1 2018-08-15 Martin Sebor <msebor@redhat.com>
3 PR tree-optimization/71625
4 * config/aarch64/aarch64-builtins.c
5 (aarch64_init_simd_builtin_types): Clear Poly8_t's TYPE_STRING_FLAG.
7 2018-08-15 Ilya Leoshkevich <iii@linux.ibm.com>
9 * config/s390/s390.c (s390_reorg): Remove loop.
11 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
14 (darwin_function_switched_text_sections): Delete.
16 (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise.
18 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
21 * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION,
22 DEBUG_RNGLISTS_SECTION) new macros. (DEBUG_PUBNAMES_SECTION,
23 DEBUG_PUBTYPES_SECTION) update to include GNU variant.
25 2018-08-15 Martin Liska <mliska@suse.cz>
27 PR tree-optimization/86925
28 * predict.c (expr_expected_value_1): When taking
29 later predictor, assign also probability.
30 Use fold_build2_initializer_loc in order to fold
31 the expression in -frounding-math.
33 2018-08-14 Allan Sandfeld Jensen <allan.jensen@qt.io>
35 * config/i386/i386.c (expand_vec_perm_movs): New method matching movs
37 (expand_vec_perm_1): Try the new method.
39 2018-08-14 Ilya Leoshkevich <iii@linux.ibm.com>
42 * lra-lives.c (remove_some_program_points_and_update_live_ranges):
43 Check whether lra_live_max_point is 0 before dividing.
45 2018-08-14 Martin Sebor <msebor@redhat.com>
47 PR tree-optimization/86650
48 * tree-vrp.c (vrp_prop::check_array_ref): Print an inform message.
49 (vrp_prop::check_mem_ref): Same.
51 2018-08-13 Liu Hao <lh_mouse@126.com>
53 * pretty-print.c (eat_esc_sequence): Swap the foreground and
54 background colors if the COMMON_LVB_REVERSE_VIDEO flag is set,
55 and clear it thereafter, as it only works for DBCS.
57 2018-08-13 Liu Hao <lh_mouse@126.com>
59 * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the
60 handle returned by _get_osf_handle().
62 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
64 * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support
67 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
69 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin):
70 Add support for gimple-folding of vec_pack() and vec_unpack()
73 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
75 * config/rs6000/rs6000.c (rs6000_builtin_valid_without_lhs): Add
76 vec_xst variants to the list.
77 (rs6000_gimple_fold_builtin): Add support for folding unaligned
78 vector loads and stores.
80 2018-08-13 David Edelsohn <dje.gcc@gmail.com>
82 * config.gcc (rs6000-ibm-aix4.x): Delete.
83 (rs6000-ibm-aix5.1): Delete.
84 (rs6000-ibm-aix5.2): Delete.
85 (rs6000-ibm-aix5.3): Delete.
86 * config/rs6000/aix43.h: Delete.
87 * config/rs6000/aix51.h: Delete.
88 * config/rs6000/aix52.h: Delete.
89 * config/rs6000/t-aix43: Delete.
91 2018-08-13 Ilya Leoshkevich <iii@linux.ibm.com>
93 * config/s390/s390.c (s390_decompose_constant_pool_ref):
95 (s390_decompose_address): Factor out constant pool ref
98 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
100 * config/nds32/nds32-predicates.c
101 (nds32_can_use_bclr_p): Change return type as bool.
102 (nds32_can_use_bset_p): Ditto.
103 (nds32_can_use_btgl_p): Ditto.
104 (nds32_can_use_bitci_p): Ditto.
105 * config/nds32/nds32-protos.h
106 (nds32_can_use_bclr_p): Change declaration.
107 (nds32_can_use_bset_p): Ditto.
108 (nds32_can_use_btgl_p): Ditto.
109 (nds32_can_use_bitci_p): Ditto.
111 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
113 * config/nds32/nds32.c (nds32_expand_prologue, nds32_expand_epilogue):
114 Support -msched-prolog-epilog option.
115 * config/nds32/nds32.opt (msched-prolog-epilog): New option.
117 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
119 * common/config/nds32/nds32-common.c
120 (nds32_option_optimization_table): Enalbe -malways-align.
122 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
124 * config.gcc (nds32*): Add nds32_isr.h and nds32_init.inc in
126 * common/config/nds32/nds32-common.c (nds32_handle_option): Handle
127 OPT_misr_secure_ case.
128 * config/nds32/nds32-isr.c: Implementation of backward compatibility.
129 * config/nds32/nds32-protos.h (nds32_isr_function_critical_p): New.
130 * config/nds32/nds32.c (nds32_attribute_table): Add critical and
132 * config/nds32/nds32.h (nds32_isr_nested_type): Add NDS32_CRITICAL.
133 (nds32_isr_info): New field security_level.
134 (TARGET_ISR_VECTOR_SIZE_4_BYTE): New macro.
135 * config/nds32/nds32.md (return_internal): Consider critical attribute.
136 * config/nds32/nds32.opt (misr-secure): New option.
137 * config/nds32/nds32_init.inc: New file.
138 * config/nds32/nds32_isr.h: New file.
140 2018-08-11 John David Anglin <danglin@gcc.gnu.org>
142 * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
143 Update comment for atomic instructions.
144 (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
145 atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
147 (atomic_loaddi): Revise fence expansion to only emit fence prior to
148 load for __ATOMIC_SEQ_CST model.
149 (atomic_loaddi_1): Remove float register target.
150 (atomic_storedi): Handle CONST_INT values.
151 (atomic_storedi_1): Remove float register source. Add special case
153 (memory_barrier): New expander and insn.
155 2018-08-11 Jakub Jelinek <jakub@redhat.com>
157 PR tree-optimization/86835
158 * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
159 new_stmt after def_gsi, make sure to insert new_square_stmt after
160 that stmt, not 2 stmts before it.
162 2018-08-10 Alexander Monakov <amonakov@ispras.ru>
165 * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
168 2018-08-10 Martin Liska <mliska@suse.cz>
171 * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
173 * builtins.c (expand_builtin_expect_with_probability):
175 (expand_builtin_expect_with_probability): New function.
176 (build_builtin_expect_predicate): Add new argumnet probability
177 for BUILT_IN_EXPECT_WITH_PROBABILITY.
178 (fold_builtin_expect):
181 * builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY):
182 * builtins.h (fold_builtin_expect): Set new argument.
183 * doc/extend.texi: Document __builtin_expect_with_probability.
184 * doc/invoke.texi: Likewise.
185 * gimple-fold.c (gimple_fold_call): Pass new argument.
186 * ipa-fnsummary.c (find_foldable_builtin_expect): Handle
187 also BUILT_IN_EXPECT_WITH_PROBABILITY.
188 * predict.c (get_predictor_value): New function.
189 (expr_expected_value): Add new argument probability. Assume
190 that predictor and probability are always non-null.
191 (expr_expected_value_1): Likewise. For __builtin_expect and
192 __builtin_expect_with_probability set probability. Handle
193 combination in binary expressions.
194 (tree_predict_by_opcode): Simplify code by simply calling
196 (pass_strip_predict_hints::execute): Add handling of
197 BUILT_IN_EXPECT_WITH_PROBABILITY.
198 * predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add
200 * tree.h (DECL_BUILT_IN_P): New function.
202 2018-08-10 Martin Liska <mliska@suse.cz>
204 PR tree-optimization/85799
205 * passes.def: Add argument for pass_strip_predict_hints.
206 * predict.c (class pass_strip_predict_hints): Add new argument
208 (strip_predictor_early): New function.
209 (pass_strip_predict_hints::execute): Call the function to
211 (strip_predict_hints): New function.
212 * predict.def: Fix comment.
214 2018-08-10 Thomas Preud'homme <thomas.preudhomme@linaro.org>
216 * Makefile.in: Clarify which tm.texi to copy over to assert the
217 right to grant a GFDL license for all.
219 2018-08-09 Jeff Law <law@redhat.com>
221 * config/m68k/m68k.c (m68k_adjust_decorated_operand): Remove
224 2018-08-09 Andreas Schwab <schwab@linux-m68k.org>
226 * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
229 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
231 * tree-vect-loop.c (vectorizable_reduction): Allow inner-loop
232 reductions for variable-length vectors.
234 2018-08-09 David Malcolm <dmalcolm@redhat.com>
237 * common.opt (fdiagnostics-show-line-numbers): New option.
238 * diagnostic-show-locus.c (class layout): Add fields
239 "m_show_line_numbers_p" and "m_linenum_width";
240 (num_digits): New function.
241 (test_num_digits): New function.
242 (layout::layout): Initialize new fields. Update m_x_offset
243 logic to handle any left margin.
244 (layout::print_source_line): Print line number when requested.
245 (layout::start_annotation_line): New member function.
246 (layout::print_annotation_line): Call it.
247 (layout::print_leading_fixits): Likewise.
248 (layout::print_trailing_fixits): Likewise. Update calls to
249 move_to_column for new parameter.
250 (layout::get_x_bound_for_row): Add "add_left_margin" param and use
251 it to potentially call start_annotation_line.
252 (layout::show_ruler): Call start_annotation_line.
253 (selftest::test_line_numbers_multiline_range): New selftest.
254 (selftest::diagnostic_show_locus_c_tests): Call test_num_digits
255 and selftest::test_line_numbers_multiline_range.
256 * diagnostic.c (diagnostic_initialize): Initialize
258 * diagnostic.h (struct diagnostic_context): Add field
259 "show_line_numbers_p".
260 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
261 -fno-diagnostics-show-line-numbers.
262 * dwarf2out.c (gen_producer_string): Add
263 OPT_fdiagnostics_show_line_numbers to the ignored options.
264 * lto-wrapper.c (merge_and_complain): Likewise to the "pick
265 one setting" options.
266 (append_compiler_options): Likewise to the dropped options.
267 (append_diag_options): Likewise to the passed-on options.
268 * opts.c (common_handle_option): Handle the new option.
269 * toplev.c (general_init): Set up global_dc->show_line_numbers_p.
271 2018-08-09 Kelvin Nilsen <kelvin@gcc.gnu.org>
273 * doc/extend.texi (PowerPC AltiVec Built-in Functions Available on
274 ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub. Add
275 third argument of type "const signed char" to descriptions of
276 __builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq,
277 __builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub,
278 __builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt,
279 __builtin_bcdsub_ov functions.
281 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
283 PR tree-optimization/86858
284 * tree-vect-loop.c (vect_is_simple_reduction): Restore
285 flow_bb_inside_loop_p calls.
287 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
289 PR tree-optimization/86871
290 * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
291 instead of gimple_assign_lhs.
293 2018-08-09 Richard Earnshaw <rearnsha@arm.com>
296 * config/aarch64/aarch64.md (add<mode>3_carryinC_zero): Add missing
297 register constraint to operand 0.
298 (add<mode>3_carryinC): Likewise.
299 (add<mode>3_carryinV_zero, add<mode>3_carryinV): Likewise.
301 2018-08-09 Martin Liska <mliska@suse.cz>
304 * common.opt: Remove extra line.
306 2018-08-09 Martin Liska <mliska@suse.cz>
308 * params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots
309 at the end of a line, make first letter capital and end up
310 a sentence with a dot.
311 (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise.
312 (PARAM_LOOP_BLOCK_TILE_SIZE): Likewise.
313 (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise.
314 (PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise.
315 (PARAM_MAX_ISL_OPERATIONS): Likewise.
316 (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise.
317 (PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise.
318 (PARAM_INDIR_CALL_TOPN_PROFILE): Likewise.
319 (PARAM_SLP_MAX_INSNS_IN_BB): Likewise.
320 (PARAM_IPA_CP_EVAL_THRESHOLD): Likewise.
321 (PARAM_IPA_CP_RECURSION_PENALTY): Likewise.
322 (PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
323 (PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise.
324 (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise.
325 (PARAM_TREE_REASSOC_WIDTH): Likewise.
326 (PARAM_HSA_GEN_DEBUG_STORES): Likewise.
327 (PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise.
328 (PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise.
330 2018-08-09 Andreas Krebbel <krebbel@linux.ibm.com>
333 * config/s390/s390.c (s390_option_override_internal): Reduce the
334 stack-clash-protection-probe-interval param if it would be too big
337 2018-08-08 Andreas Schwab <schwab@linux-m68k.org>
340 * config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define.
341 * config/m68k/m68k.c (handle_move_double): Don't call
342 m68k_final_prescan_insn.
343 (m68k_adjust_decorated_operand): Renamed from
344 m68k_final_prescan_insn, remove first and third operand and
346 (print_operand): Call it.
347 (print_operand_address): Call it.
349 2018-08-08 Nathan Sidwell <nathan@acm.org>
351 * diagnostic.c (diagnostic_report_current_module): Use
352 linemap_included_from & linemap_included_from_linemap.
354 2018-08-08 Hongbo Zhang <hongbo.zhang@linaro.org>
356 * config/aarch64/aarch64-cores.def: Add phecda core.
357 * config/aarch64/aarch64-tune.md: Regenerate.
358 * doc/invoke.texi: Add phecda core.
360 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
363 * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
365 * config/s390/s390.md ("movti"): Add more alternatives for
366 constant to GPR copies.
368 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
370 * config/s390/s390.c: Fix whitespace damage throughout the file.
371 * config/s390/s390.h: Likewise.
372 * config/s390/tpf.h: Likewise.
374 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
376 * config/s390/s390.c (s390_loadrelative_operand_p):
377 Remove TARGET_CPU_ZARCH usages.
378 (s390_rtx_costs): Likewise.
379 (s390_legitimate_constant_p): Likewise.
380 (s390_cannot_force_const_mem): Likewise.
381 (legitimate_reload_constant_p): Likewise.
382 (s390_preferred_reload_class): Likewise.
383 (legitimize_pic_address): Likewise.
384 (legitimize_tls_address): Likewise.
385 (s390_split_branches): Removed.
386 (s390_add_execute): Removed.
387 (s390_dump_pool): Remove TARGET_CPU_ZARCH usages.
388 (s390_mainpool_start): Likewise.
389 (s390_mainpool_finish): Likewise.
390 (s390_mainpool_cancel): Removed.
391 (s390_chunkify_start): Remove TARGET_CPU_ZARCH usages.
392 (s390_chunkify_cancel): Likewise.
393 (s390_return_addr_rtx): Likewise.
394 (s390_register_info): Remove split_branches_pending_p uages.
395 (s390_optimize_register_info): Likewise.
396 (s390_init_frame_layout): Remove TARGET_CPU_ZARCH and
397 split_branches_pending_p usages.
398 (s390_can_eliminate): Remove TARGET_CPU_ZARCH usages.
399 (s390_load_got): Likewise.
400 (s390_expand_split_stack_prologue): Likewise.
401 (output_asm_nops): Likewise.
402 (s390_function_profiler): Likewise.
403 (s390_emit_call): Likewise.
404 (s390_conditional_register_usage): Likewise.
405 (s390_optimize_prologue): Likewise.
406 (s390_reorg): Remove TARGET_CPU_ZARCH and
407 split_branches_pending_p usages.
408 (s390_option_override_internal): Remove TARGET_CPU_ZARCH
410 (s390_output_indirect_thunk_function): Likewise.
411 * config/s390/s390.h (TARGET_CPU_ZARCH): Removed.
412 (TARGET_CPU_ZARCH_P): Removed.
413 (struct machine_function): Remove split_branches_pending_p.
414 * config/s390/s390.md: Remove TARGET_CPU_ZARCH usages.
416 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
418 * common/config/s390/s390-common.c (processor_flags_table):
420 * config.gcc: Remove with_arch/with_tune support.
421 * config/s390/2064.md: Remove cpu attribute comparisons.
422 * config/s390/driver-native.c (s390_host_detect_local_cpu):
424 * config/s390/linux.h (ASM_SPEC):
425 Remove -march support.
426 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal):
427 Use a table to get an arch level.
428 * config/s390/s390-opts.h (enum processor_type):
431 (processor_table): Remove entries, add arch_level values.
432 (s390_issue_rate): Remove cases.
433 (s390_option_override): Adjust
434 s390_option_override_internal() call.
435 (s390_option_override_internal): Remove deprecation warning.
436 (s390_valid_target_attribute_tree): Adjust
437 s390_option_override_internal() call.
438 * config/s390/s390.h (struct s390_processor):
439 Share with s390-c.c, add arch_level field.
440 * config/s390/s390.md:
441 Remove occurrences in cpu attribute.
442 * config/s390/s390.opt: Remove -march/-mtune support.
443 * config/s390/tpf.h (ASM_SPEC): Remove -march support.
444 * doc/invoke.texi: Remove deprecation warning.
446 2018-08-08 Luis Machado <luis.machado@linaro.org>
448 * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static
450 (qdf24xx_tunings): Set vector cost structure to
453 * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
454 <register_sextend>: Set to 3.
456 2018-08-07 Richard Sandiford <richard.sandiford@arm.com>
459 * config/aarch64/iterators.md (FRECP, frecp_suffix): Delete.
460 * config/aarch64/aarch64-simd.md
461 (aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into...
462 (@aarch64_frecpe<mode>): ...here and the move FRECPX to...
463 (aarch64_frecpx<mode>): ...this new pattern.
464 * config/aarch64/aarch64-simd-builtins.def: Remove comment
465 about aarch64_frecp<FRECP:frecp_suffix><mode>.
467 2018-08-07 Martin Liska <mliska@suse.cz>
470 * predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC,
471 BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW.
472 * predict.def (PRED_MALLOC_NONNULL): New predictor.
473 * doc/extend.texi: Document that malloc attribute adds
476 2018-08-06 John David Anglin <danglin@gcc.gnu.org>
479 * config/pa/pa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
480 Define to speculation_safe_value_not_needed.
482 2018-08-06 Jeff Law <law@redhat.com>
484 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down
485 the vr_values instance to cprop_into_stmt.
486 (cprop_into_stmt): Pass vr_values instance down to cprop_operand.
487 (cprop_operand): Also query EVRP to determine if OP is a constant.
489 2018-08-06 Nathan Sidwell <nathan@acm.org>
491 * diagnostic.c (diagnostic_report_current_module): Reroll
492 included-at loop. Translate text.
494 2018-08-06 David Malcolm <dmalcolm@redhat.com>
496 * function-tests.c (selftest::test_expansion_to_rtl): Call
497 free_after_compilation.
499 2018-08-06 Alan Hayward <alan.hayward@arm.com>
501 * config/aarch64/aarch64.md: Add clobber highs to tls_desc.
503 2018-08-06 Andreas Krebbel <krebbel@linux.ibm.com>
505 * config/s390/s390.c (s390_loop_unroll_adjust): Prevent small
506 loops with memory block operations from getting unrolled.
508 2018-08-06 Ulrich Weigand <uweigand@de.ibm.com>
511 * config/spu/spu.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
512 Define to speculation_safe_value_not_needed.
514 2018-08-06 Jeff Law <law@redhat.com>
516 * reload1.c (forget_old_reloads_1): Adjust CLOBBER_HIGH
519 2018-08-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
522 * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
523 with all enabled __intN types.
525 * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
527 2018-08-06 Alan Hayward <alan.hayward@arm.com>
529 * alias.c (record_set): Check for clobber high.
530 * cfgexpand.c (expand_gimple_stmt): Likewise.
531 * combine-stack-adj.c (single_set_for_csa): Likewise.
532 * combine.c (find_single_use_1): Likewise.
533 (set_nonzero_bits_and_sign_copies): Likewise.
534 (get_combine_src_dest): Likewise.
535 (is_parallel_of_n_reg_sets): Likewise.
536 (try_combine): Likewise.
537 (record_dead_and_set_regs_1): Likewise.
538 (reg_dead_at_p_1): Likewise.
539 (reg_dead_at_p): Likewise.
540 * dce.c (deletable_insn_p): Likewise.
541 (mark_nonreg_stores_1): Likewise.
542 (mark_nonreg_stores_2): Likewise.
543 * df-scan.c (df_find_hard_reg_defs): Likewise.
544 (df_uses_record): Likewise.
545 (df_get_call_refs): Likewise.
546 * dwarf2out.c (mem_loc_descriptor): Likewise.
547 * haifa-sched.c (haifa_classify_rtx): Likewise.
548 * ira-build.c (create_insn_allocnos): Likewise.
549 * ira-costs.c (scan_one_insn): Likewise.
550 * ira.c (equiv_init_movable_p): Likewise.
551 (rtx_moveable_p): Likewise.
552 (interesting_dest_for_shprep): Likewise.
553 * jump.c (mark_jump_label_1): Likewise.
554 * postreload-gcse.c (record_opr_changes): Likewise.
555 * postreload.c (reload_cse_simplify): Likewise.
556 (struct reg_use): Add source expr.
557 (reload_combine): Check for clobber high.
558 (reload_combine_note_use): Likewise.
559 (reload_cse_move2add): Likewise.
560 (move2add_note_store): Likewise.
561 * print-rtl.c (print_pattern): Likewise.
562 * recog.c (decode_asm_operands): Likewise.
563 (store_data_bypass_p): Likewise.
564 (if_test_bypass_p): Likewise.
565 * regcprop.c (kill_clobbered_value): Likewise.
566 (kill_set_value): Likewise.
567 * reginfo.c (reg_scan_mark_refs): Likewise.
568 * reload1.c (maybe_fix_stack_asms): Likewise.
569 (eliminate_regs_1): Likewise.
570 (elimination_effects): Likewise.
571 (mark_not_eliminable): Likewise.
572 (scan_paradoxical_subregs): Likewise.
573 (forget_old_reloads_1): Likewise.
574 * reorg.c (find_end_label): Likewise.
575 (try_merge_delay_insns): Likewise.
576 (redundant_insn): Likewise.
577 (own_thread_p): Likewise.
578 (fill_simple_delay_slots): Likewise.
579 (fill_slots_from_thread): Likewise.
580 (dbr_schedule): Likewise.
581 * resource.c (update_live_status): Likewise.
582 (mark_referenced_resources): Likewise.
583 (mark_set_resources): Likewise.
584 * rtl.c (copy_rtx): Likewise.
585 * rtlanal.c (reg_referenced_p): Likewise.
586 (single_set_2): Likewise.
587 (noop_move_p): Likewise.
588 (note_stores): Likewise.
589 * sched-deps.c (sched_analyze_reg): Likewise.
590 (sched_analyze_insn): Likewise.
592 2018-08-06 Alan Hayward <alan.hayward@arm.com>
594 * cse.c (invalidate_reg): New function extracted from...
595 (invalidate): ...here.
596 (canonicalize_insn): Check for clobber high.
597 (invalidate_from_clobbers): invalidate clobber highs.
598 (invalidate_from_sets_and_clobbers): Likewise.
599 (count_reg_usage): Check for clobber high.
600 (insn_live_p): Likewise.
601 * cselib.c (cselib_expand_value_rtx_1):Likewise.
602 (cselib_invalidate_regno): Check for clobber in setter.
603 (cselib_invalidate_rtx): Pass through setter.
604 (cselib_invalidate_rtx_note_stores):
605 (cselib_process_insn): Check for clobber high.
606 * cselib.h (cselib_invalidate_rtx): Add operand.
608 2018-08-06 Alan Hayward <alan.hayward@arm.com>
610 * lra-eliminations.c (lra_eliminate_regs_1): Check for clobber high.
611 (mark_not_eliminable): Likewise.
612 * lra-int.h (struct lra_insn_reg): Add clobber high marker.
613 * lra-lives.c (process_bb_lives): Check for clobber high.
614 * lra.c (new_insn_reg): Remember clobber highs.
615 (collect_non_operand_hard_regs): Check for clobber high.
616 (lra_set_insn_recog_data): Likewise.
617 (add_regs_to_insn_regno_info): Likewise.
618 (lra_update_insn_regno_info): Likewise.
620 2018-08-06 Alan Hayward <alan.hayward@arm.com>
622 * rtl.h (reg_is_clobbered_by_clobber_high): Add declarations.
623 * rtlanal.c (reg_is_clobbered_by_clobber_high): Add function.
625 2018-08-06 Alan Hayward <alan.hayward@arm.com>
627 * emit-rtl.c (verify_rtx_sharing): Check for CLOBBER_HIGH.
628 (copy_insn_1): Likewise.
629 (gen_hard_reg_clobber_high): New gen function.
630 * genconfig.c (walk_insn_part): Check for CLOBBER_HIGH.
631 * genemit.c (gen_exp): Likewise.
632 (gen_emit_seq): Pass through info.
633 (gen_insn): Check for CLOBBER_HIGH.
634 (gen_expand): Pass through info.
635 (gen_split): Likewise.
636 (output_add_clobbers): Likewise.
637 * genrecog.c (validate_pattern): Check for CLOBBER_HIGH.
638 (remove_clobbers): Likewise.
639 * rtl.h (gen_hard_reg_clobber_high): New declaration.
641 2018-08-06 Alan Hayward <alan.hayward@arm.com>
643 * doc/rtl.texi (clobber_high): Add.
644 (parallel): Add in clobber high
645 * rtl.c (rtl_check_failed_code3): Add function.
646 * rtl.def (CLOBBER_HIGH): Add expression.
647 * rtl.h (RTL_CHECKC3): Add macro.
648 (rtl_check_failed_code3): Add declaration.
651 2018-08-05 H.J. Lu <hongjiu.lu@intel.com>
654 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
655 cfun->machine->max_used_stack_alignment if needed.
657 2018-08-04 Martin Sebor <msebor@redhat.com>
659 PR tree-optimization/86571
660 * gimple-ssa-sprintf.c (format_floating): Extend upper bound of
663 2018-08-03 Sandra Loosemore <sandra@codesourcery.com>
666 * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
669 2018-08-03 Jeff Law <law@redhat.com>
672 * config/mn10300/mn10300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
673 Define to speculation_safe_value_not_needed.
675 2018-08-03 David Malcolm <dmalcolm@redhat.com>
677 * doc/gcov.texi (-x): Remove duplicate "to".
678 * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
679 (-Wif-not-aligned): Remove duplicate "is".
680 (-flto): Remove duplicate "the".
681 (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
683 (MSP430 Options): Remove duplicate "and" from the description
684 of "-mgprel-sec=regexp".
685 (x86 Options): Remove duplicate copies of "vmldLog102" and
686 vmlsLog104 from description of "-mveclibabi=type".
