12691 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
12692 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
12693 of numerical value.
12694 (noncommutative_operator): Check the availability of barrel
12697 2018-04-09 Richard Biener <rguenther@suse.de>
12699 PR tree-optimization/85284
12700 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
12701 Only use the niter constraining form of simple_iv when the exit
12702 is always executed.
12704 2018-04-09 Tom de Vries <tom@codesourcery.com>
12707 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
12708 (define_expand "*memory_barrier"): New define_expand.
12709 (define_insn "memory_barrier"): New insn.
12711 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
12713 PR rtl-optimization/80463
12714 PR rtl-optimization/83972
12715 PR rtl-optimization/83480
12717 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
12718 correct producer for the insn.
12719 (tidy_control_flow): Fixup seqnos in case of debug insns.
12721 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
12723 PR rtl-optimization/83913
12725 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
12726 different sched-times when merging exprs.
12728 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
12730 PR rtl-optimization/83962
12732 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
12733 tidy_fallthru_edge and tidy_control_flow.
12735 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
12737 PR rtl-optimization/83530
12739 * sel-sched.c (force_next_insn): New global variable.
12740 (remove_insn_for_debug): When force_next_insn is true, also leave only
12741 next insn in the ready list.
12742 (sel_sched_region): When the region wasn't scheduled, make another pass
12743 over it with force_next_insn set to 1.
12745 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
12747 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
12749 * config/nds32/constants.md (unspec_volatile_element): Add enum values
12750 for interrupt control.
12751 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
12752 functions for interrupt control.
12753 * config/nds32/nds32-intrinsic.md: Likewise.
12754 * config/nds32/nds32_intrinsic.h: Likewise.
12755 * config/nds32/nds32.h (nds32_builtins): Likewise.
12757 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
12759 * config/nds32/nds32.c (nds32_init_machine_status,
12760 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
12761 strict_aligned_p field.
12762 (nds32_expand_to_rtl_hook): New function.
12763 (TARGET_EXPAND_TO_RTL_HOOK): Define.
12764 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
12766 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
12767 Chung-Ju Wu <jasonwucj@gmail.com>
12769 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
12770 * config/nds32/nds32-n7.md: New file.
12771 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
12772 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
12774 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
12775 * config/nds32/nds32.md (pipeline_model): Add n7.
12776 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
12777 * config/nds32/pipelines.md: Include n7 settings.
12779 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
12780 Chung-Ju Wu <jasonwucj@gmail.com>
12782 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
12783 * config/nds32/nds32-e8.md: New file.
12784 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
12785 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
12787 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
12788 * config/nds32/nds32.md (pipeline_model): Add e8.
12789 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
12790 * config/nds32/pipelines.md: Include e8 settings.
12792 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
12793 Chung-Ju Wu <jasonwucj@gmail.com>
12795 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
12796 * config/nds32/nds32-n8.md: New file.
12797 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
12798 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
12800 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
12801 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
12802 * config/nds32/nds32.md (pipeline_model): Add n8.
12803 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
12804 * config/nds32/pipelines.md: Include n8 settings.
12806 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
12807 Chung-Ju Wu <jasonwucj@gmail.com>
12809 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
12810 * config/nds32/nds32-n9-2r1w.md: New file.
12811 * config/nds32/nds32-n9-3r2w.md: New file.
12812 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
12813 nds32_register_ports): New or modify for cpu n9.
12814 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
12816 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
12817 * config/nds32/nds32-utils.c: New file.
12818 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
12819 TARGET_MUL_SLOW): Define.
12820 * config/nds32/nds32.md (pipeline_model): New attribute.
12821 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
12822 New options that support cpu n9.
12823 * config/nds32/pipelines.md: Include n9 settings.
12824 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
12826 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
12828 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
12829 information if necessary.
12830 (output_cond_branch_compare_zero): Likewise.
12831 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
12832 (nds32_target_alignment): Refine for alignment.
12833 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
12834 (FUNCTION_BOUNDARY): Modify.
12835 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
12837 * config/nds32/nds32.opt (malways-align, malign-functions): New.
12839 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
12841 * config/nds32/constants.md (unspec_volatile_element): Add values for
12842 TLB operation and data prefetch.
12843 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
12844 functions for TLB operation and data prefetch.
12845 * config/nds32/nds32-intrinsic.md: Likewise.
12846 * config/nds32/nds32_intrinsic.h: Likewise.
12847 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
12848 (nds32_print_operand): Likewise.
12849 * config/nds32/nds32.h (nds32_builtins): Likewise.
12851 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
12852 Andrew Pinski <pinsika@gcc.gnu.org>
12854 PR middle-end/82976
12855 * match.pd: Use constant_boolean_node of correct type instead of
12856 boolean_true_node or boolean_false_node for simplifying
12857 pointer comparisons to zero.
12859 2018-04-07 Jakub Jelinek <jakub@redhat.com>
12861 PR tree-optimization/80021
12862 * tree.c (verify_type_variant): Make error call in verify_variant_match
12863 translatable and remove final full stop.
12865 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
12867 * config/nds32/constants.md (unspec_volatile_element): Add
12868 UNSPEC_VOLATILE_EH_RETURN.
12869 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
12870 nds32_output_stack_pop): Support dwarf exception handling process.
12871 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
12872 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
12873 exception handling process.
12874 (nds32_compute_stack_frame): Likewise.
12875 (nds32_return_addr_rtx): Likewise.
12876 (nds32_initial_elimination_offset): Likewise.
12877 (nds32_expand_prologue): Likewise.
12878 (nds32_expand_epilogue): Likewise.
12879 (nds32_dynamic_chain_address): New function.
12880 * config/nds32/nds32.h (machine_function): Add fields for dwarf
12881 exception handling.
12882 (DYNAMIC_CHAIN_ADDRESS): Define.
12883 (EH_RETURN_DATA_REGNO): Define.
12884 (EH_RETURN_STACKADJ_RTX): Define.
12885 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
12886 patterns for dwarf exception handling.
12888 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
12890 * config/nds32/nds32.h: Clean up obsolete macros.
12892 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
12894 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
12895 Add enum values for particular instructions.
12896 * config/nds32/nds32-intrinsic.c: Implementation of expanding
12897 particular intrinsic functions.
12898 * config/nds32/nds32-intrinsic.md: Likewise.
12899 * config/nds32/nds32_intrinsic.h: Likewise.
12900 * config/nds32/nds32.h (nds32_builtins): Likewise.
12901 * config/nds32/nds32.md (type): Add pbsad and pbsada.
12902 (btst, ave): New patterns for particular instructions.
12904 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
12906 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
12907 Add enum values for atomic load/store and memory sync.
12908 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
12910 * config/nds32/nds32-intrinsic.md: Likewise.
12911 * config/nds32/nds32_intrinsic.h: Likewise.
12912 * config/nds32/nds32.h (nds32_builtins): Likewise.
12914 2018-04-07 Jakub Jelinek <jakub@redhat.com>
12916 PR tree-optimization/85257
12917 * fold-const.c (native_encode_vector): If not all elts could fit
12918 and off is -1, return 0 rather than offset.
12919 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
12920 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
12921 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
12922 adjust buffer in native_interpret_expr call.
12924 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
12926 * config/nds32/constants.md (unspec_volatile_element): Add cache
12927 control enum values.
12928 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
12929 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
12930 * config/nds32/nds32.c (nds32_cctl_names): New.
12931 (nds32_print_operand): Handle cache control register names.
12932 * config/nds32/nds32.h (nds32_builtins): New enum values.
12933 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
12935 * config/nds32/nds32.md (type): Add mmu.
12936 * config/nds32/pipelines.md (simple_insn): Add mmu.
12938 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
12940 * config/nds32/nds32.md (type): Remove call.
12941 * config/nds32/pipelines.md (simple_insn): Likewise.
12943 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
12945 * config/nds32/constants.md (unspec_volatile_element): Add
12946 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
12947 UNSPEC_VOLATILE_FMFCFG.
12948 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
12949 description for fmfcfg and fmfcsr.
12950 (bdesc_1arg): Add fmtcsr.
12951 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
12952 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
12953 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
12954 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
12955 unspec_fmfcfg): New patterns.
12956 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
12957 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
12958 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
12959 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
12960 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
12961 __nds32__fmfcfg): Define.
12963 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
12965 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
12966 intrinsic register names.
12967 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
12968 intrinsic register enum values and macros.
12970 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
12972 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
12973 for load/store addressing form.
12974 (nds32_print_operand_address): Likewise.
12976 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
12979 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
12980 based on LABEL_REF. Remove useless assertion.
12981 (pic_address_needs_scratch): Fix formatting.
12982 (sparc_legitimize_pic_address): Minor tweaks.
12983 (sparc_delegitimize_address): Adjust assertion accordingly.
12984 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
12985 into symbolic_operand.
12986 (movsi_high_pic_label_ref): Likewise.
12987 (movsi_lo_sum_pic_label_ref): Likewise.
12988 (movdi_pic_label_ref): Likewise.
12989 (movdi_high_pic_label_ref): Likewise.
12990 (movdi_lo_sum_pic_label_ref): Likewise.
12992 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
12994 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
12995 custom LIB_SPEC setup.
12997 2018-04-06 Ruslan Bukin <br@bsdpad.com>
12998 Kito Cheng <kito.cheng@gmail.com>
13000 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
13001 * config/riscv/freebsd.h: New.
13003 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
13005 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
13006 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
13009 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
13010 Kito Cheng <kito.cheng@gmail.com>
13012 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
13013 nds32_output_call, nds32_symbol_binds_local_p): New functions.
13014 * config/nds32/nds32-protos.h (nds32_output_call,
13015 nds32_output_return): Declare.
13016 * config/nds32/nds32.md: Refine all the call and return patterns.
13018 2018-04-06 Jakub Jelinek <jakub@redhat.com>
13021 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
13022 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
13024 PR rtl-optimization/84872
13025 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
13026 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
13027 EDGE_CROSSING edge.
13029 2018-04-06 Tamar Christina <tamar.christina@arm.com>
13031 * expr.c (copy_blkmode_to_reg): Revert 254862.
13032 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
13034 2018-04-06 Richard Biener <rguenther@suse.de>
13036 PR middle-end/85244
13037 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
13038 after seeing a component reference with an adjacent field. Treat
13039 refs to arrays at struct end of external decls similar to
13040 refs to unconstrained commons.
13042 2018-04-06 Jakub Jelinek <jakub@redhat.com>
13045 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
13046 look through SAVE_EXPRs with non-side-effects argument. Adjust
13048 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
13051 2018-04-06 Richard Biener <rguenther@suse.de>
13053 PR middle-end/85180
13054 * alias.c (find_base_term): New wrapper around find_base_term
13055 unwinding CSELIB_VAL_PTR changes.
13056 (find_base_term): Do not restore CSELIB_VAL_PTR during the
13059 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
13061 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
13063 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
13064 constant definitions.
13065 ("nop"): lr 0,0 -> nopr r0
13066 ("nop_lr0", "nop_lr1"): New insn definitions.
13068 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
13070 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
13071 NDS32_V3PUSH_AVAILABLE_P macro.
13073 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
13074 Chung-Ju Wu <jasonwucj@gmail.com>
13076 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
13077 (nds32*-*-*): Add float and fpu_config into supported_defaults.
13078 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
13079 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
13080 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
13081 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
13082 * config/nds32/constraints.md: New constraints and checking for hard
13083 float configuration.
13084 * config/nds32/iterators.md: New mode iterator and attribute for hard
13085 float configuration.
13086 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
13088 * config/nds32/nds32-fpu.md: New file.
13089 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
13090 deal with hard float code generation.
13091 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
13093 (abi_type, float_reg_number): New enum type.
13094 * config/nds32/nds32-predicates.c: New predicates for hard float.
13095 * config/nds32/nds32-protos.h: Declare functions for hard float.
13096 * config/nds32/nds32.c: Implementation for hard float configuration.
13097 * config/nds32/nds32.h: Definitions for hard float configuration.
13098 * config/nds32/nds32.md: Include hard float machine description and
13099 modify patterns for hard float configuration.
13100 * config/nds32/nds32.opt: New options for hard float configuration.
13101 * config/nds32/predicates.md: New predicates for hard float
13104 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
13106 * common/config/nds32/nds32-common.c
13107 (nds32_option_optimization_table): Enable -mreleax-hint by default.
13109 2018-04-05 Jakub Jelinek <jakub@redhat.com>
13111 PR middle-end/85195
13112 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
13113 CONSTRUCTOR_ELT (ctor, ...)->value.
13115 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
13118 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
13120 2018-04-05 Tom de Vries <tom@codesourcery.com>
13123 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
13126 2018-04-05 Shiva Chen <shiva0217@gmail.com>
13127 Kito Cheng <kito.cheng@gmail.com>
13129 * config/nds32/constraints.md (U33): Fine-tune checking condition.
13130 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
13131 * config/nds32/nds32.h (nds32_16bit_address_type): Add
13132 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
13134 2018-04-05 Shiva Chen <shiva0217@gmail.com>
13135 Kito Cheng <kito.cheng@gmail.com>
13137 * config/nds32/constraints.md (Ufe): New memory constraint.
13138 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
13139 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
13140 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
13142 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
13143 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
13145 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
13147 * config/nds32/nds32.md: Use optimize_size in the condition for
13148 alu-shift instructions.
13150 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
13152 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
13154 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
13156 * config/nds32/nds32.md (negsi2): Refine pattern.
13158 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
13159 Chung-Ju Wu <jasonwucj@gmail.com>
13161 * config/nds32/iterators.md (shift_rotate): New code iterator.
13162 (shift): New code attribute.
13163 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
13164 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
13165 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
13166 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
13167 bit-wise operations.
13168 (andsi3, *andsi3): Ditto.
13169 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
13170 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
13171 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
13172 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
13173 nds32_ior_operand, nds32_xor_operand): New predicates.
13175 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
13177 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
13178 (addsi3, subsi3): ... this.
13180 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
13182 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
13184 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
13186 * config/nds32/nds32.md: Adjust indention.
13188 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
13190 * config/nds32/nds32.md (feature): New attribute.
13192 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
13194 * config/nds32/nds32.md (subtype): New attribute.
13196 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
13199 * config/arm/arm-builtins.c (arm_expand_builtin): Change
13200 expansion to perform a bitwise AND of the argument followed by a
13201 boolean negation of the result.
13203 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
13205 PR rtl-optimization/84878
13206 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
13207 the basic block. Assert the use reference is not artificial and that
13208 it has an associated insn.
13210 2018-04-04 Michael Matz <matz@suse.de>
13212 * builtins.c (compute_objsize): Pass correct operand
13213 to array_at_struct_end_p.
13215 2018-04-04 Richard Biener <rguenther@suse.de>
13218 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
13219 from contexts for DINFO_LEVEL_TERSE and below.
13221 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
13223 * config/nds32/nds32-doubleword.md (move_<mode>): Require
13224 resiter_operand condition.
13225 * config/nds32/nds32.md (*move<mode>): Ditto.
13227 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
13228 Monk Chiang <sh.chiang04@gmail.com>
13230 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
13232 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
13234 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
13236 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
13237 Kito Cheng <kito.cheng@gmail.com>
13239 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
13240 nds32_cond_code_str, output_cond_branch,
13241 output_cond_branch_compare_zero, nds32_expand_cbranch,
13242 nds32_expand_cstore, nds32_expand_movcc,
13243 nds32_output_cbranchsi4_equality_zero,
13244 nds32_output_cbranchsi4_equality_reg,
13245 nds32_output_cbranchsi4_equality_reg_or_const_int,
13246 nds32_output_cbranchsi4_greater_less_zero: New functions.
13247 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
13248 nds32_expand_cstore, nds32_expand_movcc,
13249 nds32_output_cbranchsi4_equality_zero,
13250 nds32_output_cbranchsi4_equality_reg,
13251 nds32_output_cbranchsi4_equality_reg_or_const_int,
13252 nds32_output_cbranchsi4_greater_less_zero): Declare.
13253 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
13254 nds32_rimm11s_operand): New predicates.
13255 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
13256 * config/nds32/nds32.md: Rewrite all the branch and conditional move
13259 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
13261 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
13262 * config/nds32/nds32.md: Ditto.
13263 * config/nds32/pipelines.md: Ditto.
13265 2018-04-04 Richard Biener <rguenther@suse.de>
13267 PR tree-optimization/85168
13268 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
13269 propagating abnormals.
13271 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
13273 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
13275 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
13276 Kito Cheng <kito.cheng@gmail.com>
13278 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
13279 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
13280 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
13281 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
13282 * config/nds32/nds32.md (sibcall_internal): New.
13283 (sibcall_register): Remove.
13284 (sibcall_immediate): Remove.
13285 (sibcall_value_internal): New.
13286 (sibcall_value_register): Remove.
13287 (sibcall_value_immediate): Remove.
13288 * config/nds32/predicates.md (nds32_general_register_operand): New.
13289 (nds32_call_address_operand): New.
13291 2018-04-03 Jakub Jelinek <jakub@redhat.com>
13293 PR rtl-optimization/85167
13294 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
13295 bb_defs if *split_p, instead preinitialize it to NULL.
13297 PR tree-optimization/85156
13298 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
13299 evaluating the argument multiple times.
13301 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
13303 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
13305 (_mm_cvtpd_ps): Likewise.
13306 (_mm_cvttpd_epi32): Likewise.
13307 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
13308 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
13309 vector, pixel, and bool following altivec.h include.
13311 2018-04-03 Martin Sebor <msebor@redhat.com>
13313 * doc/extend.texi (Common Function Attributes): Clarify.
13314 (const attribute): Likewise.
13315 (pure attribute): Likewise.
13317 2018-04-03 Jakub Jelinek <jakub@redhat.com>
13320 * config/i386/i386.c (ix86_expand_vector_set): Use
13321 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
13323 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
13325 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
13326 instructions when changing rounding bits to preserve precision bits
13327 in the x87 control word.
13329 2018-04-03 Martin Liska <mliska@suse.cz>
13331 PR tree-optimization/82491
13332 * rtl.h (strip_offset_and_add): Replace += suboffset with
13333 poly_uint64 () + suboffset.
13335 2018-03-29 Martin Liska <mliska@suse.cz>
13336 Martin Jambor <mjambor@suse.cz>
13339 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
13340 param_type is not an integral or pointer type.
13342 2018-04-03 Richard Biener <rguenther@suse.de>
13344 * sese.h (recompute_all_dominators): Remove.
13346 2018-04-02 Martin Sebor <msebor@redhat.com>
13348 * doc/invoke.texi (-Wrestrict): Fix typos.
13350 2018-04-02 Jim Wilson <jimw@sifive.com>
13352 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
13353 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
13354 (<optab>di3, <optab>si3_extend): Likewise.
13355 (<optab>si3_mask, <optab>si3_mask_1): New.
13356 (<optab>di3_mask, <optab>di3_mask_1): New.
13357 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
13358 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
13359 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
13361 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
13363 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
13366 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
13368 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
13369 (nds32_canonicalize_comparison): New function.
13371 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
13372 Kito Cheng <kito.cheng@gmail.com>
13373 Kuan-Lin Chen <kuanlinchentw@gmail.com>
13375 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
13376 * config/nds32/constants.md (unspec_volatile_element): Add
13377 UNSPEC_VOLATILE_RELAX_GROUP.
13378 * config/nds32/nds32-relax-opt.c: New file.
13379 * config/nds32/nds32-predicates.c
13380 (nds32_symbol_load_store_p): New function.
13381 * config/nds32/nds32-protos.h
13382 (nds32_symbol_load_store_p): Declare function.
13383 (make_pass_nds32_relax_opt): Declare new rtl pass function.
13384 * config/nds32/nds32.c
13385 (nds32_register_pass): New function to register pass.
13386 (nds32_register_passes): New function to register passes.
13387 * config/nds32/nds32.md (relax_group): New pattern.
13388 * config/nds32/nds32.opt (mrelax-hint): New option.
13389 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
13391 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
13393 * config/nds32/t-nds32: Modify files dependency.
13395 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
13397 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
13398 (PROFILE_HOOK): Define its implementation.
13400 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
13402 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
13403 type and 32-bit size.
13405 2018-04-01 Jakub Jelinek <jakub@redhat.com>
13407 PR middle-end/85090
13408 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
13409 (V_128_256): New mode iterator.
13410 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
13411 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
13412 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
13414 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
13415 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
13417 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
13420 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
13421 NaN inputs correctly.
13423 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
13426 * config/rs6000/vsx.md (??r): New mode attribute.
13427 (*vsx_mov<mode>_64bit): Use it.
13428 (*vsx_mov<mode>_32bit): Likewise.
13430 2018-03-30 Martin Sebor <msebor@redhat.com>
13432 PR tree-optimization/84818
13433 * builtins.c (check_access): Use warning_n.
13435 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13438 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
13440 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
13443 2018-03-30 Julia Koval <julia.koval@intel.com>
13446 * x86-tune.def (movx, partial_reg_dependency): Enable for
13449 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
13451 PR inline-asm/84985
13452 * lra-constraints.c (process_alt_operands): Move setting
13453 this_alternative_matches below.
13455 2018-03-29 Martin Liska <mliska@suse.cz>
13458 * doc/invoke.texi: Document how LTO works with debug info.
13459 Describe auto-load support of binutils. Mention 'x86-64'
13460 as valid option value of -march option.
13462 2018-03-29 Jakub Jelinek <jakub@redhat.com>
13464 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
13467 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
13468 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
13469 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
13472 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
13475 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
13476 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
13477 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
13478 for RS6000_BTM_POWERPC64.
13479 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
13480 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
13481 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
13484 (DIVDEU): Likewise.
13486 2018-03-28 Carl Love <cel@us.ibm.com>
13489 2017-09-27 Carl Love <cel@us.ibm.com>
13491 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
13492 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
13493 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
13496 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13498 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
13499 instead of __vector bool.
13500 (_mm_max_pu8): Likewise.
13501 (_mm_min_pi16): Likewise.
13503 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
13506 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
13507 (DIVWEUO): Likewise.
13508 (DIVDEO): Likewise.
13509 (DIVDEUO): Likewise.
13510 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
13511 DIVWEUO and DIVDEUO.
13512 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
13513 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
13514 (div_extend): Likewise.
13515 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
13517 (__builtin_divweuo): Likewise.
13518 (__builtin_divdeo): Likewise.
13519 (__builtin_divdeuo): Likewise.
13521 2018-03-28 Jakub Jelinek <jakub@redhat.com>
13524 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
13525 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
13527 PR tree-optimization/82004
13528 * gimple-match-head.c (optimize_pow_to_exp): New function.
13529 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
13530 Don't fold to exp if optimize_pow_to_exp is false.
13532 2018-03-28 Martin Liska <mliska@suse.cz>
13535 * calls.c (initialize_argument_information): Fix trailing space.
13536 * common.opt: Fix typo and provide better explanation for
13537 -fsanitize-coverage option.
13538 * config/i386/i386.opt: Fix typo.
13540 2018-03-28 Jakub Jelinek <jakub@redhat.com>
13541 Martin Liska <mliska@suse.cz>
13544 * gimplify.c (asan_poison_variable): Don't do the check for
13545 gimplify_omp_ctxp here.
13546 (gimplify_decl_expr): Do it here.
13547 (gimplify_target_expr): Likewise.
13549 2018-03-28 Martin Liska <mliska@suse.cz>
13552 * config/i386/i386.c (ix86_function_arg_advance): Do not call
13553 chkp_type_bounds_count if MPX is not enabled.
13555 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
13557 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
13559 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
13562 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
13563 function to create the function decl for complex long double
13564 multiply and divide for -mabi=ieeelongdouble.
13565 (init_float128_ieee): Call it.
13567 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
13570 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
13571 -fcf-protection=branch -mibt.
13572 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
13574 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
13577 * config/arm/arm.c (arm_valid_symbolic_address): Handle
13578 arm_word_relocations.
13580 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
13583 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
13584 extern array declarations.
13586 2018-03-27 Richard Biener <rguenther@suse.de>
13588 PR middle-end/84067
13589 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
13590 explicit single_use checks.
13592 2018-03-27 Richard Biener <rguenther@suse.de>
13594 PR tree-optimization/85082
13595 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
13598 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
13600 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
13601 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
13602 Turn on fasynchronous-unwind-tables and funwind-tables.
13604 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
13607 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
13608 (*bmi_blsr_<mode>_ccz): Ditto.
13610 2018-03-26 Tom de Vries <tom@codesourcery.com>
13612 PR tree-optimization/85063
13613 * omp-general.c (offloading_function_p): New function. Factor out
13615 * omp-offload.c (pass_omp_target_link::gate): ... here.
13616 * omp-general.h (offloading_function_p): Declare.
13617 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
13618 with attribute omp declare target for offloading functions.
13620 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
13622 PR tree-optimization/84005
13623 * tree-data-ref.h (get_base_for_alignment): Declare.
13624 * tree-data-ref.c (get_base_for_alignment_1): New function.
13625 (get_base_for_alignment): Likewise.
13626 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
13627 get_base_for_alignment to find a suitable base object, instead
13628 of always using drb->base_address.
13630 2018-03-23 Jakub Jelinek <jakub@redhat.com>
13632 PR inline-asm/85022
13633 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
13634 known size by default.
13636 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
13638 PR inline-asm/85030
13639 * lra-constraints.c (process_alt_operands): Don't match BLKmode
13640 and non BLKmode operands.
13642 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13645 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
13646 Clean up attributes.
13648 2018-03-23 Richard Biener <rguenther@suse.de>
13651 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
13652 we are going to emit early debug for LTO.
13654 2018-03-23 Jakub Jelinek <jakub@redhat.com>
13656 PR inline-asm/85034
13657 * function.c (match_asm_constraints_1): Don't optimize if input
13658 doesn't satisfy general_operand predicate for output's mode.
13660 PR inline-asm/85022
13661 * alias.c (write_dependence_p): Don't require for x_canonicalized
13662 non-VOIDmode if x has VOIDmode.
13665 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
13666 just don't try to optimize it rather than assert it never happens.
13668 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
13670 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
13671 macro expansions for definition of ST_INTERNAL_<mode> and
13672 LD_INTERNAL_<mode> builtins.
13673 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
13675 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
13677 (altivec_expand_st_builtin): Likewise.
13678 (altivec_expand_builtin): Remove calls to deleted functions.
13679 (rs6000_address_for_altivec): Delete this function.
13680 * config/rs6000/vector.md: Remove expands for
13681 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
13683 2018-03-22 Sudakshina Das <sudi.das@arm.com>
13686 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
13687 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
13688 re-computing once computed.
13689 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
13690 (arm_init_machine_status): Initialize
13691 machine->static_chain_stack_bytes.
13693 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
13696 * doc/extend.texi: Add four new prototypes for vec_ld.
13697 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
13698 definitions for more logical presentation.
13699 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
13700 entries for V1TI variants of __builtin_altivec_ld builtin.
13701 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
13702 handling of V1TI variant of LVX icode pattern.
13703 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
13704 (rs6000_gimple_fold_builtin): Likewise.
13705 (altivec_init_builtins): Add code to define
13706 __builtin_altivec_lvx_v1ti function.
13708 2018-03-22 Jakub Jelinek <jakub@redhat.com>
13710 PR inline-asm/84941
13711 * function.c (match_asm_constraints_1): Don't do the optimization
13712 if input isn't a REG, SUBREG, MEM or constant.
13714 2018-03-22 Tom de Vries <tom@codesourcery.com>
13716 PR tree-optimization/84956
13717 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
13718 bb_has_abnormal_pred.
13720 2018-03-22 Jakub Jelinek <jakub@redhat.com>
13723 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
13724 DECL_INITIAL (decl) to decl at the end.
13725 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
13726 adjust the comment.
13728 2018-03-21 Joseph Myers <joseph@codesourcery.com>
13730 * doc/extend.texi (__builtin_tgmath): Document when complex
13731 integer types are treated as _Complex _Float64.
13733 2018-03-21 Tom de Vries <tom@codesourcery.com>
13735 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
13737 2018-03-21 Jakub Jelinek <jakub@redhat.com>
13739 PR tree-optimization/84960
13740 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
13741 if it is ENTRY block, move them into single succ of ENTRY in that case.
13743 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
13745 PR tree-optimization/84811
13746 * poly-int.h (poly_span_traits): Remove the T3 parameter and
13747 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
13748 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
13749 (known_subrange_p): Update accordingly. Cast each value involved
13750 in the size comparison, rather than casting the result of the
13753 2018-03-21 Jakub Jelinek <jakub@redhat.com>
13755 PR tree-optimization/84982
13756 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
13757 by flipping the least significant bit rather than all bits from
13758 bitpos to bitpos + bitsize - 1.
13760 2018-03-21 Nathan Sidwell <nathan@acm.org>
13762 * doc/extend.texi (Deprecated Features): Remove mention of
13763 long-deleted deprecations.
13765 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13768 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
13769 * configure: Regenerate.
13771 2018-03-21 Tom de Vries <tom@codesourcery.com>
13773 PR tree-optimization/83126
13774 * tree-parloops.c (num_phis): New function.
13775 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
13777 2018-03-21 Nathan Sidwell <nathan@acm.org>
13779 * doc/extend.texi (Deprecated Features): Update deprecated flags,
13780 mention anon-struct/union members and trailing attributes.
13782 2018-03-21 Bin Cheng <bin.cheng@arm.com>
13784 PR tree-optimization/84969
13785 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
13786 builtin memset partitions if they set different rhs values.
13788 2018-03-21 Jakub Jelinek <jakub@redhat.com>
13790 PR rtl-optimization/84989
13791 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
13792 VEC_DUPLICATE with scalar result mode.
13794 2018-03-21 Martin Liska <mliska@suse.cz>
13797 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
13798 not intended return statement.
13800 2018-03-21 Martin Liska <mliska@suse.cz>
13803 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
13804 (chkp_find_bound_slots_1): Limit number of iterations.
13806 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
13809 * Minor grammar fixes for x86 options.
13811 2018-03-20 Jakub Jelinek <jakub@redhat.com>
13814 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
13815 holding REG_CFA_RESTORE notes, instead turn them into a USE.
13817 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
13820 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
13821 (altivec_lvx_<mode>_1op): Likewise.
13822 (altivec_stvx_<mode>_2op): Likewise.
13823 (altivec_stvx_<mode>_1op): Likewise.
13824 (altivec_lvx_<VM2:mode>): New define_expand.
13825 (altivec_stvx_<VM2:mode>): Likewise.
13826 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
13827 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
13828 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
13829 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
13830 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
13831 (rs6000_gen_lvx): Likewise.
13832 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
13833 (altivec_expand_stv_builtin): Likewise.
13834 (altivec_expand_builtin): Likewise.
13835 * config/rs6000/vector.md: Likewise.
13837 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13840 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
13843 2018-03-20 Richard Biener <rguenther@suse.de>
13846 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
13847 sign-conversions as zero, fall back to standard scalar_stmt
13850 2018-03-20 Martin Liska <mliska@suse.cz>
13853 * predict.c (rebuild_frequencies): Handle case when we have
13854 PROFILE_ABSENT, but flag_guess_branch_prob is false.
13856 2018-03-20 Jakub Jelinek <jakub@redhat.com>
13859 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
13860 flag_section_anchors.
13861 * varasm.c (use_blocks_for_decl_p): Remove hack for
13862 dw2_force_const_mem.
13865 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
13867 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
13868 be created, use lowpart_subreg of operands[0] rather than operands[0]
13870 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
13871 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
13872 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
13873 and n constraint instead of aarch64_shift_imm_di and Usd.
13874 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
13875 (*aarch64_<optab>_reg_minus<mode>3): ... this.
13877 2018-03-20 Sudakshina Das <sudi.das@arm.com>
13880 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
13881 to favor GPR over NEON registers.
13882 (<shift>di3_neon): Likewise.
13884 2018-03-20 Tom de Vries <tom@codesourcery.com>
13887 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
13888 (nvptx_process_pars): Emit bar.sync asap and alap.
13890 2018-03-20 Tom de Vries <tom@codesourcery.com>
13893 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
13894 seen_label if seen_label is already set.
13896 2018-03-20 Jakub Jelinek <jakub@redhat.com>
13899 * config/i386/i386.c (fold_builtin_cpu): For features above 31
13900 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
13901 Use 1U instead of 1. Formatting fixes.
13904 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
13905 instead of TREE_TYPE (s1) for the return value.
13907 2018-03-19 Jakub Jelinek <jakub@redhat.com>
13909 PR tree-optimization/84946
13910 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
13911 bitsize + bitsize in poly_uint64 rather than poly_int64.
13914 * dwarf2asm.c: Include fold-const.c.
13915 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
13916 of decl rather than decl itself.
13918 PR rtl-optimization/84643
13919 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
13921 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
13924 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
13925 calling assemble_variable.
13927 2018-03-19 Sudakshina Das <sudi.das@arm.com>
13930 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
13931 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
13933 2018-03-19 Jim Wilson <jimw@sifive.com>
13936 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
13937 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
13938 (riscv_first_stack_step): Likewise.
13939 (riscv_option_override): Use STACK_BOUNDARY instead of
13940 MIN_STACK_BOUNDARY.
13941 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
13942 MIN_STACK_BOUNDARY.
13943 (BIGGEST_ALIGNMENT): Set to 128.
13944 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
13945 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
13948 2018-03-19 Richard Biener <rguenther@suse.de>
13950 PR tree-optimization/84933
13951 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
13952 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
13954 2018-03-19 Richard Biener <rguenther@suse.de>
13956 PR tree-optimization/84859
13957 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
13958 (cond_if_else_store_replacement): Perform sinking operation on
13959 single-store BBs regardless of MAX_STORES_TO_SINK setting.
13960 Generalize what a BB with a single eligible store is.
13962 2018-03-19 Richard Biener <rguenther@suse.de>
13964 PR tree-optimization/84929
13965 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
13966 chrec_is_positive against non-chrec arg.
13968 2018-03-19 Tamar Christina <tamar.christina@arm.com>
13971 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
13973 2018-03-18 Martin Liska <mliska@suse.cz>
13975 PR rtl-optimization/84635
13976 * regrename.c (build_def_use): Use matches_mode only when
13979 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
13981 PR tree-optimization/84913
13982 * tree-vect-loop.c (vectorizable_reduction): Don't try to
13983 vectorize chains of COND_EXPRs.
13985 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
13987 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
13989 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
13991 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
13993 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
13995 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
13997 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
13998 Kito Cheng <kito.cheng@gmail.com>
14000 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
14001 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
14002 (nds32_adjust_reg_alloc_order): New function.
14003 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
14005 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
14007 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
14008 nds32_print_operand, nds32_print_operand_address): Use
14009 HOST_WIDE_INT_PRINT_DEC instead.
14011 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
14013 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
14015 2018-03-17 Jakub Jelinek <jakub@redhat.com>
14018 * config/i386/i386.c (initial_ix86_tune_features,
14019 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
14020 unsigned long long.
14021 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
14022 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
14023 rather than 1u << ix86_tune. Formatting fix.
14024 (ix86_option_override_internal): Change ix86_arch_mask from
14025 unsigned int to unsigned HOST_WIDE_INT, initialize to
14026 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
14027 (ix86_function_specific_restore): Likewise.
14029 2018-03-16 Jakub Jelinek <jakub@redhat.com>
14032 * postreload.c (reload_combine_recognize_pattern): Perform
14033 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
14034 truncate_int_for_mode the result for the destination's mode.
14037 * hsa-gen.c (mem_type_for_type): Fix comment typo.
14038 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
14040 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
14043 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
14046 * lra-assigns.c (lra_split_hard_reg_for): Don't use
14047 regno_allocno_class_array and sorted_pseudos.
14048 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
14049 insns where regno is used.
14051 2018-03-16 Martin Liska <mliska@suse.cz>
14054 * multiple_target.c (create_dispatcher_calls): Redirect
14055 reference in the symbol table.
14057 2018-03-16 Martin Liska <mliska@suse.cz>
14060 * multiple_target.c (create_dispatcher_calls): Redirect also
14063 2018-03-16 Jakub Jelinek <jakub@redhat.com>
14067 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
14068 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
14069 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
14071 2018-03-16 Julia Koval <julia.koval@intel.com>
14073 * doc/invoke.texi (Skylake Server): Add CLWB.
14074 Cannonlake): Remove CLWB.
14076 2018-03-16 Jakub Jelinek <jakub@redhat.com>
14078 PR tree-optimization/84841
14079 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
14081 (FLOAT_ONE_CONST_TYPE): Define.
14082 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
14083 (sort_by_operand_rank): Put entries with higher constant_type last
14084 rather than first to match comments.
14086 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
14088 * config/nios2/nios2.md (movsi_internal): Fix thinko in
14091 2018-03-15 Jakub Jelinek <jakub@redhat.com>
14094 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
14095 check and use address of target always.
14097 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
14100 * config/i386/i386.c (indirect_thunk_needed): Update comments.
14101 (indirect_thunk_bnd_needed): Likewise.
14102 (indirect_thunks_used): Likewise.
14103 (indirect_thunks_bnd_used): Likewise.
14104 (indirect_return_needed): New.
14105 (indirect_return_bnd_needed): Likewise.
14106 (output_indirect_thunk_function): Add a bool argument for
14108 (output_indirect_thunk_function): Don't generate alias for
14109 function return thunk.
14110 (ix86_code_end): Call output_indirect_thunk_function to generate
14111 function return thunks.
14112 (ix86_output_function_return): Set indirect_return_bnd_needed
14113 and indirect_return_needed instead of indirect_thunk_bnd_needed
14114 and indirect_thunk_needed.
14116 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
14118 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
14119 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
14120 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
14122 2018-03-15 David Malcolm <dmalcolm@redhat.com>
14123 Paul Hua <paul.hua.gm@gmail.com>
14126 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
14128 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
14130 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
14131 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
14132 resp. SFmode cases.
14134 2018-03-15 Tamar Christina <tamar.christina@arm.com>
14137 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
14138 instead of GET_MODE_SIZE when comparing Units.
14140 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
14143 * varasm.c (hash_section): Return an unchangeble hash value
14144 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
14145 Return !aarch64_can_use_per_function_literal_pools_p ().
14147 2018-03-15 Jakub Jelinek <jakub@redhat.com>
14150 * optabs.c (emit_conditional_move): Pass address of cmode's copy
14151 rather than address of cmode as last argument to prepare_cmp_insn.
14153 2018-03-15 Julia Koval <julia.koval@intel.com>
14155 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
14156 F_AVX512VNNI, F_AVX512BITALG): New.
14158 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
14161 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
14162 insn for floating-point loads and stores.
14164 2018-03-14 Carl Love <cel@us.ibm.com>
14166 * config/rs6000/rs6000-c.c: Add macro definitions for
14167 ALTIVEC_BUILTIN_VEC_PERMXOR.
14168 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
14169 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
14170 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
14171 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
14173 * config/doc/extend.texi: Add prototypes for vec_permxor.
14175 2018-03-14 David Malcolm <dmalcolm@redhat.com>
14178 * diagnostic-show-locus.c (class layout_point): Convert m_line
14179 from int to linenum_type.
14180 (line_span::comparator): Use linenum "compare" function when
14181 comparing line numbers.
14182 (test_line_span): New function.
14183 (layout_range::contains_point): Convert param "row" from int to
14185 (layout_range::intersects_line_p): Likewise.
14186 (layout::will_show_line_p): Likewise.
14187 (layout::print_source_line): Likewise.
14188 (layout::should_print_annotation_line_p): Likewise.
14189 (layout::print_annotation_line): Likewise.
14190 (layout::print_leading_fixits): Likewise.
14191 (layout::annotation_line_showed_range_p): Likewise.
14192 (struct line_corrections): Likewise for field m_row.
14193 (line_corrections::line_corrections): Likewise for param "row".
14194 (layout::print_trailing_fixits): Likewise.
14195 (layout::get_state_at_point): Likewise.
14196 (layout::get_x_bound_for_row): Likewise.
14197 (layout::print_line): Likewise.
14198 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
14199 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
14200 * input.c (selftest::test_linenum_comparisons): New function.
14201 (selftest::input_c_tests): Call it.
14202 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
14203 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
14204 * selftest.h (ASSERT_GT): New macro.
14205 (ASSERT_GT_AT): New macro.
14206 (ASSERT_LT): New macro.
14207 (ASSERT_LT_AT): New macro.
14209 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
14211 PR rtl-optimization/84780
14212 * combine.c (distribute_links): Don't make a link based on pc_rtx.
14214 2018-03-14 Martin Liska <mliska@suse.cz>
14216 * tree.c (record_node_allocation_statistics): Use
14217 get_stats_node_kind.
14218 (get_stats_node_kind): New function extracted from
14219 record_node_allocation_statistics.
14220 (free_node): Use get_stats_node_kind.
14222 2018-03-14 Richard Biener <rguenther@suse.de>
14224 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
14225 that the value-set of ANTIC_IN doesn't grow.
14228 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
14230 (BB_VISITED_WITH_VISITED_SUCCS): New define.
14231 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
14233 2018-03-14 Julia Koval <julia.koval@intel.com>
14235 * config.gcc (icelake-client, icelake-server): New.
14237 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
14238 (initial_ix86_arch_features): Ditto.
14239 (PTA_SKYLAKE): Add SGX.
14240 (PTA_ICELAKE): Remove.
14241 (PTA_ICELAKE_CLIENT): New.
14242 (PTA_ICELAKE_SERVER): New.
14243 (ix86_option_override_internal): Split up icelake on icelake client and
14245 (get_builtin_code_for_version): Ditto.
14246 (fold_builtin_cpu): Ditto.
14247 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
14248 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
14249 * config/i386/i386.h (processor_type): Ditto.
14250 * doc/invoke.texi: Ditto.
14252 2018-03-14 Jakub Jelinek <jakub@redhat.com>
14255 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
14256 INTEGER_CST offset, add it together with bitpos / 8 and
14257 sign extend based on POINTER_SIZE.
14261 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
14264 * config/i386/constraints.md (Yc): New register constraint.
14265 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
14266 Use Yc constraint for alternative 2 of operand 0. Remove
14267 preferred_for_speed attribute.
14269 2018-03-14 Richard Biener <rguenther@suse.de>
14271 PR tree-optimization/84830
14272 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
14273 with the old one to avoid oscillations.
14275 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
14278 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
14280 (assign_by_spills): Return a flag of reload assignment failure.
14281 Do not process the reload assignment failures. Do not spill other
14282 reload pseudos if they has the same reg class. Update n if
14284 (lra_assign): Add a return arg. Set up from the result of
14285 assign_by_spills call.
14286 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
14287 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
14288 usage_insns if it is not NULL.
14289 (spill_hard_reg_in_range): New function.
14290 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
14291 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
14292 function prototypes.
14293 (lra_assign): Change prototype.
14294 * lra.c (lra): Add code to deal with fails by splitting hard reg
14297 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
14299 * config/riscv/riscv.opt (mrelax): New option.
14300 * config/riscv/riscv.c (riscv_file_start): Emit ".option
14301 "norelax" when riscv_mrelax is disabled.
14302 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
14304 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14307 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
14308 reassociation for int modes.
14310 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
14312 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
14313 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
14315 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
14316 * config/aarch64/aarch64-sve.md
14317 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
14318 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
14319 (*extend<mode><Vwide>2): Rename to...
14320 (aarch64_sve_extend<mode><Vwide>2): ...this.
14321 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
14322 renaming the old pattern to...
14323 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
14325 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
14326 define_expand, renaming the old pattern to...
14327 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
14328 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
14329 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
14330 account when deciding which SVE instruction the optab should use.
14331 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
14333 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
14335 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
14336 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
14337 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
14338 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
14339 (tlsdesc_small_<mode>): Turn a define_expand and use
14340 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
14341 (tlsdesc_small_advsimd_<mode>): ...this.
14342 (tlsdesc_small_sve_<mode>): New pattern.
14344 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
14346 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
14347 (UNSPEC_UMUL_HIGHPART): New constants.
14348 (MUL_HIGHPART): New int iteraor.
14349 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
14350 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
14352 (*<su>mul<mode>3_highpart): New define_insn.
14354 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
14357 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
14360 2018-03-13 Martin Liska <mliska@suse.cz>
14363 * (sem_item_optimizer::sem_item_optimizer): Initialize new
14365 (sem_item_optimizer::~sem_item_optimizer): Release it.
14366 (sem_item_optimizer::merge_classes): Register variable aliases.
14367 (sem_item_optimizer::fixup_pt_set): New function.
14368 (sem_item_optimizer::fixup_points_to_sets): Likewise.
14369 * ipa-icf.h: Declare new variables and functions.
14371 2018-03-13 Jakub Jelinek <jakub@redhat.com>
14373 PR middle-end/84834
14374 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
14375 integer_pow2p@2 and test integer_pow2p in condition.
14376 (A < 0 ? C : 0): Similarly for @1.
14378 PR middle-end/84831
14379 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
14380 characters starting at p contain '\0' character, don't look beyond
14384 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
14385 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
14388 * reg-stack.c (change_stack): Change update_end var from int to
14389 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
14390 also call set_block_for_insn on the newly added insns and rescan.
14393 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
14394 on the last operand.
14397 * tree.c (stabilize_reference_1): Return save_expr (e) for
14398 STATEMENT_LIST even if it doesn't have side-effects.
14400 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
14402 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
14404 2018-03-12 Renlin Li <renlin.li@arm.com>
14406 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
14407 aarch64_output_scalar_simd_mov_immediate.
14409 2018-03-12 Martin Sebor <msebor@redhat.com>
14411 PR tree-optimization/83456
14412 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
14413 for perfectly overlapping calls to memcpy.
14414 (gimple_fold_builtin_memory_chk): Same.
14415 (gimple_fold_builtin_strcpy): Handle no-warning.
14416 (gimple_fold_builtin_stxcpy_chk): Same.
14417 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
14419 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
14421 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
14422 parameter. Use it for SFmode.
14423 (rs6000_function_arg_advance_1): Adjust.
14424 (rs6000_function_arg): Adjust.
14425 (rs6000_gimplify_va_arg): Pass false for that new parameter.
14427 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
14429 PR rtl-optimization/84169
14430 PR rtl-optimization/84780
14431 * combine.c (can_combine_p): Check for a 2-insn combination whether
14432 the destination register is used between the two insns, too.
14434 2018-03-12 Richard Biener <rguenther@suse.de>
14436 PR tree-optimization/84803
14437 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
14438 for refs DR analysis didn't process.
14440 2018-03-12 Richard Biener <rguenther@suse.de>
14442 PR tree-optimization/84777
14443 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
14444 force-vectorize loops ignore whether we are optimizing for size.
14446 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
14448 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
14449 (TARGET_MD_ASM_ADJUST): Define.
14451 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
14452 Kito Cheng <kito.cheng@gmail.com>
14453 Chung-Ju Wu <jasonwucj@gmail.com>
14455 * config/nds32/nds32.c (nds32_compute_stack_frame,
14456 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
14457 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
14458 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
14459 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
14460 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
14461 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
14462 * config/nds32/nds32.md (prologue, epilogue): Use macro
14463 NDS32_V3PUSH_AVAILABLE_P to do checking.
14465 2018-03-11 Jakub Jelinek <jakub@redhat.com>
14468 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
14469 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
14470 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
14471 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
14472 addition of most attributes on !orig_type_die or the attribute not
14473 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
14475 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
14476 Chung-Ju Wu <jasonwucj@gmail.com>
14478 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
14479 __NDS32_VH__ macro.
14480 * config/nds32/nds32.opt (mvh): New option.
14482 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
14483 Chung-Ju Wu <jasonwucj@gmail.com>
14485 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
14487 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
14488 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
14491 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
14492 Chung-Ju Wu <jasonwucj@gmail.com>
14494 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
14496 * config/nds32/nds32-multiple.md (strlensi): New pattern.
14497 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
14499 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
14500 Kito Cheng <kito.cheng@gmail.com>
14501 Chung-Ju Wu <jasonwucj@gmail.com>
14503 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
14504 UNSPEC_FFMISM and UNSPEC_FLMISM.
14505 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
14506 for ffb, ffmism and flmism.
14507 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
14508 (unspec_ffmism): Ditto.
14509 (unspec_flmism): Ditto.
14510 (nds32_expand_builtin_impl): Check if string extension is available.
14511 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
14512 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
14514 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
14517 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
14520 * lra-assigns.c (assign_by_spills): Return a flag of reload
14521 assignment failure. Do not process the reload assignment
14522 failures. Do not spill other reload pseudos if they has the same
14524 (lra_assign): Add a return arg. Set up from the result of
14525 assign_by_spills call.
14526 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
14527 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
14528 usage_insns if it is not NULL.
14529 (spill_hard_reg_in_range): New function.
14530 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
14531 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
14532 function prototypes.
14533 (lra_assign): Change prototype.
14534 * lra.c (lra): Add code to deal with fails by splitting hard reg
14537 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
14540 * config/i386/i386.opt: Replace Enforcment with Enforcement.
14542 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
14545 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
14546 (dw_val_node): Add val_symbolic_view.
14547 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
14548 (symview_upper_bound): New.
14549 (new_line_info_table): Initialize symviews_since_reset.
14550 (dwarf2out_source_line): Count symviews_since_reset and set
14551 symview_upper_bound.
14552 (dw_val_equal_p): Handle symview.
14553 (add_AT_symview): New.
14554 (print_dw_val): Handle symview.
14555 (attr_checksum, attr_checksum_ordered): Likewise.
14556 (same_dw_val_p, size_of_die): Likewise.
14557 (value_format, output_die): Likewise.
14558 (add_high_low_attributes): Use add_AT_symview for entry_view.
14559 (dwarf2out_finish): Reset symview_upper_bound, clear
14562 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
14565 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
14566 Add strict argument and use it.
14567 (rs6000_split_multireg_move): Update for new strict argument.
14568 (mem_operand_gpr): Disallow all non-offsettable addresses.
14569 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
14571 2018-03-09 Jakub Jelinek <jakub@redhat.com>
14574 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
14575 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
14576 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
14579 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
14580 decl, use remap_type if we want to use the type.
14582 2018-03-09 Martin Sebor <msebor@redhat.com>
14584 PR tree-optimization/84526
14585 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
14587 (builtin_access::generic_overlap): Be prepared to handle non-array
14590 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
14592 PR rtl-optimization/84682
14593 * lra-constraints.c (process_address_1): Check is_address flag
14594 for address constraints.
14595 (process_alt_operands): Likewise.
14596 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
14597 preprocess_constraints.
14598 * recog.h (preprocess_constraints): Add oploc parameter.
14600 * recog.c (preprocess_constraints): Test address_operand for
14601 CT_ADDRESS constraints.
14603 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
14606 * lra-assigns.c (assign_by_spills): Return a flag of reload
14607 assignment failure. Do not process the reload assignment
14608 failures. Do not spill other reload pseudos if they has the same
14610 (lra_assign): Add a return arg. Set up from the result of
14611 assign_by_spills call.
14612 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
14613 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
14614 usage_insns if it is not NULL.
14615 (spill_hard_reg_in_range): New function.
14616 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
14617 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
14618 function prototypes.
14619 (lra_assign): Change prototype.
14620 * lra.c (lra): Add code to deal with fails by splitting hard reg
14623 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14626 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
14627 Accept complain bool parameter. Only emit errors if it is true.
14628 (arm_parse_cpu_option_name): Likewise.
14629 (arm_target_thumb_only): Adjust callers of the above.
14630 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
14631 prototype to take a default true bool parameter.
14632 (arm_parse_arch_option_name): Likewise.
14634 2018-03-09 David Malcolm <dmalcolm@redhat.com>
14635 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
14639 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
14640 * configure: Regenerate.
14641 * configure.ac ("linker --version-script option"): New.
14642 ("linker soname option"): New.
14644 2018-03-09 Richard Biener <rguenther@suse.de>
14646 PR tree-optimization/84775
14647 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
14648 immediate uses of predicate stmts and mark them modified.
14651 PR tree-optimization/84178
14652 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
14654 (version_loop_for_if_conversion): Delay update_ssa call.
14655 (tree_if_conversion): Delay update_ssa until after predicate
14658 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
14661 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
14662 when the function accesses prior frames.
14664 2018-03-08 Jakub Jelinek <jakub@redhat.com>
14667 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
14668 gen_llsym, otherwise call maybe_gen_llsym.
14670 PR inline-asm/84742
14671 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
14672 has ',' character inside of it.
14674 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14677 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
14678 as clobbering CC_REGNUM.
14680 2018-03-08 Richard Biener <rguenther@suse.de>
14682 PR middle-end/84552
14683 * tree-scalar-evolution.c: Include tree-into-ssa.h.
14684 (follow_copies_to_constant): Do not follow SSA names registered
14687 2018-03-08 Richard Biener <rguenther@suse.de>
14689 PR tree-optimization/84178
14690 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
14692 (version_loop_for_if_conversion): Delay update_ssa call.
14693 (tree_if_conversion): Delay update_ssa until after predicate
14696 2018-03-08 David Malcolm <dmalcolm@redhat.com>
14698 PR tree-optimization/84178
14699 * tree-if-conv.c (release_bb_predicate): Remove the
14700 the assertion that the stmts have NULL use_ops.
14701 Discard the statements, asserting that they haven't
14702 yet been added to a BB.
14704 2018-03-08 Richard Biener <rguenther@suse.de>
14706 PR tree-optimization/84746
14707 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
14708 (phi_translate): Pass in destination ANTIC_OUT set.
14709 (phi_translate_1): Likewise. For a simplified result lookup
14710 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
14711 (phi_translate_set): Adjust.
14712 (do_pre_regular_insertion): Likewise.
14713 (do_pre_partial_partial_insertion): Likewise.
14715 2018-03-08 Martin Liska <mliska@suse.cz>
14717 PR gcov-profile/84735
14718 * doc/gcov.texi: Document usage of profile files.
14719 * gcov-io.h: Document changes in the format.
14721 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
14725 * dwarf2out.c (struct dw_line_info_table): Update comments for
14727 (FORCE_RESET_NEXT_VIEW): New.
14728 (FORCE_RESETTING_VIEW_P): New.
14729 (RESETTING_VIEW_P): Check for -1 too.
14730 (ZERO_VIEW_P): Likewise.
14731 (new_line_info_table): Force-reset next view.
14732 (dwarf2out_begin_function): Likewise.
14733 (dwarf2out_source_line): Simplify zero_view_p initialization.
14734 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
14735 view directly. Omit view when omitting .loc at line 0.
14737 2018-03-08 Jakub Jelinek <jakub@redhat.com>
14739 PR tree-optimization/84740
14740 * tree-switch-conversion.c (process_switch): Call build_constructors
14741 only if info.phi_count is non-zero.
14743 PR tree-optimization/84739
14744 * tree-tailcall.c (find_tail_calls): Check call arguments against
14745 DECL_ARGUMENTS (current_function_decl) rather than
14746 DECL_ARGUMENTS (func) when checking for tail recursion.
14748 2018-03-07 Jakub Jelinek <jakub@redhat.com>
14750 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
14751 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
14752 Volker Reichelt's entry and add entries for people that perform
14753 GCC fuzzy testing and report numerous bugs.
14755 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
14758 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
14759 readonly data in sdata, if that is disabled.
14760 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
14761 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
14762 -mreadonly-in-sdata option.
14764 2018-03-07 Martin Sebor <msebor@redhat.com>
14766 PR tree-optimization/84468
14767 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
14768 basic block when looking for nul assignment.
14770 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
14773 * except.h (output_function_exception_table): Adjust prototype.
14774 * except.c (output_function_exception_table): Remove FNNAME parameter
14775 and add SECTION parameter. Ouput one part of the table at a time.
14776 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
14777 the first part of the exception table and emit unwind directives.
14778 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
14779 (i386_pe_seh_cold_init): Likewise.
14780 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
14781 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
14782 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
14783 (ix86_output_call_insn): Emit a nop in one more case for SEH.
14784 * config/i386/winnt.c: Include except.h.
14785 (struct seh_frame_state): Add reg_offset, after_prologue and
14786 in_cold_section fields.
14787 (i386_pe_seh_end_prologue): Set seh->after_prologue.
14788 (i386_pe_seh_cold_init): New function.
14789 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
14790 to seh->in_cold_section.
14791 (seh_emit_push): Record the offset of the push.
14792 (seh_emit_save): Record the offet of the save.
14793 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
14794 Test seh->after_prologue to disregard the epilogue.
14795 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
14796 (i386_pe_end_cold_function): New function.
14798 2018-03-07 Jakub Jelinek <jakub@redhat.com>
14801 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
14802 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
14805 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
14807 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
14808 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
14810 PR middle-end/84723
14811 * multiple_target.c: Include tree-inline.h and intl.h.
14812 (expand_target_clones): Diagnose and fail if node->definition and
14813 !tree_versionable_function_p (node->decl).
14815 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
14817 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
14819 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
14820 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
14821 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
14823 2018-03-06 Jakub Jelinek <jakub@redhat.com>
14826 * combine.c (try_combine): Use reg_or_subregno instead of handling
14827 just paradoxical SUBREGs and REGs.
14829 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
14831 * config/arc/arc.c (arc_finalize_pic): Remove function.
14832 (arc_must_save_register): We use single base PIC register, remove
14833 checks to save/restore the PIC register.
14834 (arc_expand_prologue): Likewise.
14835 * config/arc/arc-protos.h (arc_set_default_type_attributes):
14837 (arc_verify_short): Likewise.
14838 (arc_attr_type): Likewise.
14839 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
14840 (walk_stores): Likewise.
14841 (arc_address_cost): Make it static.
14842 (arc_verify_short): Likewise.
14843 (branch_dest): Likewise.
14844 (arc_attr_type): Likewise.
14845 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
14846 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
14847 (arc_final_prescan_insn): Remove inserting the nops due to
14848 hardware hazards. It is done in reorg step.
14849 (insn_length_variant_t): Remove.
14850 (insn_length_parameters_t): Likewise.
14851 (arc_insn_length_parameters): Likewise.
14852 (arc_get_insn_variants): Likewise.
14853 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
14855 2018-03-06 Jakub Jelinek <jakub@redhat.com>
14857 PR inline-asm/84683
14858 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
14861 PR tree-optimization/84687
14862 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
14864 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
14866 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14868 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
14869 Rename to ppc_speculation_barrier.
14870 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
14871 __builtin_ppc_speculation_barrier.
14873 2018-03-05 Jakub Jelinek <jakub@redhat.com>
14876 * combine.c (combine_simplify_rtx): Don't try to simplify if
14877 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
14880 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
14882 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
14883 to 32 bytes when compiling for POWER9.
14885 2018-03-05 Jakub Jelinek <jakub@redhat.com>
14888 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
14889 regparm >= 3 with no arg reg available also for calls with
14890 flag_force_indirect_call. Pass decl to ix86_function_regparm.
14893 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
14895 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
14897 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
14900 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
14902 2018-03-05 Richard Biener <rguenther@suse.de>
14904 PR tree-optimization/84486
14905 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
14906 When inserting a __builtin_assume_aligned call set the LHS
14907 SSA name alignment info accordingly.
14909 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
14911 PR tree-optimization/84114
14912 * config/aarch64/aarch64.c (aarch64_reassociation_width)
14913 Avoid reassociation of FLOAT_MODE addition.
14915 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
14917 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
14918 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
14919 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
14920 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
14921 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
14922 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
14923 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
14925 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
14926 __builtin_ia32_wbinvd): New builtins.
14927 (SPECIAL_ARGS2): New.
14928 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
14929 (SPECIAL_ARGS2): New.
14930 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
14931 (ix86_valid_target_attribute_inner_p): Ditto.
14932 (ix86_init_mmx_sse_builtins): Add special_args2.
14933 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
14934 TARGET_WBNOINVD_P): New.
14935 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
14936 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
14937 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
14938 * config/i386/immintrin.h (_wbinvd): New intrinsic.
14939 * config/i386/pconfigintrin.h: New file.
14940 * config/i386/wbnoinvdintrin.h: Ditto.
14941 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
14943 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
14945 2018-03-05 Richard Biener <rguenther@suse.de>
14947 PR tree-optimization/84670
14948 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
14950 (BB_VISITED_WITH_VISITED_SUCCS): New define.
14951 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
14952 (compute_antic_aux): Only assert the number of values in ANTIC_IN
14953 doesn't grow if all successors (recursively) were visited at least
14956 2018-03-05 Richard Biener <rguenther@suse.de>
14958 PR tree-optimization/84650
14959 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
14960 if executed in the loop pipeline.
14962 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
14964 * doc/configfiles.texi (Configuration Files): Move info about
14965 conditionalizing $target-protos.h to...
14966 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
14967 differs from $target-protos.h.
14969 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
14970 Chung-Ju Wu <jasonwucj@gmail.com>
14972 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
14973 * config/nds32/nds32-multiple.md (setmemsi): Define.
14974 * config/nds32/nds32-memory-manipulation.c
14975 (nds32_gen_dup_4_byte_to_word_value): New.
14976 (emit_setmem_word_loop): New.
14977 (emit_setmem_byte_loop): New.
14978 (nds32_expand_setmem_loop): New.
14979 (nds32_expand_setmem_loop_v3m): New.
14980 (nds32_expand_setmem_unroll): New.
14981 (nds32_expand_setmem): New.
14983 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
14984 Chung-Ju Wu <jasonwucj@gmail.com>
14986 * config/nds32/nds32-memory-manipulation.c
14987 (nds32_emit_load_store): New.
14988 (nds32_emit_post_inc_load_store): New.
14989 (nds32_emit_mem_move): New.
14990 (nds32_emit_mem_move_block): New.
14991 (nds32_expand_movmemsi_loop_unknown_size): New.
14992 (nds32_expand_movmemsi_loop_known_size): New.
14993 (nds32_expand_movmemsi_loop): New.
14994 (nds32_expand_movmemsi_unroll): New.
14995 (nds32_expand_movmemqi): Rename ...
14996 (nds32_expand_movmemsi): ... to this.
14997 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
14998 (movmemsi): ... to this.
14999 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
15000 (nds32_expand_movmemsi): ... to this.
15002 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
15003 Monk Chiang <sh.chiang04@gmail.com>
15004 Chung-Ju Wu <jasonwucj@gmail.com>
15006 * config/nds32/nds32-protos.h
15007 (nds32_expand_load_multiple): New arguments.
15008 (nds32_expand_store_multiple): Ditto.
15009 (nds32_valid_multiple_load_store): Rename ...
15010 (nds32_valid_multiple_load_store_p): ... to this.
15011 * config/nds32/nds32-memory-manipulation.c
15012 (nds32_expand_load_multiple): Refine implementation.
15013 (nds32_expand_store_multiple): Ditto.
15014 * config/nds32/nds32-multiple.md
15015 (load_multiple): Update nds32_expand_load_multiple interface.
15016 (store_multiple): Update nds32_expand_store_multiple interface.
15017 * config/nds32/nds32-predicates.c
15018 (nds32_valid_multiple_load_store): Rename ...
15019 (nds32_valid_multiple_load_store_p): ... to this and refine
15021 * config/nds32/predicates.md
15022 (nds32_load_multiple_and_update_address_operation): New predicate.
15023 (nds32_store_multiple_and_update_address_operation): New predicate.
15025 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
15026 Chung-Ju Wu <jasonwucj@gmail.com>
15028 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
15029 (combo): New attribute.
15030 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
15032 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
15034 * config/nds32/nds32.opt: Change -mcmodel= default value.
15036 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
15037 Monk Chiang <sh.chiang04@gmail.com>
15038 Chung-Ju Wu <jasonwucj@gmail.com>
15040 * config/nds32/constants.md (unspec_element): New enum.
15041 * config/nds32/constraints.md (Umw): New constraint.
15042 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
15043 * config/nds32/nds32-intrinsic.md: Likewise.
15044 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
15045 (nds32_valid_smw_lwm_base_p): New.
15046 (nds32_output_smw_single_word): New.
15047 (nds32_output_lmw_single_word): New.
15048 (nds32_expand_unaligned_load): New.
15049 (nds32_expand_unaligned_store): New.
15050 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
15051 (nds32_output_smw_single_word): Declare.
15052 (nds32_output_lmw_single_word): Declare.
15053 (nds32_expand_unaligned_load): Declare.
15054 (nds32_expand_unaligned_store): Declare.
15055 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
15056 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
15057 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
15058 NDS32_BUILTIN_UASTORE_DW.
15059 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
15062 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
15063 Kito Cheng <kito.cheng@gmail.com>
15064 Chung-Ju Wu <jasonwucj@gmail.com>
15066 * config/nds32/nds32-intrinsic.c
15067 (nds32_expand_builtin_null_ftype_reg): Delete.
15068 (nds32_expand_builtin_reg_ftype_imm): Ditto.
15069 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
15070 (nds32_read_argument): New.
15071 (nds32_legitimize_target): Ditto.
15072 (nds32_legitimize_argument): Ditto.
15073 (nds32_check_constant_argument): Ditto.
15074 (nds32_expand_unop_builtin): Ditto.
15075 (nds32_expand_unopimm_builtin): Ditto.
15076 (nds32_expand_binop_builtin): Ditto.
15077 (nds32_builtin_decl_impl): Ditto.
15078 (builtin_description): Ditto.
15079 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
15080 (nds32_init_builtins_impl): Ditto.
15081 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
15082 (nds32_builtin_decl): New.
15083 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
15084 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
15086 2018-03-02 Jeff Law <law@redhat.com>
15088 * reorg.c (stop_search_p): Handle DEBUG_INSN.
15089 (redundant_insn, fill_simple_delay_slots): Likewise.
15090 (fill_slots_from_thread): Likewise.
15091 * resource.c (mark_referenced_resources): Likewise.
15092 (mark_set_resources, find_dead_or_set_registers): Likewise.
15094 2018-03-02 Jakub Jelinek <jakub@redhat.com>
15096 * substring-locations.h (format_warning_va): Formatting fix for
15097 ATTRIBUTE_GCC_DIAG.
15098 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
15100 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
15101 * substring-locations.c: Include intl.h.
15102 (format_warning_va): Turned into small wrapper around
15103 format_warning_n_va, renamed to ...
15104 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
15105 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
15107 (format_warning_at_substring_n): New function.
15108 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
15109 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
15110 format_warning_at_substring with just a shorter name instead of
15111 const function pointer.
15112 (fmtwarn_n): New function.
15113 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
15114 appropriate, get rid of all the fmtstr temporaries, move conditionals
15115 with G_() wrapped string literals directly into fmtwarn arguments,
15116 cast dir.len to (int), formatting fixes.
15118 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
15120 * doc/invoke.texi: Remove "Cilk Plus" references.
15122 2018-03-02 Jakub Jelinek <jakub@redhat.com>
15123 Richard Biener <rguenther@suse.de>
15126 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
15127 for error or warning attributes if CALL_FROM_THUNK_P is set.
15130 2018-03-02 Jakub Jelinek <jakub@redhat.com>
15133 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
15134 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
15137 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
15138 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
15140 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
15141 instead of -1U in last predictors element's probability member.
15143 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
15146 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
15147 arguments if they are comparable.
15149 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
15151 PR tree-optimization/84634
15152 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
15153 masks and masked_loop_p with a single loop_masks, making sure it's
15154 null for bb vectorization.
15156 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
15158 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
15159 (vect_analyze_data_ref_access): Use loop->safe_len rather than
15160 loop->force_vectorize to check whether there is no alias.
15162 2018-03-02 Jakub Jelinek <jakub@redhat.com>
15165 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
15167 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
15169 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
15170 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
15171 instead of a loop around prev_real_insn.
15172 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
15175 PR inline-asm/84625
15176 * config/i386/i386.c (ix86_print_operand): Use conditional
15177 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
15180 2018-03-02 Richard Biener <rguenther@suse.de>
15182 PR tree-optimization/84427
15183 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
15184 (bitmap_set_subtract_values): Rewrite to handle multiple
15187 (prune_clobbered_mems): Likewise.
15188 (phi_translate): Take edge instead of pred/phiblock.
15189 (phi_translate_1): Likewise.
15190 (phi_translate_set): Likewise. Insert all translated
15191 exprs for a value into the set, keeping possibly multiple
15192 expressions per value.
15193 (compute_antic_aux): Adjust for phi_translate changes.
15194 When intersecting union the expressions and prune those
15195 not in the final value set, keeping possibly multiple
15196 expressions per value. Do not use value-insertion
15197 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
15198 all expressions. Add verification that the value-sets
15199 only shrink during iteration.
15200 (compute_partial_antic_aux): Adjust for the phi_translate changes.
15201 (do_pre_regular_insertion): Likewise.
15202 (do_pre_partial_partial_insertion): Likewise.
15204 2018-03-02 Richard Biener <rguenther@suse.de>
15207 * config/darwin.c (saved_debug_info_level): New static global.
15208 (darwin_asm_lto_start): Disable debug info generation for LTO out.
15209 (darwin_asm_lto_end): Restore debug info generation settings.
15211 2018-03-01 Martin Liska <mliska@suse.cz>
15214 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
15215 volatile arguments.
15217 2018-03-01 Richard Biener <rguenther@suse.de>
15220 * dwarf2out.c (gen_variable_die): Properly handle late VLA
15221 type annotation with LTO when debug was disabled at compile-time.
15223 2018-03-01 Matthew Fortune <mfortune@gmail.com>
15225 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
15227 (mips_final_postscan_insn): Likewise.
15229 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
15231 PR rtl-optimization/84528
15232 * alias.c (init_alias_target): Add commentary.
15233 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
15234 a unique base value if the frame pointer is not eliminated
15235 to the stack pointer.
15237 2018-03-01 Tom de Vries <tom@codesourcery.com>
15239 PR rtl-optimization/83327
15240 * lra-int.h (hard_regs_spilled_into): Declare.
15241 * lra.c (hard_regs_spilled_into): Define.
15242 (init_reg_info): Init hard_regs_spilled_into.
15243 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
15244 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
15245 (process_bb_lives): Handle hard_regs_spilled_into.
15246 (lra_create_live_ranges_1): Before doing liveness propagation, clear
15247 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
15249 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
15251 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
15252 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
15253 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
15254 * config/rs6000/aix72.h: New file.
15256 2018-02-28 Jakub Jelinek <jakub@redhat.com>
15258 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
15259 instead of warning_at with conditional singular and plural messages
15263 * stor-layout.c (update_alignment_for_field): For
15264 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
15265 && !DECL_PACKED (field), do the alignment update, just use
15266 only desired_align instead of MAX (type_align, desired_align)
15268 (place_field): Don't do known_align < desired_align handling
15269 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
15270 is non-NULL, instead do it after rli->prev_field handling and
15271 only if not within a bitfield word. For DECL_PACKED (field)
15272 use type_align of BITS_PER_UNIT.
15274 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
15276 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
15277 superfluous parentheses and trailing spaces.
15279 2018-02-28 Richard Biener <rguenther@suse.de>
15281 PR tree-optimization/84584
15282 * graphite-scop-detection.c (scop_detection::add_scop): Discard
15283 SCoPs with fake exit edge.
15285 2018-02-28 Martin Liska <mliska@suse.cz>
15288 * timevar.c (timer::print): Fix format to properly print 100%
15291 2018-02-28 Richard Biener <rguenther@suse.de>
15293 PR middle-end/84607
15294 * genmatch.c (capture_info::walk_match): Do not mark
15295 captured expressions without operands as expr_p given
15296 they act more like predicates and should be subject to
15297 "lost tail" side-effect preserving.
15299 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
15301 PR rtl-optimization/81611
15302 * auto-inc-dec.c (attempt_change): Move dead note from
15303 mem_insn if it's the next use of regno
15304 (find_address): Take address use of reg holding
15305 non-incremented value. Add parm to limit search to the named
15307 (merge_in_block): Attempt to use a mem insn that is the next
15308 use of the original regno.
15310 2018-02-27 Martin Sebor <msebor@redhat.com>
15313 * doc/invoke.texi (-Wmissing-attributes): New option.
15314 * print-tree.c (print_node): Handle DECL_UNINLINABLE.
15316 2018-02-27 Martin Sebor <msebor@redhat.com>
15318 PR translation/84207
15319 * diagnostic-core.h (warning_n, error_n, inform_n): Change
15320 n argument to unsigned HOST_WIDE_INT.
15321 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
15322 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
15323 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
15324 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
15326 2018-02-27 Richard Biener <rguenther@suse.de>
15328 PR tree-optimization/84512
15329 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
15330 Do not use the estimate returned from record_stmt_cost for
15331 the scalar iteration cost but sum properly using add_stmt_cost.
15333 2018-02-27 Richard Biener <rguenther@suse.de>
15335 PR tree-optimization/84466
15336 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
15337 Adjust last change to less strictly validate use operands.
15339 2018-02-27 Martin Liska <mliska@suse.cz>
15341 PR gcov-profile/84548
15342 * gcov.c (process_file): Allow partial overlap and consider it
15343 also as group functions.
15344 (output_lines): Properly calculate range of lines for a group.
15346 2018-02-27 Martin Liska <mliska@suse.cz>
15348 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
15349 'ggc' suffixes. Change first column width.
15350 (timer::print): Fix formatting of the column.
15352 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
15354 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
15355 preserve inline entry blocks for the sake of debug inline
15356 entry point markers alone.
15357 (remove_unused_locals): Suggest in comments a better place to
15358 force the preservation of inline entry blocks that are
15359 otherwise unused, but do not preserve them.
15361 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
15363 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
15365 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
15368 * config/i386/constraints.md (Bs): Replace
15369 ix86_indirect_branch_register with
15370 TARGET_INDIRECT_BRANCH_REGISTER.
15372 * config/i386/i386.md (indirect_jump): Likewise.
15373 (tablejump): Likewise.
15374 (*sibcall_memory): Likewise.
15375 (*sibcall_value_memory): Likewise.
15376 Peepholes of indirect call and jump via memory: Likewise.
15377 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
15378 (*sibcall_value_GOT_32): Likewise.
15379 * config/i386/predicates.md (indirect_branch_operand): Likewise.
15380 (GOT_memory_operand): Likewise.
15381 (call_insn_operand): Likewise.
15382 (sibcall_insn_operand): Likewise.
15383 (GOT32_symbol_operand): Likewise.
15384 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
15386 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
15388 PR rtl-optimization/83496
15389 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
15390 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
15391 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
15392 redundant insn, if any.
15393 (relax_delay_slots): Likewise.
15394 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
15396 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
15398 PR tree-optimization/83965
15399 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
15400 that grouped statements are part of a reduction chain. Return
15401 true if the statement is not marked as a reduction itself but
15402 is part of a group.
15403 (vect_recog_dot_prod_pattern): Don't check whether the statement
15404 is part of a group here.
15405 (vect_recog_sad_pattern): Likewise.
15406 (vect_recog_widen_sum_pattern): Likewise.
15408 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
15411 * final.c (rest_of_clean_state): Also look for calls inside sequences.
15413 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
15416 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
15418 (ix86_output_indirect_function_return): New prototype.
15419 (ix86_split_simple_return_pop_internal): Likewise.
15420 * config/i386/i386.c (indirect_return_via_cx): New.
15421 (indirect_return_via_cx_bnd): Likewise.
15422 (indirect_thunk_name): Handle return va CX_REG.
15423 (output_indirect_thunk_function): Create alias for
15424 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
15425 (ix86_output_indirect_jmp): Remove the bool argument.
15426 (ix86_output_indirect_function_return): New function.
15427 (ix86_split_simple_return_pop_internal): Likewise.
15428 * config/i386/i386.md (*indirect_jump): Don't pass false
15429 to ix86_output_indirect_jmp.
15430 (*tablejump_1): Likewise.
15431 (simple_return_pop_internal): Change it to define_insn_and_split.
15432 Call ix86_split_simple_return_pop_internal to split it for
15433 -mfunction-return=.
15434 (simple_return_indirect_internal): Call
15435 ix86_output_indirect_function_return instead of
15436 ix86_output_indirect_jmp.
15438 2018-02-26 Jakub Jelinek <jakub@redhat.com>
15441 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
15442 memset and value initialization afterwards.
15444 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
15446 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
15448 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
15451 * common/config/aarch64/aarch64-common.c
15452 (aarch_option_optimization_table[]): Switch
15453 off fomit-frame-pointer
15455 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
15456 Chung-Ju Wu <jasonwucj@gmail.com>
15458 * config/nds32/nds32-multiple.md (load_multiple): Disallow
15460 (store_multiple): Ditto.
15462 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
15464 * config.gcc: Add --with-cpu support for nds32 target.
15465 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
15466 * config/nds32/nds32.opt: Add -mcpu= option.
15468 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
15470 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
15471 isel=yes): Warn for these deprecated options.
15473 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
15475 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
15476 ISA_2_5_MASKS_EMBEDDED.
15478 2018-02-23 Jakub Jelinek <jakub@redhat.com>
15480 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
15481 p->max as pointers rather than using iterative_hash_expr.
15483 2018-02-23 Carl Love <cel@us.ibm.com>
15485 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
15486 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
15488 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
15489 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
15490 P8V_BUILTIN_VEC_VUNSIGNED2.
15492 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
15495 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
15496 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
15498 * lra-constraints.c (curr_insn_transform): Set up
15499 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
15501 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
15502 * lra-eliminations.c (spill_pseudos): Ditto.
15503 (process_insn_for_elimination): Ditto.
15504 * lra-lives.c (reg_early_clobber_p): Use the new macros.
15505 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
15506 LRA_NON_CLOBBERED_ALT.
15508 2018-02-22 Martin Sebor <msebor@redhat.com>
15510 PR tree-optimization/84480
15511 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
15512 to maybe_diag_stxncpy_trunc. Call it.
15513 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
15514 from gimple_fold_builtin_strcpy. Print inlining stack.
15515 (handle_builtin_stxncpy): Print inlining stack.
15516 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
15518 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
15521 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
15522 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
15523 and -fcheck-pointer-bounds are used together.