688 2018-08-03 Richard Sandiford <richard.sandiford@arm.com>
690 * internal-fn.h (first_commutative_argument): Declare.
691 * internal-fn.c (first_commutative_argument): New function.
692 * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove extra
693 restrictions for pattern statements. Use first_commutative_argument
694 to look for commutative operands in calls to internal functions.
696 2018-08-03 Aldy Hernandez <aldyh@redhat.com>
698 * Makefile.in (wide-int-range.o): New.
699 * tree-vrp.c: Move all the wide_int_* functions to...
700 * wide-int-range.cc: ...here.
701 * tree-vrp.h: Move all the wide_int_* prototypes to...
702 * wide-int-range.h: ...here.
704 2018-08-03 Tom de Vries <tdevries@suse.de>
706 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): Return
708 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Remove define.
709 * except.c (output_function_exception_table): Do early exit if
710 targetm_common.except_unwind_info (&global_options) == UI_NONE.
712 2018-08-03 Martin Liska <mliska@suse.cz>
714 * predict.c (dump_prediction): Change to 2 digits
717 2018-08-03 Siddhesh Poyarekar <siddhesh@sourceware.org>
719 * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
721 (falkor_am_1_gtov_gtov): ... a new insn reservation.
723 2018-07-19 Ilya Leoshkevich <iii@linux.ibm.com>
725 * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Replace > with >=.
726 * df-problems.c (df_remove_dead_eq_notes): Replace > with >=.
727 * dwarf2out.c (mem_loc_descriptor): Replace > with >=.
728 * lra-constraints.c (spill_hard_reg_in_range): Replace <= with <.
729 * lra-remat.c (call_used_input_regno_present_p): Replace <= with <.
731 2018-08-02 David Malcolm <dmalcolm@redhat.com>
733 * diagnostic-show-locus.c (diagnostic_show_locus): Use
734 pp_take_prefix when saving the existing prefix.
735 * diagnostic.c (diagnostic_append_note): Likewise.
736 * langhooks.c (lhd_print_error_function): Likewise.
737 * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
738 param's type. Free the existing prefix.
739 (pp_take_prefix): New function.
740 (pretty_printer::pretty_printer): Drop the prefix parameter.
741 Rename the length parameter to match the comment.
742 (pretty_printer::~pretty_printer): Free the prefix.
743 * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
745 (struct pretty_printer): Drop the "const" from "prefix" field's
746 type and clarify memory management.
747 (pp_set_prefix): Drop the "const" from the 2nd param.
748 (pp_take_prefix): New decl.
750 2018-08-02 Aaron Sawdey <acsawdey@linux.ibm.com>
752 * config/rs6000/rs6000-string.c (select_block_compare_mode): Move test
753 for word_mode_ok here instead of passing as argument.
754 (expand_block_compare): Change select_block_compare_mode() call.
755 (expand_strncmp_gpr_sequence): New function.
756 (expand_strn_compare): Make use of expand_strncmp_gpr_sequence.
758 2018-08-02 Jeff Law <law@redhat.com>
761 * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
762 Define to speculation_safe_value_not_needed.
765 * config/h8300/h8300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
766 Define to speculation_safe_value_not_needed.
768 2018-08-02 Tom de Vries <tdevries@suse.de>
771 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): New
772 function. Return UI_TARGET unconditionally.
773 (TARGET_EXCEPT_UNWIND_INFO): Redefine to nvptx_except_unwind_info.
774 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Emit commented out '.byte'.
776 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
778 * genemit.c (print_overload_test): Fix typo.
780 2018-08-02 Richard Biener <rguenther@suse.de>
782 PR tree-optimization/86816
783 * tree-ssa-tail-merge.c (tail_merge_valueize): New function
784 which checks for value availability before querying it.
785 (gvn_uses_equal): Use it.
786 (same_succ_hash): Likewise.
787 (gimple_equal_p): Likewise.
789 2018-08-02 Nick Clifton <nickc@redhat.com>
792 * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
793 Define to speculation_safe_value_not_needed.
796 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
797 Define to speculation_safe_value_not_needed.
800 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
801 Define to speculation_safe_value_not_needed.
804 * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
805 Define to speculation_safe_value_not_needed.
808 * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
809 Define to speculation_safe_value_not_needed.
812 * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
813 Define to speculation_safe_value_not_needed.
816 * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
817 Define to speculation_safe_value_not_needed.
820 * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
821 Define to speculation_safe_value_not_needed.
824 * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
825 speculation_safe_value_not_needed.
828 * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define
829 to speculation_safe_value_not_needed.
831 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
833 * doc/md.texi: Expand the documentation of instruction names
834 to mention port-local uses. Document '@' in pattern names.
835 * read-md.h (overloaded_instance, overloaded_name): New structs.
837 (md_reader::handle_overloaded_name): New member function.
838 (md_reader::get_overloads): Likewise.
839 (md_reader::m_first_overload): New member variable.
840 (md_reader::m_next_overload_ptr): Likewise.
841 (md_reader::m_overloads_htab): Likewise.
842 * read-md.c (md_reader::md_reader): Initialize m_first_overload,
843 m_next_overload_ptr and m_overloads_htab.
844 * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields.
845 (get_mode_token, get_code_token, get_int_token): New functions.
846 (map_attr_string): Add an optional argument that passes back
847 the associated iterator.
848 (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p):
849 (md_reader::handle_overloaded_name, add_overload_instance): New
851 (apply_iterators): Handle '@' names. Report an error if '@'
852 is used without iterators.
853 (initialize_iterators): Initialize the new iterator_group fields.
854 * genopinit.c (handle_overloaded_code_for)
855 (handle_overloaded_gen): New functions.
856 (main): Use them to print declarations of maybe_code_for_* and
857 maybe_gen_* functions, and inline definitions of code_for_* and gen_*.
858 * genemit.c (print_overload_arguments, print_overload_test)
859 (handle_overloaded_code_for, handle_overloaded_gen): New functions.
860 (main): Use it to print definitions of maybe_code_for_* and
861 maybe_gen_* functions.
862 * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use
863 gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high}
864 instead of explicit mode checks.
865 (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine.
866 (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov.
867 (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive.
868 (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive.
869 (aarch64_expand_compare_and_swap): Likewise
870 gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse
871 (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas.
872 (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp.
873 (aarch64_constant_pool_reload_icode): Delete.
874 (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp
875 instead of aarch64_constant_pool_reload_icode. Use
876 code_for_aarch64_reload_mov instead of explicit mode checks.
877 (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete.
878 (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of
879 get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type.
880 (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete.
881 (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of
882 get_recpe_type and gen_aarch64_frecps instead of get_recps_type.
883 (aarch64_atomic_load_op_code): Delete.
884 (aarch64_emit_atomic_load_op): Likewise.
885 (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of
886 aarch64_atomic_load_op_code. Use gen_aarch64_atomic_load
887 instead of aarch64_emit_atomic_load_op.
888 * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>)
889 (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>)
890 (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high)
891 (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@'
892 character before the pattern name.
893 * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
894 (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>)
895 (aarch64_simd_combine<mode>, aarch64_frecpe<mode>)
896 (aarch64_frecps<mode>): Likewise.
897 * config/aarch64/atomics.md (atomic_compare_and_swap<mode>)
898 (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse)
899 (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>)
900 (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>)
901 (aarch64_atomic_load<atomic_ldop><mode>): Likewise.
903 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
905 * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
906 Allow HFmode constants if TARGET_FP_F16INST.
908 2018-08-02 Jackson Woodruff <jackson.woodruff@arm.com>
911 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp):
912 No longer check last store for clobber of address register.
914 2018-08-02 Martin Liska <mliska@suse.cz>
916 PR gcov-profile/86817
917 * gcov.c (process_all_functions): New function.
919 (process_file): Move functions processing to
920 process_all_functions.
922 2018-08-02 David Malcolm <dmalcolm@redhat.com>
924 * dumpfile.c (dump_user_location_t::dump_user_location_t): Add
925 "const" to the "gimple *" and "rtx_insn *" parameters.
926 * dumpfile.h (dump_user_location_t::dump_user_location_t):
928 (dump_location_t::dump_location_t): Likewise.
930 2018-08-01 Martin Sebor <msebor@redhat.com>
932 PR tree-optimization/86650
933 * gimple-pretty-print.c (percent_G_format): Accept a "gimple *"
934 rather than a "gcall *". Directly pass the data of interest
935 to percent_K_format, rather than building a temporary CALL_EXPR
937 * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust.
938 (gimple_fold_builtin_strncat): Adjust.
939 * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace
940 gcall* argument with gimple*.
941 * gimple-ssa-warn-restrict.c (check_call): Same.
942 (wrestrict_dom_walker::before_dom_children): Same.
943 (builtin_access::builtin_access): Same.
944 (check_bounds_or_overlap): Same
945 (maybe_diag_overlap): Same.
946 (maybe_diag_offset_bounds): Same.
947 * tree-diagnostic.c (default_tree_printer): Move usage of
948 EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format
950 * tree-pretty-print.c (percent_K_format): Add argument.
951 * tree-pretty-print.h: Add argument.
952 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust.
953 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust.
954 (maybe_diag_stxncpy_trunc): Same.
955 (handle_builtin_stxncpy): Same.
956 (handle_builtin_strcat): Same.
958 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
960 * match.pd: Optimise pointer range checks.
962 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
964 PR tree-optimization/86758
965 * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try
966 to remove pattern statements.
968 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
970 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the
971 result of dfs_enumerate_from when constructing stmt_vec_infos,
972 instead of additionally calling get_loop_body.
974 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
976 * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p
978 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise.
979 When creating an iv, assert that the step is not known to be zero.
980 (vect_setup_realignment): Update call accordingly.
981 * tree-vect-stmts.c (vectorizable_store): Likewise.
982 (vectorizable_load): Likewise. Handle VMAT_INVARIANT separately.
984 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
986 * tree-vectorizer.h (vect_stmt_to_vectorize): New function.
987 * tree-vect-loop.c (vect_update_vf_for_slp): Use it.
988 (vectorizable_reduction): Likewise.
989 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
990 (vect_detect_hybrid_slp_stmts): Likewise.
991 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
993 2018-08-01 Aldy Hernandez <aldyh@redhat.com>
995 * tree-vrp (zero_nonzero_bits_from_bounds): Rename to...
996 (wide_int_set_zero_nonzero_bits): ...this.
997 (zero_nonzero_bits_from_vr): Rename to...
998 (vrp_set_zero_nonzero_bits): ...this.
999 (extract_range_from_multiplicative_op_1): Abstract wide int
1001 (wide_int_range_multiplicative_op): ...here.
1002 (extract_range_from_binary_expr_1): Extract wide int binary
1003 operations into their own functions.
1004 (wide_int_range_lshift): New.
1005 (wide_int_range_can_optimize_bit_op): New.
1006 (wide_int_range_shift_undefined_p): New.
1007 (wide_int_range_bit_xor): New.
1008 (wide_int_range_bit_ior): New.
1009 (wide_int_range_bit_and): New.
1010 (wide_int_range_trunc_mod): New.
1011 (extract_range_into_wide_ints): New.
1012 (vrp_shift_undefined_p): New.
1013 (extract_range_from_multiplicative_op): New.
1014 (vrp_can_optimize_bit_op): New.
1015 * tree-vrp.h (value_range::dump): New.
1016 (wide_int_range_multiplicative_op): New.
1017 (wide_int_range_lshift):New.
1018 (wide_int_range_shift_undefined_p): New.
1019 (wide_int_range_bit_xor): New.
1020 (wide_int_range_bit_ior): New.
1021 (wide_int_range_bit_and): New.
1022 (wide_int_range_trunc_mod): New.
1023 (zero_nonzero_bits_from_bounds): Rename to...
1024 (wide_int_set_zero_nonzero_bits): ...this.
1025 (zero_nonzero_bits_from_vr): Rename to...
1026 (vrp_set_zero_nonzero_bits): ...this.
1027 (range_easy_mask_min_max): Rename to...
1028 (wide_int_range_can_optimize_bit_op): this.
1029 * vr-values.c (simplify_bit_ops_using_ranges): Rename
1030 zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits.
1032 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
1034 * tree-vectorizer.h (vect_orig_stmt): New function.
1035 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it.
1036 * tree-vect-loop.c (vect_model_reduction_cost): Likewise.
1037 (vect_create_epilog_for_reduction): Likewise.
1038 (vectorizable_live_operation): Likewise.
1039 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
1040 (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise.
1041 * tree-vect-stmts.c (vectorizable_call): Likewise.
1042 (vectorizable_simd_clone_call, vect_remove_stores): Likewise.
1044 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
1046 * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store
1048 * tree-vect-stmts.c (vect_transform_stmt): Likewise.
1049 * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly.
1050 (vect_transform_loop): Likewise.
1051 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
1053 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
1055 * tree-vectorizer.h (vect_schedule_slp): Return void.
1056 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
1057 (vect_schedule_slp): Likewise.
1059 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
1061 * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled
1063 (vect_transform_loop): Update calls accordingly. Schedule SLP
1064 instances before the main loop, if any exist.
1066 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
1068 PR tree-optimization/86749
1069 * tree-vect-patterns.c (vect_determine_min_output_precision_1):
1070 If the lhs is used in a COND_EXPR, check that it is being used
1071 as the "then" or "else" value.
1073 2018-08-01 Tom de Vries <tdevries@suse.de>
1076 * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
1077 speculation_safe_value_not_needed.
1079 2018-08-01 Richard Biener <rguenther@suse.de>
1081 * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses
1084 2018-08-01 Martin Liska <mliska@suse.cz>
1086 * value-prof.c (gimple_divmod_fixed_value_transform): Unify
1087 format how successful transformation is dumped.
1088 (gimple_mod_pow2_value_transform): Likewise.
1089 (gimple_mod_subtract_transform): Likewise.
1090 (gimple_stringops_transform): Likewise.
1092 2018-08-01 Martin Liska <mliska@suse.cz>
1095 * value-prof.c (interesting_stringop_to_profile_p):
1096 Simplify the code and add BUILT_IN_MEMMOVE.
1097 (gimple_stringops_transform): Likewise.
1099 2018-08-01 Sam Tebbs <sam.tebbs@arm.com>
1101 * config/aarch64/aarch64-simd.md
1102 (*aarch64_get_lane_zero_extendsi<mode>): Rename to...
1103 (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and
1104 use GPI iterator instead of SI mode.
1106 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
1108 * config/rs6000/rs6000.md (speculation_barrier): Renamed from
1109 rs6000_speculation_barrier.
1110 * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for
1111 new barrier pattern name.
1113 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
1115 * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
1116 (speculation_barrier): New insn.
1118 2018-08-01 Richard Biener <rguenther@suse.de>
1121 * graphite.h: Include isl/id.h and isl/space.h to allow build
1124 2018-08-01 Jan Willem Jagersma <jwjagersma@gmail.com>
1127 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
1128 mode for COFF targets.
1129 * defaults.h (TARGET_COFF): Define.
1130 * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
1131 TARGET_COFF): Define.
1132 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
1133 * config/i386/djgpp.c (saved_debug_info_level): New static variable.
1134 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
1136 2018-07-31 Alexandre Oliva <oliva@adacore.com>
1138 * gimple-streamer-in.c (input_bb): Restore BB discriminator.
1139 * gimple-streamer-out.c (output_bb): Save it.
1140 * lto-streamer-in.c (input_struct_function_base): Restore
1141 instance discriminator if available. Create map on demand.
1142 * lto-streamer-out.c (output_struct_function_base): Save it if
1144 * final.c (decl_to_instance_map): Document LTO strategy.
1146 2018-07-31 Alexandre Oliva <oliva@adacore.com>
1147 Olivier Hainque <hainque@adacore.com>
1149 * debug.h (decl_to_instance_map_t): New type.
1150 (decl_to_instance_map): Declare.
1151 (maybe_create_decl_to_instance_map): New inline function.
1152 * final.c (bb_discriminator, last_bb_discriminator): New statics,
1153 to track basic block discriminators.
1154 (final_start_function_1): Initialize them.
1155 (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
1157 (decl_to_instance_map): New variable.
1158 (map_decl_to_instance, maybe_set_discriminator): New functions.
1159 (notice_source_line): Set discriminator.
1161 2018-07-31 Ian Lance Taylor <iant@golang.org>
1163 * targhooks.c (default_have_speculation_safe_value): Add
1166 2018-07-31 David Malcolm <dmalcolm@redhat.com>
1168 * dump-context.h: Include "pretty-print.h".
1169 (dump_context::refresh_dumps_are_enabled): New decl.
1170 (dump_context::emit_item): New decl.
1171 (class dump_context): Add fields "m_test_pp" and
1173 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
1174 (temp_dump_context::get_dumped_text): New decl.
1175 (class temp_dump_context): Add field "m_pp".
1176 * dumpfile.c (refresh_dumps_are_enabled): Convert to...
1177 (dump_context::refresh_dumps_are_enabled): ...and add a test for
1179 (set_dump_file): Update for above change.
1180 (set_alt_dump_file): Likewise.
1181 (dump_loc): New overload, taking a pretty_printer *.
1182 (dump_context::dump_loc): Call end_any_optinfo. Dump the location
1183 to any test pretty-printer.
1184 (make_item_for_dump_gimple_stmt): New function, adapted from
1185 optinfo::add_gimple_stmt.
1186 (dump_context::dump_gimple_stmt): Call it, and use the result,
1187 eliminating the direct usage of dump_file and alt_dump_file in
1188 favor of indirectly using them via emit_item.
1189 (make_item_for_dump_gimple_expr): New function, adapted from
1190 optinfo::add_gimple_expr.
1191 (dump_context::dump_gimple_expr): Call it, and use the result,
1192 eliminating the direct usage of dump_file and alt_dump_file in
1193 favor of indirectly using them via emit_item.
1194 (make_item_for_dump_generic_expr): New function, adapted from
1196 (dump_context::dump_generic_expr): Call it, and use the result,
1197 eliminating the direct usage of dump_file and alt_dump_file in
1198 favor of indirectly using them via emit_item.
1199 (make_item_for_dump_printf_va): New function, adapted from
1200 optinfo::add_printf_va.
1201 (make_item_for_dump_printf): New function.
1202 (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
1203 and use the result, eliminating the direct usage of dump_file and
1204 alt_dump_file in favor of indirectly using them via emit_item.
1205 (make_item_for_dump_dec): New function.
1206 (dump_context::dump_dec): Call it, and use the result,
1207 eliminating the direct usage of dump_file and alt_dump_file in
1208 favor of indirectly using them via emit_item.
1209 (make_item_for_dump_symtab_node): New function, adapted from
1210 optinfo::add_symtab_node.
1211 (dump_context::dump_symtab_node): Call it, and use the result,
1212 eliminating the direct usage of dump_file and alt_dump_file in
1213 favor of indirectly using them via emit_item.
1214 (dump_context::begin_scope): Reimplement, avoiding direct usage
1215 of dump_file and alt_dump_file in favor of indirectly using them
1217 (dump_context::emit_item): New member function.
1218 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
1219 Set up test pretty-printer on the underlying context. Call
1220 refresh_dumps_are_enabled.
1221 (temp_dump_context::~temp_dump_context): Call
1222 refresh_dumps_are_enabled.
1223 (temp_dump_context::get_dumped_text): New member function.
1224 (selftest::verify_dumped_text): New function.
1225 (ASSERT_DUMPED_TEXT_EQ): New macro.
1226 (selftest::test_capture_of_dump_calls): Run all tests twice, with
1227 and then without optinfo enabled. Add uses of
1228 ASSERT_DUMPED_TEXT_EQ to all tests. Add test of nested scopes.
1229 * dumpfile.h: Update comment for the dump_* API.
1230 * optinfo-emit-json.cc
1231 (selftest::test_building_json_from_dump_calls): Update for new
1232 param for temp_dump_context ctor.
1233 * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
1234 and "m_owned" field.
1235 (optinfo_item::~optinfo_item): Likewise.
1236 (optinfo::add_item): New member function.
1237 (optinfo::emit): Update comment.
1238 (optinfo::add_string): Delete.
1239 (optinfo::add_printf): Delete.
1240 (optinfo::add_printf_va): Delete.
1241 (optinfo::add_gimple_stmt): Delete.
1242 (optinfo::add_gimple_expr): Delete.
1243 (optinfo::add_tree): Delete.
1244 (optinfo::add_symtab_node): Delete.
1245 (optinfo::add_dec): Delete.
1246 * optinfo.h (class dump_context): New forward decl.
1247 (optinfo::add_item): New decl.
1248 (optinfo::add_string): Delete.
1249 (optinfo::add_printf): Delete.
1250 (optinfo::add_printf_va): Delete.
1251 (optinfo::add_gimple_stmt): Delete.
1252 (optinfo::add_gimple_expr): Delete.
1253 (optinfo::add_tree): Delete.
1254 (optinfo::add_symtab_node): Delete.
1255 (optinfo::add_dec): Delete.
1256 (optinfo::add_poly_int): Delete.
1257 (optinfo_item::optinfo_item): Remove "owned" param.
1258 (class optinfo_item): Remove field "m_owned".
1260 2018-07-31 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1263 * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
1264 MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
1265 requested variable alignment.
1266 (expand_one_ssa_partition): Likewise.
1267 (expand_one_var): Likewise.
1269 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1271 * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
1272 to speculation_safe_value_not_needed.
1274 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1276 * targhooks.h (speculation_safe_value_not_needed): New prototype.
1277 * targhooks.c (speculation_safe_value_not_needed): New function.
1278 * target.def (have_speculation_safe_value): Update documentation.
1279 * doc/tm.texi: Regenerated.
1281 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1283 * config/aarch64/iterators.md (ALLI_TI): New iterator.
1284 * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
1286 (despeculate_copy<ALLI:mode>_insn): New insn.
1287 (despeculate_copyti_insn): New insn.
1288 (despeculate_simple<ALLI:mode>): New insn
1289 (despeculate_simpleti): New insn.
1290 * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
1292 (TARGET_SPECULATION_SAFE_VALUE): Redefine to
1293 aarch64_speculation_safe_value.
1294 (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
1296 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1298 * config/aarch64/aarch64-speculation.cc: New file.
1299 * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
1300 before pass_reorder_blocks.
1301 * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
1303 * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
1304 X14 and X15 when tracking speculation.
1305 * config/aarch64/aarch64.md (register name constants): Add
1306 SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
1307 (unspec): Add UNSPEC_SPECULATION_TRACKER.
1308 (speculation_barrier): New insn attribute.
1309 (cmp<mode>): Allow SP in comparisons.
1310 (speculation_tracker): New insn.
1311 (speculation_barrier): Add speculation_barrier attribute.
1312 * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
1313 * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
1314 * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
1316 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1318 * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
1319 aarch64_track_speculation is true.
1320 (tb<optab><mode>1): Likewise.
1321 * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
1322 generate CB[N]Z when tracking speculation.
1323 (aarch64_split_compare_and_swap): Likewise.
1324 (aarch64_split_atomic_op): Likewise.
1326 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1328 * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
1330 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1332 * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
1333 (speculation_barrier): New insn.
1335 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1337 * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
1338 * config/arm/arm.md (speculation_barrier): New expand.
1339 (speculation_barrier_insn): New pattern.
1341 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1343 * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
1344 (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
1345 (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
1346 * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
1348 * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
1349 (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
1350 (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
1351 (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
1352 (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
1353 (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
1354 (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
1355 * builtins.c (expand_speculation_safe_value): New function.
1356 (expand_builtin): Call it.
1357 * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
1358 * doc/extend.texi: Document __builtin_speculation_safe_value.
1359 * doc/md.texi: Document "speculation_barrier" pattern.
1360 * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
1361 TARGET_HAVE_SPECULATION_SAFE_VALUE.
1362 * doc/tm.texi: Regenerated.
1363 * target.def (have_speculation_safe_value, speculation_safe_value): New
1365 * targhooks.c (default_have_speculation_safe_value): New function.
1366 (default_speculation_safe_value): New function.
1367 * targhooks.h (default_have_speculation_safe_value): Add prototype.
1368 (default_speculation_safe_value): Add prototype.
1370 2018-07-31 David Malcolm <dmalcolm@redhat.com>
1372 * dump-context.h (dump_context::dump_loc): New decl.
1373 * dumpfile.c (dump_context::dump_loc): New member function.
1374 (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
1375 and dump_gimple_stmt.
1376 (dump_context::dump_gimple_expr_loc): Likewise, using
1378 (dump_context::dump_generic_expr_loc): Likewise, using
1380 (dump_context::dump_printf_loc_va): Likewise, using
1382 (dump_context::begin_scope): Explicitly using the global function
1383 "dump_loc", rather than the member function.
1385 2018-07-31 Martin Sebor <msebor@redhat.com>
1387 PR tree-optimization/86741
1388 * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
1390 2018-07-31 Andreas Krebbel <krebbel@linux.ibm.com>
1392 * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
1393 depend on whether prefetch instructions will be emitted or not.
1394 Use TARGET_SETMEM_PFD for checking whether prefetch instructions
1395 will be emitted or not.
1396 * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
1397 (TARGET_SETMEM_PFD): New macros.
1399 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1401 * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
1402 (NULL_STMT_VEC_INFO): Delete.
1403 (stmt_vec_info::operator*): Likewise.
1404 (stmt_vec_info::operator gimple *): Likewise.
1405 * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
1406 of NULL_STMT_VEC_INFO.
1407 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
1408 (vect_reassociating_reduction_p): Likewise.
1409 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
1410 (vectorizable_store): Likewise.
1411 * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
1412 (vec_info::free_stmt_vec_infos): Likewise.
1414 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1416 * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
1417 * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
1419 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1421 * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
1422 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
1423 (vec_info::free_stmt_vec_info): New private member functions.
1424 (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
1425 (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
1426 * tree-parloops.c (gather_scalar_reductions): Remove calls to
1427 set_stmt_vec_info_vec and free_stmt_vec_infos.
1428 * tree-vect-loop.c (_loop_vec_info): Remove call to
1429 set_stmt_vec_info_vec.
1430 * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
1431 (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
1432 * tree-vectorizer.c (vec_info::new_stmt_vec_info)
1433 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
1434 (vec_info::free_stmt_vec_info): ...these new functions. Remove
1435 assignments in {vec_info::,}new_stmt_vec_info that are redundant
1436 with the clearing in the xcalloc.
1437 (stmt_vec_info_vec): Delete.