15524 (indirect_thunk_prefix): New enum.
15525 (indirect_thunk_need_prefix): New function.
15526 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
15527 "_nt" instead of "_bnd" for NOTRACK prefix.
15528 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
15529 (output_indirect_thunk_function): Likewise.
15531 (ix86_code_end): Update output_indirect_thunk_function calls.
15532 (ix86_output_indirect_branch_via_reg): Replace
15533 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
15534 (ix86_output_indirect_branch_via_push): Likewise.
15535 (ix86_output_function_return): Likewise.
15536 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
15537 incompatible with -fcf-protection=branch and
15538 -fcheck-pointer-bounds.
15540 2018-02-22 Steve Ellcey <sellcey@cavium.com>
15543 * config/aarch64/aarch64.c (aarch64_print_address_internal):
15544 Change gcc_assert call to output_operand_lossage.
15546 2018-02-22 Steve Ellcey <sellcey@cavium.com>
15548 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
15550 2018-02-22 DJ Delorie <dj@redhat.com>
15551 Sebastian Perta <sebastian.perta@renesas.com>
15552 Oleg Endo <olegendo@gcc.gnu.org>
15554 * config/rx/rx.c (rx_rtx_costs): New function.
15555 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
15557 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
15559 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
15561 2018-02-22 Martin Liska <mliska@suse.cz>
15564 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
15565 Add "native" as a possible value.
15567 2018-02-22 Martin Liska <mliska@suse.cz>
15570 * config/i386/i386.c (ix86_option_override_internal):
15571 Add "native" as a possible value for -march and -mtune.
15573 2018-02-22 Jakub Jelinek <jakub@redhat.com>
15576 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
15577 to all type variants.
15579 PR tree-optimization/84503
15580 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
15581 width as info->bitpos + info->bitsize - start.
15582 (merged_store_group::merge_overlapping): Simplify width computation.
15583 (check_no_overlap): New function.
15584 (imm_store_chain_info::try_coalesce_bswap): Compute expected
15585 start + width and last_order of the group, fail if check_no_overlap
15587 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
15588 to group if check_no_overlap fails.
15590 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
15592 * config/rs6000/altivec.md: Delete contraint arguments to
15593 define_expand, define_split, and define_peephole2, and in
15594 define_insn_and_split if always unused.
15595 * config/rs6000/darwin.md: Ditto.
15596 * config/rs6000/dfp.md: Ditto.
15597 * config/rs6000/rs6000.md: Ditto.
15598 * config/rs6000/sync.md: Ditto.
15599 * config/rs6000/vector.md: Ditto.
15600 * config/rs6000/vsx.md: Ditto.
15602 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
15604 * config/rs6000/altivec.md: Write output control strings as braced
15605 blocks instead of double-quoted strings.
15606 * config/rs6000/darwin.md: Ditto.
15607 * config/rs6000/rs6000.md: Ditto.
15608 * config/rs6000/vector.md: Ditto.
15609 * config/rs6000/vsx.md: Ditto.
15611 2018-02-21 Jason Merrill <jason@redhat.com>
15613 PR c++/84314 - ICE with templates and fastcall attribute.
15614 * attribs.c (build_type_attribute_qual_variant): Remove assert.
15616 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
15618 * ipa-cp.c (determine_versionability): Fix comment typos.
15620 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
15623 * ipa-cp.c (determine_versionability): Do not version functions caling
15626 2018-02-21 Martin Liska <mliska@suse.cz>
15629 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
15630 Add "native" as a possible value.
15631 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
15632 the macro when native cpu detection is available.
15634 2018-02-21 Martin Liska <mliska@suse.cz>
15637 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
15638 Add "native" as a possible value.
15639 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
15640 when native cpu detection is available.
15642 2018-02-21 Jakub Jelinek <jakub@redhat.com>
15643 Martin Sebor <msebor@redhat.com>
15645 PR tree-optimization/84478
15646 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
15648 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
15649 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
15650 support which is conservatively correct, for 2 only stay conservative
15651 for maxlen. Formatting and comment capitalization fixes. Add STRICT
15652 argument to the 2 argument get_range_strlen, adjust 6 arg
15653 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
15655 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
15656 (gimple_fold_builtin_strlen): Pass true as last argument to
15659 2018-02-20 Martin Sebor <msebor@redhat.com>
15661 PR middle-end/84095
15662 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
15663 (builtin_memref::set_base_and_offset): Same. Handle inner references.
15664 (builtin_memref::builtin_memref): Factor out parts into
15665 set_base_and_offset and call it.
15667 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
15669 PR middle-end/84406
15670 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
15671 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
15672 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
15673 search at the associated MODE_INT.
15675 2018-02-20 Jeff Law <law@redhat.com>
15677 PR middle-end/82123
15678 PR tree-optimization/81592
15679 PR middle-end/79257
15680 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
15681 for range data rather than using global data.
15682 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
15683 range data rather than using global data.
15684 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
15685 pass it to children as needed.
15686 (struct directive::fmtresult): Similarly.
15687 (struct directive::set_width): Similarly.
15688 (struct directive::set_precision): Similarly.
15689 (format_integer, format_directive, parse_directive): Similarly.
15690 (format_none): Accept unnamed vr_values parameter.
15691 (format_percent, format_floating, format_character): Similarly.
15692 (format_string, format_plain): Similarly.
15693 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
15694 the EVRP range analyzer for range data rather than using global data.
15695 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
15696 gimple-ssa-evrp-analyze.h
15697 (class sprintf_dom_walker): Add after_dom_children member function.
15698 Add evrp_range_analyzer member.
15699 (sprintf_dom_walker::before_dom_children): Call into the EVRP
15700 range analyzer as needed.
15701 (sprintf_dom_walker::after_dom_children): New member function.
15702 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
15704 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
15705 (evrp_range_analyzer::pop_to_marker): Likewise.
15707 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
15709 PR tree-optimization/84419
15710 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
15711 with the required type if its current type is compatible but
15714 2018-02-20 Jakub Jelinek <jakub@redhat.com>
15716 PR middle-end/82004
15717 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
15718 after vectorization.
15720 2018-02-20 Martin Liska <mliska@suse.cz>
15723 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
15724 possible values if we don't have a hint.
15726 2018-02-20 Martin Liska <mliska@suse.cz>
15730 * final.c (shorten_branches): Build align_tab array with one
15732 * opts.c (finish_options): Add alignment option limit check.
15733 (MAX_CODE_ALIGN): Likewise.
15734 (MAX_CODE_ALIGN_VALUE): Likewise.
15735 * doc/invoke.texi: Document maximum allowed option value for
15736 all -falign-* options.
15738 2018-02-19 Jakub Jelinek <jakub@redhat.com>
15741 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
15742 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
15743 * var-tracking.c (emit_note_insn_var_location): Remove all references
15744 to NOTE_INSN_CALL_ARG_LOCATION.
15745 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
15746 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
15747 Use copy_rtx_if_shared.
15748 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
15749 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
15750 (dwarf2out_var_location): Remove handling of
15751 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
15753 * final.c (final_scan_insn): Remove all references to
15754 NOTE_INSN_CALL_ARG_LOCATION.
15755 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
15756 before dumping final insns.
15757 * except.c (emit_note_eh_region_end): Remove all references to
15758 NOTE_INSN_CALL_ARG_LOCATION.
15759 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
15760 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
15761 * config/arc/arc.c (hwloop_optimize): Likewise.
15762 * config/arm/arm.c (create_fix_barrier): Likewise.
15763 * config/s390/s390.c (s390_chunkify_start): Likewise.
15764 * config/sh/sh.c (find_barrier): Likewise.
15765 * config/i386/i386.c (rest_of_insert_endbranch,
15766 ix86_seh_fixup_eh_fallthru): Likewise.
15767 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
15768 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
15769 * config/frv/frv.c (frv_function_prologue): Likewise.
15770 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
15772 (note_outside_basic_block_p): Remove all references to
15773 NOTE_INSN_CALL_ARG_LOCATION.
15774 * gengtype.c (adjust_field_rtx_def): Likewise.
15775 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
15777 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
15778 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
15781 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
15784 PR tree-optimization/84452
15785 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
15786 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
15789 2018-02-19 Martin Liska <mliska@suse.cz>
15792 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
15794 2018-02-19 Martin Liska <mliska@suse.cz>
15795 Richard Sandiford <richard.sandiford@linaro.org>
15797 PR tree-optimization/82491
15798 * gimple-fold.c (get_base_constructor): Make earlier bail out
15801 2018-02-19 Carl Love <cel@us.ibm.com>
15803 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
15804 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
15806 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
15807 P8V_BUILTIN_VEC_NEG.
15809 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
15811 * config/rl78/rl78.md (movdf): New define expand.
15813 2018-02-19 Martin Liska <mliska@suse.cz>
15816 * doc/invoke.texi: Fix typo.
15817 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
15819 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
15821 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
15822 handle rs6000_single_float and rs6000_double_float specially for
15825 2018-02-16 Jeff Law <law@redhat.com>
15827 * config/rx/rx.c (add_pop_cfi_notes): New function.;
15828 (pop_regs): Use it.
15830 2018-02-16 Jakub Jelinek <jakub@redhat.com>
15833 * ipa-inline.c (inline_small_functions): Fix a typo.
15835 2018-02-16 Nathan Sidwell <nathan@acm.org>
15837 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
15839 2018-02-16 Carl Love <cel@us.ibm.com>
15841 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
15842 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
15843 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
15844 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
15845 expansion to P8V_BUILTIN_VEC_FLOAT2.
15847 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
15849 PR rtl-optimization/70023
15850 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
15851 src_regno into account.
15853 2018-02-16 Carl Love <cel@us.ibm.com>
15855 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
15856 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
15857 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
15858 * config/rs6000/rs6000.c: Remove case statements for
15859 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
15860 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
15861 and P9V_BUILTIN_VEC_VINSERT4B.
15862 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
15863 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
15864 * config/rs6000/vsx.md:
15865 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
15868 2018-02-16 Carl Love <cel@us.ibm.com>
15870 * config/rs6000/altivec.h: Add builtin names vec_extract4b
15872 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
15874 * config/rs6000/rs6000-c.c: Add the definitions for
15875 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
15876 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
15877 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
15878 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
15879 definition for insert4b and define insn *insert3b_internal.
15880 * doc/extend.texi: Add documentation for vec_extract4b.
15882 2018-02-16 Nathan Sidwell <nathan@acm.org>
15884 * doc/extend.texi (Backwards Compatibility): Mention friend
15885 injection. Note for-scope is deprecated.
15886 * doc/invoke.texi (-ffriend-injection): Deprecate.
15888 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
15890 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
15891 that moved to I2, also allow destinations that are a paradoxical
15892 subreg (instead of a normal reg).
15894 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
15897 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
15900 2018-02-16 Richard Biener <rguenther@suse.de>
15902 PR tree-optimization/84037
15903 PR tree-optimization/84016
15905 * config/i386/i386.c (ix86_builtin_vectorization_cost):
15906 Adjust vec_construct for the fact we need additional higher latency
15907 128bit inserts for AVX256 and AVX512 vector builds.
15908 (ix86_add_stmt_cost): Scale vector construction cost for
15911 2018-02-16 Richard Biener <rguenther@suse.de>
15913 PR tree-optimization/84417
15914 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
15915 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
15916 (non_rewritable_lvalue_p): Likewise, use poly-ints.
15918 2018-02-16 Martin Liska <mliska@suse.cz>
15921 * internal-fn.def (ASAN_CHECK): Set proper flags.
15922 (ASAN_MARK): Likewise.
15924 2018-02-16 Julia Koval <julia.koval@intel.com>
15926 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
15927 from PTA_CANNONLAKE.
15929 2018-02-16 Jakub Jelinek <jakub@redhat.com>
15932 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
15933 Use ++iter rather than iter++ for std::list iterators.
15934 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
15935 defer deleting them until all nodes in the forest are processed. Do
15936 free even leaf nodes. Change to_process into auto_vec.
15939 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
15940 * vec.h (vec_default_construct): Use memset instead of placement new
15941 if BROKEN_VALUE_INITIALIZATION is defined.
15942 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
15943 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
15946 PR rtl-optimization/83723
15947 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
15948 * lra.c (lra_substitute_pseudo): Likewise. If true, use
15949 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
15951 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
15953 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
15955 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
15957 PR rtl-optimization/81443
15958 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
15959 from inner REGs to paradoxical SUBREGs.
15961 2018-02-16 Richard Biener <rguenther@suse.de>
15963 PR tree-optimization/84399
15964 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
15965 For operands we can analyze at their definition make sure we can
15966 analyze them at each use as well.
15968 2018-02-16 Richard Biener <rguenther@suse.de>
15970 PR tree-optimization/84190
15971 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
15972 volatile accesses if the decl isn't volatile.
15974 2018-02-15 Jason Merrill <jason@redhat.com>
15976 PR c++/84314 - ICE with templates and fastcall attribute.
15977 * attribs.c (build_type_attribute_qual_variant): Don't clobber
15978 TYPE_CANONICAL on an existing type.
15980 2018-02-15 Jakub Jelinek <jakub@redhat.com>
15982 PR tree-optimization/84383
15983 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
15984 dstoff nor call operand_equal_p if dstbase is NULL.
15986 PR tree-optimization/84334
15987 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
15988 also a CONSTANT_CLASS_P, punt.
15990 2018-02-14 Jim Wilson <jimw@sifive.com>
15992 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
15993 first SMALL_OPERAND check. New local min_second_step. Move assert
15994 to where locals are set. Add TARGET_RVC support.
15995 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
15997 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
15999 * doc/invoke.texi: Correct -Wformat-overflow code sample.
16001 2018-02-14 Martin Sebor <msebor@redhat.com>
16003 PR tree-optimization/83698
16004 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
16005 arrays constrain the offset range to their bounds.
16006 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
16007 (builtin_access::overlap): Avoid setting the size of overlap if it's
16009 (maybe_diag_overlap): Also consider arrays when deciding what values
16010 of offsets to include in diagnostics.
16012 2018-02-14 Martin Sebor <msebor@redhat.com>
16015 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
16016 that correspond to the kind of a declaration.
16018 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
16021 * config/pa/pa.md: Load address of PIC label using the linkage table
16022 if the label is nonlocal.
16024 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
16026 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
16027 warning message if user requests -maltivec=be.
16028 * doc/invoke.texi: Document deprecation of -maltivec=be.
16030 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
16033 * config/rs6000/rs6000-c.c: Update definitions for
16034 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
16035 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
16037 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
16040 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
16041 add _get_ssp intrinsics. Remove argument from
16042 __builtin_ia32_rdssp[d|q].
16043 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
16044 * config/i386/i386-builtin.def: Remove argument from
16045 __builtin_ia32_rdssp[d|q].
16046 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
16047 ix86_expand_special_args_builtin for _rdssp[d|q].
16048 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
16049 Clear register before usage.
16050 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
16051 Add documentation for new _get_ssp and _inc_ssp intrinsics.
16053 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
16055 PR tree-optimization/84357
16056 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
16057 operand 1 of an ARRAY_REF too.
16059 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
16062 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
16063 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
16065 (set_of_reg): New struct.
16066 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
16067 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
16068 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
16070 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
16071 Split into bitclr, bitset, bitinvert patterns if appropriate.
16072 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
16073 use rx_fuse_in_memory_bitop.
16074 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
16075 to named insn, correct maximum insn length.
16077 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
16080 * machmode.def: Define a complex mode for PARTIAL_INT.
16081 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
16083 * doc/rtl.texi: Document CSPImode.
16084 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
16086 (msp430_hard_regno_nregs_with_padding): Likewise.
16088 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
16091 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
16093 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
16095 PR rtl-optimization/84169
16096 * combine.c (try_combine): New variable split_i2i3. Set it to true if
16097 we generated a parallel as new i3 and we split that to new i2 and i3
16098 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
16099 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
16100 those to i2, not i1. Partially rewrite this scan code.
16102 2018-02-13 Jakub Jelinek <jakub@redhat.com>
16105 * stor-layout.c (place_field): For variable length fields, adjust
16106 offset_align afterwards not just based on the field's alignment,
16107 but also on the size.
16109 PR middle-end/84309
16110 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
16111 of exps and logs in the use_exp2 case.
16113 2018-02-13 Jeff Law <law@redhat.com>
16115 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
16116 entry for "vector".
16118 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
16121 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
16125 * common.opt (gas-loc-support, gas-locview-support): New.
16126 (ginline-points, ginternal-reset-location-views): New.
16127 * doc/invoke.texi: Document them. Use @itemx where intended.
16128 (gvariable-location-views): Adjust.
16129 * target.def (reset_location_view): New.
16130 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
16131 (TARGET_RESET_LOCATION_VIEW): New.
16132 * doc/tm.texi: Rebuilt.
16133 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
16134 (dwarf2out_default_as_locview_support): New.
16135 (output_asm_line_debug_info): Use option variables.
16136 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
16137 (output_loc_list): Likewise.
16138 (add_high_low_attributes): Check option variables.
16139 Don't output entry view attribute in strict mode.
16140 (gen_inlined_subroutine_die): Check option variables.
16141 (dwarf2out_inline_entry): Likewise.
16142 (init_sections_and_labels): Likewise.
16143 (dwarf2out_early_finish): Likewise.
16144 (maybe_reset_location_view): New, from...
16145 (dwarf2out_var_location): ... here. Call it.
16146 * debug.h (dwarf2out_default_as_loc_support): Declare.
16147 (dwarf2out_default_as_locview_support): Declare.
16148 * hooks.c (hook_int_rtx_insn_0): New.
16149 * hooks.h (hook_int_rtx_insn_0): Declare.
16150 * toplev.c (process_options): Take -gas-loc-support and
16151 -gas-locview-support from dwarf2out. Enable
16152 -gvariable-location-views by default only with locview
16153 assembler support. Enable -ginternal-reset-location-views by
16154 default only if the target defines the corresponding hook.
16155 Enable -ginline-points by default if location views are
16156 enabled; force it disabled if statement frontiers are
16158 * tree-inline.c (expand_call_inline): Check option variables.
16159 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
16161 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
16163 PR tree-optimization/84321
16164 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
16165 handling. Also check whether the anti-range contains any values
16166 that satisfy the mask; switch to a VR_RANGE if not.
16168 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
16171 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
16173 2018-02-13 Martin Jambor <mjambor@suse.cz>
16176 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
16177 of call statements, also set location of a load to a temporary.
16179 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
16181 * config/rl78/rl78.c (add_vector_labels): New function.
16182 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
16183 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
16184 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
16185 which checks that no arguments are passed.
16186 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
16187 * doc/extend.texi: Documentation for the new attribute.
16189 2018-02-13 Andreas Schwab <schwab@suse.de>
16191 * config/riscv/linux.h (CPP_SPEC): Define.
16193 2018-02-13 Jakub Jelinek <jakub@redhat.com>
16196 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
16197 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
16198 OPTION_MASK_ISA_AES as first argument to def_builtin_const
16199 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
16200 instead of OPTION_MASK_ISA_PCLMUL as first argument to
16201 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
16202 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
16203 temporarily for AES and PCLMUL builtins.
16205 PR tree-optimization/84339
16206 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
16207 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
16210 PR middle-end/84309
16211 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
16212 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
16213 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
16215 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
16217 * omp-simd-clone.h: New file.
16218 * omp-simd-clone.c: Include omp-simd-clone.h.
16219 (expand_simd_clones): No longer static.
16220 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
16221 cgraph.h and omp-simd-clone.h.
16222 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
16223 (vect_recog_widen_shift_pattern): Formatting fix.
16224 (vect_pattern_recog_1): Don't check optab for calls.
16227 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
16228 operands[2] into a REG before using gen_lowpart on it.
16230 2018-02-12 Jeff Law <law@redhat.com>
16233 * config/sh/sh.c (find_barrier): Consider a sibling call
16236 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
16237 successfully back substituting a reg.
16239 2018-02-12 Richard Biener <rguenther@suse.de>
16241 PR tree-optimization/84037
16242 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
16243 parameter, move visited init to caller.
16244 (vect_slp_analyze_operations): Separate cost from validity
16245 check, initialize visited once for all instances.
16246 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
16248 * tree-vect-stmts.c (vect_model_simple_cost): Make early
16250 (vect_model_promotion_demotion_cost): Likewise.
16251 (vectorizable_bswap): Guard cost modeling with !slp_node
16252 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
16254 (vectorizable_call): Likewise.
16255 (vectorizable_conversion): Likewise.
16256 (vectorizable_assignment): Likewise.
16257 (vectorizable_shift): Likewise.
16258 (vectorizable_operation): Likewise.
16259 (vectorizable_store): Likewise.
16260 (vectorizable_load): Likewise.
16261 (vectorizable_condition): Likewise.
16262 (vectorizable_comparison): Likewise.
16264 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
16267 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
16268 (ASAN_MARK): Fix fnspec to account for return value, change pointer
16269 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
16271 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
16273 PR middle-end/83665
16274 * params.def (inline-min-speedup): Increase from 8 to 15.
16275 (max-inline-insns-auto): Decrease from 40 to 30.
16276 * ipa-split.c (consider_split): Add some buffer for function to
16277 be considered inlining candidate.
16278 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
16281 2018-02-12 Richard Biener <rguenther@suse.de>
16283 PR tree-optimization/84037
16284 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
16285 matched stmts if we cannot swap the non-matched ones.
16287 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
16289 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
16290 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
16291 _mm_maskz_scalef_round_ss): New intrinsics.
16292 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
16293 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
16294 __builtin_ia32_scalefss_round): Remove.
16295 (__builtin_ia32_scalefsd_mask_round,
16296 __builtin_ia32_scalefss_mask_round): New intrinsics.
16297 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
16298 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
16299 ((match_operand:VF_128 2 "<round_nimm_predicate>"
16300 "<round_constraint>")): Changed to ...
16301 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
16302 "<round_scalar_constraint>")): ... this.
16303 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
16304 %0, %1, %2<round_op3>}"): Changed to ...
16305 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
16306 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
16307 %2<round_scalar_mask_op3>}"): ... this.
16308 * config/i386/subst.md (round_scalar_nimm_predicate): New.
16310 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
16312 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
16313 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
16314 (_mm_maskz_sqrt_round_ss): New intrinsics.
16315 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
16316 (__builtin_ia32_sqrtsd_mask_round)
16317 (__builtin_ia32_sqrtss_mask_round): New builtins.
16318 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
16319 (__builtin_ia32_sqrtss_round): Remove.
16320 (__builtin_ia32_sqrtsd_mask_round)
16321 (__builtin_ia32_sqrtss_mask_round): New builtins.
16322 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
16323 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
16324 ((match_operand:VF_128 1 "vector_operand"
16325 "xBm,<round_constraint>")): Changed to ...
16326 ((match_operand:VF_128 1 "vector_operand"
16327 "xBm,<round_scalar_constraint>")): ... this.
16328 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
16329 %0, %2, %<iptr>1<round_op3>}): Changed to ...
16330 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
16331 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
16332 %<iptr>1<round_scalar_mask_op3>}): ... this.
16333 ((set_attr "prefix" "<round_prefix>")): Changed to ...
16334 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
16336 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
16339 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
16340 Cast vec_cmpeq result to correct type.
16341 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
16342 Cast vec_cmpgt result to correct type.
16344 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
16346 * final.c (final_scan_insn_1): Renamed from...
16347 (final_scan_insn): ... this. New wrapper, to recover
16348 seen from the outermost call in recursive ones.
16349 * config/sparc/sparc.c (output_return): Drop seen from call.
16350 (output_sibcall): Likewise.
16351 * config/visium/visium.c (output_branch): Likewise.
16353 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
16355 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
16358 2018-02-10 Alan Modra <amodra@gmail.com>
16361 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
16362 Specify LR as an input.
16364 2018-02-10 Jakub Jelinek <jakub@redhat.com>
16367 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
16368 remove_member_access_dummy_vars): New functions.
16369 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
16370 lower_omp_1, execute_lower_omp): Use them.
16372 PR rtl-optimization/84308
16373 * shrink-wrap.c (spread_components): Release todo vector.
16375 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
16377 PR rtl-optimization/57193
16378 * ira-color.c (struct allocno_color_data): Add member
16379 conflict_allocno_hard_prefs.
16380 (update_conflict_allocno_hard_prefs): New.
16381 (bucket_allocno_compare_func): Add a preference based on
16382 conflict_allocno_hard_prefs.
16383 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
16384 (color_allocnos): Remove a dead code. Initiate
16385 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
16387 2018-02-09 Jakub Jelinek <jakub@redhat.com>
16390 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
16391 constraint from =wa to wa. Avoid a subreg on the output operand,
16392 instead use a pseudo and subreg it in a move.
16393 (p9_xxbrd_<mode>): Changed to ...
16394 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
16395 (p9_xxbrd_v2df): New expander.
16396 (p9_xxbrw_<mode>): Changed to ...
16397 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
16398 (p9_xxbrw_v4sf): New expander.
16400 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
16402 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
16404 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
16407 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
16408 multiply in 32-bit mode.
16409 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
16410 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
16413 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
16415 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
16416 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
16417 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
16418 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
16420 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
16423 * dwarf2out.c (is_trivial_indirect_ref): New function.
16424 (dwarf2out_late_global_decl): Do not generate a location
16425 attribute for variables that have a non-trivial DECL_VALUE_EXPR
16426 and that are not defined in the current unit.
16428 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
16430 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
16431 instead of a libcall for UNORDERED.
16433 2018-02-09 Tamar Christina <tamar.christina@arm.com>
16436 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
16437 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
16439 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
16442 * config/s390/s390.c (s390_set_current_function): Invoke
16443 s390_indirect_branch_settings also if fndecl didn't change.
16445 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
16447 * config/rs6000/rs6000.md (blockage): Set length to zero.
16449 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
16451 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
16453 2018-02-09 Jakub Jelinek <jakub@redhat.com>
16456 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
16457 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
16461 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
16462 PARALLEL incoming that failed vt_get_decl_and_offset check.
16464 PR middle-end/84237
16465 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
16466 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
16468 (get_variable_section): For decls in named .bss* sections pass true as
16469 second argument to bss_initializer_p.
16471 2018-02-09 Marek Polacek <polacek@redhat.com>
16472 Jakub Jelinek <jakub@redhat.com>
16475 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
16476 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
16477 Sync some changes from cxx_fold_indirect_ref.
16479 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
16481 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
16483 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
16484 (BLOCK_INLINE_ENTRY_LABEL): New.
16485 (dwarf2out_var_location): Disregard inline entry markers.
16486 (inline_entry_data): New struct.
16487 (inline_entry_data_hasher): New hashtable type.
16488 (inline_entry_data_hasher::hash): New.
16489 (inline_entry_data_hasher::equal): New.
16490 (inline_entry_data_table): New variable.
16491 (add_high_low_attributes): Add DW_AT_entry_pc and
16492 DW_AT_GNU_entry_view attributes if a pending entry is found
16493 in inline_entry_data_table. Add old entry_pc attribute only
16494 if debug nonbinding markers are disabled.
16495 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
16496 markers are enabled.
16497 (block_within_block_p, dwarf2out_inline_entry): New.
16498 (dwarf2out_finish): Check that no entries remained in
16499 inline_entry_data_table.
16500 * final.c (reemit_insn_block_notes): Handle inline entry notes.
16501 (final_scan_insn, notice_source_line): Likewise.
16502 (rest_of_clean_state): Skip inline entry markers.
16503 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
16505 * gimple.c (gimple_build_debug_inline_entry): New.
16506 * gimple.h (enum gimple_debug_subcode): Add
16507 GIMPLE_DEBUG_INLINE_ENTRY.
16508 (gimple_build_debug_inline_entry): Declare.
16509 (gimple_debug_inline_entry_p): New.
16510 (gimple_debug_nonbind_marker_p): Adjust.
16511 * insn-notes.def (INLINE_ENTRY): New.
16512 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
16513 inline entry marker notes.
16514 (print_insn): Likewise.
16515 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
16516 (INSN_DEBUG_MARKER_KIND): Likewise.
16517 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
16518 * tree-inline.c (expand_call_inline): Build and insert
16519 debug_inline_entry stmt.
16520 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
16521 inline entry blocks early, if nonbind markers are enabled.
16522 (dump_scope_block): Dump fragment info.
16523 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
16524 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
16525 (gimple_build_debug_inline_entry): New.
16526 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
16527 Enable/disable inline entry points too.
16528 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
16529 (DEBUG_INSN): Describe inline entry markers.
16531 * common.opt (gvariable-location-views): New.
16532 (gvariable-location-views=incompat5): New.
16533 * config.in: Rebuilt.
16534 * configure: Rebuilt.
16535 * configure.ac: Test assembler for view support.
16536 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
16537 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
16538 * dwarf2out.c (var_loc_view): New typedef.
16539 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
16540 (dwarf2out_locviews_in_attribute): New.
16541 (dwarf2out_locviews_in_loclist): New.
16542 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
16543 (enum dw_line_info_opcode): Add LI_adv_address.
16544 (struct dw_line_info_table): Add view.
16545 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
16546 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
16547 (zero_view_p): New variable.
16548 (ZERO_VIEW_P): New macro.
16549 (output_asm_line_debug_info): New.
16550 (struct var_loc_node): Add view.
16551 (add_AT_view_list, AT_loc_list): New.
16552 (add_var_loc_to_decl): Add view param. Test it against last.
16553 (new_loc_list): Add view params. Record them.
16554 (AT_loc_list_ptr): Handle loc and view lists.
16555 (view_list_to_loc_list_val_node): New.
16556 (print_dw_val): Handle dw_val_class_view_list.
16557 (size_of_die): Likewise.
16558 (value_format): Likewise.
16559 (loc_list_has_views): New.
16560 (gen_llsym): Set vl_symbol too.
16561 (maybe_gen_llsym, skip_loc_list_entry): New.
16562 (dwarf2out_maybe_output_loclist_view_pair): New.
16563 (output_loc_list): Output view list or entries too.
16564 (output_view_list_offset): New.
16565 (output_die): Handle dw_val_class_view_list.
16566 (output_dwarf_version): New.
16567 (output_compilation_unit_header): Use it.
16568 (output_skeleton_debug_sections): Likewise.
16569 (output_rnglists, output_line_info): Likewise.
16570 (output_pubnames, output_aranges): Update version comments.
16571 (output_one_line_info_table): Output view numbers in asm comments.
16572 (dw_loc_list): Determine current endview, pass it to new_loc_list.
16573 Call maybe_gen_llsym.
16574 (loc_list_from_tree_1): Adjust.
16575 (add_AT_location_description): Create view list attribute if
16576 needed, check it's absent otherwise.
16577 (convert_cfa_to_fb_loc_list): Adjust.
16578 (maybe_emit_file): Call output_asm_line_debug_info for test.
16579 (dwarf2out_var_location): Reset views as needed. Precompute
16580 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
16581 attribute. Set view.
16582 (new_line_info_table): Reset next view.
16583 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
16584 (dwarf2out_source_line): Likewise. Output view resets and labels to
16585 the assembler, or select appropriate line info opcodes.
16586 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
16587 (optimize_string_length): Catch it. Adjust.
16588 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
16589 dw_val_class_view_list, and remove it if no longer needed.
16590 (hash_loc_list): Hash view numbers.
16591 (loc_list_hasher::equal): Compare them.
16592 (optimize_location_lists): Check whether a view list symbol is
16593 needed, and whether the locview attribute is present, and
16594 whether they match. Remove the locview attribute if no longer
16596 (index_location_lists): Call skip_loc_list_entry for test.
16597 (dwarf2out_finish): Call output_asm_line_debug_info for test.
16598 Use output_dwarf_version.
16599 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
16600 (struct dw_val_node): Add val_view_list.
16601 * final.c (SEEN_NEXT_VIEW): New.
16602 (set_next_view_needed): New.
16603 (clear_next_view_needed): New.
16604 (maybe_output_next_view): New.
16605 (final_start_function): Rename to...
16606 (final_start_function_1): ... this. Take pointer to FIRST,
16607 add SEEN parameter. Emit param bindings in the initial view.
16608 (final_start_function): Reintroduce SEEN-less interface.
16609 (final): Rename to...
16610 (final_1): ... this. Take SEEN parameter. Output final pending
16611 next view at the end.
16612 (final): Reintroduce seen-less interface.
16613 (final_scan_insn): Output pending next view before switching
16614 sections or ending a block. Mark the next view as needed when
16615 outputting variable locations. Notify debug backend of section
16616 changes, and of location view changes.
16617 (rest_of_handle_final): Adjust.
16618 * toplev.c (process_options): Autodetect value for debug variable
16619 location views option. Warn on incompat5 without -gdwarf-5.
16620 * doc/invoke.texi (gvariable-location-views): New.
16621 (gvariable-location-views=incompat5): New.
16622 (gno-variable-location-views): New.
16624 2018-02-08 David Malcolm <dmalcolm@redhat.com>
16626 PR tree-optimization/84136
16627 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
16628 that the result of find_edge is non-NULL.
16630 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
16633 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
16634 storing integer register in SImode. Fix cost of 256 and 512
16635 byte aligned SSE register store.
16637 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
16639 * config/i386/i386.c (ix86_multiplication_cost): Fix
16640 multiplication cost for TARGET_AVX512DQ.
16642 2018-02-08 Marek Polacek <polacek@redhat.com>
16644 PR tree-optimization/84238
16645 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
16648 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
16650 PR tree-optimization/84265
16651 * tree-vect-stmts.c (vectorizable_store): Don't treat
16652 VMAT_CONTIGUOUS accesses as grouped.
16653 (vectorizable_load): Likewise.
16655 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
16657 PR tree-optimization/81635
16658 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
16659 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
16660 (test_round_for_mask): New functions.
16661 (wide_int_cc_tests): Call test_round_for_mask.
16662 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
16663 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
16664 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
16665 range returned by get_range_info.
16667 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
16670 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
16671 * symtab.c: Include builtins.h
16672 (symtab_node::output_to_lto_symbol_table_p): Move here
16673 from lto-streamer-out.c:output_symbol_p.
16674 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
16675 (output_symbol_p): Move all logic to symtab.c
16676 (produce_symtab): Update.
16678 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
16680 * config/s390/s390-opts.h (enum indirect_branch): Define.
16681 * config/s390/s390-protos.h (s390_return_addr_from_memory)
16682 (s390_indirect_branch_via_thunk)
16683 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
16684 (enum s390_indirect_branch_type): Define.
16685 * config/s390/s390.c (struct s390_frame_layout, struct
16686 machine_function): Remove.
16687 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
16688 (indirect_branch_table_label_no, indirect_branch_table_name):
16690 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
16691 (enum s390_indirect_branch_option): Define.
16692 (s390_return_addr_from_memory): New function.
16693 (s390_handle_string_attribute): New function.
16694 (s390_attribute_table): Add new attribute handler.
16695 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
16696 (s390_indirect_branch_via_thunk): New function.
16697 (s390_indirect_branch_via_inline_thunk): New function.
16698 (s390_function_ok_for_sibcall): When jumping via thunk disallow
16699 sibling call optimization for non z10 compiles.
16700 (s390_emit_call): Force indirect branch target to be a single
16701 register. Add r1 clobber for non-z10 compiles.
16702 (s390_emit_epilogue): Emit return jump via return_use expander.
16703 (s390_reorg): Handle JUMP_INSNs as execute targets.
16704 (s390_option_override_internal): Perform validity checks for the
16705 new command line options.
16706 (s390_indirect_branch_attrvalue): New function.
16707 (s390_indirect_branch_settings): New function.
16708 (s390_set_current_function): Invoke s390_indirect_branch_settings.
16709 (s390_output_indirect_thunk_function): New function.
16710 (s390_code_end): Implement target hook.
16711 (s390_case_values_threshold): Implement target hook.
16712 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
16714 * config/s390/s390.h (struct s390_frame_layout)
16715 (struct machine_function): Move here from s390.c.
16716 (TARGET_INDIRECT_BRANCH_NOBP_RET)
16717 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
16718 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
16719 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
16720 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
16721 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
16722 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
16723 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
16724 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
16725 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
16726 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
16727 (mnemonic attribute): Add values which aren't recognized
16729 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
16730 pattern for branch conversion. Fix mnemonic attribute.
16731 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
16732 indirect branch via thunk if requested.
16733 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
16734 ("*indirect_jump"): Disable for branch conversion using out of
16736 ("indirect_jump_via_thunk<mode>_z10")
16737 ("indirect_jump_via_thunk<mode>")
16738 ("indirect_jump_via_inlinethunk<mode>_z10")
16739 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
16740 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
16741 ("casesi_jump_via_inlinethunk<mode>_z10")
16742 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
16743 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
16744 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
16745 ("*indirect2_jump"): Disable for branch conversion.
16746 ("casesi_jump"): Turn into expander and expand patterns for branch
16748 ("return_use"): New expander.
16749 ("*return"): Emit return via thunk and rename it to ...
16750 ("*return<mode>"): ... this one.
16751 * config/s390/s390.opt: Add new options and and enum for the
16754 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
16756 * lra-constraints.c (match_reload): Unconditionally use
16757 gen_lowpart_SUBREG, rather than selecting between that
16758 and equivalent gen_rtx_SUBREG code.
16760 2018-02-08 Richard Biener <rguenther@suse.de>
16762 PR tree-optimization/84233
16763 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
16764 changed flag instead of boguously re-using phi_inserted.
16766 2018-02-08 Martin Jambor <mjambor@suse.cz>
16768 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
16769 static local variables.
16771 2018-02-08 Richard Biener <rguenther@suse.de>
16773 PR tree-optimization/84278
16774 * tree-vect-stmts.c (vectorizable_store): When looking for
16775 smaller vector types to perform grouped strided loads/stores
16776 make sure the mode is supported by the target.
16777 (vectorizable_load): Likewise.
16779 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
16781 * config/aarch64/aarch64.c (aarch64_components_for_bb):
16782 Increase LDP/STP opportunities by adding adjacent callee-saves.
16784 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
16786 PR rtl-optimization/84068
16787 PR rtl-optimization/83459
16788 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
16790 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
16792 PR tree-optimization/84224
16793 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
16794 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
16795 non-zero arguments.
16797 2018-02-07 Iain Sandoe <iain@codesourcery.com>
16800 * config/rs6000/altivec.md (*restore_world): Remove LR use.
16801 * config/rs6000/predicates.md (restore_world_operation): Adjust op
16802 count, remove one USE.
16804 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
16806 * doc/install.texi (Configuration): Document the
16807 --with-long-double-format={ibm,ieee} PowerPC configuration
16811 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
16812 Convert from define_expand to be define_insn_and_split. Rework
16813 float/double/_Float128 conversions to QI/HI/SImode to work with
16814 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
16815 conversions to QI/HImode types did a store and then a load to
16816 truncate the value. For conversions to VSX registers, don't split
16817 the insn, instead emit the code directly. Use the code iterator
16818 any_fix to combine signed and unsigned conversions.
16819 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
16820 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
16821 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
16822 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
16823 (fix_<mode>di2_hw): Likewise.
16824 (fixuns_<mode>di2_hw): Likewise.
16825 (fix_<mode>si2_hw): Likewise.
16826 (fixuns_<mode>si2_hw): Likewise.
16827 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
16828 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
16829 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
16830 fix<uns>_trunc<SFDF:mode>si2_p8.
16831 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
16832 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
16833 (fix<uns>_<mode>_mem): Likewise.
16834 (fctiw<u>z_<mode>_mem): Likewise.
16835 (fix<uns>_<mode>_mem): Likewise.
16836 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
16837 the register allocator from doing a direct move to the GPRs to do
16838 a store, and instead use the ISA 3.0 store byte/half-word from
16839 vector register instruction. For IEEE 128-bit floating point,
16840 also optimize stores of 32-bit ints.
16841 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
16843 2018-02-07 Alan Hayward <alan.hayward@arm.com>
16845 * genextract.c (push_pathstr_operand): New function to support
16847 (walk_rtx): Call push_pathstr_operand.
16848 (print_path): Support [a-zA-Z].
16850 2018-02-07 Richard Biener <rguenther@suse.de>
16852 PR tree-optimization/84037
16853 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
16854 (cse_and_gimplify_to_preheader): Declare.
16855 (vect_get_place_in_interleaving_chain): Likewise.
16856 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
16858 (_loop_vec_info::~_loop_vec_info): Delete it.
16859 (cse_and_gimplify_to_preheader): New function.
16860 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
16861 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
16862 (vectorizable_load): Likewise. For grouped stores always base
16863 the IV on the first element.
16864 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
16865 condition before gimplifying.
16867 2018-02-07 Jakub Jelinek <jakub@redhat.com>
16869 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
16870 *DIV_EXPR and *MOD_EXPR.
16872 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
16875 * config/i386/i386.c (ix86_option_override_internal): Mask out
16876 the CF_SET bit when checking -fcf-protection.
16878 2018-02-07 Tom de Vries <tom@codesourcery.com>
16881 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
16884 2018-02-07 Richard Biener <rguenther@suse.de>
16886 PR tree-optimization/84204
16887 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
16890 PR tree-optimization/84205
16891 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
16892 special-case isl_ast_op_zdiv_r.
16894 PR tree-optimization/84223
16895 * graphite-scop-detection.c (gather_bbs::before_dom_children):
16896 Only add conditions from within the region.
16897 (gather_bbs::after_dom_children): Adjust.
16899 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
16902 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
16903 * config/avr/avr.md: Only post-reload split REG-REG moves if
16904 either register is GENERAL_REG_P.
16906 2018-02-07 Jakub Jelinek <jakub@redhat.com>
16908 PR tree-optimization/84235
16909 * tree-ssa-scopedtables.c
16910 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
16911 if the subtraction is performed in floating point type where NaNs are
16912 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
16913 build 1. Formatting fix.
16915 2018-02-06 Jakub Jelinek <jakub@redhat.com>
16918 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
16919 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
16920 and skip it regardless of bb boundaries. Use CALL_P macro,
16921 don't test INSN_P (insn) together with CALL_P or JUMP_P check
16922 unnecessarily, formatting fix.
16924 2018-02-06 Michael Collison <michael.collison@arm.com>
16926 * config/arm/thumb2.md:
16927 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
16928 (*thumb_mov_notscc): Ditto.
16930 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
16933 * config/rs6000/rs6000.md (su code attribute): Use "u" for
16934 unsigned_fix, not "s".
16936 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
16938 * configure.ac (gcc_fn_eh_frame_ro): New function.
16939 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
16940 correct .eh_frame permissions.
16941 * configure: Regenerate.
16943 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
16945 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
16946 irrelevant options.
16948 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16950 * config/rs6000/rs6000.c (rs6000_option_override_internal):
16951 Display warning message for -mno-speculate-indirect-jumps.
16953 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
16955 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
16957 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
16959 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
16961 PR tree-optimization/84225
16962 * tree-eh.c (find_trapping_overflow): Only call
16963 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
16965 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
16968 * config/i386/i386.c: Reimplement the check of possible options
16969 -mibt/-mshstk conbination. Change error messages.
16970 * doc/invoke.texi: Fix a typo: remove extra '='.
16972 2018-02-06 Marek Polacek <polacek@redhat.com>
16974 PR tree-optimization/84228
16975 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
16977 2018-02-06 Tamar Christina <tamar.christina@arm.com>
16980 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
16981 emitted arch directives.
16982 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
16983 __ARM_FEATURE_COPROC before changing architectures.
16985 2018-02-06 Richard Biener <rguenther@suse.de>
16987 * config/i386/i386.c (print_reg): Fix typo.
16988 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
16990 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
16992 * configure: Regenerate.
16994 2018-02-05 Martin Sebor <msebor@redhat.com>
16996 PR tree-optimization/83369
16997 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
17000 2018-02-05 Martin Liska <mliska@suse.cz>
17002 * doc/invoke.texi: Cherry-pick upstream r323995.
17004 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
17006 * ira.c (ira_init_register_move_cost): Adjust comment.
17008 2018-02-05 Martin Liska <mliska@suse.cz>
17010 PR gcov-profile/84137
17011 * doc/gcov.texi: Fix typo in documentation.
17013 2018-02-05 Martin Liska <mliska@suse.cz>
17015 PR gcov-profile/83879
17016 * doc/gcov.texi: Document necessity of --dynamic-list-data when
17017 using dlopen functionality.
17019 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
17021 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
17022 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
17023 _mm_maskz_range_ss, _mm_mask_range_round_ss,
17024 _mm_maskz_range_round_ss): New intrinsics.
17025 (__builtin_ia32_rangesd128_round)
17026 (__builtin_ia32_rangess128_round): Remove.
17027 (__builtin_ia32_rangesd128_mask_round,
17028 __builtin_ia32_rangess128_mask_round): New builtins.
17029 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
17030 __builtin_ia32_rangess128_round): Remove.
17031 (__builtin_ia32_rangesd128_mask_round,
17032 __builtin_ia32_rangess128_mask_round): New builtins.
17033 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
17034 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
17035 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
17036 "<round_saeonly_constraint>")): Changed to ...
17037 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
17038 "<round_saeonly_scalar_constraint>")): ... this.
17039 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
17040 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
17041 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
17042 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
17043 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
17045 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
17047 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
17049 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
17050 Remove all values except native, 8540 and 8548.
17052 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
17054 * config/i386/i386.c (ix86_output_function_return): Pass
17055 INVALID_REGNUM, instead of -1, as invalid register number to
17056 indirect_thunk_name and output_indirect_thunk.
17058 2018-02-02 Julia Koval <julia.koval@intel.com>
17060 * config.gcc: Add -march=icelake.
17061 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
17062 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
17063 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
17064 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
17065 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
17066 (processor_target_table): Add icelake.
17067 (ix86_option_override_internal): Handle new PTAs.
17068 (get_builtin_code_for_version): Handle icelake.
17069 (M_INTEL_COREI7_ICELAKE): New.
17070 (fold_builtin_cpu): Handle icelake.
17071 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
17072 * doc/invoke.texi: Add -march=icelake.
17074 2018-02-02 Julia Koval <julia.koval@intel.com>
17076 * config/i386/i386.c (ix86_option_override_internal): Change flags type
17077 to wide_int_bitmask.
17078 * wide-int-bitmask.h: New.
17080 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
17083 * config/i386/i386.md: Replace Pmode with word_mode in
17084 builtin_setjmp_setup and builtin_longjmp to support x32.
17086 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
17090 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
17092 (rs6000_supported_cpu_names): New static variable.
17093 (linux_cpu_translation_table): Likewise.
17094 (elf_platform) <cpu>: Define new static variable and use it.
17095 Translate kernel AT_PLATFORM name to canonical name if needed.
17096 Error if platform name is unknown.
17098 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
17101 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
17103 2018-02-01 Jeff Law <law@redhat.com>
17106 * config/i386/i386.c (release_scratch_register_on_entry): Add new
17107 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
17108 the scratch if RELEASE_VIA_POP is false.
17109 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
17110 If we have to save a temporary register, decrement SIZE appropriately.
17111 Pass new arguments to release_scratch_register_on_entry.
17112 (ix86_adjust_stack_and_probe): Likewise.
17113 (ix86_emit_probe_stack_range): Pass new arguments to
17114 release_scratch_register_on_entry.
17116 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
17118 PR rtl-optimization/84157
17119 * combine.c (change_zero_ext): Use REG_P predicate in
17120 front of HARD_REGISTER_P predicate.
17122 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
17124 * config/avr/avr.c (avr_option_override): Move disabling of
17125 -fdelete-null-pointer-checks to...
17126 * common/config/avr/avr-common.c (avr_option_optimization_table):
17129 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
17131 PR tree-optimization/81635
17132 * tree-data-ref.c (split_constant_offset_1): For types that
17133 wrap on overflow, try to use range info to prove that wrapping
17136 2018-02-01 Renlin Li <renlin.li@arm.com>
17139 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
17140 TAILCALL_ADDR_REGS.
17141 (aarch64_register_move_cost): Likewise.
17142 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
17143 TAILCALL_ADDR_REGS.
17144 (REG_CLASS_NAMES): Likewise.
17145 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
17146 TAILCALL_ADDR_REGS. Remove IP registers.
17147 * config/aarch64/aarch64.md (Ucs): Update register constraint.
17149 2018-02-01 Richard Biener <rguenther@suse.de>
17151 * domwalk.h (dom_walker::dom_walker): Add additional constructor
17152 for specifying RPO order and allow NULL for that.
17153 * domwalk.c (dom_walker::dom_walker): Likewise.
17154 (dom_walker::walk): Handle NULL RPO order.
17155 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
17157 (rewrite_update_dom_walker): Likewise.
17158 (mark_def_dom_walker): Likewise.
17160 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
17162 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
17163 (aarch64_maybe_expand_sve_subreg_move): Declare.
17164 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
17165 * config/aarch64/predicates.md (aarch64_any_register_operand): New
17167 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
17168 that are semantically a reverse operation.
17169 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
17170 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
17171 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
17173 (aarch64_can_change_mode_class): For big-endian, forbid changes
17174 between two SVE modes if they have different element sizes.
17176 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
17178 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
17179 the TImode handling for big-endian targets.
17181 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
17183 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
17184 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
17186 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
17187 Remove BSWAP handing for big-endian targets and use the form of
17188 LD1RQ appropariate for the mode.
17190 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
17192 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
17193 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
17194 duplicated element.
17196 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
17199 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
17200 check for operands that need to go through aarch64_sve_reload_be.
17202 2018-02-01 Jakub Jelinek <jakub@redhat.com>
17204 PR tree-optimization/81661
17205 PR tree-optimization/84117
17206 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
17207 * tree-eh.c: Include gimplify.h.
17208 (find_trapping_overflow, replace_trapping_overflow,
17209 rewrite_to_non_trapping_overflow): New functions.
17210 * tree-vect-loop.c: Include tree-eh.h.
17211 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
17212 * tree-data-ref.c: Include tree-eh.h.
17213 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
17215 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
17217 PR rtl-optimization/84123
17218 * combine.c (change_zero_ext): Check if hard register satisfies
17219 can_change_dest_mode before calling gen_lowpart_SUBREG.
17221 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
17224 * ira.c (ira_init_register_move_cost): Remove assert.
17226 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
17228 PR rtl-optimization/84071
17229 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
17230 * doc/tm.texi: Regenerate.
17232 2018-01-31 Richard Biener <rguenther@suse.de>
17234 PR tree-optimization/84132
17235 * tree-data-ref.c (analyze_miv_subscript): Properly
17236 check whether evolution_function_is_affine_multivariate_p
17237 before calling gcd_of_steps_may_divide_p.
17239 2018-01-31 Julia Koval <julia.koval@intel.com>
17242 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
17243 * config/i386/i386.md (rdpid_rex64) New.
17244 (rdpid): Make 32bit only.
17246 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
17249 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
17250 an IDENTIFIER_NODE for FUNCTION_TYPE's.
17252 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
17255 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
17257 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
17259 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
17261 PR rtl-optimization/84071
17262 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
17263 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
17265 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
17267 * config/arc/arc.c (arc_handle_aux_attribute): New function.
17268 (arc_attribute_table): Add 'aux' attribute.
17269 (arc_in_small_data_p): Consider aux like variables.
17270 (arc_is_aux_reg_p): New function.
17271 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
17272 (arc_get_aux_arg): New function.
17273 (prepare_move_operands): Handle aux-register access.
17274 (arc_handle_aux_attribute): New function.
17275 * doc/extend.texi (ARC Variable attributes): Add subsection.
17277 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
17279 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
17280 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
17281 (arc_attribute_table): Add 'uncached' attribute.
17282 (arc_print_operand): Print '.di' flag for uncached memory
17284 (arc_in_small_data_p): Do not consider for small data the uncached
17286 (arc_is_uncached_mem_p): New function.
17287 * config/arc/predicates.md (compact_store_memory_operand): Check
17288 for uncached memory accesses.
17289 (nonvol_nonimm_operand): Likewise.
17290 * doc/extend.texi (ARC Type Attribute): New subsection.
17292 2018-01-31 Jakub Jelinek <jakub@redhat.com>
17295 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
17296 falign-loops=): Add Optimization flag.
17298 2018-01-30 Jeff Law <law@redhat.com>
17301 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
17302 INT_REGISTERS_SAVED. Check it prior to calling
17303 get_scratch_register_on_entry.
17304 (ix86_adjust_stack_and_probe): Similarly.
17305 (ix86_emit_probe_stack_range): Similarly.
17306 (ix86_expand_prologue): Corresponding changes.
17308 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
17311 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
17312 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
17314 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
17317 * lra-constraints.c (curr_insn_transform): Process AND in the
17320 2018-01-30 Jakub Jelinek <jakub@redhat.com>
17322 PR rtl-optimization/83986
17323 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
17324 dependence against last_pending_memory_flush in addition to
17325 pending_jump_insns.
17327 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
17329 PR tree-optimization/81611
17330 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
17333 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17336 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
17339 2018-01-30 Richard Biener <rguenther@suse.de>
17340 Jakub Jelinek <jakub@redhat.com>
17342 PR tree-optimization/84111
17343 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
17344 inner loops added during recursion, as they don't have up-to-date
17347 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
17350 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
17351 (can_inline_edge_by_limits_p): ... here.
17352 (can_early_inline_edge_p, check_callers,
17353 update_caller_keys, update_callee_keys, recursive_inlining,
17354 add_new_edges_to_heap, speculation_useful_p,
17355 inline_small_functions,
17356 inline_small_functions, flatten_function,
17357 inline_to_all_callers_1): Update.
17359 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
17361 * profile-count.c (profile_count::combine_with_ipa_count): Handle
17364 2018-01-30 Richard Biener <rguenther@suse.de>
17366 PR tree-optimization/83008
17367 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
17368 invariant and constant vector uses in stmts when they need
17369 more than one stmt.
17371 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
17374 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
17375 * configure: Regenerate.
17377 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
17379 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
17381 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
17382 Use gen_rtx_REG rather than gen_lowpart.
17384 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
17386 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
17387 rather than 0 when creating partial subregs.
17389 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
17391 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
17394 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
17397 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
17398 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
17399 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
17400 flags. This restores the settings used before the 2017-07-24.
17401 Turning off pre increment/decrement/modify allows IVOPTS to
17402 optimize DF/SF loops where the index is an int.
17404 2018-01-29 Richard Biener <rguenther@suse.de>
17405 Kelvin Nilsen <kelvin@gcc.gnu.org>
17408 * tree-vect-stmts.c (vectorizable_call): Don't call
17409 targetm.vectorize_builtin_md_vectorized_function if callee is
17412 2018-01-22 Carl Love <cel@us.ibm.com>
17414 * doc/extend.tex: Fix typo in second arg in
17415 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
17417 2018-01-29 Richard Biener <rguenther@suse.de>
17419 PR tree-optimization/84086
17420 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
17421 (flush_ssaname_freelist): When SSA names were released reset
17422 the SCEV hash table.
17424 2018-01-29 Richard Biener <rguenther@suse.de>
17426 PR tree-optimization/84057
17427 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
17428 removed paths when removing edges.
17430 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
17432 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
17433 -mfunction-return=@var{choice}.
17435 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
17437 PR diagnostic/84034
17438 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
17439 Handle CR like TAB.
17440 (layout::print_source_line): Likewise.
17441 (test_get_line_width_without_trailing_whitespace): Add test cases.
17443 2018-01-27 Jakub Jelinek <jakub@redhat.com>
17445 PR middle-end/84040
17446 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
17449 2018-01-26 Jim Wilson <jimw@sifive.com>
17451 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
17453 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
17456 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17458 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
17459 and CMP + SUB-immediate -> SUBS.
17461 2018-01-26 Martin Sebor <msebor@redhat.com>
17463 PR tree-optimization/83896
17464 * tree-ssa-strlen.c (get_string_len): Rename...
17465 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
17466 Avoid assuming length is constant.
17467 (handle_char_store): Use HOST_WIDE_INT for string length.
17469 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
17472 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
17473 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
17475 2018-01-26 Richard Biener <rguenther@suse.de>
17477 PR rtl-optimization/84003
17478 * dse.c (record_store): Only record redundant stores when
17479 the earlier store aliases at least all accesses the later one does.
17481 2018-01-26 Jakub Jelinek <jakub@redhat.com>
17483 PR rtl-optimization/83985
17484 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
17485 REG_CFA_RESTORE insns.
17486 (delete_unmarked_insns): Don't ignore separate shrink wrapping
17487 REG_CFA_RESTORE insns here.
17490 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
17491 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
17493 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
17495 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
17496 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
17497 (arc_init): Likewise.
17498 (arc_override_options): Likewise.
17499 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
17501 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
17503 * config/arc/arc.h (TARGET_DBNZ): Define.
17504 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
17505 properly set the tune attribute.
17506 (dbnz): Use TARGET_DBNZ guard.
17507 * config/arc/arc.opt (mtune): Add core3 option.
17509 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
17511 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
17512 recognize new pic like addresses.
17513 (arc_delegitimize_address): Clean up.
17515 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
17517 * config/arc/arc-arches.def: Option mrf16 valid for all
17519 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
17520 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
17521 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
17522 * config/arc/arc-tables.opt: Regenerate.
17523 * config/arc/arc.c (arc_conditional_register_usage): Handle
17524 reduced register file case.
17525 (arc_file_start): Set must have build attributes.
17526 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
17527 mrf16 option value.
17528 * config/arc/arc.opt (mrf16): Add new option.
17529 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
17530 * config/arc/genmultilib.awk: Handle new mrf16 option.
17531 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
17532 * config/arc/t-multilib: Regenerate.
17533 * doc/invoke.texi (ARC Options): Document mrf16 option.
17535 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
17537 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
17538 * config/arc/arc.c (arc_handle_secure_attribute): New function.
17539 (arc_attribute_table): Add 'secure_call' attribute.
17540 (arc_print_operand): Print secure call operand.
17541 (arc_function_ok_for_sibcall): Don't optimize tail calls when
17543 (arc_is_secure_call_p): New function. * config/arc/arc.md
17544 (call_i): Add support for sjli instruction.
17545 (call_value_i): Likewise.
17546 * config/arc/constraints.md (Csc): New constraint.
17548 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
17549 John Eric Martin <John.Martin@emmicro-us.com>
17551 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
17552 * config/arc/arc.c (_arc_jli_section): New struct.
17553 (arc_jli_section): New type.
17554 (rc_jli_sections): New static variable.
17555 (arc_handle_jli_attribute): New function.
17556 (arc_attribute_table): Add jli_always and jli_fixed attribute.
17557 (arc_file_end): New function.
17558 (TARGET_ASM_FILE_END): Define.
17559 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
17560 (arc_add_jli_section): New function.
17561 (jli_call_scan): Likewise.
17562 (arc_reorg): Call jli_call_scan.
17563 (arc_output_addsi): Remove 'S' from printing asm operand.
17564 (arc_is_jli_call_p): New function.
17565 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
17567 (movhi_insn): Likewise.
17568 (movsi_insn): Likewise.
17569 (movsi_set_cc_insn): Likewise.
17570 (loadqi_update): Likewise.
17571 (load_zeroextendqisi_update): Likewise.
17572 (load_signextendqisi_update): Likewise.
17573 (loadhi_update): Likewise.
17574 (load_zeroextendhisi_update): Likewise.
17575 (load_signextendhisi_update): Likewise.
17576 (loadsi_update): Likewise.
17577 (loadsf_update): Likewise.
17578 (movsicc_insn): Likewise.
17579 (bset_insn): Likewise.
17580 (bxor_insn): Likewise.
17581 (bclr_insn): Likewise.
17582 (bmsk_insn): Likewise.
17583 (bicsi3_insn): Likewise.
17584 (cmpsi_cc_c_insn): Likewise.
17585 (movsi_ne): Likewise.
17586 (movsi_cond_exec): Likewise.
17587 (clrsbsi2): Likewise.
17588 (norm_f): Likewise.
17598 (call_i): Remove 'S' asm letter, add jli instruction.
17599 (call_value_i): Likewise.
17600 * config/arc/arc.op (mjli-always): New option.
17601 * config/arc/constraints.md (Cji): New constraint.
17602 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
17604 (subsf3_fpx): Likewise.
17605 (mulsf3_fpx): Likewise.
17606 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
17608 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
17609 function attrbutes.
17610 * doc/invoke.texi (ARC): Document mjli-always option.
17612 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
17614 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
17615 avoid addition with 0 and use incw and decw where possible.
17617 2018-01-26 Richard Biener <rguenther@suse.de>
17619 PR tree-optimization/81082
17620 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
17621 association if it requires casting to unsigned.
17622 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
17623 from fold_plusminus_mult_expr to catch important cases late when
17624 range info is available.
17626 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
17628 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
17629 * configure.ac (hidden_linkonce): New test.
17630 * configure: Regenerate.
17631 * config.in: Regenerate.
17633 2018-01-26 Julia Koval <julia.koval@intel.com>
17635 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
17636 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
17637 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
17638 _mm_mask_bitshuffle_epi64_mask): Fix type.
17639 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
17640 USI_FTYPE_V4DI_V4DI_USI): Remove.
17641 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
17642 __builtin_ia32_vpshufbitqmb256_mask,
17643 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
17644 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
17645 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
17647 2018-01-26 Alan Modra <amodra@gmail.com>
17650 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
17651 UNSPEC_VBPERMQ. Sort other unspecs.
17653 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
17655 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
17657 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
17659 PR middle-end/83055
17660 * predict.c (drop_profile): Do not push/pop cfun; update also
17662 (handle_missing_profiles): Fix logic looking for zero profiles.
17664 2018-01-25 Jakub Jelinek <jakub@redhat.com>
17666 PR middle-end/83977
17667 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
17668 on functions with #pragma omp declare simd or functions with simd
17670 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
17671 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
17672 Remove trailing \n from warning_at calls.
17674 2018-01-25 Tom de Vries <tom@codesourcery.com>
17677 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
17678 for neutered workers.
17680 2018-01-24 Joseph Myers <joseph@codesourcery.com>
17683 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
17684 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
17686 2018-01-24 Jeff Law <law@redhat.com>
17689 * i386.c (get_probe_interval): Move to earlier point.
17690 (ix86_compute_frame_layout): If -fstack-clash-protection and
17691 the frame is larger than the probe interval, then use pushes
17692 to save registers rather than reg->mem moves.
17693 (ix86_expand_prologue): Remove conditional for int_registers_saved
17696 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
17699 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
17700 min/max for never referenced object.
17702 2018-01-24 Jakub Jelinek <jakub@redhat.com>
17704 PR middle-end/83977
17705 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
17707 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
17708 attributes from DECL_ATTRIBUTES (decl) without affecting
17709 DECL_ATTRIBUTES (current_function_decl).
17710 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
17711 functions with non-NULL DECL_ABSTRACT_ORIGIN.
17713 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
17715 PR tree-optimization/83979
17716 * fold-const.c (fold_comparison): Use constant_boolean_node
17717 instead of boolean_{true,false}_node.
17719 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
17721 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
17724 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
17726 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
17727 Simplify the clause that sets the length attribute.
17728 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
17729 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
17730 clause that sets the length attribute.
17731 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
17733 2018-01-24 Tom de Vries <tom@codesourcery.com>
17736 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
17737 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
17738 Add strict parameter.
17739 (prevent_branch_around_nothing): Insert dummy insn between branch to
17740 label and label with no ptx insn inbetween.
17741 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
17743 2018-01-24 Tom de Vries <tom@codesourcery.com>
17746 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
17747 for neutered threads in warp.
17748 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
17750 2018-01-24 Richard Biener <rguenther@suse.de>
17752 PR tree-optimization/83176
17753 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
17756 2018-01-24 Richard Biener <rguenther@suse.de>
17758 PR tree-optimization/82819
17759 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
17760 code generating pluses that are no-ops in the target precision.
17762 2018-01-24 Richard Biener <rguenther@suse.de>
17764 PR middle-end/84000
17765 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
17767 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
17769 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
17770 to merge probabilities.
17771 * predict.c (probably_never_executed): Also mark as cold functions
17772 with global 0 profile and guessed local profile.
17773 * profile-count.c (profile_probability::combine_with_count): New
17775 * profile-count.h (profile_probability::operator*,
17776 profile_probability::operator*=, profile_probability::operator/,
17777 profile_probability::operator/=): Reduce precision to adjusted
17778 and set value to guessed on contradictory divisions.
17779 (profile_probability::combine_with_freq): Remove.
17780 (profile_probability::combine_wiht_count): Declare.
17781 (profile_count::force_nonzero):: Set to adjusted.
17782 (profile_count::probability_in):: Set quality to adjusted.
17783 * tree-ssa-tail-merge.c (replace_block_by): Use
17784 combine_with_count.
17786 2018-01-23 Andrew Waterman <andrew@sifive.com>
17787 Jim Wilson <jimw@sifive.com>
17789 * config/riscv/riscv.c (riscv_stack_boundary): New.
17790 (riscv_option_override): Set riscv_stack_boundary. Handle
17791 riscv_preferred_stack_boundary_arg.
17792 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
17793 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
17794 (STACK_BOUNDARY): Set to riscv_stack_boundary.
17795 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
17796 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
17797 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
17799 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
17802 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
17803 of struct ix86_frame.
17804 (ix86_expand_epilogue): Likewise. Add a local variable for
17805 the reg_save_offset field in struct ix86_frame.
17807 2018-01-23 Bin Cheng <bin.cheng@arm.com>
17809 PR tree-optimization/82604
17810 * tree-loop-distribution.c (enum partition_kind): New enum item
17811 PKIND_PARTIAL_MEMSET.
17812 (partition_builtin_p): Support above new enum item.
17813 (generate_code_for_partition): Ditto.
17814 (compute_access_range): Differentiate cases that equality can be
17815 proven at all loops, the innermost loops or no loops.
17816 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
17817 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
17818 (finalize_partitions, distribute_loop): Don't fuse partition of
17819 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
17820 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
17821 parloop is enabled.
17823 2018-01-23 Martin Liska <mliska@suse.cz>
17825 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
17826 order to ignore the predictor.
17827 (PRED_POLYMORPHIC_CALL): Likewise.
17828 (PRED_RECURSIVE_CALL): Likewise.
17830 2018-01-23 Martin Liska <mliska@suse.cz>
17832 * tree-profile.c (tree_profiling): Print function header to
17833 aware reader which function we are working on.
17834 * value-prof.c (gimple_find_values_to_profile): Do not print
17835 not interesting value histograms.
17837 2018-01-23 Martin Liska <mliska@suse.cz>
17839 * profile-count.h (enum profile_quality): Add
17840 profile_uninitialized as the first value. Do not number values
17841 as they are zero based.
17842 (profile_count::verify): Update sanity check.
17843 (profile_probability::verify): Likewise.
17845 2018-01-23 Nathan Sidwell <nathan@acm.org>
17847 * doc/invoke.texi (ffor-scope): Deprecate.
17849 2018-01-23 David Malcolm <dmalcolm@redhat.com>
17851 PR tree-optimization/83510
17852 * domwalk.c (set_all_edges_as_executable): New function.
17853 (dom_walker::dom_walker): Convert bool param
17854 "skip_unreachable_blocks" to enum reachability. Move setup of
17855 edge flags to set_all_edges_as_executable and only do it when
17856 reachability is REACHABLE_BLOCKS.
17857 * domwalk.h (enum dom_walker::reachability): New enum.
17858 (dom_walker::dom_walker): Convert bool param
17859 "skip_unreachable_blocks" to enum reachability.
17860 (set_all_edges_as_executable): New decl.
17861 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
17862 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
17864 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
17865 but converting true to REACHABLE_BLOCKS.
17866 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
17868 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
17869 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
17870 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
17872 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
17873 if check_all_array_refs will be called.
17875 2018-01-23 David Malcolm <dmalcolm@redhat.com>
17877 * tree.c (selftest::test_location_wrappers): Add more test
17880 2018-01-23 David Malcolm <dmalcolm@redhat.com>
17882 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
17883 (selftest::test_bit_in_range): Likewise.
17885 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
17888 * doc/sourcebuild.texi (vect_float): Say that the selector
17889 only describes the situation when -funsafe-math-optimizations is on.
17890 (vect_float_strict): Document.
17892 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
17894 PR tree-optimization/83965
17895 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
17896 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
17897 instead of checking only for a reduction.
17898 (vect_recog_widen_sum_pattern): Likewise.
17900 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
17902 * predict.c (probably_never_executed): Only use precise profile info.
17903 (compute_function_frequency): Skip after inlining hack since we now
17904 have quality checking.
17906 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
17908 * profile-count.h (profile_probability::very_unlikely,
17909 profile_probability::unlikely, profile_probability::even): Set
17910 precision to guessed.
17912 2018-01-23 Richard Biener <rguenther@suse.de>
17914 PR tree-optimization/83963
17915 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
17916 Properly terminate dominator walk when crossing the exit edge not
17917 when visiting its source block.
17919 2018-01-23 Jakub Jelinek <jakub@redhat.com>
17922 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
17923 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
17925 2018-01-22 Jakub Jelinek <jakub@redhat.com>
17927 PR tree-optimization/83957
17928 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
17929 semicolon after for body surrounded by braces.
17931 PR tree-optimization/83081
17932 * profile-count.h (profile_probability::split): New method.
17933 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
17934 Use profile_probability::split.
17935 (do_compare_rtx_and_jump): Fix adjustment of probabilities
17936 when splitting a single conditional jump into 2.
17938 2018-01-22 David Malcolm <dmalcolm@redhat.com>
17940 PR tree-optimization/69452
17941 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
17944 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
17946 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
17947 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
17948 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
17950 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
17952 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
17954 * config/rl78/rl78.md (movdi): New define_expand.
17955 * config/rl78/rl78.c (rl78_split_movdi): New function.
17957 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
17960 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
17962 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
17963 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
17964 128-bit to produce an UNSPEC move to get the double word with the
17965 signbit and then a shift directly to do signbit.
17966 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
17967 implementation with a new version that just does either a direct
17968 move or a regular move. Move memory interface to separate insns.
17969 Move insns so they are next to the expander.
17970 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
17971 with signbit move. Split big and little endian case.
17972 (signbit<mode>2_dm_mem_le): Likewise.
17973 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
17974 (signbit<mode>2_dm2): Likewise.
17976 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
17978 * config/rl78/rl78.md (anddi3): New define_expand.
17980 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
17982 * config/rl78/rl78.md (umindi3): New define_expand.
17984 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
17986 * config/rl78/rl78.md (smindi3): New define_expand.
17988 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
17990 * config/rl78/rl78.md (smaxdi3): New define_expand.
17992 2018-01-22 Carl Love <cel@us.ibm.com>
17994 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
17995 LVX_V1TI): Add macro expansion.
17996 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
17997 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
17998 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
17999 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
18000 Change check to determine if the instruction is a byte reversing
18001 entry. Fix typo in comment.
18002 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
18003 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
18004 Add def_builtin calls for new builtins.
18005 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
18006 Add define_insn expansion.
18008 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
18010 * config/rl78/rl78.md (umaxdi3): New define_expand.
18012 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
18014 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
18015 for non-QImode registers.
18017 2018-01-22 Richard Biener <rguenther@suse.de>
18019 PR tree-optimization/83963
18020 * graphite-scop-detection.c (scop_detection::get_sese): Delay
18021 including the loop exit block.
18022 (scop_detection::merge_sese): Likewise.
18023 (scop_detection::add_scop): Do it here instead.
18025 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18027 * doc/sourcebuild.texi (arm_softfloat): Document.
18029 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
18032 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
18033 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
18034 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
18036 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
18037 David Edelsohn <dje.gcc@gmail.com>
18040 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
18041 Change "crset eq" to "crset 2".
18042 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
18043 (*call_indirect_aix<mode>_nospec): Likewise.
18044 (*call_value_indirect_aix<mode>_nospec): Likewise.
18045 (*call_indirect_elfv2<mode>_nospec): Likewise.
18046 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
18047 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
18048 change assembly output from . to $.
18049 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
18050 (indirect_jump<mode>_nospec): Change assembly output from . to $.
18051 (*tablejump<mode>_internal1_nospec): Likewise.
18053 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
18056 * config/sh/sh_optimize_sett_clrt.cc:
18057 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
18059 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
18061 PR tree-optimization/83940
18062 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
18063 offset_dt to vect_constant_def rather than vect_unknown_def_type.
18064 (vect_check_load_store_mask): Add a mask_dt_out parameter and
18065 use it to pass back the definition type.
18066 (vect_check_store_rhs): Likewise rhs_dt_out.
18067 (vect_build_gather_load_calls): Add a mask_dt argument and use
18068 it instead of a call to vect_is_simple_use.
18069 (vectorizable_store): Update calls to vect_check_load_store_mask
18070 and vect_check_store_rhs. Use the dt returned by the latter instead
18071 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
18072 instead of calls to vect_is_simple_use. Pass the scalar rather
18073 than the vector operand to vect_is_simple_use when handling
18074 second and subsequent copies of an rhs value.
18075 (vectorizable_load): Update calls to vect_check_load_store_mask
18076 and vect_build_gather_load_calls. Use the cached mask_dt and
18077 gs_info.offset_dt instead of calls to vect_is_simple_use.
18079 2018-01-20 Jakub Jelinek <jakub@redhat.com>
18081 PR middle-end/83945
18082 * tree-emutls.c: Include gimplify.h.
18083 (lower_emutls_2): New function.
18084 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
18085 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
18086 it before further processing.
18089 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
18090 UINTVAL (trueop1) instead of INTVAL (op1).
18092 2018-01-19 Jakub Jelinek <jakub@redhat.com>
18096 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
18097 INCOMING_FRAME_SP_OFFSET if not defined.
18098 (scan_trace): Add ENTRY argument. If true and
18099 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
18100 emit a note to adjust the CFA offset.
18101 (create_cfi_notes): Adjust scan_trace callers.
18102 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
18103 INCOMING_FRAME_SP_OFFSET in the CIE.
18104 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
18105 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
18107 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
18108 * doc/tm.texi: Regenerated.
18110 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
18112 PR rtl-optimization/83147
18113 * lra-constraints.c (remove_inheritance_pseudos): Use
18114 lra_substitute_pseudo_within_insn.
18116 2018-01-19 Tom de Vries <tom@codesourcery.com>
18117 Cesar Philippidis <cesar@codesourcery.com>
18120 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
18122 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
18125 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
18126 spaces for function labels.
18128 2018-01-19 Martin Liska <mliska@suse.cz>
18130 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
18131 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
18132 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
18133 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
18134 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
18135 (PRED_CONST_RETURN): Change from 69 to 65.
18136 (PRED_NULL_RETURN): Change from 91 to 71.
18137 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
18138 (PRED_LOOP_GUARD): Change from 66 to 73.
18140 2018-01-19 Martin Liska <mliska@suse.cz>
18142 * predict.c (predict_insn_def): Add new assert.
18143 (struct branch_predictor): Change type to signed integer.
18144 (test_prediction_value_range): Amend test to cover
18145 PROB_UNINITIALIZED.
18146 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
18147 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
18148 (PRED_LOOP_ITERATIONS_MAX): Likewise.
18149 (PRED_LOOP_IV_COMPARE): Likewise.
18150 * predict.h (PROB_UNINITIALIZED): Define new constant.
18152 2018-01-19 Martin Liska <mliska@suse.cz>
18154 * predict.c (dump_prediction): Add new format for
18155 analyze_brprob.py script which is enabled with -details
18157 * profile-count.h (precise_p): New function.
18159 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
18161 PR tree-optimization/83922
18162 * tree-vect-loop.c (vect_verify_full_masking): Return false if
18163 there are no statements that need masking.
18164 (vect_active_double_reduction_p): New function.
18165 (vect_analyze_loop_operations): Use it when handling phis that
18166 are not in the loop header.
18168 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
18170 PR tree-optimization/83914
18171 * tree-vect-loop.c (vectorizable_induction): Don't convert
18172 init_expr or apply the peeling adjustment for inductions
18173 that are nested within the vectorized loop.
18175 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18177 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
18180 2018-01-18 Jakub Jelinek <jakub@redhat.com>
18184 * function.h (gimplify_parameters): Add gimple_seq * argument.
18185 * function.c: Include gimple.h and options.h.
18186 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
18187 for the added local temporaries if needed.
18188 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
18189 if there are any parameter cleanups, wrap whole body into a
18190 try/finally with the cleanups.
18192 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
18195 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
18196 Use GET_MODE_CLASS for scalar floating point.
18198 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
18202 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
18203 Fix call of call_cgraph_insertion_hooks.
18205 2018-01-18 Martin Sebor <msebor@redhat.com>
18207 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
18209 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
18212 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
18215 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
18218 * common.opt: (-ffile-prefix-map): New option.
18219 * opts.c (common_handle_option): Defer it.
18220 * opts-global.c (handle_common_deferred_options): Handle it.
18221 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
18222 * file-prefix-map.h: New file.
18223 (remap_debug_filename, add_debug_prefix_map): ...here.
18224 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
18225 * final.c (debug_prefix_map, add_debug_prefix_map
18226 remap_debug_filename): Move to...
18227 * file-prefix-map.c: New file.
18228 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
18229 generalize, get rid of alloca(), use strrchr() instead of strchr().
18230 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
18231 Implement in terms of add_prefix_map().
18232 (remap_macro_filename, remap_debug_filename): Implement in term of
18234 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
18235 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
18236 * dbxout.c: Include file-prefix-map.h.
18237 * varasm.c: Likewise.
18238 * vmsdbgout.c: Likewise.
18239 * xcoffout.c: Likewise.
18240 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
18241 * doc/cppopts.texi (-fmacro-prefix-map): Document.
18242 * doc/invoke.texi (-ffile-prefix-map): Document.
18243 (-fdebug-prefix-map): Update description.
18245 2018-01-18 Martin Liska <mliska@suse.cz>
18247 * config/i386/i386.c (indirect_thunk_name): Document that also
18249 (output_indirect_thunk): Document why both instructions
18250 (pause and lfence) are generated.