1438 (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
1439 (vectorize_loops): Likewise.
1440 (vec_info::~vec_info): Remove argument from call to
1441 free_stmt_vec_infos.
1442 (vec_info::add_stmt): Remove vinfo argument from call to
1445 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1447 * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
1448 rather than a gimple stmt.
1449 * tree-vect-stmts.c (free_stmt_vec_info): Likewise. Don't free
1450 information for pattern statements when passed the original
1451 statement; instead wait to be passed the pattern statement itself.
1452 Don't call set_vinfo_for_stmt here.
1453 (free_stmt_vec_infos): Update call to free_stmt_vec_info.
1454 * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
1455 stmt_vec_infos here.
1456 * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
1457 * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
1458 stmt_vec_infos entry.
1460 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1462 * tree-vectorizer.h (vec_info::replace_stmt): Declare.
1463 * tree-vectorizer.c (vec_info::replace_stmt): New function.
1464 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
1465 * tree-vect-stmts.c (vectorizable_call): Likewise.
1466 (vectorizable_simd_clone_call): Likewise.
1468 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1470 * tree-vectorizer.h (vec_info::remove_stmt): Declare.
1471 * tree-vectorizer.c (vec_info::remove_stmt): New function.
1472 * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
1473 * tree-vect-loop.c (vect_transform_loop): Likewise.
1474 * tree-vect-slp.c (vect_schedule_slp): Likewise.
1475 * tree-vect-stmts.c (vect_remove_stores): Likewise.
1477 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1479 * tree-vectorizer.h (vec_info::lookup_dr): New member function.
1480 (vect_dr_stmt): Delete.
1481 * tree-vectorizer.c (vec_info::lookup_dr): New function.
1482 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
1484 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
1485 (vect_analyze_data_ref_dependence, vect_record_base_alignments)
1486 (vect_verify_datarefs_alignment, vect_peeling_supportable)
1487 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
1488 (vect_analyze_data_refs): Likewise.
1489 (vect_slp_analyze_data_ref_dependence): Likewise. Take a vec_info
1491 (vect_find_same_alignment_drs): Likewise.
1492 (vect_slp_analyze_node_dependences): Update calls accordingly.
1493 (vect_analyze_data_refs_alignment): Likewise. Use vec_info::lookup_dr
1494 instead of DR_VECT_AUX.
1495 (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
1496 of a vector data references. Use vec_info::lookup_dr instead of
1498 (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
1499 (vect_enhance_data_refs_alignment): Likewise. Use vec_info::lookup_dr
1500 instead of DR_VECT_AUX.
1502 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1504 * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
1506 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
1508 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
1509 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
1510 (vect_gen_prolog_loop_niters): Likewise.
1512 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1514 * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
1515 (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
1516 (vect_known_alignment_in_bytes, vect_dr_behavior)
1517 (vect_get_scalar_dr_size): Take references as dr_vec_infos
1518 instead of data_references. Update calls to other routines for
1519 which the same change has been made.
1520 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
1521 dr_vec_infos instead of stmt_vec_infos.
1522 (vect_analyze_data_ref_dependence): Update call accordingly.
1523 (vect_slp_analyze_data_ref_dependence)
1524 (vect_record_base_alignments): Use DR_VECT_AUX.
1525 (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
1526 (vect_update_misalignment_for_peel, verify_data_ref_alignment)
1527 (vector_alignment_reachable_p, vect_get_data_access_cost)
1528 (vect_peeling_supportable, vect_analyze_group_access_1)
1529 (vect_analyze_group_access, vect_analyze_data_ref_access)
1530 (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
1531 (vect_compile_time_alias, vect_small_gap_p)
1532 (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
1533 (vect_supportable_dr_alignment): Take references as dr_vec_infos
1534 instead of data_references. Update calls to other routines for
1535 which the same change has been made.
1536 (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
1537 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
1538 (vect_slp_analyze_and_verify_node_alignment)
1539 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
1540 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
1541 (vect_setup_realignment): Use dr_vec_infos. Update calls after
1543 (_vect_peel_info::dr): Replace with...
1544 (_vect_peel_info::dr_info): ...this new field.
1545 (vect_peeling_hash_get_most_frequent)
1546 (vect_peeling_hash_choose_best_peeling): Update accordingly.
1547 (vect_peeling_hash_get_lowest_cost):
1548 (vect_enhance_data_refs_alignment): Likewise. Update calls to other
1549 routines for which the same change has been made.
1550 (vect_peeling_hash_insert): Likewise. Take a dr_vec_info instead of a
1552 * tree-vect-loop-manip.c (get_misalign_in_elems)
1553 (vect_gen_prolog_loop_niters): Use dr_vec_infos. Update calls after
1555 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
1556 * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
1557 (vect_truncate_gather_scatter_offset, compare_step_with_zero)
1558 (get_group_load_store_type, get_negative_load_store_type)
1559 (vect_get_data_ptr_increment, vectorizable_store)
1560 (vectorizable_load): Likewise.
1561 (ensure_base_align): Take a dr_vec_info instead of a data_reference.
1562 Update calls to other routines for which the same change has been made.
1564 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1566 * tree-vectorizer.h (vec_info::move_dr): New member function.
1567 (dataref_aux): Rename to...
1568 (dr_vec_info): ...this and add "dr" and "stmt" fields.
1569 (_stmt_vec_info::dr_aux): Update accordingly.
1570 (_stmt_vec_info::data_ref_info): Delete.
1571 (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
1572 (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
1573 (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
1574 (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
1576 (STMT_VINFO_DATA_REF): Likewise. Turn into an lvalue.
1577 (STMT_VINFO_DR_INFO): New macro.
1578 (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
1579 (set_dr_misalignment): Update after rename of dataref_aux.
1580 (vect_dr_stmt): Move earlier in file. Return dr_aux.stmt.
1581 * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
1582 initialization of STMT_VINFO_DATA_REF.
1583 * tree-vectorizer.c (vec_info::move_dr): New function.
1584 * tree-vect-patterns.c (vect_recog_bool_pattern)
1585 (vect_recog_mask_conversion_pattern)
1586 (vect_recog_gather_scatter_pattern): Use it.
1587 * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
1588 the "dr" and "stmt" fields of dr_vec_info instead of
1589 STMT_VINFO_DATA_REF.
1591 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1593 * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
1594 (is_pattern_stmt_p): Use it.
1595 * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
1596 on pattern statements.
1598 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1600 * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
1601 original stmt as a stmt_vec_info rather than a gimple stmt.
1602 (vect_pattern_recog_1): Take the statement directly as a
1603 stmt_vec_info, rather than via a gimple_stmt_iterator.
1604 Update call to vect_mark_pattern_stmts.
1605 (vect_pattern_recog): Update calls accordingly.
1607 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1609 * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
1610 (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
1611 a vect_def_type for the first argument.
1612 * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
1613 (vect_get_vec_def_for_stmt_copy): Likewise. Return the original
1614 operand if it isn't defined by a vectorized statement.
1615 (vect_build_gather_load_calls): Remove the mask_dt argument and
1616 update calls to vect_get_vec_def_for_stmt_copy.
1617 (vectorizable_bswap): Likewise the dt argument.
1618 (vectorizable_call): Update calls to vectorizable_bswap and
1619 vect_get_vec_def_for_stmt_copy.
1620 (vectorizable_simd_clone_call, vectorizable_assignment)
1621 (vectorizable_shift, vectorizable_operation, vectorizable_condition)
1622 (vectorizable_comparison): Update calls to
1623 vect_get_vec_def_for_stmt_copy.
1624 (vectorizable_store): Likewise. Remove now-unnecessary calls to
1626 (vect_get_loop_based_defs): Remove dt argument and update call
1627 to vect_get_vec_def_for_stmt_copy.
1628 (vectorizable_conversion): Update calls to vect_get_loop_based_defs
1629 and vect_get_vec_def_for_stmt_copy.
1630 (vectorizable_load): Update calls to vect_build_gather_load_calls
1631 and vect_get_vec_def_for_stmt_copy.
1632 * tree-vect-loop.c (vect_create_epilog_for_reduction)
1633 (vectorizable_reduction, vectorizable_live_operation): Update calls
1634 to vect_get_vec_def_for_stmt_copy.
1636 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1638 * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
1639 and gimple stmt arguments with a stmt_vec_info.
1640 (vect_record_base_alignments): Update calls accordingly.
1641 * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
1642 and gimple stmt arguments with a stmt_vec_info.
1643 (vect_build_slp_tree_1): Remove vinfo argument and update call
1644 to vect_record_max_nunits.
1645 (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
1646 and vect_record_max_nunits.
1648 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1650 * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
1651 file and take a stmt_vec_info instead of a gimple stmt.
1652 (supportable_widening_operation, vect_finish_replace_stmt)
1653 (vect_finish_stmt_generation, vect_get_store_rhs)
1654 (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
1655 (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
1656 (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
1657 (vect_get_smallest_scalar_type, vect_check_gather_scatter)
1658 (vect_create_data_ref_ptr, bump_vector_ptr)
1659 (vect_permute_store_chain, vect_setup_realignment)
1660 (vect_transform_grouped_load, vect_record_grouped_load_vectors)
1661 (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
1662 (vectorizable_reduction, vectorizable_induction)
1663 (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
1664 (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
1665 than gimple stmts as arguments.
1666 * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
1667 (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
1668 (can_group_stmts_p, vect_check_gather_scatter)
1669 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
1670 (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
1671 (vect_permute_load_chain, vect_shift_permute_load_chain)
1672 (vect_transform_grouped_load)
1673 (vect_record_grouped_load_vectors): Likewise.
1674 * tree-vect-loop.c (vect_fixup_reduc_chain)
1675 (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
1676 (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
1677 (vectorizable_reduction, vectorizable_induction)
1678 (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
1679 * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
1680 (vect_get_load_store_mask): Likewise.
1681 * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
1682 (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
1683 * tree-vect-stmts.c (vect_mark_relevant)
1684 (is_simple_and_all_uses_invariant)
1685 (exist_non_indexing_operands_for_use_p, process_use)
1686 (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
1687 (vect_get_vec_def_for_operand, vect_get_vec_defs)
1688 (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
1689 (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
1690 (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
1691 (get_negative_load_store_type, get_load_store_type)
1692 (vect_check_load_store_mask, vect_check_store_rhs)
1693 (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
1694 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
1695 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
1696 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
1697 (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
1698 (vectorizable_load, vectorizable_condition, vectorizable_comparison)
1699 (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
1700 (supportable_widening_operation): Likewise.
1702 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1704 * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
1705 a stmt_vec_info instead of a gcall.
1706 (vect_check_gather_scatter): Update call accordingly.
1707 * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
1709 (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
1710 (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
1711 * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
1712 instead of a gimple stmt.
1713 (vect_transform_loop): Update calls accordingly.
1714 * tree-vect-slp.c (vect_split_slp_store_group): Take and return
1715 stmt_vec_infos instead of gimple stmts.
1716 (vect_analyze_slp_instance): Update use accordingly.
1717 * tree-vect-stmts.c (read_vector_array, write_vector_array)
1718 (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
1719 (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
1720 (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
1721 (vect_gen_widened_results_half, vect_get_loop_based_defs)
1722 (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
1723 Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
1724 down to subroutines.
1726 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1728 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
1729 of the worklist from a vector of gimple stmts to a vector of
1731 * tree-vect-stmts.c (vect_mark_relevant, process_use)
1732 (vect_mark_stmts_to_be_vectorized): Likewise
1734 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1736 * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
1737 statement before passing it to vect_analyze_stmt.
1738 (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
1739 the chain of phi vector definitions. Track the exit phi via its
1741 (vectorizable_reduction): Set cond_stmt_vinfo directly from the
1742 STMT_VINFO_REDUC_DEF.
1743 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
1744 stmt_vec_infos to handle the statement chains.
1745 (vect_get_slp_defs): Record the first statement in the node
1746 using a stmt_vec_info.
1747 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
1748 statements here and pass their stmt_vec_info down to subroutines.
1749 (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
1750 down to vect_finish_stmt_generation.
1751 (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
1752 (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
1753 stmt_vec_infos to subroutines.
1754 (vect_remove_stores): Use stmt_vec_infos to handle the statement
1757 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1759 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
1760 (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
1761 (vect_permute_store_chain, vect_setup_realignment)
1762 (vect_permute_load_chain, vect_shift_permute_load_chain)
1763 (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
1764 stmts internally, and when passing values to other vectorizer routines.
1765 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
1766 * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
1767 (vect_analyze_loop_operations, get_initial_def_for_reduction)
1768 (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
1769 (vectorizable_reduction, vectorizable_induction)
1770 (vectorizable_live_operation, vect_transform_loop_stmt)
1771 (vect_transform_loop): Likewise.
1772 * tree-vect-patterns.c (vect_reassociating_reduction_p)
1773 (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
1774 (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
1775 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
1776 (vect_slp_analyze_node_operations_1): Likewise.
1777 * tree-vect-stmts.c (vect_mark_relevant, process_use)
1778 (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
1779 (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
1780 (vect_finish_stmt_generation_1, get_group_load_store_type)
1781 (get_load_store_type, vect_build_gather_load_calls)
1782 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
1783 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
1784 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
1785 (vectorizable_store, vectorizable_load, vectorizable_condition)
1786 (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
1787 (supportable_widening_operation): Likewise.
1788 (vect_get_vector_types_for_stmt): Likewise.
1789 * tree-vectorizer.h (vect_dr_behavior): Likewise.
1791 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1793 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
1794 (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
1795 (vect_permute_store_chain, vect_permute_load_chain)
1796 (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
1797 repeated stmt_vec_info lookups.
1798 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
1799 (vect_update_ivs_after_vectorizer): Likewise.
1800 * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
1801 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
1802 * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
1803 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
1804 (vect_bb_slp_scalar_cost): Likewise.
1805 * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
1807 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1809 * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
1810 gcall rather than the generic gimple stmt to gimple_call_internal_fn.
1811 (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
1812 to get gassigns and gcalls, rather than operating on generc gimple
1814 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
1815 (vect_mark_stmts_to_be_vectorized, vectorizable_store)
1816 (vectorizable_load, vect_analyze_stmt): Likewise.
1817 * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
1819 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1821 * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
1822 return stmt_vec_infos rather than gimple stmts. Do not accept
1824 (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
1826 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
1827 Update use of get_later_stmt.
1828 (vect_get_constant_vectors): Update call accordingly.
1829 (vect_schedule_slp_instance): Likewise
1830 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
1831 (vect_slp_analyze_instance_dependence): Likewise.
1832 (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
1834 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1836 * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
1837 (stmt_info_for_cost::stmt_info): ...this new field.
1838 (add_stmt_costs): Update accordingly.
1839 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
1840 (vect_get_known_peeling_cost): Likewise.
1841 (vect_estimate_min_profitable_iters): Likewise.
1842 * tree-vect-stmts.c (record_stmt_cost): Likewise.
1844 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1846 * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
1847 from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
1848 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
1850 * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
1852 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1854 * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
1855 a gimple stmt to a stmt_vec_info.
1856 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
1858 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1860 * tree-vectorizer.h (vec_info::grouped_stores): Change from
1861 an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
1862 (_loop_vec_info::reduction_chains): Likewise.
1863 * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
1865 * tree-vect-slp.c (vect_analyze_slp): Likewise.
1867 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1869 * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
1870 a gimple stmt to a stmt_vec_info.
1871 (_stmt_vec_info::next_element): Likewise.
1872 * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
1873 (vect_slp_analyze_and_verify_node_alignment)
1874 (vect_analyze_group_access_1, vect_analyze_group_access)
1875 (vect_small_gap_p, vect_prune_runtime_alias_test_list)
1876 (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
1877 (vect_supportable_dr_alignment): Update accordingly.
1878 * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
1879 (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
1880 (vect_is_simple_reduction, vectorizable_reduction): Likewise.
1881 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
1882 * tree-vect-slp.c (vect_build_slp_tree_1)
1883 (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
1884 (vect_split_slp_store_group, vect_analyze_slp_instance)
1885 (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
1886 * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
1887 (get_group_load_store_type, get_load_store_type)
1888 (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
1889 (vect_transform_stmt, vect_remove_stores): Likewise.
1891 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1893 * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
1895 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
1896 (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
1897 (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
1898 (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
1899 (vector_alignment_reachable_p, vect_get_data_access_cost)
1900 (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
1901 (vect_peeling_supportable, vect_enhance_data_refs_alignment)
1902 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
1903 (vect_analyze_group_access_1, vect_analyze_group_access)
1904 (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
1905 (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
1906 (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
1907 result of vect_dr_stmt and use the stmt_vec_info instead of
1908 the associated gimple stmt.
1909 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
1910 (vect_gen_prolog_loop_niters): Likewise.
1911 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
1913 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1915 * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
1916 to a vec<stmt_vec_info>.
1917 * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
1918 (vect_create_new_slp_node): Take a vec<gimple *> instead of a
1920 (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
1921 to a vec<stmt_vec_info>.
1922 (bst_traits::value_type, bst_traits::value_type): Likewise.
1923 (bst_traits::hash): Update accordingly.
1924 (vect_get_and_check_slp_defs): Change the stmts parameter from
1925 a vec<gimple *> to a vec<stmt_vec_info>.
1926 (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
1927 (vect_build_slp_tree): Likewise.
1928 (vect_build_slp_tree_2): Likewise. Update uses of
1929 SLP_TREE_SCALAR_STMTS.
1930 (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
1931 (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
1932 (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
1933 (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
1934 (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
1935 (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
1936 (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
1937 (vect_get_constant_vectors, vect_get_slp_defs)
1938 (vect_transform_slp_perm_load, vect_schedule_slp_instance)
1939 (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
1940 (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
1941 instead of gimple stmts.
1942 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
1943 the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
1944 (vect_slp_analyze_instance_dependence): Update uses of
1945 SLP_TREE_SCALAR_STMTS.
1946 (vect_slp_analyze_and_verify_node_alignment): Likewise.
1947 (vect_slp_analyze_and_verify_instance_alignment): Likewise.
1948 * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
1949 (get_initial_defs_for_reduction): Likewise.
1950 (vect_create_epilog_for_reduction): Likewise.
1951 (vectorize_fold_left_reduction): Likewise.
1952 * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
1953 (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
1954 (can_vectorize_live_stmts): Likewise.
1956 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1958 * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
1959 auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
1960 (vect_force_simple_reduction): Take and return stmt_vec_infos rather
1962 * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
1964 (gather_scalar_reductions): Update after above interface changes.
1965 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
1966 (vect_is_simple_reduction): Take and return stmt_vec_infos rather
1968 (vect_force_simple_reduction): Likewise.
1969 * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
1970 LOOP_VINFO_REDUCTIONS.
1971 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
1973 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1975 * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
1976 a gimple stmt to a stmt_vec_info.
1977 * tree-vect-loop.c (vect_active_double_reduction_p)
1978 (vect_force_simple_reduction, vectorizable_reduction): Update
1981 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1983 * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
1984 vec<gimple *> to a vec<stmt_vec_info>.
1985 * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
1986 the reduction_phis argument from a vec<gimple *> to a
1988 (vectorizable_reduction): Likewise the phis local variable that
1989 is passed to vect_create_epilog_for_reduction. Update for new type
1990 of SLP_TREE_VEC_STMTS.
1991 (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
1992 (vectorizable_live_operation): Likewise.
1993 * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
1994 (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
1996 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1998 * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
1999 a gimple stmt to a stmt_vec_info.
2000 (vectorizable_condition, vectorizable_live_operation)
2001 (vectorizable_reduction, vectorizable_induction): Pass back the
2002 vectorized statement as a stmt_vec_info.
2003 * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
2004 use of STMT_VINFO_VEC_STMT.
2005 * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
2006 accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
2007 as stmt_vec_infos rather than gimple stmts.
2008 (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
2010 (vectorizable_live_operation): Likewise.
2011 (vectorizable_reduction, vectorizable_induction): Likewise,
2012 updating use of STMT_VINFO_VEC_STMT.
2013 * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
2014 of STMT_VINFO_VEC_STMT.
2015 (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
2016 (vectorizable_simd_clone_call, vectorizable_conversion)
2017 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
2018 (vectorizable_store, vectorizable_load, vectorizable_condition)
2019 (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
2020 from a gimple stmt to a stmt_vec_info.
2021 (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT. Pass a
2022 pointer to a stmt_vec_info to the vectorizable_* routines.
2024 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2026 * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
2027 a gimple stmt to a stmt_vec_info.
2028 (is_pattern_stmt_p): Update accordingly.
2029 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
2030 (vect_record_grouped_load_vectors): Likewise.
2031 * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
2032 (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
2033 (vect_model_reduction_cost): Likewise.
2034 (vect_create_epilog_for_reduction): Likewise.
2035 (vectorizable_reduction, vectorizable_induction): Likewise.
2036 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
2037 Return the stmt_vec_info for the pattern statement.
2038 (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
2039 (vect_split_statement, vect_mark_pattern_stmts): Likewise.
2040 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
2041 (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
2042 * tree-vect-stmts.c (vect_mark_relevant): Likewise.
2043 (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
2044 (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
2045 (free_stmt_vec_info, vect_is_simple_use): Likewise.
2047 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2049 * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
2050 (vect_finish_stmt_generation): Likewise.
2051 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
2052 (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
2053 (vect_build_gather_load_calls): Use the return value of the above
2054 functions instead of a separate call to vinfo_for_stmt. Use narrow
2055 scopes for the input gimple stmt and wider scopes for the associated
2056 stmt_vec_info. Use vec_info::lookup_def when setting these
2057 stmt_vec_infos from an SSA_NAME definition.
2058 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
2059 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
2060 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
2061 (vectorizable_store, vectorizable_load, vectorizable_condition)
2062 (vectorizable_comparison): Likewise.
2063 * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
2064 (vectorizable_reduction): Likewise.
2066 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2068 * tree-vectorizer.h (vect_is_simple_use): Add an optional
2069 stmt_vec_info * parameter before the optional gimple **.
2070 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
2071 (process_use, vect_get_vec_def_for_operand_1): Update callers.
2072 (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
2073 * tree-vect-loop.c (vectorizable_reduction): Likewise.
2074 (vectorizable_live_operation): Likewise.
2075 * tree-vect-patterns.c (type_conversion_p): Likewise.
2076 (vect_look_through_possible_promotion): Likewise.
2077 (vect_recog_rotate_pattern): Likewise.
2078 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
2080 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2082 * tree-vectorizer.h (stmt_vec_info): Temporarily change from
2083 a typedef to a wrapper class.
2084 (NULL_STMT_VEC_INFO): New macro.
2085 (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
2086 (stmt_vec_info::operator*): New function.
2087 (stmt_vec_info::operator gimple *): Likewise.
2088 (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
2089 (add_stmt_costs): Likewise.
2090 * tree-vect-loop-manip.c (iv_phi_p): Likewise.
2091 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
2092 (vect_get_known_peeling_cost): Likewise.
2093 (vect_estimate_min_profitable_iters): Likewise.
2094 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
2095 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
2096 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
2097 (vectorizable_store, free_stmt_vec_infos): Likewise.
2098 (new_stmt_vec_info): Change return type of xcalloc to
2101 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2103 * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
2104 * tree-vectorizer.c (vec_info::lookup_single_use): New function.
2105 * tree-vect-loop.c (vectorizable_reduction): Use it instead of
2106 a single_imm_use-based sequence.
2107 * tree-vect-stmts.c (supportable_widening_operation): Likewise.
2109 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2111 * tree-vectorizer.h (vec_info::lookup_def): Declare.
2112 * tree-vectorizer.c (vec_info::lookup_def): New function.
2113 * tree-vect-patterns.c (vect_get_internal_def): Use it.
2114 (vect_widened_op_tree): Likewise.
2115 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
2116 * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
2117 (vectorizable_reduction): Likewise.
2118 (vect_valid_reduction_input_p): Take a stmt_vec_info instead
2120 (vect_is_slp_reduction): Update calls accordingly. Use
2121 vec_info::lookup_def.
2122 (vect_is_simple_reduction): Likewise
2123 * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
2125 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2127 * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
2128 * tree-vectorizer.c (vec_info::lookup_stmt): New function.
2129 * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
2131 (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
2132 (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
2133 (vect_update_vf_for_slp, vect_analyze_loop_operations)
2134 (vect_is_slp_reduction, vectorizable_induction)
2135 (vect_transform_loop_stmt, vect_transform_loop): Likewise.
2136 * tree-vect-patterns.c (vect_init_pattern_stmt):
2137 (vect_determine_min_output_precision_1, vect_determine_precisions)
2138 (vect_pattern_recog): Likewise.
2139 * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
2140 * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
2141 * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
2142 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
2143 (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
2144 (vect_detect_hybrid_slp): Likewise. Change the walk_stmt_info
2145 info field from a loop to a loop_vec_info.
2147 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2149 * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
2150 (vec_info::add_stmt): Declare.
2151 * tree-vectorizer.c (vec_info::add_stmt): New function.
2152 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
2153 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
2154 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
2155 (vectorizable_induction): Likewise.
2156 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
2157 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
2158 (vectorizable_simd_clone_call, vectorizable_store): Likewise.
2159 (vectorizable_load): Likewise.
2160 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
2161 (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
2162 (vect_recog_gather_scatter_pattern): Likewise.
2163 (append_pattern_def_seq): Likewise. Remove a check that is
2164 performed by add_stmt itself.
2166 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2168 * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
2169 which make_ssa_name was called with new_stmt before new_stmt
2172 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2174 * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
2176 (vect_is_slp_reduction): ...here...
2177 (vect_is_simple_reduction): ...and here. Remove repetition of tests
2178 that are already known to be false.
2180 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2182 * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
2183 * tree-vect-slp.c (vect_free_slp_tree): Likewise. Don't update
2184 STMT_VINFO_NUM_SLP_USES when it's true.
2185 (vect_free_slp_instance): Add a final_p parameter and pass it to
2187 (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
2188 (vect_analyze_slp_instance): Likewise.
2189 (vect_slp_analyze_operations): Likewise.
2190 (vect_slp_analyze_bb_1): Likewise.
2191 * tree-vectorizer.c (vec_info): Likewise.
2192 * tree-vect-loop.c (vect_transform_loop): Likewise.
2194 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2196 * tree-vect-loop.c (vectorizable_reduction): Assert that the
2197 function is not called for second and subsequent members of
2200 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2202 * tree-vect-loop.c (get_initial_def_for_reduction): Move special
2203 cases for nested loops from here to ...