18252 2018-01-18 Richard Biener <rguenther@suse.de>
18254 PR tree-optimization/83887
18255 * graphite-scop-detection.c
18256 (scop_detection::get_nearest_dom_with_single_entry): Remove.
18257 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
18258 (scop_detection::merge_sese): Re-implement with a flood-fill
18259 algorithm that properly finds a SESE region if it exists.
18261 2018-01-18 Jakub Jelinek <jakub@redhat.com>
18264 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
18265 pointer_diff optimizations use view_convert instead of convert.
18267 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
18269 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
18270 Generate different code for -mno-speculate-indirect-jumps.
18271 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
18272 (*call_indirect_aix<mode>): Disable for
18273 -mno-speculate-indirect-jumps.
18274 (*call_indirect_aix<mode>_nospec): New define_insn.
18275 (*call_value_indirect_aix<mode>): Disable for
18276 -mno-speculate-indirect-jumps.
18277 (*call_value_indirect_aix<mode>_nospec): New define_insn.
18278 (*sibcall_nonlocal_sysv<mode>): Generate different code for
18279 -mno-speculate-indirect-jumps.
18280 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
18282 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
18284 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
18285 long double type, set the flags for noting the default long double
18286 type, even if we don't pass or return a long double type.
18288 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
18291 * ipa-inline.c (flatten_function): Do not overwrite final inlining
18294 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
18296 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
18297 support for merge[hl].
18298 (fold_mergehl_helper): New helper function.
18299 (tree-vector-builder.h): New #include for tree_vector_builder usage.
18300 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
18301 (altivec_vmrglw_direct): Add xxmrglw insn.
18303 2018-01-17 Andrew Waterman <andrew@sifive.com>
18305 * config/riscv/riscv.c (riscv_conditional_register_usage): If
18306 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
18308 2018-01-17 David Malcolm <dmalcolm@redhat.com>
18311 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
18312 call the lto_location_cache before reading the
18313 DECL_SOURCE_LOCATION of the types.
18315 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
18316 Richard Sandiford <richard.sandiford@linaro.org>
18318 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
18319 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
18320 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
18321 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
18322 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
18324 * config/aarch64/constraints.md (aarch64_movti_operand):
18326 * config/aarch64/predicates.md (Uti): Add new constraint.
18328 2018-01-17 Carl Love <cel@us.ibm.com>
18330 * config/rs6000/vsx.md (define_expand xl_len_r,
18331 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
18332 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
18334 (define_expand, define_insn): Move the shift left from the
18335 define_insn to the define_expand for lxvl and stxvl instructions.
18336 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
18337 and XL_LEN_R definitions to PURE.
18339 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
18341 * config/i386/i386.c (indirect_thunk_name): Declare regno
18342 as unsigned int. Compare regno with INVALID_REGNUM.
18343 (output_indirect_thunk): Ditto.
18344 (output_indirect_thunk_function): Ditto.
18345 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
18346 in the call to output_indirect_thunk_function.
18348 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
18350 PR middle-end/83884
18351 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
18352 rather than the size of inner_type to determine the stack slot size
18353 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
18355 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
18358 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
18361 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
18363 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
18364 endian Linux systems to optionally enable multilibs for selecting
18365 the long double type if the user configured an explicit type.
18366 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
18367 have no long double multilibs if not defined.
18368 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
18369 warn if the user used -mabi={ieee,ibm}longdouble and we built
18370 multilibs for long double.
18371 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
18372 appropriate multilib option.
18373 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
18375 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
18376 for building long double multilibs.
18377 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
18379 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
18381 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
18384 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
18386 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
18389 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
18392 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
18395 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
18397 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
18398 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
18400 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
18402 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
18403 different rtl trees depending on TARGET_64BIT.
18404 (rs6000_gen_lvx): Likewise.
18406 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
18408 * config/visium/visium.md (nop): Tweak comment.
18409 (hazard_nop): Likewise.
18411 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
18413 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
18414 -mspeculate-indirect-jumps.
18415 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
18416 for -mno-speculate-indirect-jumps.
18417 (*call_indirect_elfv2<mode>_nospec): New define_insn.
18418 (*call_value_indirect_elfv2<mode>): Disable for
18419 -mno-speculate-indirect-jumps.
18420 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
18421 (indirect_jump): Emit different RTL for
18422 -mno-speculate-indirect-jumps.
18423 (*indirect_jump<mode>): Disable for
18424 -mno-speculate-indirect-jumps.
18425 (*indirect_jump<mode>_nospec): New define_insn.
18426 (tablejump): Emit different RTL for
18427 -mno-speculate-indirect-jumps.
18428 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
18429 (tablejumpsi_nospec): New define_expand.
18430 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
18431 (tablejumpdi_nospec): New define_expand.
18432 (*tablejump<mode>_internal1): Disable for
18433 -mno-speculate-indirect-jumps.
18434 (*tablejump<mode>_internal1_nospec): New define_insn.
18435 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
18438 2018-01-16 Artyom Skrobov tyomitch@gmail.com
18440 * caller-save.c (insert_save): Drop unnecessary parameter. All
18443 2018-01-16 Jakub Jelinek <jakub@redhat.com>
18444 Richard Biener <rguenth@suse.de>
18447 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
18448 return early, inline manually is_gimple_sizepos. Make sure if we
18449 call gimplify_expr we don't end up with a gimple constant.
18450 * tree.c (variably_modified_type_p): Don't return true for
18451 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
18452 * gimplify.h (is_gimple_sizepos): Remove.
18454 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
18456 PR tree-optimization/83857
18457 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
18458 vectorizable_live_operation for pure SLP statements.
18459 (vectorizable_live_operation): Handle PHIs.
18461 2018-01-16 Richard Biener <rguenther@suse.de>
18463 PR tree-optimization/83867
18464 * tree-vect-stmts.c (vect_transform_stmt): Precompute
18465 nested_in_vect_loop_p since the scalar stmt may get invalidated.
18467 2018-01-16 Jakub Jelinek <jakub@redhat.com>
18470 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
18471 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
18472 If off is not INTEGER_CST, issue a may not be aligned warning
18473 rather than isn't aligned. Use isn%'t rather than isn't.
18474 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
18476 <case MULT_EXPR>: Improve the case when bottom and one of the
18477 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
18478 operand, in that case check if the other operand is multiple of
18479 bottom divided by the INTEGER_CST operand.
18481 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
18484 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
18485 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
18486 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
18487 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
18488 * config/pa/pa.c (pa_function_arg_advance): Likewise.
18489 (pa_function_arg, pa_arg_partial_bytes): Likewise.
18490 (pa_function_arg_size): New function.
18492 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
18494 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
18495 in a separate statement.
18497 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
18499 PR tree-optimization/83847
18500 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
18501 group gathers and scatters.
18503 2018-01-16 Jakub Jelinek <jakub@redhat.com>
18505 PR rtl-optimization/86620
18506 * params.def (max-sched-ready-insns): Bump minimum value to 1.
18508 PR rtl-optimization/83213
18509 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
18510 to last if both are JUMP_INSNs.
18512 PR tree-optimization/83843
18513 * gimple-ssa-store-merging.c
18514 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
18515 store_immediate_info for bswap/nop orig_stores.
18517 2018-01-15 Andrew Waterman <andrew@sifive.com>
18519 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
18521 <UDIV>: Increase cost if !TARGET_DIV.
18523 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
18525 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
18526 (define_attr "cr_logical_3op"): New.
18527 (cceq_ior_compare): Adjust.
18528 (cceq_ior_compare_complement): Adjust.
18529 (*cceq_rev_compare): Adjust.
18530 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
18531 (is_cracked_insn): Adjust.
18532 (insn_must_be_first_in_group): Adjust.
18533 * config/rs6000/40x.md: Adjust.
18534 * config/rs6000/440.md: Adjust.
18535 * config/rs6000/476.md: Adjust.
18536 * config/rs6000/601.md: Adjust.
18537 * config/rs6000/603.md: Adjust.
18538 * config/rs6000/6xx.md: Adjust.
18539 * config/rs6000/7450.md: Adjust.
18540 * config/rs6000/7xx.md: Adjust.
18541 * config/rs6000/8540.md: Adjust.
18542 * config/rs6000/cell.md: Adjust.
18543 * config/rs6000/e300c2c3.md: Adjust.
18544 * config/rs6000/e500mc.md: Adjust.
18545 * config/rs6000/e500mc64.md: Adjust.
18546 * config/rs6000/e5500.md: Adjust.
18547 * config/rs6000/e6500.md: Adjust.
18548 * config/rs6000/mpc.md: Adjust.
18549 * config/rs6000/power4.md: Adjust.
18550 * config/rs6000/power5.md: Adjust.
18551 * config/rs6000/power6.md: Adjust.
18552 * config/rs6000/power7.md: Adjust.
18553 * config/rs6000/power8.md: Adjust.
18554 * config/rs6000/power9.md: Adjust.
18555 * config/rs6000/rs64.md: Adjust.
18556 * config/rs6000/titan.md: Adjust.
18558 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
18560 * config/i386/predicates.md (indirect_branch_operand): Rewrite
18561 ix86_indirect_branch_register logic.
18563 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
18565 * config/i386/constraints.md (Bs): Update
18566 ix86_indirect_branch_register check. Don't check
18567 ix86_indirect_branch_register with GOT_memory_operand.
18569 * config/i386/predicates.md (GOT_memory_operand): Don't check
18570 ix86_indirect_branch_register here.
18571 (GOT32_symbol_operand): Likewise.
18573 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
18575 * config/i386/predicates.md (constant_call_address_operand):
18576 Rewrite ix86_indirect_branch_register logic.
18577 (sibcall_insn_operand): Likewise.
18579 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
18581 * config/i386/constraints.md (Bs): Replace
18582 ix86_indirect_branch_thunk_register with
18583 ix86_indirect_branch_register.
18585 * config/i386/i386.md (indirect_jump): Likewise.
18586 (tablejump): Likewise.
18587 (*sibcall_memory): Likewise.
18588 (*sibcall_value_memory): Likewise.
18589 Peepholes of indirect call and jump via memory: Likewise.
18590 * config/i386/i386.opt: Likewise.
18591 * config/i386/predicates.md (indirect_branch_operand): Likewise.
18592 (GOT_memory_operand): Likewise.
18593 (call_insn_operand): Likewise.
18594 (sibcall_insn_operand): Likewise.
18595 (GOT32_symbol_operand): Likewise.
18597 2018-01-15 Jakub Jelinek <jakub@redhat.com>
18599 PR middle-end/83837
18600 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
18601 type rather than type addr's type points to.
18602 (expand_omp_atomic_mutex): Likewise.
18603 (expand_omp_atomic): Likewise.
18605 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
18608 * config/i386/i386.c (output_indirect_thunk_function): Use
18609 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
18610 for __x86_return_thunk.
18612 2018-01-15 Richard Biener <rguenther@suse.de>
18614 PR middle-end/83850
18615 * expmed.c (extract_bit_field_1): Fix typo.
18617 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18620 * config/arm/iterators.md (VF): New mode iterator.
18621 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
18622 Remove integer-related logic from pattern.
18623 (neon_vabd<mode>_3): Likewise.
18625 2018-01-15 Jakub Jelinek <jakub@redhat.com>
18627 PR middle-end/82694
18628 * common.opt (fstrict-overflow): No longer an alias.
18629 (fwrapv-pointer): New option.
18630 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
18631 also for pointer types based on flag_wrapv_pointer.
18632 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
18633 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
18634 opts->x_flag_wrapv got set.
18635 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
18636 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
18637 POINTER_TYPE_OVERFLOW_UNDEFINED.
18638 * match.pd: Likewise in address comparison pattern.
18639 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
18641 2018-01-15 Richard Biener <rguenther@suse.de>
18644 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
18645 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
18646 Reset type names to their identifier if their TYPE_DECL doesn't
18647 have linkage (and thus is used for ODR and devirt).
18648 (save_debug_info_for_decl): Remove.
18649 (save_debug_info_for_type): Likewise.
18650 (add_tree_to_fld_list): Adjust.
18651 * tree-pretty-print.c (dump_generic_node): Make dumping of
18652 type names more robust.
18654 2018-01-15 Richard Biener <rguenther@suse.de>
18656 * BASE-VER: Bump to 8.0.1.
18658 2018-01-14 Martin Sebor <msebor@redhat.com>
18661 * builtins.c (check_access): Avoid warning when the no-warning bit
18664 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
18666 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
18667 * ira-color (allocno_hard_regs_compare): Likewise.
18669 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
18672 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
18673 Use .pushsection/.popsection.
18675 2018-01-14 Martin Sebor <msebor@redhat.com>
18678 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
18680 2018-01-14 Jakub Jelinek <jakub@redhat.com>
18682 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
18683 entry from extra_headers.
18684 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
18685 extra_headers, make the list bitwise identical to the i?86-*-* one.
18687 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
18689 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
18690 -mcmodel=large with -mindirect-branch=thunk,
18691 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
18692 -mfunction-return=thunk-extern.
18693 * doc/invoke.texi: Document -mcmodel=large is incompatible with
18694 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
18695 -mfunction-return=thunk and -mfunction-return=thunk-extern.
18697 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
18699 * config/i386/i386.c (print_reg): Print the name of the full
18700 integer register without '%'.
18701 (ix86_print_operand): Handle 'V'.
18702 * doc/extend.texi: Document 'V' modifier.
18704 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
18706 * config/i386/constraints.md (Bs): Disallow memory operand for
18707 -mindirect-branch-register.
18709 * config/i386/predicates.md (indirect_branch_operand): Likewise.
18710 (GOT_memory_operand): Likewise.
18711 (call_insn_operand): Likewise.
18712 (sibcall_insn_operand): Likewise.
18713 (GOT32_symbol_operand): Likewise.
18714 * config/i386/i386.md (indirect_jump): Call convert_memory_address
18715 for -mindirect-branch-register.
18716 (tablejump): Likewise.
18717 (*sibcall_memory): Likewise.
18718 (*sibcall_value_memory): Likewise.
18719 Disallow peepholes of indirect call and jump via memory for
18720 -mindirect-branch-register.
18721 (*call_pop): Replace m with Bw.
18722 (*call_value_pop): Likewise.
18723 (*sibcall_pop_memory): Replace m with Bs.
18724 * config/i386/i386.opt (mindirect-branch-register): New option.
18725 * doc/invoke.texi: Document -mindirect-branch-register option.
18727 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
18729 * config/i386/i386-protos.h (ix86_output_function_return): New.
18730 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
18731 set function_return_type.
18732 (indirect_thunk_name): Add ret_p to indicate thunk for function
18734 (output_indirect_thunk_function): Pass false to
18735 indirect_thunk_name.
18736 (ix86_output_indirect_branch_via_reg): Likewise.
18737 (ix86_output_indirect_branch_via_push): Likewise.
18738 (output_indirect_thunk_function): Create alias for function
18739 return thunk if regno < 0.
18740 (ix86_output_function_return): New function.
18741 (ix86_handle_fndecl_attribute): Handle function_return.
18742 (ix86_attribute_table): Add function_return.
18743 * config/i386/i386.h (machine_function): Add
18744 function_return_type.
18745 * config/i386/i386.md (simple_return_internal): Use
18746 ix86_output_function_return.
18747 (simple_return_internal_long): Likewise.
18748 * config/i386/i386.opt (mfunction-return=): New option.
18749 (indirect_branch): Mention -mfunction-return=.
18750 * doc/extend.texi: Document function_return function attribute.
18751 * doc/invoke.texi: Document -mfunction-return= option.
18753 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
18755 * config/i386/i386-opts.h (indirect_branch): New.
18756 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
18757 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
18758 with local indirect jump when converting indirect call and jump.
18759 (ix86_set_indirect_branch_type): New.
18760 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
18761 (indirectlabelno): New.
18762 (indirect_thunk_needed): Likewise.
18763 (indirect_thunk_bnd_needed): Likewise.
18764 (indirect_thunks_used): Likewise.
18765 (indirect_thunks_bnd_used): Likewise.
18766 (INDIRECT_LABEL): Likewise.
18767 (indirect_thunk_name): Likewise.
18768 (output_indirect_thunk): Likewise.
18769 (output_indirect_thunk_function): Likewise.
18770 (ix86_output_indirect_branch_via_reg): Likewise.
18771 (ix86_output_indirect_branch_via_push): Likewise.
18772 (ix86_output_indirect_branch): Likewise.
18773 (ix86_output_indirect_jmp): Likewise.
18774 (ix86_code_end): Call output_indirect_thunk_function if needed.
18775 (ix86_output_call_insn): Call ix86_output_indirect_branch if
18777 (ix86_handle_fndecl_attribute): Handle indirect_branch.
18778 (ix86_attribute_table): Add indirect_branch.
18779 * config/i386/i386.h (machine_function): Add indirect_branch_type
18780 and has_local_indirect_jump.
18781 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
18783 (tablejump): Likewise.
18784 (*indirect_jump): Use ix86_output_indirect_jmp.
18785 (*tablejump_1): Likewise.
18786 (simple_return_indirect_internal): Likewise.
18787 * config/i386/i386.opt (mindirect-branch=): New option.
18788 (indirect_branch): New.
18791 (thunk-inline): Likewise.
18792 (thunk-extern): Likewise.
18793 * doc/extend.texi: Document indirect_branch function attribute.
18794 * doc/invoke.texi: Document -mindirect-branch= option.
18796 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
18799 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
18801 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
18803 * ipa-inline.c (want_inline_small_function_p): Return false if
18804 inlining has already failed with CIF_FINAL_ERROR.
18805 (update_caller_keys): Call want_inline_small_function_p before
18807 (update_callee_keys): Likewise.
18809 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
18811 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
18813 (rs6000_quadword_masked_address_p): Likewise.
18814 (quad_aligned_load_p): Likewise.
18815 (quad_aligned_store_p): Likewise.
18816 (const_load_sequence_p): Add comment to describe the outer-most loop.
18817 (mimic_memory_attributes_and_flags): New function.
18818 (rs6000_gen_stvx): Likewise.
18819 (replace_swapped_aligned_store): Likewise.
18820 (rs6000_gen_lvx): Likewise.
18821 (replace_swapped_aligned_load): Likewise.
18822 (replace_swapped_load_constant): Capitalize argument name in
18823 comment describing this function.
18824 (rs6000_analyze_swaps): Add a third pass to search for vector loads
18825 and stores that access quad-word aligned addresses and replace
18826 with stvx or lvx instructions when appropriate.
18827 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
18828 New function prototype.
18829 (rs6000_quadword_masked_address_p): Likewise.
18830 (rs6000_gen_lvx): Likewise.
18831 (rs6000_gen_stvx): Likewise.
18832 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
18833 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
18834 when memory address is aligned.
18835 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
18836 this split to select lvx instruction when memory address is aligned.
18837 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
18838 instruction when memory address is aligned.
18839 (*vsx_le_perm_load_v16qi): Likewise.
18840 (four unnamed splitters): Modify to select the stvx instruction
18841 when memory is aligned.
18843 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
18845 * predict.c (determine_unlikely_bbs): Handle correctly BBs
18846 which appears in the queue multiple times.
18848 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18849 Alan Hayward <alan.hayward@arm.com>
18850 David Sherwood <david.sherwood@arm.com>
18852 * tree-vectorizer.h (vec_lower_bound): New structure.
18853 (_loop_vec_info): Add check_nonzero and lower_bounds.
18854 (LOOP_VINFO_CHECK_NONZERO): New macro.
18855 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
18856 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
18857 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
18858 fields. Make seg_len the distance travelled, not including the
18860 (dr_direction_indicator): Declare.
18861 (dr_zero_step_indicator): Likewise.
18862 (dr_known_forward_stride_p): Likewise.
18863 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
18865 (runtime_alias_check_p): Allow runtime alias checks with
18867 (operator ==): Compare access_size and align.
18868 (prune_runtime_alias_test_list): Rework for new distinction between
18869 the access_size and seg_len.
18870 (create_intersect_range_checks_index): Likewise. Cope with polynomial
18872 (get_segment_min_max): New function.
18873 (create_intersect_range_checks): Use it.
18874 (dr_step_indicator): New function.
18875 (dr_direction_indicator): Likewise.
18876 (dr_zero_step_indicator): Likewise.
18877 (dr_known_forward_stride_p): Likewise.
18878 * tree-loop-distribution.c (data_ref_segment_size): Return
18879 DR_STEP * (niters - 1).
18880 (compute_alias_check_pairs): Update call to the dr_with_seg_len
18882 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
18883 (vect_preserves_scalar_order_p): New function, split out from...
18884 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
18885 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
18886 (vect_vfa_access_size): New function.
18887 (vect_vfa_align): Likewise.
18888 (vect_compile_time_alias): Take access_size_a and access_b arguments.
18889 (dump_lower_bound): New function.
18890 (vect_check_lower_bound): Likewise.
18891 (vect_small_gap_p): Likewise.
18892 (vectorizable_with_step_bound_p): Likewise.
18893 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
18894 depencies if the vectorization factor is 1. Convert the checks
18895 for nonzero steps into checks on the bounds of DR_STEP. Try using
18896 a bunds check for variable steps if the minimum required step is
18897 relatively small. Update calls to the dr_with_seg_len
18898 constructor and to vect_compile_time_alias.
18899 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
18901 (vect_loop_versioning): Call it.
18902 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
18904 (vect_estimate_min_profitable_iters): Account for any bounds checks.
18906 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18907 Alan Hayward <alan.hayward@arm.com>
18908 David Sherwood <david.sherwood@arm.com>
18910 * doc/sourcebuild.texi (vect_scatter_store): Document.
18911 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
18913 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
18915 * genopinit.c (main): Add supports_vec_scatter_store and
18916 supports_vec_scatter_store_cached to target_optabs.
18917 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
18918 IFN_MASK_SCATTER_STORE.
18919 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
18921 * internal-fn.h (internal_store_fn_p): Declare.
18922 (internal_fn_stored_value_index): Likewise.
18923 * internal-fn.c (scatter_store_direct): New macro.
18924 (expand_scatter_store_optab_fn): New function.
18925 (direct_scatter_store_optab_supported_p): New macro.
18926 (internal_store_fn_p): New function.
18927 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
18928 IFN_MASK_SCATTER_STORE.
18929 (internal_fn_mask_index): Likewise.
18930 (internal_fn_stored_value_index): New function.
18931 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
18932 for scatter stores.
18933 * optabs-query.h (supports_vec_scatter_store_p): Declare.
18934 * optabs-query.c (supports_vec_scatter_store_p): New function.
18935 * tree-vectorizer.h (vect_get_store_rhs): Declare.
18936 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
18937 true for scatter stores.
18938 (vect_gather_scatter_fn_p): Handle scatter stores too.
18939 (vect_check_gather_scatter): Consider using scatter stores if
18940 supports_vec_scatter_store_p.
18941 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
18942 scatter stores too.
18943 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
18944 internal_fn_stored_value_index.
18945 (check_load_store_masking): Handle scatter stores too.
18946 (vect_get_store_rhs): Make public.
18947 (vectorizable_call): Use internal_store_fn_p.
18948 (vectorizable_store): Handle scatter store internal functions.
18949 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
18950 when deciding whether the end of the group has been reached.
18951 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
18952 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
18953 (mask_scatter_store<mode>): New insns.
18955 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18956 Alan Hayward <alan.hayward@arm.com>
18957 David Sherwood <david.sherwood@arm.com>
18959 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
18960 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
18961 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
18963 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
18964 Use vect_truncate_gather_scatter_offset if we can't treat the
18965 operation as a normal gather load or scatter store.
18966 (get_group_load_store_type): Take the gather_scatter_info
18967 as argument. Try using a gather load or scatter store for
18968 single-element groups.
18969 (get_load_store_type): Update calls to get_group_load_store_type
18970 and vect_use_strided_gather_scatters_p.
18972 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18973 Alan Hayward <alan.hayward@arm.com>
18974 David Sherwood <david.sherwood@arm.com>
18976 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
18977 optional tree argument.
18978 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
18980 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
18981 but continue to use the current value as a fallback.
18982 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
18983 to compare the updates.
18984 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
18985 (get_load_store_type): Use it when handling a strided access.
18986 (vect_get_strided_load_store_ops): New function.
18987 (vect_get_data_ptr_increment): Likewise.
18988 (vectorizable_load): Handle strided gather loads. Always pass
18989 a step to vect_create_data_ref_ptr and bump_vector_ptr.
18991 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18992 Alan Hayward <alan.hayward@arm.com>
18993 David Sherwood <david.sherwood@arm.com>
18995 * doc/md.texi (gather_load@var{m}): Document.
18996 (mask_gather_load@var{m}): Likewise.
18997 * genopinit.c (main): Add supports_vec_gather_load and
18998 supports_vec_gather_load_cached to target_optabs.
18999 * optabs-tree.c (init_tree_optimization_optabs): Use
19000 ggc_cleared_alloc to allocate target_optabs.
19001 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
19002 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
19004 * internal-fn.h (internal_load_fn_p): Declare.
19005 (internal_gather_scatter_fn_p): Likewise.
19006 (internal_fn_mask_index): Likewise.
19007 (internal_gather_scatter_fn_supported_p): Likewise.
19008 * internal-fn.c (gather_load_direct): New macro.
19009 (expand_gather_load_optab_fn): New function.
19010 (direct_gather_load_optab_supported_p): New macro.
19011 (direct_internal_fn_optab): New function.
19012 (internal_load_fn_p): Likewise.
19013 (internal_gather_scatter_fn_p): Likewise.
19014 (internal_fn_mask_index): Likewise.
19015 (internal_gather_scatter_fn_supported_p): Likewise.
19016 * optabs-query.c (supports_at_least_one_mode_p): New function.
19017 (supports_vec_gather_load_p): Likewise.
19018 * optabs-query.h (supports_vec_gather_load_p): Declare.
19019 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
19020 and memory_type field.
19021 (NUM_PATTERNS): Bump to 15.
19022 * tree-vect-data-refs.c: Include internal-fn.h.
19023 (vect_gather_scatter_fn_p): New function.
19024 (vect_describe_gather_scatter_call): Likewise.
19025 (vect_check_gather_scatter): Try using internal functions for
19026 gather loads. Recognize existing calls to a gather load function.
19027 (vect_analyze_data_refs): Consider using gather loads if
19028 supports_vec_gather_load_p.
19029 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
19030 (vect_get_gather_scatter_offset_type): Likewise.
19031 (vect_convert_mask_for_vectype): Likewise.
19032 (vect_add_conversion_to_patterm): Likewise.
19033 (vect_try_gather_scatter_pattern): Likewise.
19034 (vect_recog_gather_scatter_pattern): New pattern recognizer.
19035 (vect_vect_recog_func_ptrs): Add it.
19036 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
19037 internal_fn_mask_index and internal_gather_scatter_fn_p.
19038 (check_load_store_masking): Take the gather_scatter_info as an
19039 argument and handle gather loads.
19040 (vect_get_gather_scatter_ops): New function.
19041 (vectorizable_call): Check internal_load_fn_p.
19042 (vectorizable_load): Likewise. Handle gather load internal
19044 (vectorizable_store): Update call to check_load_store_masking.
19045 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
19046 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
19047 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
19048 (aarch64_gather_scale_operand_d): New predicates.
19049 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
19050 (mask_gather_load<mode>): New insns.
19052 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19053 Alan Hayward <alan.hayward@arm.com>
19054 David Sherwood <david.sherwood@arm.com>
19056 * optabs.def (fold_left_plus_optab): New optab.
19057 * doc/md.texi (fold_left_plus_@var{m}): Document.
19058 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
19059 * internal-fn.c (fold_left_direct): Define.
19060 (expand_fold_left_optab_fn): Likewise.
19061 (direct_fold_left_optab_supported_p): Likewise.
19062 * fold-const-call.c (fold_const_fold_left): New function.
19063 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
19064 * tree-parloops.c (valid_reduction_p): New function.
19065 (gather_scalar_reductions): Use it.
19066 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
19067 (vect_finish_replace_stmt): Declare.
19068 * tree-vect-loop.c (fold_left_reduction_fn): New function.
19069 (needs_fold_left_reduction_p): New function, split out from...
19070 (vect_is_simple_reduction): ...here. Accept reductions that
19071 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
19072 (vect_force_simple_reduction): Also store the reduction type in
19073 the assignment's STMT_VINFO_REDUC_TYPE.
19074 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
19075 (merge_with_identity): New function.
19076 (vect_expand_fold_left): Likewise.
19077 (vectorize_fold_left_reduction): Likewise.
19078 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
19079 scalar phi in place for it. Check for target support and reject
19080 cases that would reassociate the operation. Defer the transform
19081 phase to vectorize_fold_left_reduction.
19082 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
19083 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
19084 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
19086 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19088 * tree-if-conv.c (predicate_mem_writes): Remove redundant
19089 call to ifc_temp_var.
19091 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19092 Alan Hayward <alan.hayward@arm.com>
19093 David Sherwood <david.sherwood@arm.com>
19095 * target.def (legitimize_address_displacement): Take the original
19096 offset as a poly_int.
19097 * targhooks.h (default_legitimize_address_displacement): Update
19099 * targhooks.c (default_legitimize_address_displacement): Likewise.
19100 * doc/tm.texi: Regenerate.
19101 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
19102 as an argument, moving assert of ad->disp == ad->disp_term to...
19103 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
19104 Try calling targetm.legitimize_address_displacement before expanding
19105 the address rather than afterwards, and adjust for the new interface.
19106 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
19107 Match the new hook interface. Handle SVE addresses.
19108 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
19109 new hook interface.
19111 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19113 * Makefile.in (OBJS): Add early-remat.o.
19114 * target.def (select_early_remat_modes): New hook.
19115 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
19116 * doc/tm.texi: Regenerate.
19117 * targhooks.h (default_select_early_remat_modes): Declare.
19118 * targhooks.c (default_select_early_remat_modes): New function.
19119 * timevar.def (TV_EARLY_REMAT): New timevar.
19120 * passes.def (pass_early_remat): New pass.
19121 * tree-pass.h (make_pass_early_remat): Declare.
19122 * early-remat.c: New file.
19123 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
19125 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
19127 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19128 Alan Hayward <alan.hayward@arm.com>
19129 David Sherwood <david.sherwood@arm.com>
19131 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
19132 vfm1 with a bound_epilog parameter.
19133 (vect_do_peeling): Update calls accordingly, and move the prologue
19134 call earlier in the function. Treat the base bound_epilog as 0 for
19135 fully-masked loops and retain vf - 1 for other loops. Add 1 to
19136 this base when peeling for gaps.
19137 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
19138 with fully-masked loops.
19139 (vect_estimate_min_profitable_iters): Handle the single peeled
19140 iteration in that case.
19142 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19143 Alan Hayward <alan.hayward@arm.com>
19144 David Sherwood <david.sherwood@arm.com>
19146 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
19147 single-element interleaving even if the size is not a power of 2.
19148 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
19149 accesses for single-element interleaving if the group size is
19152 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19153 Alan Hayward <alan.hayward@arm.com>
19154 David Sherwood <david.sherwood@arm.com>
19156 * doc/md.texi (fold_extract_last_@var{m}): Document.
19157 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
19158 * optabs.def (fold_extract_last_optab): New optab.
19159 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
19160 * internal-fn.c (fold_extract_direct): New macro.
19161 (expand_fold_extract_optab_fn): Likewise.
19162 (direct_fold_extract_optab_supported_p): Likewise.
19163 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
19164 * tree-vect-loop.c (vect_model_reduction_cost): Handle
19165 EXTRACT_LAST_REDUCTION.
19166 (get_initial_def_for_reduction): Do not create an initial vector
19167 for EXTRACT_LAST_REDUCTION reductions.
19168 (vectorizable_reduction): Leave the scalar phi in place for
19169 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
19170 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
19171 epilogue code for EXTRACT_LAST_REDUCTION and defer the
19172 transform phase to vectorizable_condition.
19173 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
19175 (vect_finish_stmt_generation): ...here.
19176 (vect_finish_replace_stmt): New function.
19177 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
19178 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
19180 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
19182 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19183 Alan Hayward <alan.hayward@arm.com>
19184 David Sherwood <david.sherwood@arm.com>
19186 * doc/md.texi (extract_last_@var{m}): Document.
19187 * optabs.def (extract_last_optab): New optab.
19188 * internal-fn.def (EXTRACT_LAST): New internal function.
19189 * internal-fn.c (cond_unary_direct): New macro.
19190 (expand_cond_unary_optab_fn): Likewise.
19191 (direct_cond_unary_optab_supported_p): Likewise.
19192 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
19193 loops using EXTRACT_LAST.
19194 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
19195 (extract_last_<mode>): ...this optab.
19196 (vec_extract<mode><Vel>): Update accordingly.
19198 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19199 Alan Hayward <alan.hayward@arm.com>
19200 David Sherwood <david.sherwood@arm.com>
19202 * target.def (empty_mask_is_expensive): New hook.
19203 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
19204 * doc/tm.texi: Regenerate.
19205 * targhooks.h (default_empty_mask_is_expensive): Declare.
19206 * targhooks.c (default_empty_mask_is_expensive): New function.
19207 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
19208 if the target says that empty masks are expensive.
19209 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
19211 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
19213 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19214 Alan Hayward <alan.hayward@arm.com>
19215 David Sherwood <david.sherwood@arm.com>
19217 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
19218 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
19219 (vect_use_loop_mask_for_alignment_p): New function.
19220 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
19221 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
19222 niters_skip argument. Make sure that the first niters_skip elements
19223 of the first iteration are inactive.
19224 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
19225 Update call to vect_set_loop_masks_directly.
19226 (get_misalign_in_elems): New function, split out from...
19227 (vect_gen_prolog_loop_niters): ...here.
19228 (vect_update_init_of_dr): Take a code argument that specifies whether
19229 the adjustment should be added or subtracted.
19230 (vect_update_init_of_drs): Likewise.
19231 (vect_prepare_for_masked_peels): New function.
19232 (vect_do_peeling): Skip prologue peeling if we're using a mask
19233 instead. Update call to vect_update_inits_of_drs.
19234 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
19236 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
19237 alignment. Do not include the number of peeled iterations in
19238 the minimum threshold in that case.
19239 (vectorizable_induction): Adjust the start value down by
19240 LOOP_VINFO_MASK_SKIP_NITERS iterations.
19241 (vect_transform_loop): Call vect_prepare_for_masked_peels.
19242 Take the number of skipped iterations into account when calculating
19244 * tree-vect-stmts.c (vect_gen_while_not): New function.
19246 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19247 Alan Hayward <alan.hayward@arm.com>
19248 David Sherwood <david.sherwood@arm.com>
19250 * doc/sourcebuild.texi (vect_fully_masked): Document.
19251 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
19252 default value to 0.
19253 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
19255 (vect_analyze_loop_2): ...here. Don't check the vectorization
19256 factor against the number of loop iterations if the loop is
19259 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19260 Alan Hayward <alan.hayward@arm.com>
19261 David Sherwood <david.sherwood@arm.com>
19263 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
19264 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
19265 (dump_groups): Update accordingly.
19266 (iv_use::mem_type): New member variable.
19267 (address_p): New function.
19268 (record_use): Add a mem_type argument and initialize the new
19270 (record_group_use): Add a mem_type argument. Use address_p.
19271 Remove obsolete null checks of base_object. Update call to record_use.
19272 (find_interesting_uses_op): Update call to record_group_use.
19273 (find_interesting_uses_cond): Likewise.
19274 (find_interesting_uses_address): Likewise.
19275 (get_mem_type_for_internal_fn): New function.
19276 (find_address_like_use): Likewise.
19277 (find_interesting_uses_stmt): Try find_address_like_use before
19278 calling find_interesting_uses_op.
19279 (addr_offset_valid_p): Use the iv mem_type field as the type
19280 of the addressed memory.
19281 (add_autoinc_candidates): Likewise.
19282 (get_address_cost): Likewise.
19283 (split_small_address_groups_p): Use address_p.
19284 (split_address_groups): Likewise.
19285 (add_iv_candidate_for_use): Likewise.
19286 (autoinc_possible_for_pair): Likewise.
19287 (rewrite_groups): Likewise.
19288 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
19289 (determine_group_iv_cost): Update after split of USE_ADDRESS.
19290 (get_alias_ptr_type_for_ptr_address): New function.
19291 (rewrite_use_address): Rewrite address uses in calls that were
19292 identified by find_address_like_use.
19294 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19295 Alan Hayward <alan.hayward@arm.com>
19296 David Sherwood <david.sherwood@arm.com>
19298 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
19300 * gimple-expr.h (is_gimple_addressable: Likewise.
19301 * gimple-expr.c (is_gimple_address): Likewise.
19302 * internal-fn.c (expand_call_mem_ref): New function.
19303 (expand_mask_load_optab_fn): Use it.
19304 (expand_mask_store_optab_fn): Likewise.
19306 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19307 Alan Hayward <alan.hayward@arm.com>
19308 David Sherwood <david.sherwood@arm.com>
19310 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
19311 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
19312 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
19313 (cond_umax@var{mode}): Document.
19314 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
19315 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
19316 (cond_umin_optab, cond_umax_optab): New optabs.
19317 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
19318 (COND_IOR, COND_XOR): New internal functions.
19319 * internal-fn.h (get_conditional_internal_fn): Declare.
19320 * internal-fn.c (cond_binary_direct): New macro.
19321 (expand_cond_binary_optab_fn): Likewise.
19322 (direct_cond_binary_optab_supported_p): Likewise.
19323 (get_conditional_internal_fn): New function.
19324 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
19325 Cope with reduction statements that are vectorized as calls rather
19327 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
19328 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
19329 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
19330 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
19331 (UNSPEC_COND_EOR): New unspecs.
19332 (optab): Add mappings for them.
19333 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
19334 (sve_int_op, sve_fp_op): New int attributes.
19336 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19337 Alan Hayward <alan.hayward@arm.com>
19338 David Sherwood <david.sherwood@arm.com>
19340 * optabs.def (while_ult_optab): New optab.
19341 * doc/md.texi (while_ult@var{m}@var{n}): Document.
19342 * internal-fn.def (WHILE_ULT): New internal function.
19343 * internal-fn.h (direct_internal_fn_supported_p): New override
19344 that takes two types as argument.
19345 * internal-fn.c (while_direct): New macro.
19346 (expand_while_optab_fn): New function.
19347 (convert_optab_supported_p): Likewise.
19348 (direct_while_optab_supported_p): New macro.
19349 * wide-int.h (wi::udiv_ceil): New function.
19350 * tree-vectorizer.h (rgroup_masks): New structure.
19351 (vec_loop_masks): New typedef.
19352 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
19353 and fully_masked_p.
19354 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
19355 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
19356 (vect_max_vf): New function.
19357 (slpeel_make_loop_iterate_ntimes): Delete.
19358 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
19359 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
19360 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
19361 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
19362 internal-fn.h, stor-layout.h and optabs-query.h.
19363 (vect_set_loop_mask): New function.
19364 (add_preheader_seq): Likewise.
19365 (add_header_seq): Likewise.
19366 (interleave_supported_p): Likewise.