2204 (vect_create_epilog_for_reduction): ...here. Only call
2205 vect_is_simple_use for inner-loop reductions.
2207 2018-07-31 Martin Liska <mliska@suse.cz>
2209 PR gcov-profile/85338
2210 PR gcov-profile/85350
2211 PR gcov-profile/85372
2212 * profile.c (struct location_triplet): New.
2213 (struct location_triplet_hash): Likewise.
2214 (output_location): Do not output a BB that
2215 is already recorded for a line.
2216 (branch_prob): Use streamed_locations.
2218 2018-07-31 Martin Liska <mliska@suse.cz>
2220 PR gcov-profile/85370
2221 * coverage.c (coverage_begin_function): Do not mark target
2222 clones as artificial functions.
2224 2018-07-31 Martin Liska <mliska@suse.cz>
2226 PR gcov-profile/83813
2227 PR gcov-profile/84758
2228 PR gcov-profile/85217
2229 PR gcov-profile/85332
2230 * profile.c (branch_prob): Do not record GOTO expressions
2231 for GIMPLE statements which locations are already streamed.
2233 2018-07-31 Olivier Hainque <hainque@adacore.com>
2235 * gcc.c (handle_spec_function): Accept a soft_matched_part
2236 argument, as do_spec_1. Pass it down to ...
2237 (eval_spec_function): Accept a soft_matched_part argument,
2238 and pass it down to ...
2239 (do_spec_2): Accept a soft_matched_part argument, and pass
2240 it down to do_spec_1.
2241 (do_spec_1): Pass soft_matched_part to handle_spec_function.
2242 (handle_braces): Update call to handle_spec_function.
2243 (driver::set_up_specs): Update calls to do_spec_2.
2244 (compare_debug_dump_opt_spec_function): Likewise.
2245 (compare_debug_self_opt_spec_function): Likewise.
2247 2018-07-31 Olivier Hainque <hainque@adacore.com>
2249 * common.opt (nolibc): New option.
2250 * doc/invoke.texi (Link Options): Document it.
2251 * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
2252 * config/alpha/linux.h: Likewise.
2253 * config/arc/elf.h: Likewise.
2254 * config/arm/uclinux-elf.h: Likewise.
2255 * config/arm/unknown-elf.h: Likewise.
2256 * config/avr/avrlibc.h: Likewise.
2257 * config/bfin/bfin.h: Likewise.
2258 * config/bfin/linux.h: Likewise.
2259 * config/bfin/uclinux.h: Likewise.
2260 * config/darwin.h: Likewise.
2261 * config/darwin10.h: Likewise.
2262 * config/darwin12.h: Likewise.
2263 * config/gnu-user.h: Likewise.
2264 * config/lm32/uclinux-elf.h: Likewise.
2265 * config/pa/pa-hpux11.h: Likewise.
2266 * config/pa/pa64-hpux.h: Likewise.
2267 * config/sparc/sparc.h: Likewise.
2269 2018-07-31 Olivier Hainque <hainque@adacore.com>
2271 * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
2272 undefined variables.
2274 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
2277 * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
2280 2018-07-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
2282 * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
2283 terminated string literal.
2285 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
2287 PR rtl-optimization/85160
2288 * combine.c (is_just_move): New function.
2289 (try_combine): Allow combining two instructions into two if neither of
2290 the original instructions was a move.
2292 2018-07-30 Alexander Monakov <amonakov@ispras.ru>
2295 * doc/extend.texi (Global Register Variables): Discourage use of type
2297 (Local Register Variables): Likewise.
2299 2018-07-30 Richard Sandiford <richard.sandiford@arm.com>
2301 PR tree-optimization/86506
2302 * hwint.h (ceil_log2): Resync with hwint.c implementation.
2304 2018-07-30 Ilya Leoshkevich <iii@linux.ibm.com>
2307 * lra-constraints.c (spill_hard_reg_in_range): When selecting the
2308 hard_regno, make sure no insn between `from` and `to` clobbers it.
2310 2018-07-30 Cesar Philippidis <cesar@codesourcery.com>
2311 Tom de Vries <tdevries@suse.de>
2313 * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
2314 (PTX_DEFAULT_RUNTIME_DIM): ... this.
2315 (nvptx_goacc_validate_dims): Set default worker and gang dims to
2316 PTX_DEFAULT_RUNTIME_DIM.
2317 (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
2319 2018-07-29 John David Anglin <danglin@gcc.gnu.org>
2321 * config/pa/pa.c (pa_output_addr_vec): Align address table.
2322 * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
2323 * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
2325 2018-07-27 Michael Meissner <meissner@linux.ibm.com>
2327 * config/rs6000/constraints.md (wG constraint): Delete, no longer
2329 * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
2330 predicate to reflect toc fusion has been deleted.
2331 (toc_fusion_mem_raw): Delete, no longer used.
2332 (toc_fusion_mem_wrapped): Likewise.
2333 * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
2335 * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
2336 Delete, no longer used.
2337 * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
2338 meant to be used for toc fusion.
2339 (rs6000_debug_print_mode): Delete toc fusion debugging.
2340 (rs6000_debug_reg_global): Likewise.
2341 (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
2342 fusion and secondary reload support that were never used.
2343 (rs6000_option_override_internal): Delete TOC fusion, that was only
2344 partially defined, and it did not work unless you also used the
2346 (rs6000_legitimate_address_p): Delete TOC fusion support.
2347 (rs6000_opt_masks): Likewise.
2348 (fusion_wrap_memory_address): Delete function, no longer used.
2349 (fusion_split_address); Delete TOC fusion support.
2350 * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
2351 longer used with toc fusion being deleted.
2352 (TARGET_TOC_FUSION_FP): Likewise.
2353 * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
2355 (toc fusion spliter): Delete TOC fusion support.
2356 (toc_fusionload_<mode>): Likewise.
2357 (toc_fusionload_di): Likewise.
2358 (fusion_gpr_load_<mode>): Delete generator function, this insn no
2359 longer needs to be named. Rename predicate to delete TOC fusion.
2360 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
2361 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
2362 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
2363 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
2364 (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
2366 2018-07-27 Kelvin Nilsen <kelvin@gcc.gnu.org>
2368 * doc/extend.texi (Basic PowerPC Built-in Functions Available on
2369 ISA 2.05): Replace __uint128_t with __uint128 and __int128_t with
2370 __int128 in built-in function prototypes.
2371 (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
2372 (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
2374 2018-07-27 Martin Sebor <msebor@redhat.com>
2376 PR tree-optimization/86696
2377 * tree-ssa-strlen.c (get_min_string_length): Handle all integer
2378 types, including enums.
2379 (handle_char_store): Be prepared for the above function to fail.
2381 2018-07-26 Qing Zhao <qing.zhao@oracle.com>
2383 * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
2384 when optimization level is lower than 2 or optimize for size.
2386 2018-07-26 Martin Sebor <msebor@redhat.com>
2388 PR tree-optimization/86043
2389 PR tree-optimization/86042
2390 * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
2391 (get_string_cst_length): Rename...
2392 (get_min_string_length): ...to this. Add argument.
2393 (handle_char_store): Extend to handle multi-character stores by
2395 * tree.c (initializer_zerop): Use new argument. Handle MEM_REF.
2396 * tree.h (initializer_zerop): Add argument.
2398 2018-07-26 Jakub Jelinek <jakub@redhat.com>
2401 * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
2402 declare target to variables if they have always,{to,from,tofrom} map
2405 2018-07-26 Martin Liska <mliska@suse.cz>
2408 * lto-wrapper.c: Add linker_output as prefix
2409 for ltrans_output_file.
2411 2018-07-26 Segher Boessenkool <segher@kernel.crashing.org>
2413 PR rtl-optimization/85805
2414 * combine.c (reg_nonzero_bits_for_combine): Only use the last set
2415 value for hard registers if that was written in the same mode.
2417 2018-07-26 Martin Liska <mliska@suse.cz>
2419 PR gcov-profile/86536
2420 * gcov.c (format_gcov): Use printf format %.*f directly
2421 and do not handle special values.
2423 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
2425 * common/config/arc/arc-common.c (arc_option_optimization_table):
2426 Update default optimizations for size.
2428 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
2430 * config/arc/arc.md (movsf_insn): Add short instruction selection.
2431 * config/arc/constraints.md (CfZ): New constraint.
2432 * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
2433 (subsf3_fpu): Likewise.
2434 (cmpsf_fpu): Likewise.
2435 (cmpsf_fpu_uneq): Likewise.
2437 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
2439 * config/arc/arc.c (compact_memory_operand_p): Check for uncached
2441 (arc_is_uncached_mem_p): uncached applies to both the variable and
2444 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
2446 * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
2449 2018-07-25 David Malcolm <dmalcolm@redhat.com>
2451 * optinfo-emit-json.cc (class optrecord_json_writer): Convert
2452 field "m_scopes" from vec to auto_vec.
2454 2018-07-25 Martin Liska <mliska@suse.cz>
2456 * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
2459 2018-07-25 Richard Biener <rguenther@suse.de>
2462 * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
2463 special wrt context_die late.
2464 (gen_subprogram_die): Re-use DIEs in local scope.
2466 2018-07-25 Richard Sandiford <richard.sandiford@arm.com>
2468 PR tree-optimization/86644
2469 * hwint.c (ceil_log2): Fix comment. Return 0 for 0.
2471 2018-07-25 Martin Liska <mliska@suse.cz>
2474 * dumpfile.c: And excluded values with TDF_ALL_VALUES.
2475 * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
2477 2018-07-25 Martin Liska <mliska@suse.cz>
2480 * params.def: Explain ASan abbreviation and provide
2481 a documentation link.
2483 2018-07-24 Martin Sebor <msebor@redhat.com>
2485 PR tree-optimization/86622
2486 PR tree-optimization/86532
2487 * builtins.h (string_length): Declare.
2488 * builtins.c (c_strlen): Correct handling of non-constant offsets.
2489 (check_access): Be prepared for non-constant length ranges.
2490 (string_length): Make extern.
2491 * expr.c (string_constant): Only handle the minor non-constant
2492 array index. Use string_constant to compute the length of
2493 a generic string constant.
2495 2018-07-24 Richard Sandiford <richard.sandiford@arm.com>
2497 PR tree-optimization/86618
2498 * tree-vect-stmts.c (vectorizable_call): Don't take the address
2499 of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
2501 2018-07-24 David Malcolm <dmalcolm@redhat.com>
2503 PR tree-optimization/86636
2504 * json.cc (json::object::set): Fix comment. Add assertions.
2505 (json::array::append): Move here from json.h. Add comment and an
2507 (json::string::string): Likewise.
2508 * json.h (json::array::append): Move to json.cc.
2509 (json::string::string): Likewise.
2510 * optinfo-emit-json.cc
2511 (optrecord_json_writer::impl_location_to_json): Assert that we
2512 aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
2513 wrapper around it. Expand the location once, rather than three
2515 (optrecord_json_writer::inlining_chain_to_json): Fix the check for
2516 UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
2518 (optrecord_json_writer::optinfo_to_json): Likewise, in four
2519 places. Fix some overlong lines.
2521 2018-07-24 Matthew Malcomson <matthew.malcomson@arm.com>
2523 * config/aarch64/aarch64-simd.md
2524 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
2525 (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
2526 (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
2527 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
2528 (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
2529 (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
2530 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
2531 (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
2532 (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
2534 2018-07-24 Jakub Jelinek <jakub@redhat.com>
2537 * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
2538 and size > HOST_BITS_PER_WIDE_INT. For size > HOST_BITS_PER_WIDE_INT
2539 and abs_d == d, do the power of two handling if profitable.
2541 2018-07-24 Richard Biener <rguenther@suse.de>
2543 * match.pd: Add BIT_FIELD_REF canonicalizations.
2545 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
2548 * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
2550 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
2552 * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
2553 terminated STRING_CST object.
2555 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
2557 hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
2559 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
2561 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
2562 * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
2563 * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
2564 the elements into a register.
2565 (rs6000_split_v4si_init_di_reg): Delete.
2566 (rs6000_split_v4si_init): Delete.
2567 * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
2568 (vsx_init_v4si): Rewrite as a define_expand.
2570 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
2572 * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
2573 zero_extend argument from memory): New.
2575 2018-07-22 Martin Sebor <msebor@redhat.com>
2578 * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
2579 diagnosing calls with unknown arguments unless -Walloca-larger-than
2580 is restricted to less than PTRDIFF_MAX bytes.
2582 2018-07-22 Gerald Pfeifer <gerald@pfeifer.com>
2584 * doc/gcov.texi (Invoking Gcov): Editorial changes.
2586 2018-07-20 David Malcolm <dmalcolm@redhat.com>
2588 * pretty-print.c (text_info::set_location): Remove redundant
2589 "line_table" parameter from call to rich_location::set_range.
2591 2018-07-20 Martin Sebor <msebor@redhat.com>
2594 * builtins.c (expand_builtin_alloca): Adjust.
2595 * calls.c (alloc_max_size): Simplify.
2596 * cgraphunit.c (cgraph_node::expand): Adjust.
2597 * common.opt (larger_than_size, warn_frame_larger_than): Remove
2599 (frame_larger_than_size): Same.
2600 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
2601 to take a HOST_WIDE_INT argument and accept a byte-size suffix.
2603 * doc/invoke.texi (GCC Command Options): Document option arguments.
2604 Explain byte-size arguments and suffixes.
2605 (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
2606 (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
2607 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
2608 * doc/options.texi (UInteger): Expand.
2609 (Host_Wide_Int, ByteSize): Document new properties.
2610 * final.c (final_start_function_1): Include sizes in an error message.
2611 * function.c (frame_offset_overflow): Same.
2612 * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
2613 (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
2614 Diagnose unbounded alloca calls only for limits of less than
2616 (alloca_call_type): Adjust. Diagnose possibly out-of-bounds alloca
2617 calls and VLA size only for limits of less than PTRDIFF_MAX. Same
2619 (pass_walloca::execute): Adjust. Diagnose alloca calls in loops
2620 only for limits of less than PTRDIFF_MAX.
2621 * langhooks-def.h (lhd_handle_option): Change function argument
2623 * langhooks.c (lhd_handle_option): Same.
2624 * langhooks.h (handle_option): Same.
2625 * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
2627 (var_type, var_type_struct): Same.
2628 (var_set): Handle ByteSize flag.
2629 * optc-gen.awk: Add comments to output to ease debugging. Make
2630 use of HOST_WIDE_INT where appropriate.
2631 * opts-gen-save.awk: Use %lx to format unsigned long.
2632 * opth-gen.awk: Change function argument to HOST_WIDE_INT.
2633 * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
2634 arguments. Parse bytes-size suffixes.
2635 (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
2636 (enum_value_to_arg): Same.
2637 (decode_cmdline_option): Handle cl_host_wide_int. Adjust.
2638 (handle_option): Adjust.
2639 (generate_option): Change function argument to HOST_WIDE_INT.
2640 (cmdline_handle_error): Adjust.
2641 (read_cmdline_option): Change function argument to HOST_WIDE_INT.
2642 (set_option): Change function argument to HOST_WIDE_INT.
2643 (option_enabled): Handle cl_host_wide_int.
2644 (get_option_state): Handle CLVC_SIZE.
2645 (control_warning_option): Same.
2646 * opts.c (common_handle_option): Change function argument to
2647 HOST_WIDE_INT. Remove handling of OPT_Walloca_larger_than_ and
2648 OPT_Wvla_larger_than_.
2649 * opts.h (enum cl_var_type): Add an enumerator.
2650 * stor-layout.c (layout_decl): Print a more meaningful warning.
2651 * toplev.c (output_stack_usage): Adjust.
2653 2018-07-20 Qing Zhao <qing.zhao@oracle.com>
2655 * builtins.c (expand_builtin_memcmp): Delete the last parameter for
2656 call to inline_expand_builtin_string_cmp.
2657 (expand_builtin_strcmp): Likewise.
2658 (expand_builtin_strncmp): Likewise.
2659 (inline_string_cmp): Delete the last parameter, change char_type_node
2660 to unsigned_char_type_node for strcmp/strncmp, add conversions to the
2662 (inline_expand_builtin_string_cmp): Delete the last parameter, give up
2663 the inlining expansion on target where the type of the call has same or
2664 narrower precision than unsigned char.
2666 2018-07-20 David Malcolm <dmalcolm@redhat.com>
2668 * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
2669 (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
2670 * common.opt (fsave-optimization-record): New option.
2671 * coretypes.h (struct kv_pair): Move here from dumpfile.c.
2672 * doc/invoke.texi (-fsave-optimization-record): New option.
2673 * dumpfile.c: Include "optinfo-emit-json.h".
2674 (struct kv_pair): Move to coretypes.h.
2675 (optgroup_options): Make non-static.
2676 (dump_context::end_scope): Call
2677 optimization_records_maybe_pop_dump_scope.
2678 * dumpfile.h (optgroup_options): New decl.
2679 * json.cc: New file.
2681 * optinfo-emit-json.cc: New file.
2682 * optinfo-emit-json.h: New file.
2683 * optinfo.cc: Include "optinfo-emit-json.h".
2684 (optinfo::emit): Call optimization_records_maybe_record_optinfo.
2685 (optinfo_enabled_p): Check optimization_records_enabled_p.
2686 (optinfo_wants_inlining_info_p): Likewise.
2687 * optinfo.h: Update comment.
2688 * profile-count.c (profile_quality_as_string): New function.
2689 * profile-count.h (profile_quality_as_string): New decl.
2690 (profile_count::quality): New accessor.
2691 * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
2692 and optinfo_emit_json_cc_tests.
2693 * selftest.h (selftest::json_cc_tests): New decl.
2694 (selftest::optinfo_emit_json_cc_tests): New decl.
2695 * toplev.c: Include "optinfo-emit-json.h".
2696 (compile_file): Call optimization_records_finish.
2697 (do_compile): Call optimization_records_start.
2698 * tree-ssa-live.c: Include optinfo.h.
2699 (remove_unused_scope_block_p): Retain inlining information if
2700 optinfo_wants_inlining_info_p returns true.
2702 2018-07-20 Richard Biener <rguenther@suse.de>
2705 * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
2706 to cover -flto-partition=none.
2708 2018-07-20 Martin Liska <mliska@suse.cz>
2710 * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
2711 (get_decl_source_range): Remove unused function.
2713 2018-07-20 Richard Biener <rguenther@suse.de>
2715 * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
2716 (struct vn_phi_s): Likewise.
2717 (struct vn_reference_s): Likewise.
2718 * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
2719 for searching the slot of an entry known to be in the hash itself.
2720 (vn_phi_hasher::equal): Likewise.
2721 (vn_reference_hasher::equal): Likewise.
2722 (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
2724 (optimistic_info, current_info): Remove, keeping only valid_info.
2725 (vn_reference_lookup_1): Remove fallback lookup.
2726 (vn_reference_lookup_2): Likewise.
2727 (vn_nary_op_lookup_1): Likewise.
2728 (vn_phi_lookup): Likewise.
2729 (vn_nary_build_or_lookup_1): Make sure to not chain the built
2731 (vn_reference_insert): Adjust, chain the inserted hash element
2732 at last_inserted_ref.
2733 (vn_reference_insert_pieces): Likewise.
2734 (visit_reference_op_call): Likewise.
2735 (vn_nary_op_insert_into): Chain the inserted hash element at
2737 (vn_nary_op_insert_pieces): Adjust.
2738 (vn_nary_op_insert): Likewise.
2739 (vn_nary_op_insert_stmt): Likewise.
2740 (vn_phi_insert): Adjust, chain the inserted hash element at
2742 (process_scc): Remove clearing and copying the optimistic
2743 table. Instead remove elements inserted during an optimistic
2744 iteration from the single table we maintain.
2745 (init_scc_vn): Adjust.
2746 (free_scc_vn): Likewise.
2747 (sccvn_dom_walker::record_cond): Likewise.
2748 (sccvn_dom_walker::after_dom_children): Likewise.
2750 2018-07-19 Martin Sebor <msebor@redhat.com>
2752 PR tree-optimization/84047
2753 PR tree-optimization/83776
2754 * tree-vrp.c (vrp_prop::check_mem_ref): New function.
2755 (check_array_bounds): Call it.
2757 2018-07-19 Martin Sebor <msebor@redhat.com>
2759 * align.h (align_flags): Use member initialization.
2761 2018-07-19 David Malcolm <dmalcolm@redhat.com>
2763 * Makefile.in (OBJS): Add optinfo.o.
2764 * coretypes.h (class symtab_node): New forward decl.
2765 (struct cgraph_node): New forward decl.
2766 (class varpool_node): New forward decl.
2767 * dump-context.h: New file.
2768 * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
2770 (refresh_dumps_are_enabled): Use optinfo_enabled_p.
2771 (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
2772 (set_alt_dump_file): Likewise.
2773 (dump_context::~dump_context): New dtor.
2774 (dump_gimple_stmt): Move implementation to...
2775 (dump_context::dump_gimple_stmt): ...this new member function.
2776 Add the stmt to any pending optinfo, creating one if need be.
2777 (dump_gimple_stmt_loc): Move implementation to...
2778 (dump_context::dump_gimple_stmt_loc): ...this new member function.
2779 Start a new optinfo and add the stmt to it.
2780 (dump_gimple_expr): Move implementation to...
2781 (dump_context::dump_gimple_expr): ...this new member function.
2782 Add the stmt to any pending optinfo, creating one if need be.
2783 (dump_gimple_expr_loc): Move implementation to...
2784 (dump_context::dump_gimple_expr_loc): ...this new member function.
2785 Start a new optinfo and add the stmt to it.
2786 (dump_generic_expr): Move implementation to...
2787 (dump_context::dump_generic_expr): ...this new member function.
2788 Add the tree to any pending optinfo, creating one if need be.
2789 (dump_generic_expr_loc): Move implementation to...
2790 (dump_context::dump_generic_expr_loc): ...this new member
2791 function. Add the tree to any pending optinfo, creating one if
2793 (dump_printf): Move implementation to...
2794 (dump_context::dump_printf_va): ...this new member function. Add
2795 the text to any pending optinfo, creating one if need be.
2796 (dump_printf_loc): Move implementation to...
2797 (dump_context::dump_printf_loc_va): ...this new member function.
2798 Start a new optinfo and add the stmt to it.
2799 (dump_dec): Move implementation to...
2800 (dump_context::dump_dec): ...this new member function. Add the
2801 value to any pending optinfo, creating one if need be.
2802 (dump_context::dump_symtab_node): New member function.
2803 (dump_context::get_scope_depth): New member function.
2804 (dump_context::begin_scope): New member function.
2805 (dump_context::end_scope): New member function.
2806 (dump_context::ensure_pending_optinfo): New member function.
2807 (dump_context::begin_next_optinfo): New member function.
2808 (dump_context::end_any_optinfo): New member function.
2809 (dump_context::s_current): New global.
2810 (dump_context::s_default): New global.
2811 (dump_scope_depth): Delete global.
2812 (dumpfile_ensure_any_optinfo_are_flushed): New function.
2813 (dump_symtab_node): New function.
2814 (get_dump_scope_depth): Reimplement in terms of dump_context.
2815 (dump_begin_scope): Likewise.
2816 (dump_end_scope): Likewise.
2817 (selftest::temp_dump_context::temp_dump_context): New ctor.
2818 (selftest::temp_dump_context::~temp_dump_context): New dtor.
2819 (selftest::verify_item): New function.
2820 (ASSERT_IS_TEXT): New macro.
2821 (ASSERT_IS_TREE): New macro.
2822 (ASSERT_IS_GIMPLE): New macro.
2823 (selftest::test_capture_of_dump_calls): New test.
2824 (selftest::dumpfile_c_tests): Call it.
2825 * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
2826 (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
2827 (dump_gimple_stmt, dump_dec): Gather these related decls and add a
2828 descriptive comment.
2829 (dump_function, print_combine_total_stats, enable_rtl_dump_file)
2830 (dump_node, dump_bb): Move these unrelated decls.
2831 (class dump_manager): Add leading comment.
2832 * optinfo.cc: New file.
2833 * optinfo.h: New file.
2835 2018-07-19 Michael Collison <michael.collison@arm.com>
2836 Richard Henderson <rth@redhat.com>
2838 * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
2839 (subti): Handle op1 zero.
2840 (subvti4, usub4ti4): New.
2841 (*sub<GPI>3_compare1_imm): New.
2842 (sub<GPI>3_carryinCV): New.
2843 (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
2844 (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
2846 2018-07-19 Michael Collison <michael.collison@arm.com>
2847 Richard Henderson <rth@redhat.com>
2849 * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
2850 (addti3): Create simpler code if low part is already known to be 0.
2851 (addvti4, uaddvti4): New.
2852 (*add<GPI>3_compareC_cconly_imm): New.
2853 (*add<GPI>3_compareC_cconly): New.
2854 (*add<GPI>3_compareC_imm): New.
2855 (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
2856 handle constants within this pattern..
2857 (*add<GPI>3_compareV_cconly_imm): New.
2858 (*add<GPI>3_compareV_cconly): New.
2859 (*add<GPI>3_compareV_imm): New.
2860 (add<GPI>3_compareV): New.
2861 (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
2862 (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
2863 (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
2864 ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
2865 with 'comparison' operator.
2866 (*add<GPI>3_compareV_cconly_imm): Ditto.
2867 (*add<GPI>3_compareV_cconly): Ditto.
2868 (*add<GPI>3_compareV_imm): Ditto.
2869 (add<GPI>3_compareV): Ditto.
2870 (add<mode>3_carryinC): Ditto.
2871 (*add<mode>3_carryinC_zero): Ditto.
2872 (*add<mode>3_carryinC): Ditto.
2873 (add<mode>3_carryinV): Ditto.
2874 (*add<mode>3_carryinV_zero): Ditto.
2875 (*add<mode>3_carryinV): Ditto.
2877 2018-07-19 Michael Collison <michael.collison@arm.com>
2878 Richard Henderson <rth@redhat.com>
2880 * config/aarch64/aarch64-modes.def (CC_V): New.
2881 * config/aarch64/aarch64-protos.h
2882 (aarch64_addti_scratch_regs): Declare
2883 (aarch64_subvti_scratch_regs): Declare.
2884 (aarch64_expand_subvti): Declare.