19367 (vect_maybe_permute_loop_masks): Likewise.
19368 (vect_set_loop_masks_directly): Likewise.
19369 (vect_set_loop_condition_masked): Likewise.
19370 (vect_set_loop_condition_unmasked): New function, split out from
19371 slpeel_make_loop_iterate_ntimes.
19372 (slpeel_make_loop_iterate_ntimes): Rename to..
19373 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
19374 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
19375 (vect_do_peeling): Update call accordingly.
19376 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
19378 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
19379 mask_compare_type, can_fully_mask_p and fully_masked_p.
19380 (release_vec_loop_masks): New function.
19381 (_loop_vec_info): Use it to free the loop masks.
19382 (can_produce_all_loop_masks_p): New function.
19383 (vect_get_max_nscalars_per_iter): Likewise.
19384 (vect_verify_full_masking): Likewise.
19385 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
19386 retries, and free the mask rgroups before retrying. Check loop-wide
19387 reasons for disallowing fully-masked loops. Make the final decision
19388 about whether use a fully-masked loop or not.
19389 (vect_estimate_min_profitable_iters): Do not assume that peeling
19390 for the number of iterations will be needed for fully-masked loops.
19391 (vectorizable_reduction): Disable fully-masked loops.
19392 (vectorizable_live_operation): Likewise.
19393 (vect_halve_mask_nunits): New function.
19394 (vect_double_mask_nunits): Likewise.
19395 (vect_record_loop_mask): Likewise.
19396 (vect_get_loop_mask): Likewise.
19397 (vect_transform_loop): Handle the case in which the final loop
19398 iteration might handle a partial vector. Call vect_set_loop_condition
19399 instead of slpeel_make_loop_iterate_ntimes.
19400 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
19401 (check_load_store_masking): New function.
19402 (prepare_load_store_mask): Likewise.
19403 (vectorizable_store): Handle fully-masked loops.
19404 (vectorizable_load): Likewise.
19405 (supportable_widening_operation): Use vect_halve_mask_nunits for
19407 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
19408 (vect_gen_while): New function.
19409 * config/aarch64/aarch64.md (umax<mode>3): New expander.
19410 (aarch64_uqdec<mode>): New insn.
19412 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19413 Alan Hayward <alan.hayward@arm.com>
19414 David Sherwood <david.sherwood@arm.com>
19416 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
19417 (reduc_xor_scal_optab): New optabs.
19418 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
19419 (reduc_xor_scal_@var{m}): Document.
19420 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
19421 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
19422 internal functions.
19423 * fold-const-call.c (fold_const_call): Handle them.
19424 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
19425 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
19426 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
19427 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
19428 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
19429 (UNSPEC_XORV): New unspecs.
19430 (optab): Add entries for them.
19431 (BITWISEV): New int iterator.
19432 (bit_reduc_op): New int attributes.
19434 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19435 Alan Hayward <alan.hayward@arm.com>
19436 David Sherwood <david.sherwood@arm.com>
19438 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
19439 * internal-fn.def (VEC_SHL_INSERT): New internal function.
19440 * optabs.def (vec_shl_insert_optab): New optab.
19441 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
19442 (duplicate_and_interleave): Likewise.
19443 * tree-vect-loop.c: Include internal-fn.h.
19444 (neutral_op_for_slp_reduction): New function, split out from
19445 get_initial_defs_for_reduction.
19446 (get_initial_def_for_reduction): Handle option 2 for variable-length
19447 vectors by loading the neutral value into a vector and then shifting
19448 the initial value into element 0.
19449 (get_initial_defs_for_reduction): Replace the code argument with
19450 the neutral value calculated by neutral_op_for_slp_reduction.
19451 Use gimple_build_vector for constant-length vectors.
19452 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
19453 but the first group_size elements have a neutral value.
19454 Use duplicate_and_interleave otherwise.
19455 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
19456 Update call to get_initial_defs_for_reduction. Handle SLP
19457 reductions for variable-length vectors by creating one vector
19458 result for each scalar result, with the elements associated
19459 with other scalar results stubbed out with the neutral value.
19460 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
19461 Require IFN_VEC_SHL_INSERT for double reductions on
19462 variable-length vectors, or SLP reductions that have
19463 a neutral value. Require can_duplicate_and_interleave_p
19464 support for variable-length unchained SLP reductions if there
19465 is no neutral value, such as for MIN/MAX reductions. Also require
19466 the number of vector elements to be a multiple of the number of
19467 SLP statements when doing variable-length unchained SLP reductions.
19468 Update call to vect_create_epilog_for_reduction.
19469 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
19470 and remove initial values.
19471 (duplicate_and_interleave): Make public.
19472 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
19473 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
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-slp.c: Include gimple-fold.h and internal-fn.h
19480 (can_duplicate_and_interleave_p): New function.
19481 (vect_get_and_check_slp_defs): Take the vector of statements
19482 rather than just the current one. Remove excess parentheses.
19483 Restriction rejectinon of vect_constant_def and vect_external_def
19484 for variable-length vectors to boolean types, or types for which
19485 can_duplicate_and_interleave_p is false.
19486 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
19487 (duplicate_and_interleave): New function.
19488 (vect_get_constant_vectors): Use gimple_build_vector for
19489 constant-length vectors and suitable variable-length constant
19490 vectors. Use duplicate_and_interleave for other variable-length
19491 vectors. Don't defer the update when inserting new statements.
19493 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19494 Alan Hayward <alan.hayward@arm.com>
19495 David Sherwood <david.sherwood@arm.com>
19497 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
19498 min_profitable_iters doesn't go negative.
19500 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19501 Alan Hayward <alan.hayward@arm.com>
19502 David Sherwood <david.sherwood@arm.com>
19504 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
19505 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
19506 * optabs.def (vec_mask_load_lanes_optab): New optab.
19507 (vec_mask_store_lanes_optab): Likewise.
19508 * internal-fn.def (MASK_LOAD_LANES): New internal function.
19509 (MASK_STORE_LANES): Likewise.
19510 * internal-fn.c (mask_load_lanes_direct): New macro.
19511 (mask_store_lanes_direct): Likewise.
19512 (expand_mask_load_optab_fn): Handle masked operations.
19513 (expand_mask_load_lanes_optab_fn): New macro.
19514 (expand_mask_store_optab_fn): Handle masked operations.
19515 (expand_mask_store_lanes_optab_fn): New macro.
19516 (direct_mask_load_lanes_optab_supported_p): Likewise.
19517 (direct_mask_store_lanes_optab_supported_p): Likewise.
19518 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
19520 (vect_load_lanes_supported): Likewise.
19521 * tree-vect-data-refs.c (strip_conversion): New function.
19522 (can_group_stmts_p): Likewise.
19523 (vect_analyze_data_ref_accesses): Use it instead of checking
19524 for a pair of assignments.
19525 (vect_store_lanes_supported): Take a masked_p parameter.
19526 (vect_load_lanes_supported): Likewise.
19527 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
19528 vect_store_lanes_supported and vect_load_lanes_supported.
19529 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
19530 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
19531 parameter. Don't allow gaps for masked accesses.
19532 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
19533 and vect_load_lanes_supported.
19534 (get_load_store_type): Take a masked_p parameter and update
19535 call to get_group_load_store_type.
19536 (vectorizable_store): Update call to get_load_store_type.
19537 Handle IFN_MASK_STORE_LANES.
19538 (vectorizable_load): Update call to get_load_store_type.
19539 Handle IFN_MASK_LOAD_LANES.
19541 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19542 Alan Hayward <alan.hayward@arm.com>
19543 David Sherwood <david.sherwood@arm.com>
19545 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
19547 * config/aarch64/aarch64-protos.h
19548 (aarch64_sve_struct_memory_operand_p): Declare.
19549 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
19550 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
19551 (VPRED, vpred): Handle SVE structure modes.
19552 * config/aarch64/constraints.md (Utx): New constraint.
19553 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
19554 (aarch64_sve_struct_nonimmediate_operand): New predicates.
19555 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
19556 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
19557 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
19558 structure modes. Split into pieces after RA.
19559 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
19560 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
19562 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
19563 SVE structure modes.
19564 (aarch64_classify_address): Likewise.
19565 (sizetochar): Move earlier in file.
19566 (aarch64_print_operand): Handle SVE register lists.
19567 (aarch64_array_mode): New function.
19568 (aarch64_sve_struct_memory_operand_p): Likewise.
19569 (TARGET_ARRAY_MODE): Redefine.
19571 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19572 Alan Hayward <alan.hayward@arm.com>
19573 David Sherwood <david.sherwood@arm.com>
19575 * target.def (array_mode): New target hook.
19576 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
19577 * doc/tm.texi: Regenerate.
19578 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
19579 * hooks.c (hook_optmode_mode_uhwi_none): New function.
19580 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
19581 targetm.array_mode.
19582 * stor-layout.c (mode_for_array): Likewise. Support polynomial
19585 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19586 Alan Hayward <alan.hayward@arm.com>
19587 David Sherwood <david.sherwood@arm.com>
19589 * fold-const.c (fold_binary_loc): Check the argument types
19590 rather than the result type when testing for a vector operation.
19592 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19594 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
19595 * doc/tm.texi: Regenerate.
19597 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19598 Alan Hayward <alan.hayward@arm.com>
19599 David Sherwood <david.sherwood@arm.com>
19601 * doc/invoke.texi (-msve-vector-bits=): Document new option.
19602 (sve): Document new AArch64 extension.
19603 * doc/md.texi (w): Extend the description of the AArch64
19604 constraint to include SVE vectors.
19605 (Upl, Upa): Document new AArch64 predicate constraints.
19606 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
19608 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
19609 (msve-vector-bits=): New option.
19610 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
19611 SVE when these are disabled.
19612 (sve): New extension.
19613 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
19614 modes. Adjust their number of units based on aarch64_sve_vg.
19615 (MAX_BITSIZE_MODE_ANY_MODE): Define.
19616 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
19617 aarch64_addr_query_type.
19618 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
19619 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
19620 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
19621 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
19622 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
19623 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
19624 (aarch64_simd_imm_zero_p): Delete.
19625 (aarch64_check_zero_based_sve_index_immediate): Declare.
19626 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
19627 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
19628 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
19629 (aarch64_sve_float_mul_immediate_p): Likewise.
19630 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
19631 rather than an rtx.
19632 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
19633 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
19634 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
19635 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
19636 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
19637 (aarch64_regmode_natural_size): Likewise.
19638 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
19639 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
19641 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
19642 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
19643 for VG and the SVE predicate registers.
19644 (V_ALIASES): Add a "z"-prefixed alias.
19645 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
19646 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
19647 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
19648 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
19649 (REG_CLASS_NAMES): Add entries for them.
19650 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
19651 and the predicate registers.
19652 (aarch64_sve_vg): Declare.
19653 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
19654 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
19655 (REGMODE_NATURAL_SIZE): Define.
19656 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
19658 * config/aarch64/aarch64.c: Include cfgrtl.h.
19659 (simd_immediate_info): Add a constructor for series vectors,
19660 and an associated step field.
19661 (aarch64_sve_vg): New variable.
19662 (aarch64_dbx_register_number): Handle VG and the predicate registers.
19663 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
19664 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
19665 (VEC_ANY_DATA, VEC_STRUCT): New constants.
19666 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
19667 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
19668 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
19669 (aarch64_get_mask_mode): New functions.
19670 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
19671 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
19672 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
19673 predicate modes and predicate registers. Explicitly restrict
19674 GPRs to modes of 16 bytes or smaller. Only allow FP registers
19675 to store a vector mode if it is recognized by
19676 aarch64_classify_vector_mode.
19677 (aarch64_regmode_natural_size): New function.
19678 (aarch64_hard_regno_caller_save_mode): Return the original mode
19680 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
19681 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
19682 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
19683 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
19685 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
19686 does not overlap dest if the function is frame-related. Handle
19688 (aarch64_split_add_offset): New function.
19689 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
19690 them aarch64_add_offset.
19691 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
19692 and update call to aarch64_sub_sp.
19693 (aarch64_add_cfa_expression): New function.
19694 (aarch64_expand_prologue): Pass extra temporary registers to the
19695 functions above. Handle the case in which we need to emit new
19696 DW_CFA_expressions for registers that were originally saved
19697 relative to the stack pointer, but now have to be expressed
19698 relative to the frame pointer.
19699 (aarch64_output_mi_thunk): Pass extra temporary registers to the
19701 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
19702 IP0 and IP1 values for SVE frames.
19703 (aarch64_expand_vec_series): New function.
19704 (aarch64_expand_sve_widened_duplicate): Likewise.
19705 (aarch64_expand_sve_const_vector): Likewise.
19706 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
19707 Handle SVE constants. Use emit_move_insn to move a force_const_mem
19708 into the register, rather than emitting a SET directly.
19709 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
19710 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
19711 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
19712 (offset_9bit_signed_scaled_p): New functions.
19713 (aarch64_replicate_bitmask_imm): New function.
19714 (aarch64_bitmask_imm): Use it.
19715 (aarch64_cannot_force_const_mem): Reject expressions involving
19716 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
19717 (aarch64_classify_index): Handle SVE indices, by requiring
19718 a plain register index with a scale that matches the element size.
19719 (aarch64_classify_address): Handle SVE addresses. Assert that
19720 the mode of the address is VOIDmode or an integer mode.
19721 Update call to aarch64_classify_symbol.
19722 (aarch64_classify_symbolic_expression): Update call to
19723 aarch64_classify_symbol.
19724 (aarch64_const_vec_all_in_range_p): New function.
19725 (aarch64_print_vector_float_operand): Likewise.
19726 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
19727 "vN" for FP registers with SVE modes. Handle (const ...) vectors
19728 and the FP immediates 1.0 and 0.5.
19729 (aarch64_print_address_internal): Handle SVE addresses.
19730 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
19731 (aarch64_regno_regclass): Handle predicate registers.
19732 (aarch64_secondary_reload): Handle big-endian reloads of SVE
19734 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
19735 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
19736 (aarch64_convert_sve_vector_bits): New function.
19737 (aarch64_override_options): Use it to handle -msve-vector-bits=.
19738 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
19739 rather than an rtx.
19740 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
19741 Handle SVE vector and predicate modes. Accept VL-based constants
19742 that need only one temporary register, and VL offsets that require
19743 no temporary registers.
19744 (aarch64_conditional_register_usage): Mark the predicate registers
19745 as fixed if SVE isn't available.
19746 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
19747 Return true for SVE vector and predicate modes.
19748 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
19749 rather than an unsigned int. Handle SVE modes.
19750 (aarch64_preferred_simd_mode): Update call accordingly. Handle
19752 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
19754 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
19755 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
19756 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
19757 (aarch64_sve_float_mul_immediate_p): New functions.
19758 (aarch64_sve_valid_immediate): New function.
19759 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
19760 Explicitly reject structure modes. Check for INDEX constants.
19761 Handle PTRUE and PFALSE constants.
19762 (aarch64_check_zero_based_sve_index_immediate): New function.
19763 (aarch64_simd_imm_zero_p): Delete.
19764 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
19765 vector modes. Accept constants in the range of CNT[BHWD].
19766 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
19767 ask for an Advanced SIMD mode.
19768 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
19769 (aarch64_simd_vector_alignment): Handle SVE predicates.
19770 (aarch64_vectorize_preferred_vector_alignment): New function.
19771 (aarch64_simd_vector_alignment_reachable): Use it instead of
19773 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
19774 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
19776 (MAX_VECT_LEN): Delete.
19777 (expand_vec_perm_d): Add a vec_flags field.
19778 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
19779 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
19780 (aarch64_evpc_ext): Don't apply a big-endian lane correction
19782 (aarch64_evpc_rev): Rename to...
19783 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
19784 (aarch64_evpc_rev_global): New function.
19785 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
19786 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
19788 (aarch64_evpc_sve_tbl): New function.
19789 (aarch64_expand_vec_perm_const_1): Update after rename of
19790 aarch64_evpc_rev. Handle SVE permutes too, trying
19791 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
19792 than aarch64_evpc_tbl.
19793 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
19794 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
19795 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
19796 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
19797 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
19798 (aarch64_expand_sve_vcond): New functions.
19799 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
19800 of aarch64_vector_mode_p.
19801 (aarch64_dwarf_poly_indeterminate_value): New function.
19802 (aarch64_compute_pressure_classes): Likewise.
19803 (aarch64_can_change_mode_class): Likewise.
19804 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
19805 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
19806 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
19807 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
19808 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
19809 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
19810 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
19811 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
19813 (Dn, Dl, Dr): Accept const as well as const_vector.
19814 (Dz): Likewise. Compare against CONST0_RTX.
19815 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
19816 of "vector" where appropriate.
19817 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
19818 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
19819 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
19820 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
19821 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
19822 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
19823 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
19824 (v_int_equiv): Extend to SVE modes.
19825 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
19827 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
19828 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
19829 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
19830 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
19831 (SVE_COND_FP_CMP): New int iterators.
19832 (perm_hilo): Handle the new unpack unspecs.
19833 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
19835 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
19836 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
19837 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
19838 (aarch64_equality_operator, aarch64_constant_vector_operand)
19839 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
19840 (aarch64_sve_nonimmediate_operand): Likewise.
19841 (aarch64_sve_general_operand): Likewise.
19842 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
19843 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
19844 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
19845 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
19846 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
19847 (aarch64_sve_float_arith_immediate): Likewise.
19848 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
19849 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
19850 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
19851 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
19852 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
19853 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
19854 (aarch64_sve_float_arith_operand): Likewise.
19855 (aarch64_sve_float_arith_with_sub_operand): Likewise.
19856 (aarch64_sve_float_mul_operand): Likewise.
19857 (aarch64_sve_vec_perm_operand): Likewise.
19858 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
19859 (aarch64_mov_operand): Accept const_poly_int and const_vector.
19860 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
19861 as well as const_vector.
19862 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
19863 in file. Use CONST0_RTX and CONSTM1_RTX.
19864 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
19865 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
19866 Use aarch64_simd_imm_zero.
19867 * config/aarch64/aarch64-sve.md: New file.
19868 * config/aarch64/aarch64.md: Include it.
19869 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
19870 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
19871 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
19872 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
19873 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
19874 (sve): New attribute.
19875 (enabled): Disable instructions with the sve attribute unless
19877 (movqi, movhi): Pass CONST_POLY_INT operaneds through
19878 aarch64_expand_mov_immediate.
19879 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
19880 CNT[BHSD] immediates.
19881 (movti): Split CONST_POLY_INT moves into two halves.
19882 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
19883 Split additions that need a temporary here if the destination
19884 is the stack pointer.
19885 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
19886 (*add<mode>3_poly_1): New instruction.
19887 (set_clobber_cc): New expander.
19889 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19891 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
19892 parameter and use it instead of GET_MODE_SIZE (innermode). Use
19893 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
19894 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
19895 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
19896 Change innermode from fixed_mode_size to machine_mode.
19897 (simplify_subreg): Update call accordingly. Handle a constant-sized
19898 subreg of a variable-length CONST_VECTOR.
19900 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19901 Alan Hayward <alan.hayward@arm.com>
19902 David Sherwood <david.sherwood@arm.com>
19904 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
19905 (add_offset_to_base): New function, split out from...
19906 (create_mem_ref): ...here. When handling a scale other than 1,
19907 check first whether the address is valid without the offset.
19908 Add it into the base if so, leaving the index and scale as-is.
19910 2018-01-12 Jakub Jelinek <jakub@redhat.com>
19913 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
19914 fold_for_warn before checking if arg2 is INTEGER_CST.
19916 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
19918 * config/rs6000/predicates.md (load_multiple_operation): Delete.
19919 (store_multiple_operation): Delete.
19920 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
19921 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
19922 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
19923 guarded by TARGET_STRING.
19924 (rs6000_output_load_multiple): Delete.
19925 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
19926 OPTION_MASK_STRING / TARGET_STRING handling.
19927 (print_operand) <'N', 'O'>: Add comment that these are unused now.
19928 (const rs6000_opt_masks) <"string">: Change mask to 0.
19929 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
19930 (MASK_STRING): Delete.
19931 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
19933 (load_multiple): Delete.
19940 (store_multiple): Delete.
19947 (movmemsi_8reg): Delete.
19948 (corresponding unnamed define_insn): Delete.
19949 (movmemsi_6reg): Delete.
19950 (corresponding unnamed define_insn): Delete.
19951 (movmemsi_4reg): Delete.
19952 (corresponding unnamed define_insn): Delete.
19953 (movmemsi_2reg): Delete.
19954 (corresponding unnamed define_insn): Delete.
19955 (movmemsi_1reg): Delete.
19956 (corresponding unnamed define_insn): Delete.
19957 * config/rs6000/rs6000.opt (mno-string): New.
19958 (mstring): Replace by deprecation warning stub.
19959 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
19961 2018-01-12 Jakub Jelinek <jakub@redhat.com>
19963 * regrename.c (regrename_do_replace): If replacing the same
19964 reg multiple times, try to reuse last created gen_raw_REG.
19967 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
19968 main to workaround a bug in GDB.
19970 2018-01-12 Tom de Vries <tom@codesourcery.com>
19973 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
19975 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
19977 PR rtl-optimization/80481
19978 * ira-color.c (get_cap_member): New function.
19979 (allocnos_conflict_by_live_ranges_p): Use it.
19980 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
19981 (setup_slot_coalesced_allocno_live_ranges): Ditto.
19983 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
19986 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
19987 (*saddl_se_1): Ditto.
19988 (*ssubsi_1): Ditto.
19989 (*ssubl_se_1): Ditto.
19991 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
19993 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
19994 rather than wi::to_widest for DR_INITs.
19995 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
19996 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
19997 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
19999 (vect_analyze_group_access_1): Note that here.
20001 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
20003 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
20004 polynomial type sizes.
20006 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
20008 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
20009 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
20010 (gimple_add_tmp_var): Likewise.
20012 2018-01-12 Martin Liska <mliska@suse.cz>
20014 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
20015 (gimple_alloc_sizes): Likewise.
20016 (dump_gimple_statistics): Use PRIu64 in printf format.
20017 * gimple.h: Change uint64_t to int.
20019 2018-01-12 Martin Liska <mliska@suse.cz>
20021 * tree-core.h: Use uint64_t instead of int.
20022 * tree.c (tree_node_counts): Likewise.
20023 (tree_node_sizes): Likewise.
20024 (dump_tree_statistics): Use PRIu64 in printf format.
20026 2018-01-12 Martin Liska <mliska@suse.cz>
20028 * Makefile.in: As qsort_chk is implemented in vec.c, add
20029 vec.o to linkage of gencfn-macros.
20030 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
20031 passing the info to record_node_allocation_statistics.
20032 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
20034 * ggc-common.c (struct ggc_usage): Add operator== and use
20035 it in operator< and compare function.
20036 * mem-stats.h (struct mem_usage): Likewise.
20037 * vec.c (struct vec_usage): Remove operator< and compare
20038 function. Can be simply inherited.
20040 2018-01-12 Martin Jambor <mjambor@suse.cz>
20043 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
20044 * tree-ssa-math-opts.c: Include domwalk.h.
20045 (convert_mult_to_fma_1): New function.
20046 (fma_transformation_info): New type.
20047 (fma_deferring_state): Likewise.
20048 (cancel_fma_deferring): New function.
20049 (result_of_phi): Likewise.
20050 (last_fma_candidate_feeds_initial_phi): Likewise.
20051 (convert_mult_to_fma): Added deferring logic, split actual
20052 transformation to convert_mult_to_fma_1.
20053 (math_opts_dom_walker): New type.
20054 (math_opts_dom_walker::after_dom_children): New method, body moved
20055 here from pass_optimize_widening_mul::execute, added deferring logic
20057 (pass_optimize_widening_mul::execute): Moved most of code to
20058 math_opts_dom_walker::after_dom_children.
20059 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
20060 * config/i386/i386.c (ix86_option_override_internal): Added
20061 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
20063 2018-01-12 Richard Biener <rguenther@suse.de>
20066 * dwarf2out.c (gen_variable_die): Do not reset old_die for
20067 inline instance vars.
20069 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
20072 * config/rx/rx.c (rx_is_restricted_memory_address):
20073 Handle SUBREG case.
20075 2018-01-12 Richard Biener <rguenther@suse.de>
20077 PR tree-optimization/80846
20078 * target.def (split_reduction): New target hook.
20079 * targhooks.c (default_split_reduction): New function.
20080 * targhooks.h (default_split_reduction): Declare.
20081 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
20082 target requests first reduce vectors by combining low and high
20084 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
20085 (get_vectype_for_scalar_type_and_size): Export.
20086 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
20087 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
20088 * doc/tm.texi: Regenerate.
20089 * config/i386/i386.c (ix86_split_reduction): Implement
20090 TARGET_VECTORIZE_SPLIT_REDUCTION.
20092 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
20095 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
20096 in PIC mode except for TARGET_VXWORKS_RTP.
20097 * config/sparc/sparc.c: Include cfgrtl.h.
20098 (TARGET_INIT_PIC_REG): Define.
20099 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
20100 (sparc_pic_register_p): New predicate.
20101 (sparc_legitimate_address_p): Use it.
20102 (sparc_legitimize_pic_address): Likewise.
20103 (sparc_delegitimize_address): Likewise.
20104 (sparc_mode_dependent_address_p): Likewise.
20105 (gen_load_pcrel_sym): Remove 4th parameter.
20106 (load_got_register): Adjust call to above. Remove obsolete stuff.
20107 (sparc_expand_prologue): Do not call load_got_register here.
20108 (sparc_flat_expand_prologue): Likewise.
20109 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
20110 (sparc_use_pseudo_pic_reg): New function.
20111 (sparc_init_pic_reg): Likewise.
20112 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
20113 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
20115 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
20117 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
20118 Add item for branch_cost.
20120 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
20122 PR rtl-optimization/83565
20123 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
20124 not extend the result to a larger mode for rotate operations.
20125 (num_sign_bit_copies1): Likewise.
20127 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
20130 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
20132 Use values-Xc.o for -pedantic.
20133 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
20135 2018-01-12 Martin Liska <mliska@suse.cz>
20138 * ipa-devirt.c (final_warning_record::grow_type_warnings):
20140 (possible_polymorphic_call_targets): Use it.
20141 (ipa_devirt): Likewise.
20143 2018-01-12 Martin Liska <mliska@suse.cz>
20145 * profile-count.h (enum profile_quality): Use 0 as invalid
20146 enum value of profile_quality.
20148 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
20150 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
20151 -mext-string options.
20153 2018-01-12 Richard Biener <rguenther@suse.de>
20155 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
20156 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
20157 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
20159 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
20161 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
20163 * configure.ac (--with-long-double-format): Add support for the
20164 configuration option to change the default long double format on
20166 * config.gcc (powerpc*-linux*-*): Likewise.
20167 * configure: Regenerate.
20168 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
20169 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
20170 used without modification.
20172 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
20174 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
20175 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
20176 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
20177 MISC_BUILTIN_SPEC_BARRIER.
20178 (rs6000_init_builtins): Likewise.
20179 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
20181 (speculation_barrier): New define_insn.
20182 * doc/extend.texi: Document __builtin_speculation_barrier.
20184 2018-01-11 Jakub Jelinek <jakub@redhat.com>
20187 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
20188 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
20189 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
20191 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
20192 integral modes instead of "ss" and "sd".
20193 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
20194 vectors with 32-bit and 64-bit elements.
20195 (vecdupssescalarmodesuffix): New mode attribute.
20196 (vec_dup<mode>): Use it.
20198 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
20201 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
20202 frame if argument is passed on stack.
20204 2018-01-11 Jakub Jelinek <jakub@redhat.com>
20207 * ree.c (combine_reaching_defs): Optimize also
20208 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
20209 reg2=any_extend(exp); reg1=reg2;, formatting fix.
20211 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
20213 PR middle-end/83189
20214 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
20216 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
20218 PR middle-end/83718
20219 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
20220 after they are computed.
20222 2018-01-11 Bin Cheng <bin.cheng@arm.com>
20224 PR tree-optimization/83695
20225 * gimple-loop-linterchange.cc
20226 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
20227 reset cached scev information after interchange.
20228 (pass_linterchange::execute): Remove call to scev_reset_htab.
20230 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20232 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
20233 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
20234 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
20235 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
20236 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
20237 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
20238 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
20239 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
20240 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
20241 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
20242 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
20243 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
20244 (V_lane_reg): Likewise.
20245 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
20247 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
20248 (vfmal_lane_low<mode>_intrinsic,
20249 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
20250 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
20251 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
20252 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
20253 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
20254 vfmsl_lane_high<mode>_intrinsic): New define_insns.
20256 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20258 * config/arm/arm-cpus.in (fp16fml): New feature.
20259 (ALL_SIMD): Add fp16fml.
20260 (armv8.2-a): Add fp16fml as an option.
20261 (armv8.3-a): Likewise.
20262 (armv8.4-a): Add fp16fml as part of fp16.
20263 * config/arm/arm.h (TARGET_FP16FML): Define.
20264 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
20266 * config/arm/arm-modes.def (V2HF): Define.
20267 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
20268 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
20269 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
20270 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
20271 vfmsl_low, vfmsl_high): New set of builtins.
20272 * config/arm/iterators.md (PLUSMINUS): New code iterator.
20273 (vfml_op): New code attribute.
20274 (VFMLHALVES): New int iterator.
20275 (VFML, VFMLSEL): New mode attributes.
20276 (V_reg): Define mapping for V2HF.
20277 (V_hi, V_lo): New mode attributes.
20278 (VF_constraint): Likewise.
20279 (vfml_half, vfml_half_selector): New int attributes.
20280 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
20282 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
20283 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
20285 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
20286 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
20287 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
20288 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
20290 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
20291 Document new effective target and option set.
20293 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20295 * config/arm/arm-cpus.in (armv8_4): New feature.
20296 (ARMv8_4a): New fgroup.
20297 (armv8.4-a): New arch.
20298 * config/arm/arm-tables.opt: Regenerate.
20299 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
20300 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
20301 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
20302 Add matching rules for -march=armv8.4-a and extensions.
20303 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
20305 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
20308 * config/rx/rx.md (BW): New mode attribute.
20309 (sync_lock_test_and_setsi): Add mode suffix to insn output.
20311 2018-01-11 Richard Biener <rguenther@suse.de>
20313 PR tree-optimization/83435
20314 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
20315 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
20316 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
20318 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
20319 Alan Hayward <alan.hayward@arm.com>
20320 David Sherwood <david.sherwood@arm.com>
20322 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
20324 (aarch64_classify_address): Initialize it. Track polynomial offsets.
20325 (aarch64_print_address_internal): Use it to check for a zero offset.
20327 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
20328 Alan Hayward <alan.hayward@arm.com>
20329 David Sherwood <david.sherwood@arm.com>
20331 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
20332 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
20333 Return a poly_int64 rather than a HOST_WIDE_INT.
20334 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
20335 rather than a HOST_WIDE_INT.
20336 * config/aarch64/aarch64.h (aarch64_frame): Protect with
20337 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
20338 hard_fp_offset, frame_size, initial_adjust, callee_offset and
20339 final_offset from HOST_WIDE_INT to poly_int64.
20340 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
20341 to_constant when getting the number of units in an Advanced SIMD
20343 (aarch64_builtin_vectorized_function): Check for a constant number
20345 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
20347 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
20348 attribute instead of GET_MODE_NUNITS.
20349 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
20350 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
20351 GET_MODE_SIZE for fixed-size registers.
20352 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
20353 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
20354 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
20355 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
20356 (aarch64_print_operand, aarch64_print_address_internal)
20357 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
20358 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
20359 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
20360 Handle polynomial GET_MODE_SIZE.
20361 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
20362 wider than SImode without modification.
20363 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
20364 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
20365 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
20366 passing and returning SVE modes.
20367 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
20368 rather than GEN_INT.
20369 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
20370 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
20371 (aarch64_allocate_and_probe_stack_space): Likewise.
20372 (aarch64_layout_frame): Cope with polynomial offsets.
20373 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
20374 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
20375 polynomial offsets.
20376 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
20377 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
20378 poly_int64 rather than a HOST_WIDE_INT.
20379 (aarch64_get_separate_components, aarch64_process_components)
20380 (aarch64_expand_prologue, aarch64_expand_epilogue)
20381 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
20382 (aarch64_anchor_offset): New function, split out from...
20383 (aarch64_legitimize_address): ...here.
20384 (aarch64_builtin_vectorization_cost): Handle polynomial
20385 TYPE_VECTOR_SUBPARTS.
20386 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
20388 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
20389 number of elements from the PARALLEL rather than the mode.
20390 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
20391 rather than GET_MODE_BITSIZE.
20392 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
20393 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
20394 (aarch64_expand_vec_perm_const_1): Handle polynomial
20395 d->perm.length () and d->perm elements.
20396 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
20397 Apply to_constant to d->perm elements.
20398 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
20399 polynomial CONST_VECTOR_NUNITS.
20400 (aarch64_move_pointer): Take amount as a poly_int64 rather
20402 (aarch64_progress_pointer): Avoid temporary variable.
20403 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
20404 the mode attribute instead of GET_MODE.
20406 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
20407 Alan Hayward <alan.hayward@arm.com>
20408 David Sherwood <david.sherwood@arm.com>
20410 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
20411 x exists before using it.
20412 (aarch64_add_constant_internal): Rename to...
20413 (aarch64_add_offset_1): ...this. Replace regnum with separate
20414 src and dest rtxes. Handle the case in which they're different,
20415 including when the offset is zero. Replace scratchreg with an rtx.
20416 Use 2 additions if there is no spare register into which we can
20417 move a 16-bit constant.
20418 (aarch64_add_constant): Delete.
20419 (aarch64_add_offset): Replace reg with separate src and dest
20420 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
20421 Use aarch64_add_offset_1.
20422 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
20423 an rtx rather than an int. Take the delta as a poly_int64
20424 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
20425 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
20426 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
20427 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
20428 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
20429 and aarch64_add_sp.
20430 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
20431 aarch64_add_constant.
20433 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
20435 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
20436 Use scalar_float_mode.
20438 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
20440 * config/aarch64/aarch64-simd.md
20441 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
20442 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
20443 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
20444 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
20445 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
20446 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
20447 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
20448 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
20449 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
20450 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
20452 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
20455 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
20456 targ_options->x_arm_arch_string is non NULL.
20458 2018-01-11 Tamar Christina <tamar.christina@arm.com>
20460 * config/aarch64/aarch64.h
20461 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
20463 2018-01-11 Sudakshina Das <sudi.das@arm.com>
20466 * expmed.c (emit_store_flag_force): Swap if const op0
20467 and change VOIDmode to mode of op0.
20469 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
20471 PR rtl-optimization/83761
20472 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
20473 than bytes to mode_for_size.
20475 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
20477 PR middle-end/83189
20478 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
20479 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
20482 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
20484 PR middle-end/83575
20485 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
20486 when in layout mode.
20487 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
20488 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
20491 2018-01-10 Michael Collison <michael.collison@arm.com>
20493 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
20494 * config/aarch64/aarch64-option-extension.def: Add
20495 AARCH64_OPT_EXTENSION of 'fp16fml'.
20496 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
20497 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
20498 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
20499 * config/aarch64/constraints.md (Ui7): New constraint.
20500 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
20501 (VFMLA_SEL_W): Ditto.
20504 (VFMLA16_LOW): New int iterator.
20505 (VFMLA16_HIGH): Ditto.
20506 (UNSPEC_FMLAL): New unspec.
20507 (UNSPEC_FMLSL): Ditto.
20508 (UNSPEC_FMLAL2): Ditto.
20509 (UNSPEC_FMLSL2): Ditto.
20510 (f16mac): New code attribute.
20511 * config/aarch64/aarch64-simd-builtins.def
20512 (aarch64_fmlal_lowv2sf): Ditto.
20513 (aarch64_fmlsl_lowv2sf): Ditto.
20514 (aarch64_fmlalq_lowv4sf): Ditto.
20515 (aarch64_fmlslq_lowv4sf): Ditto.
20516 (aarch64_fmlal_highv2sf): Ditto.
20517 (aarch64_fmlsl_highv2sf): Ditto.
20518 (aarch64_fmlalq_highv4sf): Ditto.
20519 (aarch64_fmlslq_highv4sf): Ditto.
20520 (aarch64_fmlal_lane_lowv2sf): Ditto.
20521 (aarch64_fmlsl_lane_lowv2sf): Ditto.
20522 (aarch64_fmlal_laneq_lowv2sf): Ditto.
20523 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
20524 (aarch64_fmlalq_lane_lowv4sf): Ditto.
20525 (aarch64_fmlsl_lane_lowv4sf): Ditto.
20526 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
20527 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
20528 (aarch64_fmlal_lane_highv2sf): Ditto.
20529 (aarch64_fmlsl_lane_highv2sf): Ditto.
20530 (aarch64_fmlal_laneq_highv2sf): Ditto.
20531 (aarch64_fmlsl_laneq_highv2sf): Ditto.
20532 (aarch64_fmlalq_lane_highv4sf): Ditto.
20533 (aarch64_fmlsl_lane_highv4sf): Ditto.
20534 (aarch64_fmlalq_laneq_highv4sf): Ditto.
20535 (aarch64_fmlsl_laneq_highv4sf): Ditto.
20536 * config/aarch64/aarch64-simd.md:
20537 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
20538 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
20539 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
20540 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
20541 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
20542 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
20543 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
20544 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
20545 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
20546 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
20547 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
20548 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
20549 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
20550 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
20551 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
20552 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
20553 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
20554 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
20555 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
20556 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
20557 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
20558 (vfmlsl_low_u32): Ditto.
20559 (vfmlalq_low_u32): Ditto.
20560 (vfmlslq_low_u32): Ditto.
20561 (vfmlal_high_u32): Ditto.
20562 (vfmlsl_high_u32): Ditto.
20563 (vfmlalq_high_u32): Ditto.
20564 (vfmlslq_high_u32): Ditto.
20565 (vfmlal_lane_low_u32): Ditto.
20566 (vfmlsl_lane_low_u32): Ditto.
20567 (vfmlal_laneq_low_u32): Ditto.
20568 (vfmlsl_laneq_low_u32): Ditto.
20569 (vfmlalq_lane_low_u32): Ditto.
20570 (vfmlslq_lane_low_u32): Ditto.
20571 (vfmlalq_laneq_low_u32): Ditto.
20572 (vfmlslq_laneq_low_u32): Ditto.
20573 (vfmlal_lane_high_u32): Ditto.
20574 (vfmlsl_lane_high_u32): Ditto.
20575 (vfmlal_laneq_high_u32): Ditto.
20576 (vfmlsl_laneq_high_u32): Ditto.
20577 (vfmlalq_lane_high_u32): Ditto.
20578 (vfmlslq_lane_high_u32): Ditto.
20579 (vfmlalq_laneq_high_u32): Ditto.
20580 (vfmlslq_laneq_high_u32): Ditto.
20581 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
20582 (AARCH64_FL_FOR_ARCH8_4): New.
20583 (AARCH64_ISA_F16FML): New ISA flag.