2885 (aarch64_gen_unlikely_cbranch): Declare
2886 * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
2887 for signed overflow using CC_Vmode.
2888 (aarch64_get_condition_code_1): Handle CC_Vmode.
2889 (aarch64_gen_unlikely_cbranch): New function.
2890 (aarch64_addti_scratch_regs): New function.
2891 (aarch64_subvti_scratch_regs): New function.
2892 (aarch64_expand_subvti): New function.
2894 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
2896 * config/aarch64/aarch64-option-extensions.def: New entry for profile
2898 * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
2899 * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
2902 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
2905 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
2906 address check not strict.
2908 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
2910 * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
2912 (store_pair_lanes<mode>): Likewise.
2913 * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
2914 enum value 'ADDR_QUERY_LDP_STP_N'.
2915 * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
2916 (aarch64_print_address_internal): Add declaration.
2917 (aarch64_print_ldpstp_address): Remove.
2918 (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
2919 (aarch64_print_operand): Change printing of 'y'.
2920 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
2921 new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
2922 'true' rather than '1'.
2923 * config/aarch64/constraints.md (Uml): Likewise.
2924 (Uml): Rename to Umn.
2927 2018-07-19 Richard Biener <rguenther@suse.de>
2929 * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
2931 * tree-ssa-sccvn.c: Remove alloc-pool.h use.
2932 (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
2933 (vn_reference_hasher): Likewise.
2934 (struct vn_tables_s): Remove obstack and alloc-pool members.
2935 (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
2936 (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
2937 (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
2938 (vn_reference_insert_pieces): Likewise.
2939 (alloc_vn_nary_op_noinit): Adjust.
2940 (vn_nary_op_insert_stmt): Allocate phiargs in-place.
2941 (vn_phi_eq): Adjust.
2942 (shared_lookup_phiargs): Remove.
2943 (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
2944 (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
2945 (visit_reference_op_call): Likewise.
2946 (copy_nary, copy_phi, copy_reference): Remove.
2947 (process_scc): Rewind the obstack when iterating. Do not
2948 copy the elements to valid_info but just move them from one
2949 hashtable to the other.
2950 (allocate_vn_table): Adjust.
2951 (free_vn_table): Likewise.
2952 (init_scc_vn): Likewise.
2953 (free_scc_vn): Likewise.
2955 2018-07-19 H.J. Lu <hongjiu.lu@intel.com>
2958 * config/i386/i386.c (rest_of_insert_endbranch): Lookup
2959 indirect_return as function type attribute.
2960 (ix86_attribute_table): Change indirect_return to function
2962 * doc/extend.texi: Update indirect_return attribute.
2964 2018-07-19 Aldy Hernandez <aldyh@redhat.com>
2966 * wide-int.h (widest2_int): New.
2967 * gimple-fold.c (arith_overflowed_p): Use it.
2968 * tree.h (widest2_int_cst): New.
2969 * tree-vrp.c (wide_int_binop_overflow): Rename from
2970 vrp_int_const_binop.
2971 Rewrite to work on trees.
2972 (extract_range_from_multiplicative_op_1): Abstract code to...
2973 (wide_int_range_min_max): ...here.
2974 (wide_int_range_cross_product): ...and here.
2975 (extract_range_from_binary_expr_1): Abstract overflow code to...
2976 (wide_int_range_mult_wrapping): ...here.
2977 * tree-vrp.h (wide_int_range_cross_product): New.
2978 (wide_int_range_mult_wrapping): New.
2980 2018-07-19 Andrew Senkevich <andrew.senkevich@intel.com>
2981 Julia Koval <julia.koval@intel.com>
2983 * config/i386/x86-tune-costs.h (skylake_memcpy,
2984 skylake_memset): Replace rep_prefix with unrolling for size 512.
2986 2018-07-18 Kugan Vivekanandarajah <kuganv@linaro.org>
2989 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle comparision with EQ_EXPR
2992 2018-07-18 Kelvin Nilsen <kelvin@gcc.gnu.org>
2994 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
2995 this subsection to "PowerPC AltiVec/VSX Built-in Functions".
2996 (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
2997 previously known as "PowerPC AltiVec Built-in Functions". Move
2998 some material to new subsubsections "PowerPC AltiVec Built-in
2999 Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
3001 (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
3002 (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
3003 (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
3004 (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
3006 2018-07-18 Richard Biener <rguenther@suse.de>
3008 PR tree-optimization/86557
3009 * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
3012 2018-07-18 Ilya Leoshkevich <iii@linux.ibm.com>
3014 * config/s390/s390.c (s390_function_profiler): Generate CFI.
3016 2018-07-17 Jeff Law <law@redhat.com>
3018 * config/arm/arm.c (get_label_padding): Update for recent
3019 changes to label_to_alignment.
3021 PR tree-optimization/86010
3022 * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
3024 * config/mips/mips.c (vr4130_align_insns): Update for recent
3025 changes to label_to_alignment.
3027 * config/frv/frv.c (frv_label_align): Update for recent changes
3028 to label_to_alignment.
3030 * config/nios2/nios2.c (nios2_label_align): Update for recent
3031 changes which dropped ALIGN_LABELS_LOG.
3033 2018-07-17 Andreas Schwab <schwab@linux-m68k.org>
3035 * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
3036 (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
3038 2018-07-17 Claudiu Zissulescu <claziss@synopsys.com>
3040 * config/arc/arc.c (arc_label_align): Use align_labels instead of
3041 deprecated align_labels_log.
3043 2018-07-17 Richard Biener <rguenther@suse.de>
3046 * dwarf2out.c (init_sections_and_labels): Always generate
3047 a debug_line_str_section for early LTO debug.
3048 (dwarf2out_finish): Reset debug_line_str_hash output early.
3049 Bump counter for extra dwarf5 .debug_loc labels to not conflict
3051 (dwarf2out_early_finish): Output debug_line_str.
3053 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
3055 * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
3056 index register on z196 or later.
3058 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
3060 * config/s390/s390.c (s390_default_align): Set default function
3062 (s390_override_options_after_change): Call s390_default align.
3063 (s390_option_override_internal): Call s390_default_align.
3064 (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
3066 2018-07-17 Jakub Jelinek <jakub@redhat.com>
3069 * omp-low.c (create_task_copyfn): Copy over also fields corresponding
3070 to _looptemp_ clauses, other than the first two.
3072 2018-07-17 Martin Liska <mliska@suse.cz>
3074 * opts.c: Do not enable OPT_falign_* for -Os.
3076 2018-07-17 Martin Liska <mliska@suse.cz>
3078 * align.h (MAX_CODE_ALIGN): New.
3079 (MAX_CODE_ALIGN_VALUE): New.
3080 * common/config/i386/i386-common.c (ix86_handle_option):
3081 (MAX_CODE_ALIGN): Moved to align.h.
3082 * final.c (MAX_CODE_ALIGN): Likewise.
3083 * opts.c (parse_and_check_align_values):
3084 (MAX_CODE_ALIGN): Likewise.
3085 (MAX_CODE_ALIGN_VALUE): Likewise.
3087 2018-07-17 Martin Liska <mliska@suse.cz>
3089 * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
3090 in order to fulfil coding style.
3091 * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
3092 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
3093 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
3094 * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
3095 * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
3096 * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
3097 * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
3098 * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
3099 (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
3101 2018-07-17 Martin Liska <mliska@suse.cz>
3103 * align.h: New file.
3104 * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions directly.
3105 * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
3106 align_flags of label_to_alignment.
3107 * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into align_flags
3109 * config/m68k/m68k.c: Do not use removed align_labels_value and
3111 * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
3112 (LOOP_ALIGN): Likewise.
3113 (LABEL_ALIGN): Likewise.
3114 * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
3115 Remove not used macro.
3116 (rs6000_loop_align): Change return type to align_flags.
3117 (rs6000_loop_align_max_skip): Remove.
3118 * config/rs6000/rs6000-protos.h (rs6000_loop_align):
3119 Change return type to align_flags.
3120 * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
3121 Remove not used macro.
3122 (rs6000_loop_align): Change return type to align_flags.
3123 (rs6000_loop_align_max_skip): Remove.
3124 * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
3125 * config/rx/rx-protos.h (rx_align_for_label): Make it
3127 * config/rx/rx.c (rx_align_for_label): Change return type
3129 (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
3131 into align_flags class.
3132 (LABEL_ALIGN): Likewise.
3133 (LOOP_ALIGN): Likewise.
3134 * config/s390/s390.c (s390_label_align): Use align_flags
3136 (s390_asm_output_function_label): Likewise.
3137 * config/sh/sh.c (sh_override_options_after_change):
3138 Use align_flags class directly without macros.
3139 (find_barrier): Likewise.
3140 (barrier_align): Likewise.
3141 (sh_loop_align): Likewise.
3142 * config/spu/spu.c (spu_option_override):
3143 Use align_flags_tuple::get_value instead of removed macros.
3144 (spu_sched_init): Likewise.
3145 * config/spu/spu.h (GTY): Likewise.
3146 * config/visium/visium.c (visium_option_override):
3147 Set "8" as default secondary alignment.
3148 * config/visium/visium.h (SUBALIGN_LOG): Define to 3
3149 in order to guarantee secondary alignment of 8.
3150 * coretypes.h: Include align.h header file.
3151 * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
3152 TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
3153 and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
3154 * doc/tm.texi.in: Likewise.
3155 * final.c (struct label_alignment): Remove not used structure.
3156 (LABEL_ALIGN): Change type to align_flags.
3157 (LOOP_ALIGN): Likewise.
3158 (JUMP_ALIGN): Likewise.
3159 (default_loop_align_max_skip): Remove.
3160 (default_label_align_max_skip): Likewise.
3161 (default_jump_align_max_skip): Likewise.
3162 (default_label_align_after_barrier_max_skip):
3163 (LABEL_TO_ALIGNMENT): Change to access label_align vector.
3164 (LABEL_TO_MAX_SKIP): Remove.
3165 (label_to_alignment): Return align_flags type instead of integer.
3166 (label_to_max_skip): Remove.
3167 (align_fuzz): Use align_flags type.
3168 (compute_alignments): Use align_flags type and use align_flags::max
3169 to combine multiple alignments.
3170 (grow_label_align): Grow vec instead of C array.
3171 (update_alignments): Assign just LABEL_TO_ALIGNMENT.
3172 (shorten_branches): Use align_flags type and use align_flags::max
3173 to combine multiple alignments.
3174 (final_scan_insn_1): Remove usage of secondary alignment that comes
3175 from label alignment, but instead use proper secondary alignment
3176 which is computed in grow_label_align.
3177 * flags.h (struct align_flags_tuple): Move to align.h.
3178 (struct align_flags): Likewise.
3179 (state_align_loops): Rename to align_loops.
3180 (state_align_jumps): Rename to align_jumps.
3181 (state_align_labels): Rename to align_labels.
3182 (state_align_functions): Rename to align_functions.
3183 (align_loops_log): Remove.
3184 (align_jumps_log): Remove.
3185 (align_labels_log): Remove.
3186 (align_functions_log): Remove.
3187 (align_loops_max_skip): Remove.
3188 (align_jumps_max_skip): Remove.
3189 (align_labels_max_skip): Remove.
3190 (align_functions_max_skip): Remove.
3191 (align_loops_value): Remove.
3192 (align_jumps_value): Remove.
3193 (align_labels_value): Remove.
3194 (align_functions_value): Remove.
3195 * output.h (label_to_alignment): Change return type to align_flags.
3196 (label_to_max_skip): Remove.
3197 * target.def: Remove loop_align_max_skip, label_align_max_skip,
3198 jump_align_max_skip macros.
3199 * targhooks.h (default_loop_align_max_skip): Remove.
3200 (default_label_align_max_skip): Likewise.
3201 (default_jump_align_max_skip): Likewise.
3202 (default_label_align_after_barrier_max_skip): Remove.
3203 * toplev.c (read_log_maxskip): Use ::normalize function.
3204 (parse_N_M): Remove not used argument and also call ::normalize.
3205 (parse_alignment_opts): Do not pass unused arguments.
3206 * varasm.c (assemble_start_function): Use directly align_functions
3207 instead of removed macros.
3208 * system.h: Do not poison removed macros.
3210 2018-07-17 Jakub Jelinek <jakub@redhat.com>
3213 * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
3214 and cond temporaries don't have reference type if iterator has
3215 pointer type. For init use &for_pre_body instead of pre_p if
3216 for_pre_body is non-empty.
3218 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
3220 * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
3221 double-double modes to SFmode directly directly.
3222 (trunc<mode>sf2_fprs): Delete.
3224 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
3226 * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
3227 for conversions between IFmode and the decimal floating point modes.
3228 (init_float128_ieee): Use the correct names for conversions between
3229 KFmode and the decimal floating point modes.
3231 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
3233 * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
3234 for the conversions between TDmode and IFmode.
3235 (init_float128_ieee): Use more correct names for the conversions
3236 between TDmode and KFmode.
3238 2018-07-16 Jakub Jelinek <jakub@redhat.com>
3240 PR tree-optimization/86526
3241 * builtins.c (expand_builtin_memcmp): Formatting fixes.
3242 (inline_expand_builtin_string_cmp): Likewise.
3243 (inline_string_cmp): Likewise. Use c_readstr instead of
3244 builtin_memcpy_read_str. Add unit_mode temporary.
3246 2018-07-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
3249 * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
3250 * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
3252 2018-07-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
3254 * doc/extend.texi (PowerPC AltiVec Built-in Functions):
3255 Alphabetize prototypes of built-in functions, separating out
3256 built-in functions that are listed in this section but should be
3257 described elsewhere.
3259 2018-07-16 Uros Bizjak <ubizjak@gmail.com>
3262 * expmed.c (emit_store_flag): Do not emit setcc followed by a
3263 conditional move when trapping comparison was split to a
3264 non-trapping one (and vice versa).
3266 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
3268 * config/s390/s390.c (s390_function_profiler): Generate nops
3269 instead of profiler call sequences.
3270 * config/s390/s390.opt: Add the new option.
3272 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
3274 * config/s390/s390.c (s390_function_profiler): Generate
3275 __mcount_loc section.
3276 * config/s390/s390.opt: Add the new option.
3278 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
3280 * common.opt: Add the new warning.
3281 * config/s390/s390.c (s390_function_profiler): Emit "brasl
3282 %r0,__fentry__" when -mfentry is specified.
3283 (s390_option_override_internal): Disallow -mfentry for 31-bit
3285 * config/s390/s390.opt: Add the new option.
3287 2018-07-16 Richard Biener <rguenther@suse.de>
3290 * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
3291 for function-local FUNCTION_DECL and RESULT_DECL immediately.
3293 2018-07-16 Martin Liska <mliska@suse.cz>
3296 * ipa-pure-const.c (malloc_candidate_p): Revert ::get
3299 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
3301 * config/arc/arcHS.md: Update ARCHS scheduling rules.
3303 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
3305 * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
3307 * config/arc/arc-cpus.def (hs4x): New cpu.
3309 * config/arc/arc-tables.opt: Regenerate.
3310 * config/arc/arc.c (arc_sched_issue_rate): New function.
3311 (TARGET_SCHED_ISSUE_RATE): Define.
3312 (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
3313 * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
3315 (attr tune): Add ARCHS4x tune values.
3316 (attr tune_dspmpy): Define.
3317 (*tst): Correct instruction type.
3318 * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
3319 * config/arc/arcHS4x.md: New file.
3320 * config/arc/fpu.md: Update instruction type attributes.
3321 * config/arc/t-multilib: Regenerate.
3323 2018-07-16 Tom de Vries <tdevries@suse.de>
3326 * var-tracking.c (vt_initialize): Fix pre_dec handling.
3328 2018-07-16 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3330 * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
3333 2018-07-16 Eric Botcazou <ebotcazou@adacore.com>
3335 PR tree-optimization/86514
3336 * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
3337 conversion to a boolean type from a type with greater precision.
3339 2018-07-16 Tom de Vries <tdevries@suse.de>
3341 * var-tracking.c (vt_initialize): Print adjusted insn slim if
3342 dump_flags request TDF_SLIM.
3344 2018-07-16 Aldy Hernandez <aldyh@redhat.com>
3346 * fold-const.c (int_const_binop_1): Abstract...
3347 (wide_int_binop): ...wide int code here.
3348 (poly_int_binop): ...poly int code here.
3349 Abstract the rest of int_const_binop_1 into int_const_binop.
3350 * fold-const.h (wide_int_binop): New.
3351 * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
3352 Remove useless PLUS/MINUS_EXPR case.
3353 (zero_nonzero_bits_from_vr): Move wide int code...
3354 (zero_nonzero_bits_from_bounds): ...here.
3355 (extract_range_from_binary_expr_1): Move mask optimization code...
3356 (range_easy_mask_min_max): ...here.
3357 * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
3358 (range_easy_mask_min_max): New.
3360 2018-07-15 Jeff Law <law@redhat.com>
3363 * config/sh/sh.c (output_mi_thunk): Remove dead conditional
3366 2018-07-14 Jim Wilson <jimw@sifive.com>
3368 * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
3370 2018-07-14 Paul Koning <ni1d@arrl.net>
3372 * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
3374 2018-07-13 Jan Hubicka <hubicka@ucw.cz>
3376 * lto-streamer-out.c (copy_function_or_variable): Dump info about
3379 2018-07-13 Bill Schmidt <wschmidt@linux.ibm.com>
3380 Steve Munroe <munroesj52@gmail.com>
3382 * config/rs6000/emmintrin.h (_mm_and_si128): New function.
3383 (_mm_andnot_si128): Likewise.
3384 (_mm_or_si128): Likewise.
3385 (_mm_xor_si128): Likewise.
3387 2018-07-13 Qing Zhao <qing.zhao@oracle.com>
3390 * builtins.c (expand_builtin_memcmp): Inline the calls first
3391 when result_eq is false.
3392 (expand_builtin_strcmp): Inline the calls first.
3393 (expand_builtin_strncmp): Likewise.
3394 (inline_string_cmp): New routine. Expand a string compare
3395 call by using a sequence of char comparison.
3396 (inline_expand_builtin_string_cmp): New routine. Inline expansion
3397 a call to str(n)cmp/memcmp.
3398 * doc/invoke.texi (--param builtin-string-cmp-inline-length):
3400 * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
3402 2018-07-13 Richard Earnshaw <rearnsha@arm.com>
3404 * config/arm/driver-arm.c: Include arm-native.h.
3405 (host_detect_local_cpu): Use auto-generated data tables.
3406 (vendors, arm_cpu_table): Delete. Move part information to ...
3407 * config/arm/arm-cpus.in: ... here.
3408 * config/arm/parsecpu.awk (gen_native): New function.
3409 (vendor, part): New CPU fields.
3410 (END): Add support for building the native CPU detection tables.
3411 * config/arm/t-arm (arm-native.h): Add build rule.
3412 (driver-arm.o): Add dependency on arm-native.h.
3414 2018-07-13 Richard Biener <rguenther@suse.de>
3417 * match.pd (addr1 - addr2): Allow either of the operand to
3420 2018-07-13 Tom de Vries <tdevries@suse.de>
3422 * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
3425 2018-07-13 Jackson Woodruff <jackson.woodruff@arm.com>
3427 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
3428 arrays instead of numbered variables.
3430 2018-07-13 Eric Botcazou <ebotcazou@adacore.com>
3432 * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
3433 * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
3435 2018-07-13 Richard Biener <rguenther@suse.de>
3438 * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
3439 instead of get_context_die.
3441 2018-07-13 Kugan Vivekanandarajah <kuganv@linaro.org>
3442 Richard Biener <rguenther@suse.de>
3445 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
3446 that the loop latch destination where phi is defined.
3448 2018-07-12 Kito Cheng <kito.cheng@gmail.com>
3450 * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
3451 (riscv_expand_epilogue): Add assertion to check interrupt mode.
3452 (riscv_set_current_function): Extract getting interrupt type to new
3454 (riscv_get_interrupt_type): New function.
3455 (riscv_merge_decl_attributes): New function, checking interrupt type is
3457 (TARGET_MERGE_DECL_ATTRIBUTES): Define.
3459 2018-07-12 Paul Koning <ni1d@arrl.net>
3461 * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
3464 2018-07-12 Paul Koning <ni1d@arrl.net>
3466 * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
3467 zero reference, add doloop_end instead.
3468 * doc/md.texi (decrement_and_branch_until_zero): Remove.
3469 (Looping patterns): Remove decrement_and_branch_until_zero. Add
3470 detail for doloop_end.
3472 2018-07-12 Martin Sebor <msebor@redhat.com>
3475 * attribs.c (decl_attributes): Reject conflicting attributes before
3476 calling attribute handlers.
3478 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
3480 * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
3482 (gcc::dump_manager::get_dump_file_name): likewise.
3483 (dump_begin): Likewise.
3484 * dumpfile.h (dump_begin): Update prototype.
3485 (gcc::dump_manager::get_dump_file_name,
3486 gcc::dump_manager::get_dump_file_name): Update prototype.
3488 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
3490 * internal-fn.h (vectorizable_internal_fn_p): New function.
3491 * tree-vect-slp.c (compatible_calls_p): Likewise.
3492 (vect_build_slp_tree_1): Remove nops argument. Handle calls
3493 to internal functions.
3494 (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
3496 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
3498 * fold-const.h (inverse_conditions_p): Declare.
3499 * fold-const.c (inverse_conditions_p): New function.
3500 * match.pd: Use inverse_conditions_p. Add folds of view_converts
3501 that test the inverse condition of a conditional internal function.
3502 * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
3503 * internal-fn.c (internal_fn_mask_index): Handle conditional
3505 (vectorized_internal_fn_supported_p): New function.
3506 * tree-if-conv.c: Include internal-fn.h and fold-const.h.
3507 (any_pred_load_store): Replace with...
3508 (need_to_predicate): ...this new variable.
3509 (redundant_ssa_names): New variable.
3510 (ifcvt_can_use_mask_load_store): Move initial checks to...
3511 (ifcvt_can_predicate): ...this new function. Handle tree codes
3512 for which a conditional internal function exists.
3513 (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
3514 instead of ifcvt_can_use_mask_load_store. Update after variable
3516 (predicate_load_or_store): New function, split out from
3517 predicate_mem_writes.
3518 (check_redundant_cond_expr): New function.
3519 (value_available_p): Likewise.
3520 (predicate_rhs_code): Likewise.
3521 (predicate_mem_writes): Rename to...
3522 (predicate_statements): ...this. Use predicate_load_or_store
3523 and predicate_rhs_code.
3524 (combine_blocks, tree_if_conversion): Update after above name changes.
3525 (ifcvt_local_dce): Handle redundant_ssa_names.
3526 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
3527 general conditional functions.
3528 * tree-vect-stmts.c (vectorizable_call): Likewise.
3530 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
3531 Alan Hayward <alan.hayward@arm.com>
3532 David Sherwood <david.sherwood@arm.com>
3534 * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
3535 * internal-fn.c (can_interpret_as_conditional_op_p): New function.
3536 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
3537 plus and minus and convert them into IFN_COND_FMA-based sequences.
3538 (convert_mult_to_fma): Handle conditional plus and minus.
3540 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
3542 * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
3543 * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
3544 (cond_fnms_optab): New optabs.
3545 * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
3547 (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
3548 * internal-fn.h (get_conditional_internal_fn): Declare.
3549 (get_unconditional_internal_fn): Likewise.
3550 * internal-fn.c (cond_ternary_direct): New macro.
3551 (expand_cond_ternary_optab_fn): Likewise.
3552 (direct_cond_ternary_optab_supported_p): Likewise.
3553 (FOR_EACH_COND_FN_PAIR): Likewise.
3554 (get_conditional_internal_fn): New function.
3555 (get_unconditional_internal_fn): Likewise.
3556 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
3557 (gimple_match_op::gimple_match_op): Add a new overload for 5
3559 (gimple_match_op::set_op): Likewise.
3560 (gimple_resimplify5): Declare.
3561 * genmatch.c (decision_tree::gen): Generate simplifications for
3563 * gimple-match-head.c (gimple_simplify): Define an overload for
3564 5 operands. Handle calls with 5 arguments in the top-level overload.
3565 (convert_conditional_op): Handle conversions from unconditional
3566 internal functions to conditional ones.
3567 (gimple_resimplify5): New function.
3568 (build_call_internal): Pass a fifth operand.
3569 (maybe_push_res_to_seq): Likewise.
3570 (try_conditional_simplification): Try converting conditional
3571 internal functions to unconditional internal functions.
3572 Handle 3-operand unconditional forms.
3573 * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
3574 Define ternary equivalents of the current rules for binary conditional
3576 * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
3578 * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
3579 (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
3580 (optab): Handle them.
3581 (SVE_COND_FP_TERNARY): New int iterator.
3582 (sve_fmla_op, sve_fmad_op): New int attributes.
3583 * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
3584 (*cond_<optab><mode>_2, *cond_<optab><mode_4)
3585 (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
3587 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
3589 * target.def (preferred_else_value): New target hook.
3590 * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
3591 * doc/tm.texi: Regenerate.
3592 * targhooks.h (default_preferred_else_value): Declare.
3593 * targhooks.c (default_preferred_else_value): New function.
3594 * internal-fn.h (conditional_internal_fn_code): Declare.
3595 * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
3596 (get_conditional_internal_fn): Use it.
3597 (conditional_internal_fn_code): New function.
3598 * gimple-match.h (gimple_match_cond): New struct.
3599 (gimple_match_op): Add a cond member function.
3600 (gimple_match_op::gimple_match_op): Update all forms to take a
3602 * genmatch.c (expr::gen_transform): Use the same condition as res_op
3603 for the suboperation, but don't specify a particular else_value.
3604 * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
3605 (visit_nary_op, visit_reference_op_load): Pass
3606 gimple_match_cond::UNCOND to the gimple_match_op constructor.
3607 * gimple-match-head.c: Include tree-eh.h
3608 (convert_conditional_op): New function.
3609 (maybe_resimplify_conditional_op): Likewise.
3610 (gimple_resimplify1): Call maybe_resimplify_conditional_op.
3611 (gimple_resimplify2): Likewise.
3612 (gimple_resimplify3): Likewise.
3613 (gimple_resimplify4): Likewise.
3614 (maybe_push_res_to_seq): Return null for conditional operations.
3615 (try_conditional_simplification): New function.
3616 (gimple_simplify): Call it. Pass conditions to the gimple_match_op
3618 * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
3620 * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
3622 (TARGET_PREFERRED_ELSE_VALUE): Redefine.
3624 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
3626 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
3628 (hash_tree): Do not hash DECL_FCONTEXT
3629 * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
3630 Do not stream DECL_FCONTEXT.
3631 * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
3632 * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
3634 2018-07-12 Richard Biener <rguenther@suse.de>
3637 * dwarf2out.c (gen_block_die): Only output blocks when they have
3638 at least one !DECL_IGNORED_P variable.
3640 2018-07-12 Richard Biener <rguenther@suse.de>
3643 * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
3644 Remove -mieee-fp handling.
3646 2018-07-12 Richard Biener <rguenther@suse.de>
3648 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
3649 left-over from last patch.
3651 2018-07-12 Jakub Jelinek <jakub@redhat.com>
3653 PR tree-optimization/86492
3654 * gimple-ssa-store-merging.c
3655 (imm_store_chain_info::coalesce_immediate_stores): Call
3656 check_no_overlap even for the merge_overlapping case. Formatting fix.
3658 2018-07-12 Richard Biener <rguenther@suse.de>
3661 * fold-const.c (fold_binary_op_with_conditional_arg): Do not
3662 move possibly trapping operations into the conditional.
3664 2018-07-12 Richard Biener <rguenther@suse.de>
3666 * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
3667 (vn_lookup_simplify_result): Remove recursion limit applied
3669 (vn_nary_build_or_lookup_1): Adjust.
3670 (try_to_simplify): Likewise.
3671 * gimple-match-head.c (gimple_resimplify1): Instead apply one
3673 (gimple_resimplify2): Likewise.
3674 (gimple_resimplify3): Likewise.
3675 (gimple_resimplify4): Likewise.
3677 2018-07-11 Jakub Jelinek <jakub@redhat.com>
3679 * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
3680 Use __mmask64 type instead of __mmask8 for __M argument.
3681 * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
3682 _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
3684 (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
3685 __mmask16 for __M argument.
3686 (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
3687 _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
3688 to __mmask16 instead of __mmask8.
3689 * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
3690 _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
3691 _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
3692 _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
3693 instead of __mmask16 for __U argument.
3694 * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
3695 __mmask16 instead of __mmask8 for __U argument.
3696 (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
3698 (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
3700 (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
3702 (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
3704 (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
3706 (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
3708 (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
3710 (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
3712 (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
3714 (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
3716 (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
3717 _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
3718 return type as well as __M argument type and all casts from __mmask8
3720 (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
3721 _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
3722 return type as well as __M argument type and all casts from __mmask8
3724 (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
3725 _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
3726 return type as well as __M argument type and all casts from __mmask8
3728 (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
3729 _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
3730 return type as well as __M argument type and all casts from __mmask8
3732 * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
3733 _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
3736 2018-07-11 Grazvydas Ignotas <notasas@gmail.com>
3738 * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
3739 _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
3742 2018-07-11 Paul Koning <ni1d@arrl.net>
3744 * doc/md.texi (define_subst): Document how multiple occurrences of
3745 the same argument in the replacement pattern are handled.
3747 2018-07-11 Paul Koning <ni1d@arrl.net>
3749 * doc/extend.texi (Common Variable Attributes): Move "mode" into
3751 (Common Type Attributes): Add "mode" attribute.
3753 2018-07-11 Jan Hubicka <hubicka@ucw.cz>
3755 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
3756 stream DECL_ORIGINAL_TYPE.
3757 (DFS::DFS_write_tree_body): Drop hack handling local external decls.
3758 (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
3759 * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
3760 Do not walk original type.
3761 * tree-streamer-out.c (streamer_write_chain): Drop hack handling
3763 (write_ts_decl_non_common_tree_pointers): Do not stream
3765 * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
3766 (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
3768 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
3770 * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
3771 thread twice from the same starting edge.
3773 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
3775 * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
3776 * gimple.c (gimple_call_nonnull_result_p): ...here...
3777 (gimple_call_nonnull_arg): ...and here.
3778 * gimple.h (gimple_call_nonnull_result_p): New.
3779 (gimple_call_nonnull_arg): New.
3781 2018-07-11 Richard Earnshaw <rearnsha@arm.com>
3783 * config/arm/arm-cpus.in: Move information from fpu field of each
3784 cpu definition to the isa field.
3785 * config/arm/parsecpu.awk (fpu): Delete match rule.
3786 (gen_comm_data): Don't add bits from the CPU's FPU entry.
3788 2018-07-11 Richard Biener <rguenther@suse.de>
3791 * dwarf2out.c (init_sections_and_labels): Use
3792 output_asm_line_debug_info consistently.
3793 (dwarf2out_early_finish): Likewise.
3794 (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
3797 2018-07-11 Richard Biener <rguenther@suse.de>
3799 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
3800 Rework father_bb setting in a way to avoid propagating constants
3801 multiple times on a loop body.
3803 2018-07-10 Mark Wielaard <mark@klomp.org>
3806 * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
3808 2018-07-10 Richard Biener <rguenther@suse.de>
3810 * hash-map.h (hash_map::iterator::operator*): Return
3811 references to key and value.
3813 2018-07-10 Jakub Jelinek <jakub@redhat.com>
3816 * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
3817 to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
3818 BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
3819 (gimplify_omp_for): For composite loops, move outer
3820 OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
3821 OMP_FOR/OMP_SIMD if there are any non-trivial wrappers. For class
3822 iterators add any needed clauses. Allow OMP_FOR_ORIG_DECLS to contain
3823 TREE_LIST for both the original class iterator and the "last" helper
3824 var. Gimplify OMP_FOR_PRE_BODY before the outermost composite
3825 loop, remember has_decl_expr from outer composite loops for the
3826 innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
3828 2018-07-09 Martin Sebor <msebor@redhat.com>
3832 * builtins.c (c_strlen): Avoid out-of-bounds warnings when
3833 accessing implicitly initialized array elements.
3834 * expr.c (string_constant): Handle string initializers of
3835 character arrays within aggregates.
3836 * gimple-fold.c (fold_array_ctor_reference): Add argument.
3837 Store element offset. As a special case, handle zero size.
3838 (fold_nonarray_ctor_reference): Same.
3839 (fold_ctor_reference): Add argument. Store subobject offset.
3840 * gimple-fold.h (fold_ctor_reference): Add argument.
3842 2018-07-09 Paul Koning <ni1d@arrl.net>
3844 * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
3845 (pdp11_insn_cost): New function.
3846 (pdp11_md_asm_adjust): New function.
3847 (TARGET_INVALID_WITHIN_DOLOOP): Define.
3848 (pdp11_rtx_costs): Update to match machine better.
3849 (output_addr_const_pdp11): Correct format mismatch warnings.
3850 * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
3851 * config/pdp11/pdp11.md: General change to add base_cost and/or
3852 length attributes for use by new pdp11_insn_cost function.
3853 (MIN_BRANCH): Correct definition.
3855 (doloop_end): Use standard pattern name for looping pattern.
3856 (doloop_end_nocc): New.
3857 (movsf): Add another constraint alternative.
3858 (zero_extendqihi2): Add constraint alternatives for not in place
3860 (zero_extendhisi2): Remove.
3861 (shift patterns): Add CC handling variants.
3865 (define_peephole2): New peephole to recognize mov that sets CC for
3868 2018-07-09 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3870 * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
3871 SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
3874 2018-07-09 Jakub Jelinek <jakub@redhat.com>
3877 * real.c (real_nextafter): Return true if result is denormal.
3879 2018-07-09 Martin Liska <mliska@suse.cz>
3881 * common.opt: Add back wrongly removed attribute.
3883 2018-07-09 Richard Biener <rguenther@suse.de>
3886 * dwarf2out.c (gen_block_die): For an early generated DIE
3887 always output high/low PC attributes.
3889 2018-07-09 Tom de Vries <tdevries@suse.de>
3891 * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
3892 * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
3893 onto VAR_DECL with abstract origin.
3895 2018-07-07 Jim Wilson <jimw@sifive.com>
3897 * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
3899 2018-07-07 Tom de Vries <tdevries@suse.de>
3901 * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
3903 2018-07-07 Aldy Hernandez <aldyh@redhat.com>
3905 * tree-vrp.c (vrp_int_const_binop): Change overflow type to
3907 (combine_bound): Use wide-int overflow calculation instead of
3909 * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
3911 * fold-const.c (int_const_binop_2): Same.
3912 (extract_muldiv_1): Same.
3913 (fold_div_compare): Same.
3914 (fold_abs_const): Same.
3916 * poly-int.h (add): Same.
3920 * predict.c (predict_iv_comparison): Same.
3921 * profile-count.c (slow_safe_scale_64bit): Same.
3922 * simplify-rtx.c (simplify_const_binary_operation): Same.
3923 * tree-chrec.c (tree_fold_binomial): Same.
3924 * tree-data-ref.c (split_constant_offset_1): Same.
3925 * tree-if-conv.c (idx_within_array_bound): Same.
3926 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
3927 * tree-ssa-phiopt.c (minmax_replacement): Same.
3928 * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
3929 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
3930 * vr-values.c (vr_values::adjust_range_with_scev): Same.
3931 * wide-int.cc (wi::add_large): Same.
3932 (wi::mul_internal): Same.
3933 (wi::sub_large): Same.
3934 (wi::divmod_internal): Same.
3935 * wide-int.h: Change overflow type to overflow_type for neg, add,
3936 mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
3937 mod_trunc, mod_ceil, mod_round, add_large, sub_large,
3938 mul_internal, divmod_internal.
3939 (overflow_type): New enum.
3940 (accumulate_overflow): New.
3942 2018-07-06 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
3944 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
3945 (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
3947 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
3949 * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
3950 argument is checked for zero before entering loop, avoid checking again.
3952 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
3954 * gimplify.h (generic_expr_could_trap_p): Set as global function.
3955 * gimplify.h (generic_expr_could_trap_p): Likwise.
3956 * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
3958 2018-07-06 Jakub Jelinek <jakub@redhat.com>
3960 PR tree-optimization/86401
3961 * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
3962 ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
3963 (fold_bit_and_mask): ... here. New helper function for match.pd.
3964 * fold-const.h (fold_bit_and_mask): Declare.
3965 * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
3967 2018-07-06 Peter Bergner <bergner@linux.ibm.com>
3970 * target.def (translate_mode_attribute): New hook.
3971 * targhooks.h (default_translate_mode_attribute): Declare.
3972 * targhooks.c (default_translate_mode_attribute): New function.
3973 * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
3974 * doc/tm.texi: Regenerate.
3975 * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
3976 (rs6000_translate_mode_attribute): New function.
3978 2018-07-06 Paul Koning <ni1d@arrl.net>
3980 * doc/md.texi (define_split): Document DONE and FAIL.
3981 (define_peephole2): Ditto.
3983 2018-07-05 Jeff Law <law@redhat.com>
3985 PR tree-optimization/86010
3986 * tree-ssa-dse.c (compute_trims): More aggressively trim at
3987 both the head and tail of mem* and str* calls.
3989 2018-07-05 Jim Wilson <jimw@sifive.com>
3991 * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
3993 2018-07-05 Indu Bhagat <indu.bhagat@oracle.com>
3995 * config/aarch64/aarch64-simd.md: correct flags text for
3996 MIN_EXPR replacement.
3998 2018-07-05 James Clarke <jrtc27@jrtc27.com>
4000 * configure: Regenerated.
4002 2018-07-05 Carl Love <cel@us.ibm.com>
4004 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
4005 float argument to VSX_BUILTIN_DOUBLEH_V4SF.
4006 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
4007 VSX_BUILTIN_DOUBLEL_V4SF.
4009 2018-07-05 Martin Sebor <msebor@redhat.com>
4012 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
4013 than its domain to compute its the upper bound of a char array.
4015 2018-07-05 Nathan Sidwell <nathan@acm.org>
4017 Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
4018 * doc/cpp.texi: Update comment.
4019 * doc/tm.texi: Rebuilt.
4020 * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
4021 (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
4022 * doc/extend.texi (Backwards Compatibility): Clarify it is system
4023 headers affected by extern "C".
4024 * system.h: Poison NO_IMPLICIT_EXTERN_C.
4025 * config/alpha/alpha.h, config/arm/uclinux-elf.h,
4026 config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
4027 config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
4028 config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
4029 config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
4030 config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
4031 config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
4032 config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
4033 config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
4034 config/sparc/openbsd64.h, config/sparc/sp-elf.h,
4035 config/sparc/sp64-elf.h, config/spu/spu.h,
4036 config/stormy16/stormy16.h, config/v850/v850.h,
4037 config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
4038 define NO_IMPLICIT_EXTERN_C.
4039 * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
4041 2018-07-05 Tamar Christina <tamar.christina@arm.com>
4044 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
4045 instead of GET_MODE_SIZE when comparing Units.
4047 2018-07-05 Tamar Christina <tamar.christina@arm.com>
4050 * rtlanal.c (set_noop_p): Constrain on mode change,
4051 include hard-reg-set.h
4053 2018-07-05 Tamar Christina <tamar.christina@arm.com>
4055 * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
4057 2018-07-05 Jakub Jelinek <jakub@redhat.com>
4060 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
4063 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
4065 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
4067 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
4070 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
4072 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
4074 2018-07-04 Eric Botcazou <ebotcazou@adacore.com>
4077 * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
4079 2018-07-04 Aldy Hernandez <aldyh@redhat.com>
4081 * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
4082 neg_*_op* variables.
4084 2018-07-04 Martin Liska <mliska@suse.cz>
4086 * tree-switch-conversion.c: Define
4087 max_ratio_for_speed and max_ratio_for_size constants.
4089 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
4090 Martin Liska <mliska@suse.cz>
4095 * common.opt: Rename align options with 'str_' prefix.
4096 * common/config/i386/i386-common.c (set_malign_value): New
4098 (ix86_handle_option): Use it to set -falign-* options/
4099 * config/aarch64/aarch64-protos.h (struct tune_params): Change
4100 type from int to string.
4101 * config/aarch64/aarch64.c: Update default values from int
4103 * config/alpha/alpha.c (alpha_override_options_after_change):
4105 * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
4106 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
4107 max skip conditionally.
4108 * config/i386/freebsd.h (SUBALIGN_LOG): New.
4109 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
4110 max skip conditionally.
4111 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
4112 max skip conditionally.
4113 * config/i386/gnu-user.h (SUBALIGN_LOG): New.
4114 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
4115 max skip conditionally.
4116 * config/i386/i386.c (struct ptt): Change type from int to
4118 (ix86_default_align): Set default values.
4119 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
4120 max skip conditionally.
4121 * config/i386/iamcu.h (SUBALIGN_LOG): New.
4122 (ASM_OUTPUT_MAX_SKIP_ALIGN):
4123 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
4124 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
4125 max skip conditionally.
4126 * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
4127 (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
4128 * config/i386/x86-64.h (SUBALIGN_LOG): New.
4129 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
4130 max skip conditionally.
4131 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
4132 * config/ia64/ia64.c (ia64_option_override): Set default values
4133 for alignment options.
4134 * config/m68k/m68k.c: Handle new str_align_* options.
4135 * config/mips/mips.c (mips_set_compression_mode): Change
4137 (mips_option_override): Set default values for options.
4138 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
4140 * config/rs6000/rs6000.c (rs6000_option_override_internal):
4142 * config/rx/rx.c (rx_option_override): Likewise.
4143 * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
4144 (LABEL_ALIGN): Use align_labels_log.
4145 (LOOP_ALIGN): Use align_loops_align.
4146 * config/s390/s390.c (s390_asm_output_function_label): Use new
4148 * config/sh/sh.c (sh_override_options_after_change):
4149 Change type of constants.
4150 * config/spu/spu.c (spu_sched_init): Likewise.
4151 * config/sparc/sparc.c (sparc_option_override): Set default
4153 * config/visium/visium.c (visium_option_override): Likewise.
4154 * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
4155 emit p2align format with last argument if it's not needed.
4156 * doc/invoke.texi: Document extended format of -falign-*.
4157 * final.c: Use align_labels alignment.
4158 * flags.h (struct target_flag_state): Change type to use
4160 (struct align_flags_tuple): New.
4161 (struct align_flags): Likewise.
4162 (align_loops_log): Redefine macro to use new types.
4163 (align_loops_max_skip): Redefine macro to use new types.
4164 (align_jumps_log): Redefine macro to use new types.
4165 (align_jumps_max_skip): Redefine macro to use new types.
4166 (align_labels_log): Redefine macro to use new types.
4167 (align_labels_max_skip): Redefine macro to use new types.
4168 (align_functions_log): Redefine macro to use new types.
4169 (align_loops): Redefine macro to use new types.
4170 (align_jumps): Redefine macro to use new types.
4171 (align_labels): Redefine macro to use new types.
4172 (align_functions): Redefine macro to use new types.
4173 (align_functions_max_skip): Redefine macro to use new types.
4174 (align_loops_value): New macro.
4175 (align_jumps_value): New macro.
4176 (align_labels_value): New macro.
4177 (align_functions_value): New macro.
4178 * function.c (invoke_set_current_function_hook): Propagate
4179 alignment values from flags to global variables default in
4181 * ipa-icf.c (sem_function::equals_wpa): Use
4182 cl_optimization_option_eq instead of memcmp.
4183 * lto-streamer.h (cl_optimization_stream_out): Support streaming
4185 (cl_optimization_stream_in): Likewise.
4186 * optc-save-gen.awk: Support strings in cl_optimization.
4187 * opth-gen.awk: Likewise.
4188 * opts.c (finish_options): Remove error checking of invalid
4190 (MAX_CODE_ALIGN): Remove.
4191 (MAX_CODE_ALIGN_VALUE): Likewise.
4192 (parse_and_check_align_values): New function.
4193 (check_alignment_argument): Likewise.
4194 (common_handle_option): Use check_alignment_argument.
4195 * opts.h (parse_and_check_align_values): Declare.
4196 * toplev.c (init_alignments): Remove.
4197 (read_log_maxskip): New.
4198 (parse_N_M): Likewise.
4199 (parse_alignment_opts): Likewise.
4200 (backend_init_target): Remove usage of init_alignments.
4201 * toplev.h (parse_alignment_opts): Declare.
4202 * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
4204 * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
4205 * tree.c (cl_option_hasher::equal): New.
4206 * varasm.c: Use new global macros.
4208 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
4210 * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
4211 Use a simpler align directive also if MAXSKIP = ALIGN-1.
4212 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4213 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4214 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4215 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
4216 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
4217 is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
4218 define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
4219 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4220 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4221 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4222 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4224 2018-07-04 Martin Liska <mliska@suse.cz>
4225 Jonathan Wakely <jwakely@redhat.com>
4227 * coverage.c: Use correct type.
4228 * doc/invoke.texi: Language correction.
4230 2018-07-03 H.J. Lu <hongjiu.lu@intel.com>
4233 * config/i386/i386.c (rest_of_insert_endbranch): Also generate
4234 ENDBRANCH for non-tail call which may return via indirect branch.
4235 * doc/extend.texi: Document indirect_return attribute.
4237 2018-07-03 Martin Sebor <msebor@redhat.com>
4239 PR tree-optimization/86274
4240 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
4242 (format_floating): Correct handling of infinities and NaNs.
4244 2018-07-03 Martin Sebor <msebor@redhat.com>
4246 * print-tree.c (print_real_cst): New function.
4247 (print_node_brief): Call it.
4248 (print_node): Ditto.
4250 2018-07-03 Jeff Law <law@redhat.com>
4252 * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
4253 into a single pattern.
4255 * config/h8300/h8300.md (ors code_iterator): New.
4256 (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
4257 a single pattern and single splitter.
4258 (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
4259 (iorqi3_1, xorqi3_1): Likewise.
4260 (iorqi3, xorqi3 expanders): Similarly.
4262 * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
4263 (movmd_internal) into a single pattern using the P mode iterator.
4264 (movmd splitters): Similarly.
4265 (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
4266 (movsd splitters): Similarly.
4268 * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
4269 ADDB, ADDW and ADDL into a single ADD attribute which selects the
4270 right table based on the size of the operand.
4271 * config/h8300/h8300.md (length_table): Corresponding changes. All
4272 references to "addb", "addw" and "addl" changed to "add".
4273 (btst patterns): Merge two variants into a single pattern.
4274 (tstqi, tsthi): Likewise.
4275 (addhi3_incdec, addsi3_incdec): Likewise.
4276 (subhi3_h8300hs, subsi3_h8300hs): Likewise.
4277 (mulhi3, mulsi3): Likewise.
4278 (udivhi3, udivsi3): Likewise.
4279 (divhi3, divsi3): Likewise.
4280 (andorqi3, andorhi3, andorsi3): Likewise.
4282 2018-07-03 Uros Bizjak <ubizjak@gmail.com>
4285 * config/i386/sse.md (uavg<mode>3_ceil): New expander.
4286 (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
4288 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
4290 PR tree-optimization/85694
4291 * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
4292 (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
4294 * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
4295 (<u>avg<mode>3_ceil): New patterns.
4297 2018-07-03 David Malcolm <dmalcolm@redhat.com>
4299 * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
4300 scan-tree-dump directive.
4301 * gcc.dg/vect/slp-perm-2.c: Likewise.
4302 * gcc.dg/vect/slp-perm-3.c: Likewise.
4303 * gcc.dg/vect/slp-perm-5.c: Likewise.
4304 * gcc.dg/vect/slp-perm-6.c: Likewise.
4305 * gcc.dg/vect/slp-perm-7.c: Likewise.
4306 * gcc.dg/vect/slp-perm-8.c: Likewise.
4308 2018-07-03 Marek Polacek <polacek@redhat.com>
4311 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
4313 2018-07-03 Richard Biener <rguenther@suse.de>
4316 * tree-ssa-structalias.c (find_func_clobbers): Properly
4317 handle indirect calls.
4319 2018-07-03 Jeff Law <law@redhat.com>
4321 * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
4322 (shifts): New code iterator.
4323 (movqi, movhi, movsi, movsf expanders): Consolidate into a single
4324 expander. Fix HImode handling on H8/SX.
4325 (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
4326 (subqi3, subhi3, subsi3 expanders): Likewise.
4327 (andqi3, andhi3, andsi3 expanders): Likewise.
4328 (iorqi3, iorhi3, iorsi3 expanders): Likewise.
4329 (xorqi3, xorhi3, xorsi3 expanders): Likewise.
4330 (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
4331 (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
4332 (zero_extendqihi2, zero_extendqisi2): Likewise.
4333 (extendqihi2, extendqisi2): Likewise.
4334 (rotlqi3, rotlhi3, rotlsi3): Likewise.
4335 (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
4336 (rotlqi3_1, rotlhi3_1): Likewise.
4337 (logicalhi3_sn, logicalsi3_sn): Likewise.
4338 (logicalhi3, logicalsi3): Likewise.
4340 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
4342 * tree-vect-patterns.c (vect_recog_rotate_pattern)
4343 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
4344 (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
4345 (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
4346 type to append_pattern_def_seq instead of creating a stmt_vec_info
4348 (build_mask_conversion): Likewise. Remove vinfo argument.
4349 (vect_add_conversion_to_patterm): Likewise, renaming to...
4350 (vect_add_conversion_to_pattern): ...this.
4351 (vect_recog_mask_conversion_pattern): Update call to
4352 build_mask_conversion. Pass the vector type to
4353 append_pattern_def_seq here too.
4354 (vect_recog_gather_scatter_pattern): Update call to
4355 vect_add_conversion_to_pattern.
4357 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
4359 * tree-vect-patterns.c (new_pattern_def_seq): Delete.
4360 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
4361 (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
4362 (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
4363 STMT_VINFO_PATTERN_DEF_SEQ to null here.
4364 (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
4365 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
4366 append_pattern_def_seq instead of new_pattern_def_seq.
4367 (vect_recog_divmod_pattern): Do both of the above.
4368 (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
4371 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
4373 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
4374 (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
4375 (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
4376 (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
4377 (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
4378 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
4379 (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
4380 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
4381 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
4382 (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
4383 parameter with a single stmt_vec_info.
4384 (vect_recog_func_ptr): Likewise.
4385 (vect_recog_gather_scatter_pattern): Likewise, folding in...
4386 (vect_try_gather_scatter_pattern): ...this.
4387 (vect_pattern_recog_1): Remove stmts_to_replace and just pass
4388 the stmt_vec_info of the statement to be matched. Don't clear
4389 STMT_VINFO_RELATED_STMT.
4390 (vect_pattern_recog): Update call accordingly.
4392 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
4394 PR tree-optimization/85694
4395 * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
4396 (uavgM3_ceil): Document new optabs.
4397 * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
4398 * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
4400 * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
4401 (savg_ceil_optab): New optabs.
4402 * tree-vect-patterns.c (vect_recog_average_pattern): New function.
4403 (vect_vect_recog_func_ptrs): Add it.
4404 * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
4405 constant directly from the associated lhs.
4407 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
4409 * tree-vect-patterns.c (vect_split_statement): New function.
4410 (vect_convert_input): Use it to try to split an existing cast.
4412 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
4414 * poly-int.h (print_hex): New function.
4415 * dumpfile.h (dump_dec, dump_hex): Declare.
4416 * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
4417 * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
4418 min_input_precision, operation_precision and operation_sign.
4419 * tree-vect-patterns.c (vect_get_range_info): New function.
4420 (vect_same_loop_or_bb_p, vect_single_imm_use)
4421 (vect_operation_fits_smaller_type): Delete.
4422 (vect_look_through_possible_promotion): Add an optional
4423 single_use_p parameter.
4424 (vect_recog_over_widening_pattern): Rewrite to use new
4425 stmt_vec_info infomration. Handle one operation at a time.
4426 (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
4427 (vect_truncatable_operation_p, vect_set_operation_type)
4428 (vect_set_min_input_precision): New functions.
4429 (vect_determine_min_output_precision_1): Likewise.
4430 (vect_determine_min_output_precision): Likewise.
4431 (vect_determine_precisions_from_range): Likewise.
4432 (vect_determine_precisions_from_users): Likewise.
4433 (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
4434 (vect_vect_recog_func_ptrs): Put over_widening first.
4436 (vect_pattern_recog): Call vect_determine_precisions.
4438 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
4440 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
4441 statements that have been replaced by further pattern statements.