20584 (TARGET_F16FML): New feature flag for fp16fml.
20585 (doc/invoke.texi): Document new fp16fml option.
20587 2018-01-10 Michael Collison <michael.collison@arm.com>
20589 * config/aarch64/aarch64-builtins.c:
20590 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
20591 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
20592 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
20593 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
20594 (AARCH64_ISA_SHA3): New ISA flag.
20595 (TARGET_SHA3): New feature flag for sha3.
20596 * config/aarch64/iterators.md (sha512_op): New int attribute.
20597 (CRYPTO_SHA512): New int iterator.
20598 (UNSPEC_SHA512H): New unspec.
20599 (UNSPEC_SHA512H2): Ditto.
20600 (UNSPEC_SHA512SU0): Ditto.
20601 (UNSPEC_SHA512SU1): Ditto.
20602 * config/aarch64/aarch64-simd-builtins.def
20603 (aarch64_crypto_sha512hqv2di): New builtin.
20604 (aarch64_crypto_sha512h2qv2di): Ditto.
20605 (aarch64_crypto_sha512su0qv2di): Ditto.
20606 (aarch64_crypto_sha512su1qv2di): Ditto.
20607 (aarch64_eor3qv8hi): Ditto.
20608 (aarch64_rax1qv2di): Ditto.
20609 (aarch64_xarqv2di): Ditto.
20610 (aarch64_bcaxqv8hi): Ditto.
20611 * config/aarch64/aarch64-simd.md:
20612 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
20613 (aarch64_crypto_sha512su0qv2di): Ditto.
20614 (aarch64_crypto_sha512su1qv2di): Ditto.
20615 (aarch64_eor3qv8hi): Ditto.
20616 (aarch64_rax1qv2di): Ditto.
20617 (aarch64_xarqv2di): Ditto.
20618 (aarch64_bcaxqv8hi): Ditto.
20619 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
20620 (vsha512h2q_u64): Ditto.
20621 (vsha512su0q_u64): Ditto.
20622 (vsha512su1q_u64): Ditto.
20623 (veor3q_u16): Ditto.
20624 (vrax1q_u64): Ditto.
20625 (vxarq_u64): Ditto.
20626 (vbcaxq_u16): Ditto.
20627 * config/arm/types.md (crypto_sha512): New type attribute.
20628 (crypto_sha3): Ditto.
20629 (doc/invoke.texi): Document new sha3 option.
20631 2018-01-10 Michael Collison <michael.collison@arm.com>
20633 * config/aarch64/aarch64-builtins.c:
20634 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
20635 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
20636 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
20637 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
20638 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
20639 (AARCH64_ISA_SM4): New ISA flag.
20640 (TARGET_SM4): New feature flag for sm4.
20641 * config/aarch64/aarch64-simd-builtins.def
20642 (aarch64_sm3ss1qv4si): Ditto.
20643 (aarch64_sm3tt1aq4si): Ditto.
20644 (aarch64_sm3tt1bq4si): Ditto.
20645 (aarch64_sm3tt2aq4si): Ditto.
20646 (aarch64_sm3tt2bq4si): Ditto.
20647 (aarch64_sm3partw1qv4si): Ditto.
20648 (aarch64_sm3partw2qv4si): Ditto.
20649 (aarch64_sm4eqv4si): Ditto.
20650 (aarch64_sm4ekeyqv4si): Ditto.
20651 * config/aarch64/aarch64-simd.md:
20652 (aarch64_sm3ss1qv4si): Ditto.
20653 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
20654 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
20655 (aarch64_sm4eqv4si): Ditto.
20656 (aarch64_sm4ekeyqv4si): Ditto.
20657 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
20658 (sm3part_op): Ditto.
20659 (CRYPTO_SM3TT): Ditto.
20660 (CRYPTO_SM3PART): Ditto.
20661 (UNSPEC_SM3SS1): New unspec.
20662 (UNSPEC_SM3TT1A): Ditto.
20663 (UNSPEC_SM3TT1B): Ditto.
20664 (UNSPEC_SM3TT2A): Ditto.
20665 (UNSPEC_SM3TT2B): Ditto.
20666 (UNSPEC_SM3PARTW1): Ditto.
20667 (UNSPEC_SM3PARTW2): Ditto.
20668 (UNSPEC_SM4E): Ditto.
20669 (UNSPEC_SM4EKEY): Ditto.
20670 * config/aarch64/constraints.md (Ui2): New constraint.
20671 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
20672 * config/arm/types.md (crypto_sm3): New type attribute.
20673 (crypto_sm4): Ditto.
20674 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
20675 (vsm3tt1aq_u32): Ditto.
20676 (vsm3tt1bq_u32): Ditto.
20677 (vsm3tt2aq_u32): Ditto.
20678 (vsm3tt2bq_u32): Ditto.
20679 (vsm3partw1q_u32): Ditto.
20680 (vsm3partw2q_u32): Ditto.
20681 (vsm4eq_u32): Ditto.
20682 (vsm4ekeyq_u32): Ditto.
20683 (doc/invoke.texi): Document new sm4 option.
20685 2018-01-10 Michael Collison <michael.collison@arm.com>
20687 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
20688 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
20689 (AARCH64_FL_FOR_ARCH8_4): New.
20690 (AARCH64_FL_V8_4): New flag.
20691 (doc/invoke.texi): Document new armv8.4-a option.
20693 2018-01-10 Michael Collison <michael.collison@arm.com>
20695 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
20696 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
20697 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
20698 * config/aarch64/aarch64-option-extension.def: Add
20699 AARCH64_OPT_EXTENSION of 'sha2'.
20700 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
20701 (crypto): Disable sha2 and aes if crypto disabled.
20702 (crypto): Enable aes and sha2 if enabled.
20703 (simd): Disable sha2 and aes if simd disabled.
20704 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
20706 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
20707 (TARGET_SHA2): New feature flag for sha2.
20708 (TARGET_AES): New feature flag for aes.
20709 * config/aarch64/aarch64-simd.md:
20710 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
20711 conditional on TARGET_AES.
20712 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
20713 (aarch64_crypto_sha1hsi): Make pattern conditional
20715 (aarch64_crypto_sha1hv4si): Ditto.
20716 (aarch64_be_crypto_sha1hv4si): Ditto.
20717 (aarch64_crypto_sha1su1v4si): Ditto.
20718 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
20719 (aarch64_crypto_sha1su0v4si): Ditto.
20720 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
20721 (aarch64_crypto_sha256su0v4si): Ditto.
20722 (aarch64_crypto_sha256su1v4si): Ditto.
20723 (doc/invoke.texi): Document new aes and sha2 options.
20725 2018-01-10 Martin Sebor <msebor@redhat.com>
20727 PR tree-optimization/83781
20728 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
20731 2018-01-11 Martin Sebor <msebor@gmail.com>
20732 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
20734 PR tree-optimization/83501
20735 PR tree-optimization/81703
20737 * tree-ssa-strlen.c (get_string_cst): Rename...
20738 (get_string_len): ...to this. Handle global constants.
20739 (handle_char_store): Adjust.
20741 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
20742 Jim Wilson <jimw@sifive.com>
20744 * config/riscv/riscv-protos.h (riscv_output_return): New.
20745 * config/riscv/riscv.c (struct machine_function): New naked_p field.
20746 (riscv_attribute_table, riscv_output_return),
20747 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
20748 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
20749 (riscv_compute_frame_info): Only compute frame->mask if not a naked
20751 (riscv_expand_prologue): Add early return for naked function.
20752 (riscv_expand_epilogue): Likewise.
20753 (riscv_function_ok_for_sibcall): Return false for naked function.
20754 (riscv_set_current_function): New.
20755 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
20756 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
20757 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
20758 * doc/extend.texi (RISC-V Function Attributes): New.
20760 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
20762 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
20763 check for 128-bit long double before checking TCmode.
20764 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
20765 128-bit long doubles before checking TFmode or TCmode.
20766 (FLOAT128_IBM_P): Likewise.
20768 2018-01-10 Martin Sebor <msebor@redhat.com>
20770 PR tree-optimization/83671
20771 * builtins.c (c_strlen): Unconditionally return zero for the empty
20773 Use -Warray-bounds for warnings.
20774 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
20775 for non-constant array indices with COMPONENT_REF, arrays of
20776 arrays, and pointers to arrays.
20777 (gimple_fold_builtin_strlen): Determine and set length range for
20778 non-constant character arrays.
20780 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
20782 PR middle-end/81897
20783 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
20786 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
20788 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
20790 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
20793 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
20794 VECTOR_MEM_ALTIVEC_OR_VSX_P.
20795 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
20796 indexed_or_indirect_operand predicate.
20797 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
20798 (*vsx_le_perm_load_v8hi): Likewise.
20799 (*vsx_le_perm_load_v16qi): Likewise.
20800 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
20801 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
20802 (*vsx_le_perm_store_v8hi): Likewise.
20803 (*vsx_le_perm_store_v16qi): Likewise.
20804 (eight unnamed splitters): Likewise.
20806 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
20808 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
20809 * config/rs6000/emmintrin.h: Likewise.
20810 * config/rs6000/mmintrin.h: Likewise.
20811 * config/rs6000/xmmintrin.h: Likewise.
20813 2018-01-10 David Malcolm <dmalcolm@redhat.com>
20816 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
20818 * tree.c (tree_nop_conversion): Return true for location wrapper
20820 (maybe_wrap_with_location): New function.
20821 (selftest::check_strip_nops): New function.
20822 (selftest::test_location_wrappers): New function.
20823 (selftest::tree_c_tests): Call it.
20824 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
20825 (maybe_wrap_with_location): New decl.
20826 (EXPR_LOCATION_WRAPPER_P): New macro.
20827 (location_wrapper_p): New inline function.
20828 (tree_strip_any_location_wrapper): New inline function.
20830 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
20833 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
20834 stack_realign_offset for the largest alignment of stack slot
20836 (ix86_find_max_used_stack_alignment): New function.
20837 (ix86_finalize_stack_frame_flags): Use it. Set
20838 max_used_stack_alignment if we don't realign stack.
20839 * config/i386/i386.h (machine_function): Add
20840 max_used_stack_alignment.
20842 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
20844 * config/arm/arm.opt (-mbranch-cost): New option.
20845 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
20848 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
20851 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
20852 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
20854 2018-01-10 Richard Biener <rguenther@suse.de>
20857 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
20858 early out so it also covers the case where we have a non-NULL
20861 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
20863 PR tree-optimization/83753
20864 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
20865 for non-strided grouped accesses if the number of elements is 1.
20867 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
20870 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
20871 * i386.h (TARGET_USE_GATHER): Define.
20872 * x86-tune.def (X86_TUNE_USE_GATHER): New.
20874 2018-01-10 Martin Liska <mliska@suse.cz>
20877 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
20878 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
20880 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
20881 CLEANUP_NO_PARTITIONING is not set.
20883 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
20885 * doc/rtl.texi: Remove documentation of (const ...) wrappers
20886 for vectors, as a partial revert of r254296.
20887 * rtl.h (const_vec_p): Delete.
20888 (const_vec_duplicate_p): Don't test for vector CONSTs.
20889 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
20890 * expmed.c (make_tree): Likewise.
20893 * common.md (E, F): Use CONSTANT_P instead of checking for
20895 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
20896 checking for CONST_VECTOR.
20898 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
20900 PR middle-end/83575
20901 * predict.c (force_edge_cold): Handle in more sane way edges
20902 with no prediction.
20904 2018-01-09 Carl Love <cel@us.ibm.com>
20906 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
20908 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
20909 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
20910 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
20911 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
20912 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
20913 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
20914 * config/rs6000/rs6000-protos.h: Add extern defition for
20915 rs6000_generate_float2_double_code.
20916 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
20918 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
20919 (float2_v2df): Add define_expand.
20921 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
20924 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
20925 op_mode in the force_to_mode call.
20927 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
20929 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
20930 instead of checking each element individually.
20931 (aarch64_evpc_uzp): Likewise.
20932 (aarch64_evpc_zip): Likewise.
20933 (aarch64_evpc_ext): Likewise.
20934 (aarch64_evpc_rev): Likewise.
20935 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
20936 instead of checking each element individually. Return true without
20938 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
20939 whether all selected elements come from the same input, instead of
20940 checking each element individually. Remove calls to gen_rtx_REG,
20941 start_sequence and end_sequence and instead assert that no rtl is
20944 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
20946 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
20947 order of HIGH and CONST checks.
20949 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
20951 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
20952 if the destination isn't an SSA_NAME.
20954 2018-01-09 Richard Biener <rguenther@suse.de>
20956 PR tree-optimization/83668
20957 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
20959 (canonicalize_loop_form): ... here, renamed from ...
20960 (canonicalize_loop_closed_ssa_form): ... this and amended to
20961 swap successor edges for loop exit blocks to make us use
20962 the RPO order we need for initial schedule generation.
20964 2018-01-09 Joseph Myers <joseph@codesourcery.com>
20966 PR tree-optimization/64811
20967 * match.pd: When optimizing comparisons with Inf, avoid
20968 introducing or losing exceptions from comparisons with NaN.
20970 2018-01-09 Martin Liska <mliska@suse.cz>
20973 * asan.c (shadow_mem_size): Add gcc_assert.
20975 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
20977 Don't save registers in main().
20980 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
20981 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
20982 * config/avr/avr.c (avr_set_current_function): Don't error if
20983 naked, OS_task or OS_main are specified at the same time.
20984 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
20986 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
20988 * common/config/avr/avr-common.c (avr_option_optimization_table):
20989 Switch on -mmain-is-OS_task for optimizing compilations.
20991 2018-01-09 Richard Biener <rguenther@suse.de>
20993 PR tree-optimization/83572
20994 * graphite.c: Include cfganal.h.
20995 (graphite_transform_loops): Connect infinite loops to exit
20996 and remove fake edges at the end.
20998 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
21000 * ipa-inline.c (edge_badness): Revert accidental checkin.
21002 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
21005 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
21006 symbols; not inline clones.
21008 2018-01-09 Jakub Jelinek <jakub@redhat.com>
21011 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
21012 hard registers. Formatting fixes.
21014 PR preprocessor/83722
21015 * gcc.c (try_generate_repro): Pass
21016 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
21017 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
21020 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
21021 Kito Cheng <kito.cheng@gmail.com>
21023 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
21024 (riscv_leaf_function_p): Delete.
21025 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
21027 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
21029 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
21031 (do_ifelse): New function.
21032 (do_isel): New function.
21033 (do_sub3): New function.
21034 (do_add3): New function.
21035 (do_load_mask_compare): New function.
21036 (do_overlap_load_compare): New function.
21037 (expand_compare_loop): New function.
21038 (expand_block_compare): Call expand_compare_loop() when appropriate.
21039 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
21040 option description.
21041 (-mblock-compare-inline-loop-limit): New option.
21043 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
21046 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
21047 Reverse order of second and third operands in first alternative.
21048 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
21049 of first and second elements in UNSPEC_VPERMR vector.
21050 (altivec_expand_vec_perm_le): Likewise.
21052 2018-01-08 Jeff Law <law@redhat.com>
21054 PR rtl-optimizatin/81308
21055 * tree-switch-conversion.c (cfg_altered): New file scoped static.
21056 (process_switch): If group_case_labels makes a change, then set
21058 (pass_convert_switch::execute): If a switch is converted, then
21059 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
21061 PR rtl-optimization/81308
21062 * recog.c (split_all_insns): Conditionally cleanup the CFG after
21065 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
21067 PR target/83663 - Revert r255946
21068 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
21069 generation for cases where splatting a value is not useful.
21070 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
21071 across a vec_duplicate and a paradoxical subreg forming a vector
21072 mode to a vec_concat.
21074 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21076 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
21077 -march=armv8.3-a variants.
21078 * config/arm/t-multilib: Likewise.
21079 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
21081 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
21083 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
21085 (cceq_ior_compare_complement): Give it a name so I can use it, and
21086 change boolean_or_operator predicate to boolean_operator so it can
21087 be used to generate a crand.
21088 (eqne): New code iterator.
21089 (bd/bd_neg): New code_attrs.
21090 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
21091 a single define_insn.
21092 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
21093 decrement (bdnzt/bdnzf/bdzt/bdzf).
21094 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
21095 with the new names of the branch decrement patterns, and added the
21096 names of the branch decrement conditional patterns.
21098 2018-01-08 Richard Biener <rguenther@suse.de>
21100 PR tree-optimization/83563
21101 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
21104 2018-01-08 Richard Biener <rguenther@suse.de>
21106 PR middle-end/83713
21107 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
21109 2018-01-08 Richard Biener <rguenther@suse.de>
21111 PR tree-optimization/83685
21112 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
21113 references to abnormals.
21115 2018-01-08 Richard Biener <rguenther@suse.de>
21118 * dwarf2out.c (output_indirect_strings): Handle empty
21119 skeleton_debug_str_hash.
21120 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
21122 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
21124 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
21125 (emit_store_direct): Likewise.
21126 (arc_trampoline_adjust_address): Likewise.
21127 (arc_asm_trampoline_template): New function.
21128 (arc_initialize_trampoline): Use asm_trampoline_template.
21129 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
21130 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
21131 * config/arc/arc.md (flush_icache): Delete pattern.
21133 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
21135 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
21136 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
21139 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
21142 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
21143 by not USED_FOR_TARGET.
21144 (make_pass_resolve_sw_modes): Likewise.
21146 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
21148 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
21151 2018-01-08 Richard Biener <rguenther@suse.de>
21153 PR middle-end/83580
21154 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
21156 2018-01-08 Richard Biener <rguenther@suse.de>
21158 PR middle-end/83517
21159 * match.pd ((t * 2) / 2) -> t): Add missing :c.
21161 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
21163 PR middle-end/81897
21164 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
21165 basic blocks with a small number of successors.
21166 (convert_control_dep_chain_into_preds): Improve handling of
21168 (dump_predicates): Split apart into...
21169 (dump_pred_chain): ...here...
21170 (dump_pred_info): ...and here.
21171 (can_one_predicate_be_invalidated_p): Add debugging printfs.
21172 (can_chain_union_be_invalidated_p): Improve check for invalidation
21174 (uninit_uses_cannot_happen): Avoid unnecessary if
21175 convert_control_dep_chain_into_preds yielded nothing.
21177 2018-01-06 Martin Sebor <msebor@redhat.com>
21179 PR tree-optimization/83640
21180 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
21181 subtracting negative offset from size.
21182 (builtin_access::overlap): Adjust offset bounds of the access to fall
21183 within the size of the object if possible.
21185 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
21187 PR rtl-optimization/83699
21188 * expmed.c (extract_bit_field_1): Restrict the vector usage of
21189 extract_bit_field_as_subreg to cases in which the extracted
21190 value is also a vector.
21192 * lra-constraints.c (process_alt_operands): Test for the equivalence
21193 substitutions when detecting a possible reload cycle.
21195 2018-01-06 Jakub Jelinek <jakub@redhat.com>
21198 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
21199 by default if flag_selective_schedling{,2}. Formatting fixes.
21201 PR rtl-optimization/83682
21202 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
21203 if it has non-VECTOR_MODE element mode.
21204 (vec_duplicate_p): Likewise.
21206 PR middle-end/83694
21207 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
21208 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
21210 2018-01-05 Jakub Jelinek <jakub@redhat.com>
21213 * config/i386/i386-builtin.def
21214 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
21215 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
21216 Require also OPTION_MASK_ISA_AVX512F in addition to
21217 OPTION_MASK_ISA_GFNI.
21218 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
21219 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
21220 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
21221 to OPTION_MASK_ISA_GFNI.
21222 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
21223 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
21224 OPTION_MASK_ISA_AVX512BW.
21225 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
21226 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
21227 addition to OPTION_MASK_ISA_GFNI.
21228 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
21229 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
21230 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
21231 to OPTION_MASK_ISA_GFNI.
21232 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
21233 a requirement for all ISAs rather than any of them with a few
21235 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
21237 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
21238 bitmasks to be enabled with 3 exceptions, instead of requiring any
21239 enabled ISA with lots of exceptions.
21240 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
21241 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
21242 Change avx512bw in isa attribute to avx512f.
21243 * config/i386/sgxintrin.h: Add license boilerplate.
21244 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
21245 to __AVX512F__ and __AVX512VL to __AVX512VL__.
21246 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
21247 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
21249 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
21250 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
21251 temporarily sse2 rather than sse if not enabled already.
21254 * config/i386/sse.md (VI248_VLBW): Rename to ...
21255 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
21256 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
21257 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
21258 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
21259 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
21260 mode iterator instead of VI248_VLBW.
21262 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
21264 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
21265 (record_modified): Skip clobbers; add debug output.
21266 (param_change_prob): Use sreal frequencies.
21268 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
21270 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
21271 punt for user-aligned variables.
21273 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
21275 * tree-chrec.c (chrec_contains_symbols): Return true for
21278 2018-01-05 Sudakshina Das <sudi.das@arm.com>
21281 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
21282 of (x|y) == x for BICS pattern.
21284 2018-01-05 Jakub Jelinek <jakub@redhat.com>
21286 PR tree-optimization/83605
21287 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
21288 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
21291 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
21293 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
21294 * config/epiphany/rtems.h: New file.
21296 2018-01-04 Jakub Jelinek <jakub@redhat.com>
21297 Uros Bizjak <ubizjak@gmail.com>
21300 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
21301 QIreg_operand instead of register_operand predicate.
21302 * config/i386/i386.c (ix86_rop_should_change_byte_p,
21303 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
21304 comments instead of -fmitigate[-_]rop.
21306 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
21309 * cgraphunit.c (symbol_table::compile): Switch to text_section
21310 before calling assembly_start debug hook.
21311 * run-rtl-passes.c (run_rtl_passes): Likewise.
21314 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
21316 * tree-vrp.c (extract_range_from_binary_expr_1): Check
21317 range_int_cst_p rather than !symbolic_range_p before calling
21318 extract_range_from_multiplicative_op_1.
21320 2018-01-04 Jeff Law <law@redhat.com>
21322 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
21323 redundant test in assertion.
21325 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
21327 * doc/rtl.texi: Document machine_mode wrapper classes.
21329 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
21331 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
21332 using tree_to_uhwi.
21334 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
21336 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
21337 the VEC_PERM_EXPR fold to fail.
21339 2018-01-04 Jakub Jelinek <jakub@redhat.com>
21342 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
21343 to switched_sections.
21345 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
21348 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
21349 test for d.testing.
21351 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
21354 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
21355 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
21357 2018-01-04 Jakub Jelinek <jakub@redhat.com>
21360 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
21361 is BLKmode and bitpos not zero or mode change is needed.
21363 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
21366 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
21369 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
21372 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
21373 instead of MULT rtx. Update all corresponding splitters.
21374 (*saddl_se): Ditto.
21375 (*ssub<modesuffix>): Ditto.
21376 (*ssubl_se): Ditto.
21377 (*cmp_sadd_di): Update split patterns.
21378 (*cmp_sadd_si): Ditto.
21379 (*cmp_sadd_sidi): Ditto.
21380 (*cmp_ssub_di): Ditto.
21381 (*cmp_ssub_si): Ditto.
21382 (*cmp_ssub_sidi): Ditto.
21383 * config/alpha/predicates.md (const23_operand): New predicate.
21384 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
21385 Look for ASHIFT, not MULT inner operand.
21386 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
21388 2018-01-04 Martin Liska <mliska@suse.cz>
21390 PR gcov-profile/83669
21391 * gcov.c (output_intermediate_file): Add version to intermediate
21393 * doc/gcov.texi: Document new field 'version' in intermediate
21394 file format. Fix location of '-k' option of gcov command.
21396 2018-01-04 Martin Liska <mliska@suse.cz>
21399 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
21401 2018-01-04 Jakub Jelinek <jakub@redhat.com>
21403 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
21405 2018-01-03 Martin Sebor <msebor@redhat.com>
21407 PR tree-optimization/83655
21408 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
21409 checking calls with invalid arguments.
21411 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21413 * tree-vect-stmts.c (vect_get_store_rhs): New function.
21414 (vectorizable_mask_load_store): Delete.
21415 (vectorizable_call): Return false for masked loads and stores.
21416 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
21417 instead of gimple_assign_rhs1.
21418 (vectorizable_load): Handle IFN_MASK_LOAD.
21419 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
21421 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21423 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
21425 (vectorizable_mask_load_store): ...here.
21426 (vectorizable_load): ...and here.
21428 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21430 * tree-vect-stmts.c (vect_build_all_ones_mask)
21431 (vect_build_zero_merge_argument): New functions, split out from...
21432 (vectorizable_load): ...here.
21434 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21436 * tree-vect-stmts.c (vect_check_store_rhs): New function,
21438 (vectorizable_mask_load_store): ...here.
21439 (vectorizable_store): ...and here.
21441 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21443 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
21445 (vectorizable_mask_load_store): ...here.
21447 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21449 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
21450 (vect_model_store_cost): Take a vec_load_store_type instead of a
21452 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
21453 (vect_model_store_cost): Take a vec_load_store_type instead of a
21455 (vectorizable_mask_load_store): Update accordingly.
21456 (vectorizable_store): Likewise.
21457 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
21459 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21461 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
21462 IFN_MASK_LOAD calls here rather than...
21463 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
21465 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21466 Alan Hayward <alan.hayward@arm.com>
21467 David Sherwood <david.sherwood@arm.com>
21469 * expmed.c (extract_bit_field_1): For vector extracts,
21470 fall back to extract_bit_field_as_subreg if vec_extract
21473 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21474 Alan Hayward <alan.hayward@arm.com>
21475 David Sherwood <david.sherwood@arm.com>
21477 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
21478 they are variable or constant sized.
21479 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
21480 slots for constant-sized data.
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 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
21487 handling COND_EXPRs with boolean comparisons, try to find a better
21488 basis for the mask type than the boolean itself.
21490 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21492 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
21493 is calculated and how it can be overridden.
21494 * genmodes.c (max_bitsize_mode_any_mode): New variable.
21495 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
21497 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
21500 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21501 Alan Hayward <alan.hayward@arm.com>
21502 David Sherwood <david.sherwood@arm.com>
21504 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
21505 Remove the mode argument.
21506 (aarch64_simd_valid_immediate): Remove the mode and inverse
21508 * config/aarch64/iterators.md (bitsize): New iterator.
21509 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
21510 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
21511 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
21512 aarch64_simd_valid_immediate.
21513 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
21514 (aarch64_reg_or_bic_imm): Likewise.
21515 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
21516 with an insn_type enum and msl with a modifier_type enum.
21517 Replace element_width with a scalar_mode. Change the shift
21518 to unsigned int. Add constructors for scalar_float_mode and
21519 scalar_int_mode elements.
21520 (aarch64_vect_float_const_representable_p): Delete.
21521 (aarch64_can_const_movi_rtx_p)
21522 (aarch64_simd_scalar_immediate_valid_for_move)
21523 (aarch64_simd_make_constant): Update call to
21524 aarch64_simd_valid_immediate.
21525 (aarch64_advsimd_valid_immediate_hs): New function.
21526 (aarch64_advsimd_valid_immediate): Likewise.
21527 (aarch64_simd_valid_immediate): Remove mode and inverse
21528 arguments. Rewrite to use the above. Use const_vec_duplicate_p
21529 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
21530 and aarch64_float_const_representable_p on the result.
21531 (aarch64_output_simd_mov_immediate): Remove mode argument.
21532 Update call to aarch64_simd_valid_immediate and use of
21533 simd_immediate_info.
21534 (aarch64_output_scalar_simd_mov_immediate): Update call
21537 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21538 Alan Hayward <alan.hayward@arm.com>
21539 David Sherwood <david.sherwood@arm.com>
21541 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
21542 (mode_nunits): Likewise CONST_MODE_NUNITS.
21543 * machmode.def (ADJUST_NUNITS): Document.
21544 * genmodes.c (mode_data::need_nunits_adj): New field.
21545 (blank_mode): Update accordingly.
21546 (adj_nunits): New variable.
21547 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
21549 (emit_mode_size_inline): Set need_bytesize_adj for all modes
21550 listed in adj_nunits.
21551 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
21552 listed in adj_nunits. Don't emit case statements for such modes.
21553 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
21554 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
21555 nothing if adj_nunits is nonnull.
21556 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
21557 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
21558 (emit_mode_fbit): Update use of print_maybe_const_decl.
21559 (emit_move_size): Likewise. Treat the array as non-const
21561 (emit_mode_adjustments): Handle adj_nunits.
21563 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21565 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
21566 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
21567 (VECTOR_MODES): Use it.
21568 (make_vector_modes): Take the prefix as an argument.
21570 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21571 Alan Hayward <alan.hayward@arm.com>
21572 David Sherwood <david.sherwood@arm.com>
21574 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
21575 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
21576 for MODE_VECTOR_BOOL.
21577 * machmode.def (VECTOR_BOOL_MODE): Document.
21578 * genmodes.c (VECTOR_BOOL_MODE): New macro.
21579 (make_vector_bool_mode): New function.
21580 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
21582 * lto-streamer-in.c (lto_input_mode_table): Likewise.
21583 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
21585 * stor-layout.c (int_mode_for_mode): Likewise.
21586 * tree.c (build_vector_type_for_mode): Likewise.
21587 * varasm.c (output_constant_pool_2): Likewise.
21588 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
21589 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
21590 for MODE_VECTOR_BOOL.
21591 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
21592 of mode class checks.
21593 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
21594 instead of a list of mode class checks.
21595 (expand_vector_scalar_condition): Likewise.
21596 (type_for_widest_vector_mode): Handle BImode as an inner mode.
21598 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21599 Alan Hayward <alan.hayward@arm.com>
21600 David Sherwood <david.sherwood@arm.com>
21602 * machmode.h (mode_size): Change from unsigned short to
21604 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
21605 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
21606 or if measurement_type is not polynomial.
21607 (fixed_size_mode::includes_p): Check for constant-sized modes.
21608 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
21609 return a poly_uint16 rather than an unsigned short.
21610 (emit_mode_size): Change the type of mode_size from unsigned short
21611 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
21612 (emit_mode_adjustments): Cope with polynomial vector sizes.
21613 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
21615 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
21617 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
21618 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
21619 * caller-save.c (setup_save_areas): Likewise.
21620 (replace_reg_with_saved_mem): Likewise.
21621 * calls.c (emit_library_call_value_1): Likewise.
21622 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
21623 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
21624 (gen_lowpart_for_combine): Likewise.
21625 * convert.c (convert_to_integer_1): Likewise.
21626 * cse.c (equiv_constant, cse_insn): Likewise.
21627 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
21628 (cselib_subst_to_values): Likewise.
21629 * dce.c (word_dce_process_block): Likewise.
21630 * df-problems.c (df_word_lr_mark_ref): Likewise.
21631 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
21632 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
21633 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
21634 (rtl_for_decl_location): Likewise.
21635 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
21636 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
21637 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
21638 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
21639 (expand_expr_real_1): Likewise.
21640 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
21641 (pad_below): Likewise.
21642 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
21643 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
21644 * ira.c (get_subreg_tracking_sizes): Likewise.
21645 * ira-build.c (ira_create_allocno_objects): Likewise.
21646 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
21647 (ira_sort_regnos_for_alter_reg): Likewise.
21648 * ira-costs.c (record_operand_costs): Likewise.
21649 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
21650 (resolve_simple_move): Likewise.
21651 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
21652 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
21653 (lra_constraints): Likewise.
21654 (CONST_POOL_OK_P): Reject variable-sized modes.
21655 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
21656 (add_pseudo_to_slot, lra_spill): Likewise.
21657 * omp-low.c (omp_clause_aligned_alignment): Likewise.
21658 * optabs-query.c (get_best_extraction_insn): Likewise.
21659 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
21660 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
21661 (expand_mult_highpart, valid_multiword_target_p): Likewise.
21662 * recog.c (offsettable_address_addr_space_p): Likewise.
21663 * regcprop.c (maybe_mode_change): Likewise.
21664 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
21665 * regrename.c (build_def_use): Likewise.
21666 * regstat.c (dump_reg_info): Likewise.
21667 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
21668 (find_reloads, find_reloads_subreg_address): Likewise.
21669 * reload1.c (eliminate_regs_1): Likewise.
21670 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
21671 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
21672 (simplify_binary_operation_1, simplify_subreg): Likewise.
21673 * targhooks.c (default_function_arg_padding): Likewise.
21674 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
21675 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
21676 (verify_gimple_assign_ternary): Likewise.
21677 * tree-inline.c (estimate_move_cost): Likewise.
21678 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
21679 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
21680 (get_address_cost_ainc): Likewise.
21681 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
21682 (vect_supportable_dr_alignment): Likewise.
21683 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
21684 (vectorizable_reduction): Likewise.
21685 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
21686 (vectorizable_operation, vectorizable_load): Likewise.
21687 * tree.c (build_same_sized_truth_vector_type): Likewise.
21688 * valtrack.c (cleanup_auto_inc_dec): Likewise.
21689 * var-tracking.c (emit_note_insn_var_location): Likewise.
21690 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
21691 (ADDR_VEC_ALIGN): Likewise.
21693 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21694 Alan Hayward <alan.hayward@arm.com>
21695 David Sherwood <david.sherwood@arm.com>
21697 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
21699 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
21700 or if measurement_type is polynomial.
21701 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
21702 * combine.c (make_extraction): Likewise.
21703 * dse.c (find_shift_sequence): Likewise.
21704 * dwarf2out.c (mem_loc_descriptor): Likewise.
21705 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
21706 (extract_bit_field, extract_low_bits): Likewise.
21707 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
21708 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
21709 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
21710 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
21711 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
21712 * reload.c (find_reloads): Likewise.
21713 * reload1.c (alter_reg): Likewise.
21714 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
21715 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
21716 * tree-if-conv.c (predicate_mem_writes): Likewise.
21717 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
21718 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
21719 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
21720 * valtrack.c (dead_debug_insert_temp): Likewise.
21721 * varasm.c (mergeable_constant_section): Likewise.
21722 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
21724 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21725 Alan Hayward <alan.hayward@arm.com>
21726 David Sherwood <david.sherwood@arm.com>
21728 * expr.c (expand_assignment): Cope with polynomial mode sizes
21729 when assigning to a CONCAT.
21731 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21732 Alan Hayward <alan.hayward@arm.com>
21733 David Sherwood <david.sherwood@arm.com>
21735 * machmode.h (mode_precision): Change from unsigned short to
21737 (mode_to_precision): Return a poly_uint16 rather than an unsigned
21739 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
21740 or if measurement_type is not polynomial.
21741 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
21742 in which the mode is already known to be a scalar_int_mode.
21743 * genmodes.c (emit_mode_precision): Change the type of mode_precision
21744 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
21746 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
21747 for GET_MODE_PRECISION.
21748 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
21749 for GET_MODE_PRECISION.
21750 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
21752 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
21753 (expand_field_assignment, make_extraction): Likewise.
21754 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
21755 (get_last_value): Likewise.
21756 * convert.c (convert_to_integer_1): Likewise.
21757 * cse.c (cse_insn): Likewise.
21758 * expr.c (expand_expr_real_1): Likewise.
21759 * lra-constraints.c (simplify_operand_subreg): Likewise.
21760 * optabs-query.c (can_atomic_load_p): Likewise.
21761 * optabs.c (expand_atomic_load): Likewise.
21762 (expand_atomic_store): Likewise.
21763 * ree.c (combine_reaching_defs): Likewise.
21764 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
21765 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
21766 * tree.h (type_has_mode_precision_p): Likewise.
21767 * ubsan.c (instrument_si_overflow): Likewise.
21769 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21770 Alan Hayward <alan.hayward@arm.com>
21771 David Sherwood <david.sherwood@arm.com>
21773 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
21774 polynomial numbers of units.
21775 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
21776 (valid_vector_subparts_p): New function.
21777 (build_vector_type): Remove temporary shim and take the number
21778 of units as a poly_uint64 rather than an int.
21779 (build_opaque_vector_type): Take the number of units as a
21780 poly_uint64 rather than an int.
21781 * tree.c (build_vector_from_ctor): Handle polynomial
21782 TYPE_VECTOR_SUBPARTS.
21783 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
21784 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
21785 (build_vector_from_val): If the number of units is variable,
21786 use build_vec_duplicate_cst for constant operands and
21787 VEC_DUPLICATE_EXPR otherwise.
21788 (make_vector_type): Remove temporary is_constant ().
21789 (build_vector_type, build_opaque_vector_type): Take the number of
21790 units as a poly_uint64 rather than an int.
21791 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
21793 * cfgexpand.c (expand_debug_expr): Likewise.
21794 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
21795 (store_constructor, expand_expr_real_1): Likewise.
21796 (const_scalar_mask_from_tree): Likewise.
21797 * fold-const-call.c (fold_const_reduction): Likewise.
21798 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
21799 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
21800 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
21801 (fold_relational_const): Likewise.
21802 (native_interpret_vector): Likewise. Change the size from an
21803 int to an unsigned int.
21804 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
21805 TYPE_VECTOR_SUBPARTS.
21806 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
21807 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
21808 duplicating a non-constant operand into a variable-length vector.
21809 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
21810 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
21811 * ipa-icf.c (sem_variable::equals): Likewise.
21812 * match.pd: Likewise.
21813 * omp-simd-clone.c (simd_clone_subparts): Likewise.
21814 * print-tree.c (print_node): Likewise.
21815 * stor-layout.c (layout_type): Likewise.
21816 * targhooks.c (default_builtin_vectorization_cost): Likewise.
21817 * tree-cfg.c (verify_gimple_comparison): Likewise.
21818 (verify_gimple_assign_binary): Likewise.
21819 (verify_gimple_assign_ternary): Likewise.
21820 (verify_gimple_assign_single): Likewise.
21821 * tree-pretty-print.c (dump_generic_node): Likewise.
21822 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
21823 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
21824 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
21825 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
21826 (vect_shift_permute_load_chain): Likewise.
21827 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
21828 (expand_vector_condition, optimize_vector_constructor): Likewise.
21829 (lower_vec_perm, get_compute_type): Likewise.
21830 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
21831 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
21832 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
21833 (vect_recog_mask_conversion_pattern): Likewise.
21834 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
21835 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
21836 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
21837 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
21838 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
21839 (vectorizable_shift, vectorizable_operation, vectorizable_store)
21840 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
21841 (supportable_widening_operation): Likewise.
21842 (supportable_narrowing_operation): Likewise.
21843 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
21845 * varasm.c (output_constant): Likewise.
21847 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21848 Alan Hayward <alan.hayward@arm.com>
21849 David Sherwood <david.sherwood@arm.com>
21851 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
21852 so that both the length == 3 and length != 3 cases set up their
21853 own permute vectors. Add comments explaining why we know the
21854 number of elements is constant.
21855 (vect_permute_load_chain): Likewise.
21857 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21858 Alan Hayward <alan.hayward@arm.com>
21859 David Sherwood <david.sherwood@arm.com>
21861 * machmode.h (mode_nunits): Change from unsigned char to
21863 (ONLY_FIXED_SIZE_MODES): New macro.