4442 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
4444 2018-07-03 Richard Biener <rguenther@suse.de>
4446 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
4447 always set *dt. Dump vectype in vectype overload.
4448 * dumpfile.h (dump_gimple_expr): New function.
4449 (dump_gimple_expr_loc): Likewise.
4450 * dumpfile.c (dump_gimple_expr): New function.
4451 (dump_gimple_expr_loc): Likewise.
4453 2018-07-02 Jeff Law <law@redhat.com>
4455 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
4456 the H8/300, H8/300H and H8/S variants into a single pattern.
4457 (movhi_h8300, movqi_h8300hs): Similarly.
4458 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
4459 (QHI mode iterator): New.
4461 * config/h8300/h8300.md: Remove trailing whitespace.
4463 2018-07-02 Jim Wilson <jimw@sifive.com>
4465 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
4466 instead of emit_insn for interrupt returns.
4467 * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
4468 (riscv_sret, riscv_uret): Likewise.
4470 2018-07-02 David Malcolm <dmalcolm@redhat.com>
4472 * pretty-print.c (selftest::test_pp_format): Move save and restore
4473 of quotes to class auto_fix_quotes, and add an instance.
4474 * selftest.c: Include "intl.h".
4475 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
4476 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
4477 * selftest.h (selftest::auto_fix_quotes): New class.
4479 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
4481 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
4482 (aarch64_sve_prepare_conditional_op): Remove.
4483 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
4484 Allow aarch64_simd_reg_or_zero as select operand; remove
4485 the aarch64_sve_prepare_conditional_op call.
4486 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
4487 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
4488 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
4489 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
4490 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
4491 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
4492 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
4493 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
4494 and a splitters to match all of the *_any patterns.
4495 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
4497 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
4498 (SVE_COND_FP_BINARY_REV): Remove.
4499 (sve_int_op_rev, sve_fp_op_rev): New.
4500 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
4501 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
4502 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
4503 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
4504 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
4505 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
4506 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
4507 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
4508 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
4510 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
4511 Remove match_dup 1 from the inner unspec.
4512 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
4514 * config/aarch64/aarch64.md (movprfx): New attr.
4515 (length): Default movprfx to 8.
4516 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
4517 (*madd<SVE_I>, *msub<SVE_I): Likewise.
4518 (*<su>mul<SVE_I>3_highpart): Likewise.
4519 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
4520 (*v<ASHIFT><SVE_I>3): Likewise.
4521 (*<su><MAXMIN><SVE_I>3): Likewise.
4522 (*<su><MAXMIN><SVE_F>3): Likewise.
4523 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
4524 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
4525 (*div<SVE_F>4): Likewise.
4527 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
4529 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
4532 2018-07-02 Richard Biener <rguenther@suse.de>
4534 PR tree-optimization/86363
4535 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
4536 memset argument refers to a non-variable address.
4538 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
4540 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
4541 {PLUS,MINUS}_EXPR code to...
4542 (adjust_symbolic_bound): ...here,
4543 (combine_bound): ...here,
4544 (set_value_range_with_overflow): ...and here.
4546 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
4548 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
4550 (extract_range_from_abs_expr): ...here.
4552 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
4554 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
4555 -fno-omit-frame-pointer when not optimizing.
4557 2018-07-02 Martin Liska <mliska@suse.cz>
4560 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
4561 (propagate_nothrow): Likewise.
4563 2018-07-02 Martin Liska <mliska@suse.cz>
4566 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
4568 2018-07-02 David Malcolm <dmalcolm@redhat.com>
4570 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
4571 function in r262149, changing "loc" param from source_location to
4572 const dump_location_t &.
4573 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
4574 declaration, as above.
4576 2018-07-01 Paul Koning <ni1d@arrl.net>
4578 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
4579 -munit-asm, -mgnu-asm, -mdec-asm.
4580 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
4581 (pdp11_output_labelref): New.
4582 (pdp11_output_def): New.
4583 (pdp11_output_addr_vec_elt): New.
4584 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
4585 %# and %@ format codes.
4586 (pdp11_option_override): New.
4587 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
4588 (pdp11_output_ident): New.
4589 (pdp11_asm_named_section): New.
4590 (pdp11_asm_init_sections): New.
4591 (pdp11_file_start): New.
4592 (pdp11_file_end): New.
4593 (output_ascii): Use .ascii/.asciz for -mdec-asm.
4594 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
4595 %o, like %c but octal.
4596 (pdp11_option_override): New.
4597 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
4599 (DATA_SECTION_ASM_OP): Ditto.
4600 (READONLY_DATA_SECTION_ASM_OP): New.
4601 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
4602 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
4603 (ASM_OUTPUT_LABELREF): Ditto.
4604 (ASM_OUTPUT_DEF): Ditto.
4605 (ASM_OUTPUT_EXTERNAL): New.
4606 (ASM_OUTPUT_SOURCE_FILENAME): New.
4607 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
4608 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
4609 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
4610 %# and %@ format codes.
4611 * config/pdp11/pdp11.opt (mgnu-asm): New.
4612 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
4613 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
4614 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
4616 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
4618 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
4619 dereferencing path[] beyond its length.
4621 (debug_all_paths): New.
4622 (rewire_first_differing_edge): New.
4623 (adjust_paths_after_duplication): New.
4624 (duplicate_thread_path): Call adjust_paths_after_duplication.
4626 (thread_through_all_blocks): Add new argument to
4627 duplicate_thread_path.
4629 2018-06-30 Jim Wilson <jimw@sifive.com>
4631 * config/riscv/predicates.md (p2m1_shift_operand): New.
4632 (high_mask_shift_operand): New.
4633 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
4634 pattern using p2m1_shift_operand.
4635 (lshsi3_zero_extend_3+2): New combiner pattern using
4636 high_mask_shift_operand.
4638 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
4640 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
4642 (vect_recog_rotate_pattern): ...here.
4643 (vect_convert_input): Try to insert casts of invariants in the
4645 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
4646 preheader to be empty.
4648 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
4650 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
4651 vector type. If given, install it in the new statement's
4653 (vect_element_precision): New function.
4654 (vect_unpromoted_value): New struct.
4655 (vect_unpromoted_value::vect_unpromoted_value): New function.
4656 (vect_unpromoted_value::set_op): Likewise.
4657 (vect_look_through_possible_promotion): Likewise.
4658 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
4659 (vect_widened_op_tree, vect_convert_input): Likewise.
4660 (vect_convert_inputs, vect_convert_output): Likewise.
4661 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
4662 to handle the optional cast of the multiplication result and
4663 vect_widened_op_tree to detect the widened multiplication itself.
4664 Do not require the input and output of promotion casts to have
4665 the same sign, but base the signedness of the operation on the
4666 input rather than the result. If the pattern includes two
4667 promotions, check that those promotions have the same sign.
4668 Do not restrict the MULT_EXPR handling to a double-width result;
4669 handle quadruple-width results and wider. Use vect_convert_inputs
4670 to convert the inputs to the common type.
4671 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
4672 to handle the optional cast of the ABS result. Also allow a sign
4673 change or a sign extension between the ABS and MINUS.
4674 Use vect_widened_op_tree to detect the widened subtraction and use
4675 vect_convert_inputs to convert the inputs to the common type.
4676 (vect_handle_widen_op_by_const): Delete.
4677 (vect_recog_widen_op_pattern): New function.
4678 (vect_recog_widen_mult_pattern): Use it.
4679 (vect_recog_widen_shift_pattern): Likewise.
4680 (vect_recog_widen_sum_pattern): Use
4681 vect_look_through_possible_promotion to handle the promoted
4684 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
4686 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
4687 the containing gimple_seq *.
4688 * gimple-iterator.h (gsi_for_stmt): Declare it.
4689 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
4690 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
4691 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
4692 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
4693 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
4695 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
4697 (vect_mark_pattern_stmts): ...here. Handle cases in which the
4698 statement being replaced is part of an existing pattern
4699 definition sequence, inserting the new pattern statements before
4701 (vect_pattern_recog_1): Don't return a bool. If the statement
4702 is already part of a pattern, instead apply pattern matching
4703 to the pattern definition statements. Don't clear the
4704 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
4705 (vect_pattern_recog): Don't break after the first match;
4706 continue processing the pattern definition statements instead.
4707 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
4709 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
4711 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
4712 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
4713 (vect_recog_widen_sum_pattern): Use it.
4715 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
4717 * tree-vect-loop.c (vectorizable_reduction): Assert that the
4718 phi is not a pattern statement and has not been replaced by
4719 a pattern statement.
4720 * tree-vect-patterns.c (type_conversion_p): Don't check
4721 STMT_VINFO_IN_PATTERN_P.
4722 (vect_recog_vector_vector_shift_pattern): Likewise.
4723 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
4724 the pattern statement rather than the original statement; check
4725 directly for a WIDEN_MULT_EXPR here.
4726 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
4727 vect_is_simple_use to return the pattern statement rather
4728 than the original statement; use is_pattern_stmt_p to check
4729 for such a pattern statement.
4730 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
4731 to return the pattern statement rather than the original statement;
4732 don't do the same transformation here.
4733 (vect_is_simple_use): If the defining statement has been replaced
4734 by a pattern statement, return the pattern statement instead.
4735 Remove the corresponding (local) transformation from the vectype
4738 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
4740 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
4741 end and default to null.
4742 * tree-vect-loop.c (vect_create_epilog_for_reduction)
4743 (vectorizable_reduction): Update calls accordingly, dropping the
4744 gimple ** argument if the passed-back statement isn't needed.
4745 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
4746 (vect_recog_rotate_pattern): Likewise.
4747 (vect_recog_mask_conversion_pattern): Likewise.
4748 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
4749 (vect_mask_constant_operand_p): Likewise.
4750 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
4751 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
4752 (get_group_load_store_type, get_load_store_type): Likewise.
4753 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
4754 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
4755 (vectorizable_conversion, vectorizable_assignment): Likewise.
4756 (vectorizable_shift, vectorizable_operation): Likewise.
4757 (vectorizable_store, vect_is_simple_cond): Likewise.
4758 (vectorizable_condition, vectorizable_comparison): Likewise.
4759 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
4760 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
4761 and move it to the end. Cope with null def_stmt_outs.
4763 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
4765 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
4767 2018-06-29 Jeff Law <law@redhat.com>
4769 * config/v850/v850.c (v850_legitimate_address_p): Handle large
4770 displacements for TARGET_V850E2V3 and newer.
4771 (TARGET_LRA_P): Remove. Defaults to LRA now.
4772 * config/v850/v850.md (sign23byte_load): Remove.
4773 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
4774 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
4776 2018-06-29 Martin Liska <mliska@suse.cz>
4779 * coverage.c (coverage_init): Mangle full path name.
4780 * doc/invoke.texi: Document the change.
4781 * gcov-io.c (mangle_path): New.
4782 * gcov-io.h (mangle_path): Likewise.
4783 * gcov.c (mangle_name): Use mangle_path for path mangling.
4785 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4787 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
4788 if starting source register is not even.
4790 2018-06-29 Martin Liska <mliska@suse.cz>
4792 PR tree-optimization/86263
4793 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
4794 Make edge redirection.
4796 2018-06-29 David Malcolm <dmalcolm@redhat.com>
4798 * dumpfile.c (dump_loc): Add indentation based on scope depth.
4799 (dump_scope_depth): New variable.
4800 (get_dump_scope_depth): New function.
4801 (dump_begin_scope): New function.
4802 (dump_end_scope): New function.
4803 * dumpfile.h (get_dump_scope_depth): New declaration.
4804 (dump_begin_scope): New declaration.
4805 (dump_end_scope): New declaration.
4806 (class auto_dump_scope): New class.
4807 (AUTO_DUMP_SCOPE): New macro.
4808 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
4811 2018-06-29 Richard Biener <rguenther@suse.de>
4813 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
4814 compute_all_dependences succeeds.
4815 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
4816 exceed --param loop-max-datarefs-for-datadeps.
4818 2018-06-29 Jakub Jelinek <jakub@redhat.com>
4820 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
4822 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
4825 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
4826 alternative 0 in preferred_for_speed attribute.
4828 2018-06-28 Paul Koning <ni1d@arrl.net>
4830 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
4831 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
4832 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
4833 * config/pdp11/pdp11.md: Correct "length" attribute calculation
4834 for shift insn patterns.
4836 2018-06-28 David Malcolm <dmalcolm@redhat.com>
4838 * cgraph.c (cgraph_node::get_body): Replace assignments to
4839 "dump_file" with calls to set_dump_file.
4840 * dumpfile.c (alt_dump_file): Make static, and group with...
4841 (alt_flags): ...this definition.
4842 (dumps_are_enabled): New variable.
4843 (refresh_dumps_are_enabled): New function.
4844 (set_dump_file): New function.
4845 (set_alt_dump_file): New function.
4846 (gcc::dump_manager::dump_start): Replace assignments to
4847 "dump_file" and "alt_dump_file" with calls to set_dump_file and
4849 (gcc::dump_manager::dump_finish): Likewise.
4850 * dumpfile.h (alt_dump_file): Delete decl.
4851 (dumps_are_enabled): New variable decl.
4852 (set_dump_file): New function decl.
4853 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
4855 * tree-nested.c (lower_nested_functions): Replace assignments to
4856 "dump_file" with calls to set_dump_file.
4858 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
4860 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
4861 goto_locus of each outgoing edge of each basic block.
4863 2018-06-28 Richard Biener <rguenther@suse.de>
4865 * dwarf2out.c (decl_scope_table): Remove.
4866 (push_decl_scope): Likewise.
4867 (pop_decl_scope): Likewise.
4868 (gen_type_die_for_member): Do not call push/pop_decl_scope.
4869 (gen_struct_or_union_type_die): Likewise.
4870 (gen_tagged_type_die): Likewise.
4871 (dwarf2out_init): Do not initialize decl_scope_table.
4872 (dwarf2out_c_finalize): Do not free it.
4874 2018-06-28 Richard Biener <rguenther@suse.de>
4876 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
4877 deciding whether to not re-use a DIE.
4879 2018-06-28 Richard Biener <rguenther@suse.de>
4881 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
4882 DW_AT_abstract_origin attribute.
4884 2018-06-28 Martin Liska <mliska@suse.cz>
4886 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
4887 Use newly introduced constants.
4888 * tree-switch-conversion.h (struct jump_table_cluster):
4889 Define max_ratio_for_size and max_ratio_for_speed.
4891 2018-06-28 Martin Liska <mliska@suse.cz>
4893 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
4894 Add new checking assert to catch invalid state.
4895 (jump_table_cluster::can_be_handled): Handle single case
4897 (jump_table_cluster::is_beneficial): Bail out for such case.
4898 (bit_test_cluster::find_bit_tests):
4899 Add new checking assert to catch invalid state.
4900 (bit_test_cluster::can_be_handled): Handle single case
4902 (bit_test_cluster::is_beneficial): Bail out for such case.
4903 (switch_decision_tree::analyze_switch_statement):
4906 2018-06-28 Martin Liska <mliska@suse.cz>
4908 * common.opt: Introduce -completion option.
4909 * gcc.c (driver_handle_option): Handle it.
4910 (driver::main): Print completions if completion
4912 * opt-suggestions.c (option_proposer::get_completions):
4914 (option_proposer::suggest_completion): Likewise.
4915 (option_proposer::find_param_completions): Likewise.
4916 (verify_autocompletions): Likewise.
4917 (test_completion_valid_options): Likewise.
4918 (test_completion_valid_params): Likewise.
4919 (in_completion_p): Likewise.
4920 (empty_completion_p): Likewise.
4921 (test_completion_partial_match): Likewise.
4922 (test_completion_garbage): Likewise.
4923 (opt_proposer_c_tests): Likewise.
4924 * opt-suggestions.h: Declare new functions.
4925 * opts.c (common_handle_option): Handle OPT__completion_.
4926 * selftest-run-tests.c (selftest::run_tests): Add
4927 opt_proposer_c_tests.
4928 * selftest.c (assert_str_startswith): New.
4929 * selftest.h (assert_str_startswith): Likewise.
4930 (opt_proposer_c_tests): New.
4931 (ASSERT_STR_STARTSWITH): Likewise.
4933 2018-06-28 Martin Liska <mliska@suse.cz>
4935 * Makefile.in: Add opt-suggestions.o.
4936 * gcc-main.c: Include opt-suggestions.h.
4937 * gcc.c (driver::driver): Likewise.
4938 (driver::~driver): Remove m_option_suggestions.
4939 (driver::build_option_suggestions): Moved to option_proposer.
4940 (driver::suggest_option): Likewise.
4941 (driver::handle_unrecognized_options): Use option_proposer.
4942 * gcc.h (class driver): Add new memver m_option_proposer.
4943 * opt-suggestions.c: New file.
4944 * opt-suggestions.h: New file.
4946 2018-06-28 Martin Liska <mliska@suse.cz>
4948 * vec.h (class auto_string_vec): New (moved from auto_argvec).
4949 (auto_string_vec::~auto_string_vec): Likewise.
4951 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
4953 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
4954 prevent_decl_creation_for_types fields up and add reset_location field.
4955 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
4956 statement if id->reset_location is true.
4957 (copy_edges_for_bb): Do not set goto_locus on the new edges if
4958 id->reset_location is true.
4959 (copy_phis_for_bb): Force input_location on the arguments if
4960 id->reset_location is true.
4961 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
4962 is set on the function to be inlined.
4964 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
4966 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
4968 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
4970 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
4971 registers for Pmode.
4972 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
4973 hard registers for the clobbered pseudo.
4975 2018-06-27 Paul Koning <ni1d@arrl.net>
4977 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
4978 mutually exclusive options.
4979 * config/pdp11/constraints.md (h): New constraint.
4980 (O): Update definition to match shift code generation.
4981 (D): New constraint.
4982 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
4984 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
4986 (output_jump): Change arguments.
4987 (pdp11_fixed_cc_regs): New function.
4988 (pdp11_cc_mode): Ditto.
4989 (pdp11_expand_shift): Ditto.
4990 (pdp11_assemble_shift): Ditto.
4991 (pdp11_small_shift): Ditto.
4992 (pdp11_branch_cost): Remove.
4993 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
4995 (pdp11_register_move_cost): Update for CC registers.
4996 (pdp11_rtx_costs): Add case for LSHIFTRT.
4997 (pdp11_output_jump): Add CCNZ mode conditional branches.
4998 (notice_update_cc_on_set): Remove.
4999 (pdp11_cc_mode): New function.
5000 (simple_memory_operand): Correct pre/post decrement case.
5001 (no_side_effect_operand): New function.
5002 (pdp11_regno_reg_class): Add CC_REGS class.
5003 (pdp11_fixed_cc_regs): New function.
5004 (pdp11_small_shift): New function.
5005 (pdp11_expand_shift): New function to expand shift insns.
5006 (pdp11_assemble_shift): New function to output shifts.
5007 (pdp11_branch_cost): Remove.
5008 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
5009 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
5010 (WCHAR_TYPE): Ditto.
5011 (PTRDIFF_TYPE): Ditto.
5012 (ADJUST_INSN_LENGTH): New macro.
5013 (FIXED_REGISTERS): Add CC registers.
5014 (CALL_USED_REGISTERS): Ditto.
5016 (REG_CLASS_NAMES): Ditto.
5017 (REG_CLASS_CONTENTS): Ditto.
5018 (SELECT_CC_MODE): Use new function.
5019 (TARGET_FLAGS_REGNUM): New macro.
5020 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
5021 (cc0_reg_rtx): Remove.
5022 (CC_STATUS_MDEP): Remove.
5023 (CC_STATUS_MDEFP_INIT): Remove.
5024 (CC_IN_FPU): Remove.
5025 (NOTICE_UPDATE_CC): Remove.
5026 (REGISTER_NAMES): Add CC registers.
5027 (BRANCH_COST): Change to constant 1.
5028 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
5030 * config/pdp11/pdp11.opt (mbcopy): Remove.
5031 (mbcopy-builtin): Remove.
5032 (mbranch-cheap): Remove.
5033 (mbranch-expensive): Remove.
5034 * config/pdp11/predicates.md (expand_shift_operand): Update to
5035 match shift code generation.
5036 (ccnz_operator): New predicate.
5037 * doc/invoke.texi (PDP-11 Options): Remove deleted options
5038 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
5039 Remove non-existent option -mabshi, -mno-abshi. Document mutually
5041 * doc/md.texi (PDP-11): Document new D and h constraints. Update
5042 description of O constraint.
5044 2018-06-27 Jeff Law <law@redhat.com>
5045 Austin Law <austinklaw@gmail.com>
5047 * config/v850/v850.md (addsi3_set_flags): New pattern.
5048 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
5049 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
5050 (zero_extendhisi2_v850_set_flags): Likewise.
5051 (zero_extendqisi2_v850_set_flags): Likewise.
5052 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
5053 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
5054 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
5056 * config/v850/v850-protos.h (notice_update_cc): Remove.
5057 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
5058 (v850_print_operand): Handle 'D' and "d".
5059 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
5060 Add handling of arithmetic/logical operations compared against zero.
5061 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
5062 Do not look at v850_compare_op, instead get mode from last argument.
5063 (v850_gen_compare): Remove
5064 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
5065 after reload for prologue insns.
5066 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
5068 (construct_save_jarl): Likewise.
5069 (TARGET_FLAGS_REGNUM): Define.
5070 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
5071 (NOTICE_UPDATE_CC): Remove.
5072 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
5073 than cc0. Conditionalize on reload_completed.
5074 (cmpsi_insn, setfcc_insn): Likewise.
5075 (tst1 splitter): Turn into define_and_split which sets the flags
5077 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
5078 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
5079 (cstoresf4, cstoredf4): Clobber the flags.
5080 (cmpsi, cmpsf, cmpdf): Remove expanders.
5081 (setf_insn): Remove pattern.
5082 (addsi3): Turn into define_and_split which clobbers the flags after
5083 reload and a suitable pattern (addsi3_clobber_flags) for use after
5085 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
5086 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
5087 (ashrsi3, ashrsi3_v850e2): Likewise.
5088 (bins): Clobber the flags.
5089 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
5090 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
5091 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
5092 (call_value_internal_short, call_value_internal_long): Likewise.
5093 (callt_save_interrupt, callt_return_interrupt): Likewise.
5094 (save_interrupt, return_interrupt): Likewise.
5095 (callt_save_all_interrupt, save_all_interrupt): Likewise.
5096 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
5097 (restore_all_interrupt, _restore_all_interrupt): Likewise.
5098 (All FP comparisons): Only allow after reload has completed.
5100 (divh, divhu): Tweak output template.
5101 (branch_z_normal, branch_z_invert): Remove
5102 (branch_nz_normal, branch_nz_invert): Likewise.
5103 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
5105 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
5106 * config/v850/v850.c (notice_update_cc): Remove.
5107 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
5108 (CC_NO_CARRY): Likewise.
5109 (NOTICE_UPDATE_CC): Define to nothing.
5110 * config/v850/v850.md: Remove block comment on cc0 handling
5111 Remove "cc" attribute from all patterns. Remove cc_status handling
5112 from all patterns. Minor formatting fixes.
5114 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5116 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
5117 (cortex-a76.cortex-a55): Likewise.
5118 * config/aarch64/aarch64-tune.md: Regenerate.
5119 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
5120 cortex-a76.cortex-a55.
5122 2018-06-27 Jeff Law <law@redhat.com>
5124 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
5125 (MULTILIB_DIRNAMES): Similarly.
5127 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
5129 * gimple.h (gimple_return_retbnd): Delete.
5130 (gimple_return_set_retbnd): Likewise.
5131 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
5132 gimple_return_set_retbnd.
5133 * gimple-pretty-print.c (dump_gimple_return): Remove call to
5134 gimple_return_retbnd and adjust.
5135 * tree-inline.h (struct copy_body_data): Remove retbnd field.
5136 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
5137 Explicitly return NULL in a couple more cases. Move assertion
5138 on debug statements and remove unreachable code.
5139 (reset_debug_binding): Do not test id->retbnd.
5140 (expand_call_inline): Do not set it.
5142 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
5144 * configure.ac: Add --disable-gcov option.
5145 * configure: Regenerate.
5146 * Makefile.in: Honour @enable_gcov@.
5147 * doc/install.texi: Document --disable-gcov.
5149 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5151 * config/arm/arm-cpus.in (cortex-a76): New entry.
5152 (cortex-a76.cortex-a55): Likewise.
5153 * config/arm/arm-tables.opt: Regenerate.
5154 * config/arm/arm-tune.md: Likewise.
5155 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
5156 * doc/invoke.texi (ARM Options): Document cortex-a76 and
5157 cortex-a76.cortex-a55.
5159 2018-06-27 Tamar Christina <tamar.christina@arm.com>
5162 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
5164 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
5166 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
5168 (EPILOGUE_USES): Likewise.
5170 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
5172 * tree-inline.c (remap_location): New function extracted from...
5173 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
5174 (copy_phis_for_bb): ...here. Call remap_location.
5175 (copy_cfg_body): Adjust call to copy_edges_for_bb.
5177 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
5179 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
5180 unaligned vsx for 16B memset.
5182 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
5185 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
5186 ieee128_float_type_node to long_double_type_node unless
5187 TARGET_LONG_DOUBLE_128 is set.
5189 2018-06-26 David Malcolm <dmalcolm@redhat.com>
5191 * cfgloop.c (get_loop_location): Convert return type from
5192 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
5193 by implicit construction from rtx_insn *, and using
5194 dump_user_location_t::from_function_decl for the fallback case.
5195 * cfgloop.h (get_loop_location): Convert return type from
5196 location_t to dump_user_location_t.
5197 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
5198 dump_printf_loc to pass in a dump_location_t rather than a
5199 location_t, via the gimple stmt.
5200 * coverage.c (get_coverage_counts): Update calls to
5201 dump_printf_loc to pass in dump_location_t rather than a
5203 * doc/optinfo.texi (Dump types): Convert example of
5204 dump_printf_loc from taking "locus" to taking "insn". Update
5205 description of the "_loc" calls to cover dump_location_t.
5206 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
5208 (dump_user_location_t::dump_user_location_t): New constructors,
5209 from gimple *stmt and rtx_insn *.
5210 (dump_user_location_t::from_function_decl): New function.
5211 (dump_loc): Make static.
5212 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
5213 const dump_location_t &.
5214 (dump_generic_expr_loc): Delete.