21864 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
21865 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
21866 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
21868 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
21869 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
21870 or if measurement_type is not polynomial.
21871 * genmodes.c (ZERO_COEFFS): New macro.
21872 (emit_mode_nunits_inline): Make mode_nunits_inline return a
21874 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
21875 Use ZERO_COEFFS when emitting initializers.
21876 * data-streamer.h (bp_pack_poly_value): New function.
21877 (bp_unpack_poly_value): Likewise.
21878 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
21879 for GET_MODE_NUNITS.
21880 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
21881 for GET_MODE_NUNITS.
21882 * tree.c (make_vector_type): Remove temporary shim and make
21883 the real function take the number of units as a poly_uint64
21884 rather than an int.
21885 (build_vector_type_for_mode): Handle polynomial nunits.
21886 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
21887 * emit-rtl.c (const_vec_series_p_1): Likewise.
21888 (gen_rtx_CONST_VECTOR): Likewise.
21889 * fold-const.c (test_vec_duplicate_folding): Likewise.
21890 * genrecog.c (validate_pattern): Likewise.
21891 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
21892 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
21893 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
21894 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
21895 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
21896 * rtlanal.c (subreg_get_info): Likewise.
21897 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
21898 (vect_grouped_load_supported): Likewise.
21899 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
21900 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
21901 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
21902 (simplify_const_unary_operation, simplify_binary_operation_1)
21903 (simplify_const_binary_operation, simplify_ternary_operation)
21904 (test_vector_ops_duplicate, test_vector_ops): Likewise.
21905 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
21906 instead of CONST_VECTOR_NUNITS.
21907 * varasm.c (output_constant_pool_2): Likewise.
21908 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
21909 explicit-encoded elements in the XVEC for variable-length vectors.
21911 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21913 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
21915 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21916 Alan Hayward <alan.hayward@arm.com>
21917 David Sherwood <david.sherwood@arm.com>
21919 * coretypes.h (fixed_size_mode): Declare.
21920 (fixed_size_mode_pod): New typedef.
21921 * builtins.h (target_builtins::x_apply_args_mode)
21922 (target_builtins::x_apply_result_mode): Change type to
21923 fixed_size_mode_pod.
21924 * builtins.c (apply_args_size, apply_result_size, result_vector)
21925 (expand_builtin_apply_args_1, expand_builtin_apply)
21926 (expand_builtin_return): Update accordingly.
21928 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21930 * cse.c (hash_rtx_cb): Hash only the encoded elements.
21931 * cselib.c (cselib_hash_rtx): Likewise.
21932 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
21933 CONST_VECTOR encoding.
21935 2018-01-03 Jakub Jelinek <jakub@redhat.com>
21936 Jeff Law <law@redhat.com>
21939 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
21940 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
21941 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
21942 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
21945 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
21946 explicitly probe *sp in a noreturn function if there were any callee
21947 register saves or frame pointer is needed.
21949 2018-01-03 Jakub Jelinek <jakub@redhat.com>
21952 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
21953 BLKmode for ternary, binary or unary expressions.
21956 * var-tracking.c (delete_vta_debug_insn): New inline function.
21957 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
21958 insns from get_insns () to NULL instead of each bb separately.
21959 Use delete_vta_debug_insn. No longer static.
21960 (vt_debug_insns_local, variable_tracking_main_1): Adjust
21961 delete_vta_debug_insns callers.
21962 * rtl.h (delete_vta_debug_insns): Declare.
21963 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
21964 instead of variable_tracking_main.
21966 2018-01-03 Martin Sebor <msebor@redhat.com>
21968 PR tree-optimization/83603
21969 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
21970 arguments past the endof the argument list in functions declared
21971 without a prototype.
21972 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
21973 Avoid checking when arguments are null.
21975 2018-01-03 Martin Sebor <msebor@redhat.com>
21978 * doc/extend.texi (attribute const): Fix a typo.
21979 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
21980 issuing -Wsuggest-attribute for void functions.
21982 2018-01-03 Martin Sebor <msebor@redhat.com>
21984 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
21985 offset_int::from instead of wide_int::to_shwi.
21986 (maybe_diag_overlap): Remove assertion.
21987 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
21988 * gimple-ssa-sprintf.c (format_directive): Same.
21989 (parse_directive): Same.
21990 (sprintf_dom_walker::compute_format_length): Same.
21991 (try_substitute_return_value): Same.
21993 2018-01-03 Jeff Law <law@redhat.com>
21995 PR middle-end/83654
21996 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
21997 non-constant residual for zero at runtime and avoid probing in
21998 that case. Reorganize code for trailing problem to mirror handling
22001 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
22003 PR tree-optimization/83501
22004 * tree-ssa-strlen.c (get_string_cst): New.
22005 (handle_char_store): Call get_string_cst.
22007 2018-01-03 Martin Liska <mliska@suse.cz>
22009 PR tree-optimization/83593
22010 * tree-ssa-strlen.c: Include tree-cfg.h.
22011 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
22012 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
22013 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
22015 (strlen_dom_walker::before_dom_children): Call
22016 gimple_purge_dead_eh_edges. Dump tranformation with details
22018 (strlen_dom_walker::before_dom_children): Update call by adding
22019 new argument cleanup_eh.
22020 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
22022 2018-01-03 Martin Liska <mliska@suse.cz>
22025 * cif-code.def (VARIADIC_THUNK): New enum value.
22026 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
22029 2018-01-03 Jan Beulich <jbeulich@suse.com>
22031 * sse.md (mov<mode>_internal): Tighten condition for when to use
22032 vmovdqu<ssescalarsize> for TI and OI modes.
22034 2018-01-03 Jakub Jelinek <jakub@redhat.com>
22036 Update copyright years.
22038 2018-01-03 Martin Liska <mliska@suse.cz>
22041 * ipa-visibility.c (function_and_variable_visibility): Skip
22042 functions with noipa attribure.
22044 2018-01-03 Jakub Jelinek <jakub@redhat.com>
22046 * gcc.c (process_command): Update copyright notice dates.
22047 * gcov-dump.c (print_version): Ditto.
22048 * gcov.c (print_version): Ditto.
22049 * gcov-tool.c (print_version): Ditto.
22050 * gengtype.c (create_file): Ditto.
22051 * doc/cpp.texi: Bump @copying's copyright year.
22052 * doc/cppinternals.texi: Ditto.
22053 * doc/gcc.texi: Ditto.
22054 * doc/gccint.texi: Ditto.
22055 * doc/gcov.texi: Ditto.
22056 * doc/install.texi: Ditto.
22057 * doc/invoke.texi: Ditto.
22059 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22061 * vector-builder.h (vector_builder::m_full_nelts): Change from
22062 unsigned int to poly_uint64.
22063 (vector_builder::full_nelts): Update prototype accordingly.
22064 (vector_builder::new_vector): Likewise.
22065 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
22066 (vector_builder::operator ==): Likewise.
22067 (vector_builder::finalize): Likewise.
22068 * int-vector-builder.h (int_vector_builder::int_vector_builder):
22069 Take the number of elements as a poly_uint64 rather than an
22071 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
22072 from unsigned int to poly_uint64.
22073 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
22074 (vec_perm_indices::new_vector): Likewise.
22075 (vec_perm_indices::length): Likewise.
22076 (vec_perm_indices::nelts_per_input): Likewise.
22077 (vec_perm_indices::input_nelts): Likewise.
22078 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
22079 number of elements per input as a poly_uint64 rather than an
22080 unsigned int. Use the original encoding for variable-length
22081 vectors, rather than clamping each individual element.
22082 For the second and subsequent elements in each pattern,
22083 clamp the step and base before clamping their sum.
22084 (vec_perm_indices::series_p): Handle polynomial element counts.
22085 (vec_perm_indices::all_in_range_p): Likewise.
22086 (vec_perm_indices_to_tree): Likewise.
22087 (vec_perm_indices_to_rtx): Likewise.
22088 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
22089 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
22090 (tree_vector_builder::new_binary_operation): Handle polynomial
22091 element counts. Return false if we need to know the number
22092 of elements at compile time.
22093 * fold-const.c (fold_vec_perm): Punt if the number of elements
22094 isn't known at compile time.
22096 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22098 * vec-perm-indices.h (vec_perm_builder): Change element type
22099 from HOST_WIDE_INT to poly_int64.
22100 (vec_perm_indices::element_type): Update accordingly.
22101 (vec_perm_indices::clamp): Handle polynomial element_types.
22102 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
22103 (vec_perm_indices::all_in_range_p): Likewise.
22104 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
22106 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
22107 polynomial vec_perm_indices element types.
22108 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
22109 * fold-const.c (fold_vec_perm): Likewise.
22110 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
22111 * tree-vect-generic.c (lower_vec_perm): Likewise.
22112 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
22113 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
22114 element type to HOST_WIDE_INT.
22116 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22117 Alan Hayward <alan.hayward@arm.com>
22118 David Sherwood <david.sherwood@arm.com>
22120 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
22121 rather than an int. Use plus_constant.
22122 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
22123 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
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 * calls.c (emit_call_1, expand_call): Change struct_value_size from
22130 a HOST_WIDE_INT to a poly_int64.
22132 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22133 Alan Hayward <alan.hayward@arm.com>
22134 David Sherwood <david.sherwood@arm.com>
22136 * calls.c (load_register_parameters): Cope with polynomial
22137 mode sizes. Require a constant size for BLKmode parameters
22138 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
22139 forces a parameter to be padded at the lsb end in order to
22140 fill a complete number of words, require the parameter size
22141 to be ordered wrt UNITS_PER_WORD.
22143 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22144 Alan Hayward <alan.hayward@arm.com>
22145 David Sherwood <david.sherwood@arm.com>
22147 * reload1.c (spill_stack_slot_width): Change element type
22148 from unsigned int to poly_uint64_pod.
22149 (alter_reg): Treat mode sizes as polynomial.
22151 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22152 Alan Hayward <alan.hayward@arm.com>
22153 David Sherwood <david.sherwood@arm.com>
22155 * reload.c (complex_word_subreg_p): New function.
22156 (reload_inner_reg_of_subreg, push_reload): Use it.
22158 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22159 Alan Hayward <alan.hayward@arm.com>
22160 David Sherwood <david.sherwood@arm.com>
22162 * lra-constraints.c (process_alt_operands): Reject matched
22163 operands whose sizes aren't ordered.
22164 (match_reload): Refer to this check here.
22166 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22167 Alan Hayward <alan.hayward@arm.com>
22168 David Sherwood <david.sherwood@arm.com>
22170 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
22171 that the mode size is in the set {1, 2, 4, 8, 16}.
22173 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22174 Alan Hayward <alan.hayward@arm.com>
22175 David Sherwood <david.sherwood@arm.com>
22177 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
22178 Use plus_constant instead of gen_rtx_PLUS.
22180 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22181 Alan Hayward <alan.hayward@arm.com>
22182 David Sherwood <david.sherwood@arm.com>
22184 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
22185 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
22186 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
22187 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
22188 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
22189 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
22190 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
22191 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
22192 * config/i386/i386.c (ix86_push_rounding): ...this new function.
22193 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
22195 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
22196 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
22197 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
22198 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
22199 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
22200 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
22201 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
22202 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
22203 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
22204 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
22206 * expr.c (emit_move_resolve_push): Treat the input and result
22207 of PUSH_ROUNDING as a poly_int64.
22208 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
22209 (emit_push_insn): Likewise.
22210 * lra-eliminations.c (mark_not_eliminable): Likewise.
22211 * recog.c (push_operand): Likewise.
22212 * reload1.c (elimination_effects): Likewise.
22213 * rtlanal.c (nonzero_bits1): Likewise.
22214 * calls.c (store_one_arg): Likewise. Require the padding to be
22215 known at compile time.
22217 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22218 Alan Hayward <alan.hayward@arm.com>
22219 David Sherwood <david.sherwood@arm.com>
22221 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
22222 Use plus_constant instead of gen_rtx_PLUS.
22224 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22225 Alan Hayward <alan.hayward@arm.com>
22226 David Sherwood <david.sherwood@arm.com>
22228 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
22229 rather than an int.
22231 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22232 Alan Hayward <alan.hayward@arm.com>
22233 David Sherwood <david.sherwood@arm.com>
22235 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
22236 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
22237 via stack temporaries. Treat the mode size as polynomial too.
22239 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22240 Alan Hayward <alan.hayward@arm.com>
22241 David Sherwood <david.sherwood@arm.com>
22243 * expr.c (expand_expr_real_2): When handling conversions involving
22244 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
22245 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
22246 as a poly_uint64 too.
22248 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22249 Alan Hayward <alan.hayward@arm.com>
22250 David Sherwood <david.sherwood@arm.com>
22252 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
22254 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22255 Alan Hayward <alan.hayward@arm.com>
22256 David Sherwood <david.sherwood@arm.com>
22258 * combine.c (can_change_dest_mode): Handle polynomial
22259 REGMODE_NATURAL_SIZE.
22260 * expmed.c (store_bit_field_1): Likewise.
22261 * expr.c (store_constructor): Likewise.
22262 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
22263 and polynomial REGMODE_NATURAL_SIZE.
22264 (gen_lowpart_common): Likewise.
22265 * reginfo.c (record_subregs_of_mode): Likewise.
22266 * rtlanal.c (read_modify_subreg_p): Likewise.
22268 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22269 Alan Hayward <alan.hayward@arm.com>
22270 David Sherwood <david.sherwood@arm.com>
22272 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
22273 numbers of elements.
22275 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22276 Alan Hayward <alan.hayward@arm.com>
22277 David Sherwood <david.sherwood@arm.com>
22279 * match.pd: Cope with polynomial numbers of vector elements.
22281 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22282 Alan Hayward <alan.hayward@arm.com>
22283 David Sherwood <david.sherwood@arm.com>
22285 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
22286 in a POINTER_PLUS_EXPR.
22288 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22289 Alan Hayward <alan.hayward@arm.com>
22290 David Sherwood <david.sherwood@arm.com>
22292 * omp-simd-clone.c (simd_clone_subparts): New function.
22293 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
22294 (ipa_simd_modify_function_body): Likewise.
22296 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22297 Alan Hayward <alan.hayward@arm.com>
22298 David Sherwood <david.sherwood@arm.com>
22300 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
22301 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
22302 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
22303 (expand_vector_condition, vector_element): Likewise.
22304 (subparts_gt): New function.
22305 (get_compute_type): Use subparts_gt.
22306 (count_type_subparts): Delete.
22307 (expand_vector_operations_1): Use subparts_gt instead of
22308 count_type_subparts.
22310 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22311 Alan Hayward <alan.hayward@arm.com>
22312 David Sherwood <david.sherwood@arm.com>
22314 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
22315 (vect_compile_time_alias): ...this new function. Do the calculation
22316 on poly_ints rather than trees.
22317 (vect_prune_runtime_alias_test_list): Update call accordingly.
22319 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22320 Alan Hayward <alan.hayward@arm.com>
22321 David Sherwood <david.sherwood@arm.com>
22323 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
22325 (vect_schedule_slp_instance): Likewise.
22327 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22328 Alan Hayward <alan.hayward@arm.com>
22329 David Sherwood <david.sherwood@arm.com>
22331 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
22332 constant and extern definitions for variable-length vectors.
22333 (vect_get_constant_vectors): Note that the number of units
22334 is known to be constant.
22336 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22337 Alan Hayward <alan.hayward@arm.com>
22338 David Sherwood <david.sherwood@arm.com>
22340 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
22341 of units as polynomial. Choose between WIDE and NARROW based
22344 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22345 Alan Hayward <alan.hayward@arm.com>
22346 David Sherwood <david.sherwood@arm.com>
22348 * tree-vect-stmts.c (simd_clone_subparts): New function.
22349 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
22351 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22352 Alan Hayward <alan.hayward@arm.com>
22353 David Sherwood <david.sherwood@arm.com>
22355 * tree-vect-stmts.c (vectorizable_call): Treat the number of
22356 vectors as polynomial. Use build_index_vector for
22357 IFN_GOMP_SIMD_LANE.
22359 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22360 Alan Hayward <alan.hayward@arm.com>
22361 David Sherwood <david.sherwood@arm.com>
22363 * tree-vect-stmts.c (get_load_store_type): Treat the number of
22364 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
22365 for variable-length vectors.
22366 (vectorizable_mask_load_store): Treat the number of units as
22367 polynomial, asserting that it is constant if the condition has
22368 already been enforced.
22369 (vectorizable_store, vectorizable_load): Likewise.
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-vect-loop.c (vectorizable_live_operation): Treat the number
22376 of units as polynomial. Punt if we can't tell at compile time
22377 which vector contains the final result.
22379 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22380 Alan Hayward <alan.hayward@arm.com>
22381 David Sherwood <david.sherwood@arm.com>
22383 * tree-vect-loop.c (vectorizable_induction): Treat the number
22384 of units as polynomial. Punt on SLP inductions. Use an integer
22385 VEC_SERIES_EXPR for variable-length integer reductions. Use a
22386 cast of such a series for variable-length floating-point
22389 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22390 Alan Hayward <alan.hayward@arm.com>
22391 David Sherwood <david.sherwood@arm.com>
22393 * tree.h (build_index_vector): Declare.
22394 * tree.c (build_index_vector): New function.
22395 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
22396 of units as polynomial, forcibly converting it to a constant if
22397 vectorizable_reduction has already enforced the condition.
22398 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
22399 to create a {1,2,3,...} vector.
22400 (vectorizable_reduction): Treat the number of units as polynomial.
22401 Choose vectype_in based on the largest scalar element size rather
22402 than the smallest number of units. Enforce the restrictions
22405 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22406 Alan Hayward <alan.hayward@arm.com>
22407 David Sherwood <david.sherwood@arm.com>
22409 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
22410 number of units as polynomial.
22412 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22413 Alan Hayward <alan.hayward@arm.com>
22414 David Sherwood <david.sherwood@arm.com>
22416 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
22417 * target.def (autovectorize_vector_sizes): Return the vector sizes
22418 by pointer, using vector_sizes rather than a bitmask.
22419 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
22420 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
22421 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
22423 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
22424 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
22425 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
22426 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
22427 * omp-general.c (omp_max_vf): Likewise.
22428 * omp-low.c (omp_clause_aligned_alignment): Likewise.
22429 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
22430 * tree-vect-loop.c (vect_analyze_loop): Likewise.
22431 * tree-vect-slp.c (vect_slp_bb): Likewise.
22432 * doc/tm.texi: Regenerate.
22433 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
22435 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
22436 the vector size as a poly_uint64 rather than an unsigned int.
22437 (current_vector_size): Change from an unsigned int to a poly_uint64.
22438 (get_vectype_for_scalar_type): Update accordingly.
22439 * tree.h (build_truth_vector_type): Take the size and number of
22440 units as a poly_uint64 rather than an unsigned int.
22441 (build_vector_type): Add a temporary overload that takes
22442 the number of units as a poly_uint64 rather than an unsigned int.
22443 * tree.c (make_vector_type): Likewise.
22444 (build_truth_vector_type): Take the number of units as a poly_uint64
22445 rather than an unsigned int.
22447 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22448 Alan Hayward <alan.hayward@arm.com>
22449 David Sherwood <david.sherwood@arm.com>
22451 * target.def (get_mask_mode): Take the number of units and length
22452 as poly_uint64s rather than unsigned ints.
22453 * targhooks.h (default_get_mask_mode): Update accordingly.
22454 * targhooks.c (default_get_mask_mode): Likewise.
22455 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
22456 * doc/tm.texi: Regenerate.
22458 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22459 Alan Hayward <alan.hayward@arm.com>
22460 David Sherwood <david.sherwood@arm.com>
22462 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
22463 * omp-general.c (omp_max_vf): Likewise.
22464 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
22465 (expand_omp_simd): Handle polynomial safelen.
22466 * omp-low.c (omplow_simd_context): Add a default constructor.
22467 (omplow_simd_context::max_vf): Change from int to poly_uint64.
22468 (lower_rec_simd_input_clauses): Update accordingly.
22469 (lower_rec_input_clauses): Likewise.
22471 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22472 Alan Hayward <alan.hayward@arm.com>
22473 David Sherwood <david.sherwood@arm.com>
22475 * tree-vectorizer.h (vect_nunits_for_cost): New function.
22476 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
22477 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
22478 (vect_analyze_slp_cost): Likewise.
22479 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
22480 (vect_model_load_cost): Likewise.
22482 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22483 Alan Hayward <alan.hayward@arm.com>
22484 David Sherwood <david.sherwood@arm.com>
22486 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
22487 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
22488 from an unsigned int * to a poly_uint64_pod *.
22489 (calculate_unrolling_factor): New function.
22490 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
22492 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22493 Alan Hayward <alan.hayward@arm.com>
22494 David Sherwood <david.sherwood@arm.com>
22496 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
22497 from an unsigned int to a poly_uint64.
22498 (_loop_vec_info::slp_unrolling_factor): Likewise.
22499 (_loop_vec_info::vectorization_factor): Change from an int
22501 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
22502 (vect_get_num_vectors): New function.
22503 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
22504 (vect_get_num_copies): Use vect_get_num_vectors.
22505 (vect_analyze_data_ref_dependences): Change max_vf from an int *
22506 to an unsigned int *.
22507 (vect_analyze_data_refs): Change min_vf from an int * to a
22509 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
22510 than an unsigned HOST_WIDE_INT.
22511 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
22512 (vect_analyze_data_ref_dependence): Change max_vf from an int *
22513 to an unsigned int *.
22514 (vect_analyze_data_ref_dependences): Likewise.
22515 (vect_compute_data_ref_alignment): Handle polynomial vf.
22516 (vect_enhance_data_refs_alignment): Likewise.
22517 (vect_prune_runtime_alias_test_list): Likewise.
22518 (vect_shift_permute_load_chain): Likewise.
22519 (vect_supportable_dr_alignment): Likewise.
22520 (dependence_distance_ge_vf): Take the vectorization factor as a
22521 poly_uint64 rather than an unsigned HOST_WIDE_INT.
22522 (vect_analyze_data_refs): Change min_vf from an int * to a
22524 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
22525 vfm1 as a poly_uint64 rather than an int. Make the same change
22526 for the returned bound_scalar.
22527 (vect_gen_vector_loop_niters): Handle polynomial vf.
22528 (vect_do_peeling): Likewise. Update call to
22529 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
22530 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
22532 * tree-vect-loop.c (vect_determine_vectorization_factor)
22533 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
22534 (vect_get_known_peeling_cost): Likewise.
22535 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
22536 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
22537 (vect_transform_loop): Likewise. Use the lowest possible VF when
22538 updating the upper bounds of the loop.
22539 (vect_min_worthwhile_factor): Make static. Return an unsigned int
22540 rather than an int.
22541 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
22542 polynomial unroll factors.
22543 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
22544 (vect_make_slp_decision): Likewise.
22545 (vect_supported_load_permutation_p): Likewise, and polynomial
22547 (vect_analyze_slp_cost): Handle polynomial vf.
22548 (vect_slp_analyze_node_operations): Likewise.
22549 (vect_slp_analyze_bb_1): Likewise.
22550 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
22551 than an unsigned HOST_WIDE_INT.
22552 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
22553 (vectorizable_load): Handle polynomial vf.
22554 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
22556 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
22558 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22559 Alan Hayward <alan.hayward@arm.com>
22560 David Sherwood <david.sherwood@arm.com>
22562 * match.pd: Handle bit operations involving three constants
22563 and try to fold one pair.
22565 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22567 * tree-vect-loop-manip.c: Include gimple-fold.h.
22568 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
22569 niters_maybe_zero parameters. Handle other cases besides a step of 1.
22570 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
22571 Add a path that uses a step of VF instead of 1, but disable it
22573 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
22574 and niters_no_overflow parameters. Update calls to
22575 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
22576 Create a new SSA name if the latter choses to use a ste other
22577 than zero, and return it via niters_vector_mult_vf_var.
22578 * tree-vect-loop.c (vect_transform_loop): Update calls to
22579 vect_do_peeling, vect_gen_vector_loop_niters and
22580 slpeel_make_loop_iterate_ntimes.
22581 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
22582 (vect_gen_vector_loop_niters): Update declarations after above changes.
22584 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
22586 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
22587 128-bit round to integer instructions.
22588 (ceil<mode>2): Likewise.
22589 (btrunc<mode>2): Likewise.
22590 (round<mode>2): Likewise.
22592 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
22594 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
22595 unaligned VSX load/store on P8/P9.
22596 (expand_block_clear): Allow the use of unaligned VSX
22597 load/store on P8/P9.
22599 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
22601 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
22603 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
22604 swap associated with both a load and a store.
22606 2018-01-02 Andrew Waterman <andrew@sifive.com>
22608 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
22609 * config/riscv/riscv.md (clear_cache): Use it.
22611 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
22613 * web.c: Remove out-of-date comment.
22615 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
22617 * expr.c (fixup_args_size_notes): Check that any existing
22618 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
22619 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
22620 (emit_single_push_insn): ...here.
22622 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
22624 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
22625 (const_vector_encoded_nelts): New function.
22626 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
22627 (const_vector_int_elt, const_vector_elt): Declare.
22628 * emit-rtl.c (const_vector_int_elt_1): New function.
22629 (const_vector_elt): Likewise.
22630 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
22631 of CONST_VECTOR_ELT.
22633 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
22635 * expr.c: Include rtx-vector-builder.h.
22636 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
22637 directly on the tree encoding.
22638 (const_vector_from_tree): Likewise.
22639 * optabs.c: Include rtx-vector-builder.h.
22640 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
22641 sequence of "u" values.
22642 * vec-perm-indices.c: Include rtx-vector-builder.h.
22643 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
22644 directly on the vec_perm_indices encoding.
22646 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
22648 * doc/rtl.texi (const_vector): Describe new encoding scheme.
22649 * Makefile.in (OBJS): Add rtx-vector-builder.o.
22650 * rtx-vector-builder.h: New file.
22651 * rtx-vector-builder.c: Likewise.
22652 * rtl.h (rtx_def::u2): Add a const_vector field.
22653 (CONST_VECTOR_NPATTERNS): New macro.
22654 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
22655 (CONST_VECTOR_DUPLICATE_P): Likewise.
22656 (CONST_VECTOR_STEPPED_P): Likewise.
22657 (CONST_VECTOR_ENCODED_ELT): Likewise.
22658 (const_vec_duplicate_p): Check for a duplicated vector encoding.
22659 (unwrap_const_vec_duplicate): Likewise.
22660 (const_vec_series_p): Check for a non-duplicated vector encoding.
22661 Say that the function only returns true for integer vectors.
22662 * emit-rtl.c: Include rtx-vector-builder.h.
22663 (gen_const_vec_duplicate_1): Delete.
22664 (gen_const_vector): Call gen_const_vec_duplicate instead of
22665 gen_const_vec_duplicate_1.
22666 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
22667 (gen_const_vec_duplicate): Use rtx_vector_builder.
22668 (gen_const_vec_series): Likewise.
22669 (gen_rtx_CONST_VECTOR): Likewise.
22670 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
22671 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
22672 Build a new vector rather than modifying a CONST_VECTOR in-place.
22673 (handle_special_swappables): Update call accordingly.
22674 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
22675 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
22676 Build a new vector rather than modifying a CONST_VECTOR in-place.
22677 (handle_special_swappables): Update call accordingly.
22679 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
22681 * simplify-rtx.c (simplify_const_binary_operation): Use
22682 CONST_VECTOR_ELT instead of XVECEXP.
22684 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
22686 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
22687 the selector elements to be different from the data elements
22688 if the selector is a VECTOR_CST.
22689 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
22690 ssizetype for the selector.
22692 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
22694 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
22695 before testing each element individually.
22696 * tree-vect-generic.c (lower_vec_perm): Likewise.
22698 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
22700 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
22701 * selftest-run-tests.c (selftest::run_tests): Call it.
22702 * vector-builder.h (vector_builder::operator ==): New function.
22703 (vector_builder::operator !=): Likewise.
22704 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
22705 (vec_perm_indices::all_from_input_p): New function.
22706 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
22707 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
22708 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
22709 instead of reading the VECTOR_CST directly. Detect whether both
22710 vector inputs are the same before constructing the vec_perm_indices,
22711 and update the number of inputs argument accordingly. Use the
22712 utility functions added above. Only construct sel2 if we need to.
22714 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
22716 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
22717 the broadcast of the low byte.
22718 (expand_mult_highpart): Use an explicit encoding for the permutes.
22719 * optabs-query.c (can_mult_highpart_p): Likewise.
22720 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
22721 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
22722 (vectorizable_bswap): Likewise.
22723 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
22724 explicit encoding for the power-of-2 permutes.
22725 (vect_permute_store_chain): Likewise.
22726 (vect_grouped_load_supported): Likewise.
22727 (vect_permute_load_chain): Likewise.
22729 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
22731 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
22732 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
22733 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
22734 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
22735 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
22736 (vect_gen_perm_mask_any): Likewise.
22738 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
22740 * int-vector-builder.h: New file.
22741 * vec-perm-indices.h: Include int-vector-builder.h.
22742 (vec_perm_indices): Redefine as an int_vector_builder.
22743 (auto_vec_perm_indices): Delete.
22744 (vec_perm_builder): Redefine as a stand-alone class.
22745 (vec_perm_indices::vec_perm_indices): New function.
22746 (vec_perm_indices::clamp): Likewise.
22747 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
22748 (vec_perm_indices::new_vector): New function.
22749 (vec_perm_indices::new_expanded_vector): Update for new
22750 vec_perm_indices class.
22751 (vec_perm_indices::rotate_inputs): New function.
22752 (vec_perm_indices::all_in_range_p): Operate directly on the
22753 encoded form, without computing elided elements.
22754 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
22755 encoding. Update for new vec_perm_indices class.
22756 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
22757 the given vec_perm_builder.
22758 (expand_vec_perm_var): Update vec_perm_builder constructor.
22759 (expand_mult_highpart): Use vec_perm_builder instead of
22760 auto_vec_perm_indices.
22761 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
22762 vec_perm_indices instead of auto_vec_perm_indices. Use a single
22763 or double series encoding as appropriate.
22764 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
22765 vec_perm_indices instead of auto_vec_perm_indices.
22766 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
22767 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
22768 (vect_permute_store_chain): Likewise.
22769 (vect_grouped_load_supported): Likewise.
22770 (vect_permute_load_chain): Likewise.
22771 (vect_shift_permute_load_chain): Likewise.
22772 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
22773 (vect_transform_slp_perm_load): Likewise.
22774 (vect_schedule_slp_instance): Likewise.
22775 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
22776 (vectorizable_mask_load_store): Likewise.
22777 (vectorizable_bswap): Likewise.
22778 (vectorizable_store): Likewise.
22779 (vectorizable_load): Likewise.
22780 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
22781 vec_perm_indices instead of auto_vec_perm_indices. Use
22782 tree_to_vec_perm_builder to read the vector from a tree.
22783 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
22784 vec_perm_builder instead of a vec_perm_indices.
22785 (have_whole_vector_shift): Use vec_perm_builder and
22786 vec_perm_indices instead of auto_vec_perm_indices. Leave the
22787 truncation to calc_vec_perm_mask_for_shift.
22788 (vect_create_epilog_for_reduction): Likewise.
22789 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
22790 from auto_vec_perm_indices to vec_perm_indices.
22791 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
22792 instead of changing individual elements.
22793 (aarch64_vectorize_vec_perm_const): Use new_vector to install
22794 the vector in d.perm.
22795 * config/arm/arm.c (expand_vec_perm_d::perm): Change
22796 from auto_vec_perm_indices to vec_perm_indices.
22797 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
22798 instead of changing individual elements.
22799 (arm_vectorize_vec_perm_const): Use new_vector to install
22800 the vector in d.perm.
22801 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
22802 Update vec_perm_builder constructor.
22803 (rs6000_expand_interleave): Likewise.
22804 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
22805 (rs6000_expand_interleave): Likewise.
22807 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
22809 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
22810 to qimode could truncate the indices.
22811 * optabs.c (expand_vec_perm_var): Likewise.
22813 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
22815 * Makefile.in (OBJS): Add vec-perm-indices.o.
22816 * vec-perm-indices.h: New file.
22817 * vec-perm-indices.c: Likewise.
22818 * target.h (vec_perm_indices): Replace with a forward class
22820 (auto_vec_perm_indices): Move to vec-perm-indices.h.
22821 * optabs.h: Include vec-perm-indices.h.
22822 (expand_vec_perm): Delete.
22823 (selector_fits_mode_p, expand_vec_perm_var): Declare.
22824 (expand_vec_perm_const): Declare.
22825 * target.def (vec_perm_const_ok): Replace with...
22826 (vec_perm_const): ...this new hook.
22827 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
22828 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
22829 * doc/tm.texi: Regenerate.
22830 * optabs.def (vec_perm_const): Delete.
22831 * doc/md.texi (vec_perm_const): Likewise.
22832 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
22833 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
22834 expand_vec_perm for constant permutation vectors. Assert that
22835 the mode of variable permutation vectors is the integer equivalent
22836 of the mode that is being permuted.
22837 * optabs-query.h (selector_fits_mode_p): Declare.
22838 * optabs-query.c: Include vec-perm-indices.h.
22839 (selector_fits_mode_p): New function.
22840 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
22841 is defined, instead of checking whether the vec_perm_const_optab
22842 exists. Use targetm.vectorize.vec_perm_const instead of
22843 targetm.vectorize.vec_perm_const_ok. Check whether the indices
22844 fit in the vector mode before using a variable permute.
22845 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
22846 vec_perm_indices instead of an rtx.
22847 (expand_vec_perm): Replace with...
22848 (expand_vec_perm_const): ...this new function. Take the selector
22849 as a vec_perm_indices rather than an rtx. Also take the mode of
22850 the selector. Update call to shift_amt_for_vec_perm_mask.
22851 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
22852 Use vec_perm_indices::new_expanded_vector to expand the original
22853 selector into bytes. Check whether the indices fit in the vector
22854 mode before using a variable permute.
22855 (expand_vec_perm_var): Make global.
22856 (expand_mult_highpart): Use expand_vec_perm_const.
22857 * fold-const.c: Includes vec-perm-indices.h.
22858 * tree-ssa-forwprop.c: Likewise.
22859 * tree-vect-data-refs.c: Likewise.
22860 * tree-vect-generic.c: Likewise.
22861 * tree-vect-loop.c: Likewise.
22862 * tree-vect-slp.c: Likewise.
22863 * tree-vect-stmts.c: Likewise.
22864 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
22866 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
22867 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
22868 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
22869 (aarch64_vectorize_vec_perm_const): ...this new function.
22870 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
22871 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
22872 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
22873 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
22874 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
22875 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
22876 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
22878 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
22879 check for NEON modes.
22880 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
22881 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
22882 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
22883 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
22885 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
22886 the old VEC_PERM_CONST conditions.
22887 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
22888 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
22889 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
22890 (ia64_vectorize_vec_perm_const_ok): Merge into...
22891 (ia64_vectorize_vec_perm_const): ...this new function.
22892 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
22893 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
22894 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
22895 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
22896 * config/mips/mips.c (mips_expand_vec_perm_const)
22897 (mips_vectorize_vec_perm_const_ok): Merge into...
22898 (mips_vectorize_vec_perm_const): ...this new function.
22899 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
22900 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
22901 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
22902 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
22903 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
22904 (rs6000_expand_vec_perm_const): Delete.
22905 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
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.
22915 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
22916 an element count and rtx array.
22917 (rs6000_expand_extract_even): Update call accordingly.
22918 (rs6000_expand_interleave): Likewise.
22919 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
22920 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
22921 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
22922 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
22923 (rs6000_expand_vec_perm_const): Delete.
22924 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
22925 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
22926 (altivec_expand_vec_perm_const_le): Take each operand individually.
22927 Operate on constant selectors rather than rtxes.
22928 (altivec_expand_vec_perm_const): Likewise. Update call to
22929 altivec_expand_vec_perm_const_le.
22930 (rs6000_expand_vec_perm_const): Delete.
22931 (rs6000_vectorize_vec_perm_const_ok): Delete.
22932 (rs6000_vectorize_vec_perm_const): New function. Remove stray
22933 reference to the SPE evmerge intructions.
22934 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
22935 an element count and rtx array.
22936 (rs6000_expand_extract_even): Update call accordingly.
22937 (rs6000_expand_interleave): Likewise.
22938 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
22939 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
22941 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
22943 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
22945 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
22946 vector mode and that that mode matches the mode of the data
22948 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
22949 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
22950 directly using expand_vec_perm_1 when forcing selectors into
22952 (expand_vec_perm_var): New function, split out from expand_vec_perm.
22954 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
22956 * optabs-query.h (can_vec_perm_p): Delete.
22957 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
22958 * optabs-query.c (can_vec_perm_p): Split into...
22959 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
22960 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
22961 particular selector is valid.
22962 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
22963 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
22964 (vect_grouped_load_supported): Likewise.
22965 (vect_shift_permute_load_chain): Likewise.
22966 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
22967 (vect_transform_slp_perm_load): Likewise.
22968 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
22969 (vectorizable_bswap): Likewise.
22970 (vect_gen_perm_mask_checked): Likewise.
22971 * fold-const.c (fold_ternary_loc): Likewise. Don't take
22972 implementations of variable permutation vectors into account
22973 when deciding which selector to use.
22974 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
22975 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
22976 with a false third argument.
22977 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
22978 to test whether the constant selector is valid and can_vec_perm_var_p
22979 to test whether a variable selector is valid.
22981 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
22983 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
22984 * optabs-query.c (can_vec_perm_p): Likewise.
22985 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
22986 instead of vec_perm_indices.
22987 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
22988 (vect_gen_perm_mask_checked): Likewise,
22989 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
22990 (vect_gen_perm_mask_checked): Likewise,
22992 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
22994 * optabs-query.h (qimode_for_vec_perm): Declare.
22995 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
22996 (qimode_for_vec_perm): ...this new function.
22997 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
22999 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
23001 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
23002 does not have a conditional at the top.
23004 2018-01-02 Richard Biener <rguenther@suse.de>
23006 * ipa-inline.c (big_speedup_p): Fix expression.
23008 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
23011 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
23014 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
23018 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
23019 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
23020 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
23021 cond_taken_branch_cost 3->4.
23023 2018-01-01 Jakub Jelinek <jakub@redhat.com>
23025 PR tree-optimization/83581
23026 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
23027 TODO_cleanup_cfg if any changes have been made.
23029 PR middle-end/83608
23030 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
23031 convert_modes if target mode has the right side, but different mode
23034 PR middle-end/83609
23035 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
23036 last argument when extracting from CONCAT. If either from_real or
23037 from_imag is NULL, use expansion through memory. If result is not
23038 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
23039 the parts directly to inner mode, if even that fails, use expansion
23042 PR middle-end/83623
23043 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
23044 check for bswap in mode rather than HImode and use that in expand_unop
23047 Copyright (C) 2018 Free Software Foundation, Inc.
23049 Copying and distribution of this file, with or without modification,
23050 are permitted in any medium without royalty provided the copyright
23051 notice and this notice are preserved.