5215 (dump_printf_loc): Convert param "loc" from location_t to
5216 const dump_location_t &.
5217 (selftest::test_impl_location): New function.
5218 (selftest::dumpfile_c_tests): New function.
5219 * dumpfile.h: Include "profile-count.h".
5220 (class dump_user_location_t): New class.
5221 (struct dump_impl_location_t): New struct.
5222 (class dump_location_t): New class.
5223 (dump_printf_loc): Convert 2nd param from source_location to
5224 const dump_location_t &.
5225 (dump_generic_expr_loc): Delete.
5226 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
5227 const dump_location_t &.
5228 * gimple-fold.c (fold_gimple_assign): Update call to
5229 dump_printf_loc to pass in a dump_location_t rather than a
5230 location_t, via the gimple stmt.
5231 (gimple_fold_call): Likewise.
5232 * gimple-loop-interchange.cc
5233 (loop_cand::analyze_iloop_reduction_var): Update for change to
5234 check_reduction_path.
5235 (tree_loop_interchange::interchange): Update for change to
5237 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
5238 change in return-type of find_loop_location.
5239 (graphite_regenerate_ast_isl): Likewise.
5240 * graphite-optimize-isl.c (optimize_isl): Likewise.
5241 * graphite.c (graphite_transform_loops): Likewise.
5242 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
5243 pass in a dump_location_t rather than a location_t, via the
5245 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
5246 * ipa.c (walk_polymorphic_call_targets): Likewise.
5247 * loop-unroll.c (report_unroll): Convert "locus" param from
5248 location_t to dump_location_t.
5249 (decide_unrolling): Update for change to get_loop_location's
5251 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
5252 location_t to dump_user_location_t.
5253 (grid_find_single_omp_among_assignments_1): Updates calls to
5254 dump_printf_loc to pass in a dump_location_t rather than a
5255 location_t, via the gimple stmt.
5256 (grid_parallel_clauses_gridifiable): Convert "tloc" from
5257 location_t to dump_location_t. Updates calls to dump_printf_loc
5258 to pass in a dump_location_t rather than a location_t, via the
5260 (grid_inner_loop_gridifiable_p): Likewise.
5261 (grid_dist_follows_simple_pattern): Likewise.
5262 (grid_gfor_follows_tiling_pattern): Likewise.
5263 (grid_target_follows_gridifiable_pattern): Likewise.
5264 (grid_attempt_target_gridification): Convert initialization
5265 of local "grid" from memset to zero-initialization; FIXME: does
5266 this require C++11? Update call to dump_printf_loc to pass in a
5267 optinfo_location rather than a location_t, via the gimple stmt.
5268 * profile.c (read_profile_edge_counts): Updates call to
5269 dump_printf_loc to pass in a dump_location_t rather than a
5271 (compute_branch_probabilities): Likewise.
5272 * selftest-run-tests.c (selftest::run_tests): Call
5274 * selftest.h (dumpfile_c_tests): New decl.
5275 * tree-loop-distribution.c (pass_loop_distribution::execute):
5276 Update for change in return type of find_loop_location.
5277 * tree-parloops.c (parallelize_loops): Likewise.
5278 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
5279 "locus" from location_t to dump_user_location_t.
5280 (canonicalize_loop_induction_variables): Likewise.
5281 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
5282 for change in return type of find_loop_location.
5283 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
5284 to dump_printf_loc to pass in a dump_location_t rather than a
5285 location_t, via the stmt.
5286 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
5288 * tree-vect-loop-manip.c (find_loop_location): Convert return
5289 type from source_location to dump_user_location_t.
5290 (vect_do_peeling): Update for above change.
5291 (vect_loop_versioning): Update for change in type of
5293 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
5294 from location_t to dump_user_location_t.
5295 (vect_estimate_min_profitable_iters): Update for change in type
5297 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
5298 location_t to dump_location_t.
5299 (vect_slp_bb): Update for change in type of vect_location.
5300 * tree-vectorizer.c (vect_location): Convert from source_location
5301 to dump_user_location_t.
5302 (try_vectorize_loop_1): Update for change in vect_location's type.
5303 (vectorize_loops): Likewise.
5304 (increase_alignment): Likewise.
5305 * tree-vectorizer.h (vect_location): Convert from source_location
5306 to dump_user_location_t.
5307 (find_loop_location): Convert return type from source_location to
5308 dump_user_location_t.
5309 (check_reduction_path): Convert 1st param from location_t to
5310 dump_user_location_t.
5311 * value-prof.c (check_counter): Update call to dump_printf_loc to
5312 pass in a dump_user_location_t rather than a location_t; update
5313 call to error_at for change in type of "locus".
5314 (check_ic_target): Update call to dump_printf_loc to
5315 pass in a dump_user_location_t rather than a location_t, via the
5318 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
5320 * config/s390/s390.h (enum processor_flags): Do not use
5321 default tune parameter when -march was specified.
5323 2018-06-26 Jakub Jelinek <jakub@redhat.com>
5326 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
5327 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
5330 2018-06-26 Richard Biener <rguenther@suse.de>
5332 PR tree-optimization/86287
5334 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
5335 (vect_analyze_loop): Initialize n_stmts.
5337 2018-06-26 Richard Biener <rguenther@suse.de>
5340 * fold-const.c (fold_convertible_p): Pointer extension
5343 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
5346 * dwarf2out.c (loc_list_has_views): Adjust comments.
5347 (dw_loc_list): Split single cross-partition range with
5350 2018-06-25 Jeff Law <law@redhat.com>
5352 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
5353 on -mbig-switch by default.
5355 * config/v850/predicates.md (const_float_1_operand): Fix match_code
5357 (const_float_0_operand): Remove unused predicate.
5358 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
5359 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
5360 (recipsf2): New expander. Original pattern now called
5362 (recipdf2, recipdf2_insn): Similarly.
5363 (rsqrtsf2, rsqrtsf2_insn): Similarly
5364 (rsqrtdf2, rsqrtdf2_insn): Similarly
5366 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
5368 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
5369 Simplify logic for FreeBSD (twice).
5371 2018-06-25 Martin Sebor <msebor@redhat.com>
5373 PR tree-optimization/86204
5374 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
5375 a strnlen result if it's less than the length of the string.
5377 2018-06-25 Martin Sebor <msebor@redhat.com>
5379 PR tree-optimization/85700
5380 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
5381 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
5382 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
5384 2018-06-25 Martin Sebor <msebor@redhat.com>
5386 * doc/extend.texi (Zero-length arrays): Update and clarify.
5388 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
5390 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
5391 added IEEE/IBM long double multilib support on PowerPC little
5392 endian Linux systems.
5393 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
5394 (MULTILIB_DEFAULTS): Likewise.
5395 * config/rs6000/rs6000.c (rs6000_option_override_internal):
5397 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
5398 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
5399 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
5401 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
5404 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
5405 (REORDER_45): Likewise.
5407 2018-06-25 Jeff Law <law@redhat.com>
5409 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
5410 dividend to 32 bits. Adjust length.
5411 (udivmodhi4): Cleanup output template. Fix length.
5413 2018-06-25 Carl Love <cel@us.ibm.com>
5415 * config/rs6000/vsx.md: Change word selector to prefered location.
5417 2018-06-25 Richard Biener <rguenther@suse.de>
5419 PR tree-optimization/86304
5420 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
5421 epilogue-if-converted loops as well.
5423 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
5425 * lto-section-out.c (lto_begin_section): Do not print section
5426 name for noaddr and unnumbered dumps.
5428 2018-06-25 Richard Biener <rguenther@suse.de>
5430 * tree-vectorizer.h (struct vec_info_shared): New structure
5431 with parts split out from struct vec_info and loop_nest from
5432 struct _loop_vec_info.
5433 (struct vec_info): Adjust accordingly.
5434 (struct _loop_vec_info): Likewise.
5435 (LOOP_VINFO_LOOP_NEST): Adjust.
5436 (LOOP_VINFO_DATAREFS): Likewise.
5437 (LOOP_VINFO_DDRS): Likewise.
5438 (struct _bb_vec_info): Likewise.
5439 (BB_VINFO_DATAREFS): Likewise.
5440 (BB_VINFO_DDRS): Likewise.
5441 (struct _stmt_vec_info): Add dr_aux member.
5442 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
5443 (DR_MISALIGNMENT_UNINITIALIZED): New.
5444 (set_dr_misalignment): Adjust.
5445 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
5446 (vect_analyze_loop): Adjust prototype.
5447 (vect_analyze_loop_form): Likewise.
5448 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
5449 Compute dependences lazily.
5450 (vect_record_base_alignments): Use shared datarefs/ddrs.
5451 (vect_verify_datarefs_alignment): Likewise.
5452 (vect_analyze_data_refs_alignment): Likewise.
5453 (vect_analyze_data_ref_accesses): Likewise.
5454 (vect_analyze_data_refs): Likewise.
5455 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
5456 constructor parameter for shared part.
5457 (vect_analyze_loop_form): Pass in shared part and adjust.
5458 (vect_analyze_loop_2): Pass in storage for the number of
5459 stmts. Move loop nest finding to the caller. Compute
5461 (vect_analyze_loop): Pass in shared part.
5462 (vect_transform_loop): Verify shared datarefs are unchanged.
5463 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
5464 constructor parameter for shared part.
5465 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
5466 (vect_slp_bb): Verify shared datarefs are unchanged before
5468 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
5470 (new_stmt_vec_info): Initialize DR_AUX misalignment to
5471 DR_MISALIGNMENT_UNINITIALIZED.
5472 * tree-vectorizer.c (vec_info::vec_info): Add constructor
5473 parameter for shared part.
5474 (vec_info::~vec_info): Adjust.
5475 (vec_info_shared::vec_info_shared): New.
5476 (vec_info_shared::~vec_info_shared): Likewise.
5477 (vec_info_shared::save_datarefs): Likewise.
5478 (vec_info_shared::check_datarefs): Likewise.
5479 (try_vectorize_loop_1): Construct shared part live for analyses
5480 of a single loop for multiple vector sizes.
5481 * tree-parloops.c (gather_scalar_reductions): Adjust.
5483 2018-06-25 Richard Biener <rguenther@suse.de>
5485 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
5486 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
5487 (vect_analyze_data_refs): Remove similar code from here and
5488 simplify accordingly.
5490 2018-06-25 Richard Biener <rguenther@suse.de>
5492 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
5493 for reverse storage order accesses rather than asserting
5494 they cannot happen here.
5496 2018-06-25 Tom de Vries <tdevries@suse.de>
5499 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
5500 Use data16 instead of .byte for insn prefix.
5502 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
5505 * parser.c (make_char_string_pack): Pass this literal chars
5506 through cpp_interpret_string.
5507 (cp_parser_userdef_numeric_literal): Check the result of
5508 make_char_string_pack.
5510 2018-06-24 Maya Rashish <coypu@sdf.org>
5512 * ginclude/stddef.h: Simplify conditions around avoiding
5513 re-definition of __size_t.
5515 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
5517 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
5518 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
5520 2018-06-22 Maya Rashish <coypu@sdf.org>
5522 * doc/invoke.texi (mno-fancy-math-387): Update for changes
5523 made to OpenBSD and NetBSD through the years.
5525 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
5527 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
5528 behavior of vec_pack (vector double, vector double) to match
5529 behavior of vec_float2 (vector double, vector double).
5531 2018-06-22 Olivier Hainque <hainque@adacore.com>
5533 * gimplify.c (gimplify_function_tree): Prevent creation
5534 of a trampoline for the address of the current function
5535 passed to entry/exit instrumentation hooks.
5537 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
5540 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
5543 2018-06-22 Martin Liska <mliska@suse.cz>
5545 PR tree-optimization/86263
5546 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
5547 Bail out if is_enabled is false.
5548 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
5550 (jump_table_cluster::is_enabled): New function.
5552 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
5554 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
5555 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
5556 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
5557 (lto_input_ts_binfo_tree_pointers): Likewise.
5558 * tree-streamer-out.c (streamer_write_tree_bitfields,
5559 write_ts_binfo_tree_pointers): Likewise.
5560 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
5562 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
5564 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
5566 2018-06-22 Martin Liska <mliska@suse.cz>
5568 * symbol-summary.h (get): Make it pure and inline move
5569 functionality from ::get function.
5570 (get): Remove and inline into ::get and ::get_create.
5571 (get_create): Move code from ::get function.
5573 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5576 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
5577 -x assembler-with-cpp.
5579 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5581 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
5582 _FILE_OFFSET_BITS=64 for C++.
5584 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
5586 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
5587 conversion insn that shows up when pr85657-3.c is compiled using
5588 IEEE 128-bit long double.
5589 (neg<mode>2_internal): Use the correct mode to check whether the
5590 mode is IBM extended.
5591 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
5592 multiply and divide external functions from being created more
5595 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
5597 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
5599 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
5600 the edge can be forwarded.
5601 (cfg_layout_merge_blocks): Likewise.
5603 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
5605 * except.c (finish_eh_generation): Commit edge insertions only after
5606 the EH edges have been redirected from post-landing to landing pads.
5608 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
5610 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
5611 create_tmp_var_for to create the FRAME decl.
5612 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
5614 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
5616 * tree-inline.c (copy_edges_for_bb): Minor tweak.
5617 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
5618 debug statement when resetting its value.
5619 (expand_call_inline): Copy the locus of the call onto the assignment
5620 of the return value, if any. Use local variable in more cases.
5622 2018-06-21 Martin Liska <mliska@suse.cz>
5624 * ipa-pure-const.c (propagate_nothrow): Use
5625 funct_state_summaries->get.
5626 (dump_malloc_lattice): Likewise.
5627 (propagate_malloc): Likewise.
5629 2018-06-21 Richard Biener <rguenther@suse.de>
5631 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
5632 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
5633 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
5635 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
5636 BLOCK_ABSTRACT_ORIGIN unconditionally.
5638 2018-06-21 David Malcolm <dmalcolm@redhat.com>
5640 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
5642 * ipa-reference.c (ipa_reference_c_finalize): Delete
5643 ipa_ref_opt_sum_summaries and set it to NULL.
5645 2018-06-21 Tom de Vries <tdevries@suse.de>
5647 PR tree-optimization/85859
5648 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
5649 test with comment from bb_no_side_effects_p.
5651 2018-06-21 Richard Biener <rguenther@suse.de>
5653 PR tree-optimization/86232
5654 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
5655 max for constant niter.
5657 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
5659 * config/aarch64/aarch64-simd.md
5660 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
5662 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
5664 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
5665 Make opernads of the unspec commutative.
5667 2018-06-21 Richard Biener <rguenther@suse.de>
5669 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
5670 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
5671 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
5672 (vect_analyze_data_ref_dependence): Re-order checks to deal with
5674 (vect_record_base_alignments): Do not record base alignment
5675 for gathers or scatters.
5676 (vect_compute_data_ref_alignment): Drop return value that is always
5677 true. Bail out early for gathers or scatters.
5678 (vect_enhance_data_refs_alignment): Bail out early for gathers
5680 (vect_find_same_alignment_drs): Likewise.
5681 (vect_analyze_data_refs_alignment): Remove dead code.
5682 (vect_slp_analyze_and_verify_node_alignment): Likewise.
5683 (vect_analyze_data_refs): For possible gathers or scatters do
5684 not create an alternate DR, just check their possible validity
5685 and mark them. Adjust DECL_NONALIASED handling to not rely
5687 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
5688 update inits of gathers or scatters.
5689 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
5690 Also copy gather/scatter flag to pattern vinfo.
5692 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
5694 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
5695 behavior of vec_packsu (vector unsigned long long, vector unsigned
5696 long long) to match behavior of vec_packs with same signature.
5698 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
5699 Thomas Schwinge <thomas@codesourcery.com>
5700 Cesar Philippidis <cesar@codesourcery.com>
5702 * gimplify.c (gimplify_scan_omp_clauses): Add support for
5703 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
5704 (gimplify_adjust_omp_clauses): Likewise.
5705 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
5706 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
5707 (gimplify_omp_target_update): Update handling of acc update and
5709 * omp-low.c (install_var_field): Remove unused parameter
5710 base_pointers_restrict.
5711 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
5712 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
5714 (omp_target_base_pointers_restrict_p): Delete.
5715 (scan_omp_target): Update call to scan_sharing_clauses.
5716 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
5718 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
5719 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
5720 (convert_local_omp_clauses): Likewise.
5721 * tree-pretty-print.c (dump_omp_clause): Likewise.
5722 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
5724 (omp_clause_code_name): Likewise.
5726 2018-06-20 Jakub Jelinek <jakub@redhat.com>
5729 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
5732 PR tree-optimization/86231
5733 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
5734 anti-range don't overwrite *vr0min before using it to compute *vr0max.
5736 2018-06-20 Tom de Vries <tdevries@suse.de>
5738 PR tree-optimization/86097
5739 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
5740 iv type if signedness of iv type is not the same as that of *nit.
5742 2018-06-20 Jakub Jelinek <jakub@redhat.com>
5744 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
5745 EDGE_EH edges, verify they are all EDGE_EH.
5747 2018-06-20 Maya Rashish <coypu@sdf.org>
5749 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
5751 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5753 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
5754 * config/aarch64/aarch64.c (xgene1_tunings): Add
5755 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
5756 (aarch64_mode_valid_for_sched_fusion_p):
5757 Allow 16-byte modes.
5758 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
5759 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
5761 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
5763 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
5764 * config/aarch64/iterators.md (VQ2): New mode iterator.
5766 2018-06-20 Martin Liska <mliska@suse.cz>
5768 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
5769 Change default ratio from 10 to 8.
5771 2018-06-20 Martin Liska <mliska@suse.cz>
5773 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
5775 (bit_test_cluster::find_bit_tests): Likewise.
5776 (switch_decision_tree::analyze_switch_statement): Find clusters.
5777 * tree-switch-conversion.h (struct jump_table_cluster): Document
5780 2018-06-20 Martin Liska <mliska@suse.cz>
5782 * tree-switch-conversion.c (switch_conversion::collect):
5783 Record m_uniq property.
5784 (switch_conversion::expand): Bail out for special conditions.
5785 (group_cluster::~group_cluster): New.
5786 (group_cluster::group_cluster): Likewise.
5787 (group_cluster::dump): Likewise.
5788 (jump_table_cluster::emit): New.
5789 (switch_decision_tree::fix_phi_operands_for_edges): New.
5790 (struct case_node): Remove struct.
5791 (jump_table_cluster::can_be_handled): New.
5792 (case_values_threshold): Moved to header.
5793 (reset_out_edges_aux): Likewise.
5794 (jump_table_cluster::is_beneficial): New.
5795 (bit_test_cluster::can_be_handled): Likewise.
5796 (add_case_node): Remove.
5797 (bit_test_cluster::is_beneficial): New.
5798 (case_bit_test::cmp): New.
5799 (bit_test_cluster::emit): New.
5800 (expand_switch_as_decision_tree_p): Remove.
5801 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
5802 (fix_phi_operands_for_edge): Likewise.
5803 (switch_decision_tree::analyze_switch_statement): New.
5804 (compute_cases_per_edge): Move ...
5805 (switch_decision_tree::compute_cases_per_edge): ... here.
5806 (try_switch_expansion): Likewise.
5807 (switch_decision_tree::try_switch_expansion): Likewise.
5808 (record_phi_operand_mapping): Likewise.
5809 (switch_decision_tree::record_phi_operand_mapping): Likewise.
5810 (emit_case_decision_tree): Likewise.
5811 (switch_decision_tree::emit): Likewise.
5812 (balance_case_nodes): Likewise.
5813 (switch_decision_tree::balance_case_nodes): Likewise.
5814 (dump_case_nodes): Likewise.
5815 (switch_decision_tree::dump_case_nodes): Likewise.
5816 (emit_jump): Likewise.
5817 (switch_decision_tree::emit_jump): Likewise.
5818 (emit_cmp_and_jump_insns): Likewise.
5819 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
5820 (emit_case_nodes): Likewise.
5821 (switch_decision_tree::emit_case_nodes): Likewise.
5822 (conditional_probability): Remove.
5823 * tree-switch-conversion.h (enum cluster_type): New.
5825 (struct cluster): Likewise.
5826 (cluster::cluster): Likewise.
5827 (struct simple_cluster): Likewise.
5828 (simple_cluster::simple_cluster): Likewise.
5829 (struct group_cluster): Likewise.
5830 (struct jump_table_cluster): Likewise.
5831 (struct bit_test_cluster): Likewise.
5832 (struct min_cluster_item): Likewise.
5833 (struct case_tree_node): Likewise.
5834 (case_tree_node::case_tree_node): Likewise.
5835 (jump_table_cluster::case_values_threshold): Likewise.
5836 (struct case_bit_test): Likewise.
5837 (struct switch_decision_tree): Likewise.
5838 (struct switch_conversion): Likewise.
5839 (switch_decision_tree::reset_out_edges_aux): Likewise.
5841 2018-06-20 Martin Liska <mliska@suse.cz>
5843 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
5844 (hoist_edge_and_branch_if_true): Likewise.
5845 (expand_switch_using_bit_tests_p): Likewise.
5846 (struct case_bit_test): Likewise.
5847 (case_bit_test_cmp): Likewise.
5848 (emit_case_bit_tests): Likewise.
5849 (switch_conversion::switch_conversion): New class.
5850 (struct switch_conv_info): Remove old struct.
5851 (collect_switch_conv_info): More to ...
5852 (switch_conversion::collect): ... this.
5853 (check_range): Likewise.
5854 (switch_conversion::check_range): Likewise.
5855 (check_all_empty_except_final): Likewise.
5856 (switch_conversion::check_all_empty_except_final): Likewise.
5857 (check_final_bb): Likewise.
5858 (switch_conversion::check_final_bb): Likewise.
5859 (create_temp_arrays): Likewise.
5860 (switch_conversion::create_temp_arrays): Likewise.
5861 (free_temp_arrays): Likewise.
5862 (gather_default_values): Likewise.
5863 (switch_conversion::gather_default_values): Likewise.
5864 (build_constructors): Likewise.
5865 (switch_conversion::build_constructors): Likewise.
5866 (constructor_contains_same_values_p): Likewise.
5867 (switch_conversion::contains_same_values_p): Likewise.
5868 (array_value_type): Likewise.
5869 (switch_conversion::array_value_type): Likewise.
5870 (build_one_array): Likewise.
5871 (switch_conversion::build_one_array): Likewise.
5872 (build_arrays): Likewise.
5873 (switch_conversion::build_arrays): Likewise.
5874 (gen_def_assigns): Likewise.
5875 (switch_conversion::gen_def_assigns): Likewise.
5876 (prune_bbs): Likewise.
5877 (switch_conversion::prune_bbs): Likewise.
5878 (fix_phi_nodes): Likewise.
5879 (switch_conversion::fix_phi_nodes): Likewise.
5880 (gen_inbound_check): Likewise.
5881 (switch_conversion::gen_inbound_check): Likewise.
5882 (process_switch): Use the newly created class.
5883 (switch_conversion::expand): New.
5884 (switch_conversion::~switch_conversion): New.
5885 * tree-switch-conversion.h: New file.
5887 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
5889 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
5890 tree-vect-patterns.c.
5891 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
5892 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
5893 (vect_recog_sad_pattern): Likewise.
5894 (vect_recog_widen_sum_pattern): Likewise.
5895 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
5896 (vect_recog_widen_shift_pattern): Remove the type_in argument.
5897 (vect_recog_rotate_pattern): Likewise.
5898 (vect_recog_mult_pattern): Likewise.
5899 (vect_recog_vector_vector_shift_pattern): Likewise.
5900 (vect_recog_divmod_pattern): Likewise.
5901 (vect_recog_mixed_size_cond_pattern): Likewise.
5902 (vect_recog_bool_pattern): Likewise.
5903 (vect_recog_mask_conversion_pattern): Likewise.
5904 (vect_try_gather_scatter_pattern): Likewise.
5905 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
5906 (vect_recog_over_widening_pattern): Likewise.
5907 (vect_recog_gather_scatter_pattern): Likewise.
5908 (vect_recog_func_ptr): Move from tree-vectorizer.h
5909 (vect_vect_recog_func_ptrs): Move further down the file.
5910 (vect_recog_func): Likewise. Remove the third argument.
5911 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
5912 (vect_pattern_recog_1): Expect the pattern function to do any
5913 necessary target tests. Also expect it to provide a vector type.
5914 Remove the type_in handling.
5916 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
5918 * tree-vect-patterns.c (vect_pattern_detected): New function.
5919 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
5920 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
5921 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
5922 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
5923 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
5924 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
5925 (vect_recog_mask_conversion_pattern)
5926 (vect_try_gather_scatter_pattern): Likewise.
5928 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
5930 * tree-vect-patterns.c (vect_get_internal_def): New function.
5931 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
5932 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
5933 (search_type_for_mask_1): Use it.
5935 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
5937 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
5938 redundant WIDEN_SUM_EXPR handling.
5939 (vect_recog_sad_pattern): Likewise.
5941 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
5943 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
5944 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
5945 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
5946 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
5947 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
5949 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
5951 * tree-vect-stmts.c (vectorizable_call): Make sure that we
5952 use the stmt_vec_info of the original bb statement for the
5953 new zero assignment, even if the call is part of a pattern.
5955 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
5957 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
5958 that the sequence is attached to the original statement rather
5959 than the pattern statement.
5960 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
5961 PATTERN_DEF_SEQ from the original statement rather than
5962 the main pattern statement.
5963 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
5964 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
5965 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
5967 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
5969 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
5970 definition statements before the early exit for statements that aren't
5972 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
5974 (vect_transform_loop): ...here. Process pattern definition
5975 statements without first checking whether the main pattern
5976 statement is live or relevant.
5978 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
5980 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
5981 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
5983 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
5985 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
5986 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
5987 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
5988 (expand_block_compare): Change select_block_compare_mode call.
5989 (expand_strncmp_align_check): Use new functions, fix comment.
5990 (emit_final_str_compare_gpr): New function.
5991 (expand_strn_compare): Refactor and clean up code.
5992 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
5994 2018-06-19 Tony Reix <tony.reix@atos.com>
5995 Damien Bergamini <damien.bergamini@atos.com>
5996 David Edelsohn <dje.gcc@gmail.com>
5998 * collect2.c (static_obj): New variable.
5999 (static_libs): New variable.
6000 (is_in_list): Uncomment declaration.
6001 (main): Track AIX libraries linked statically.
6002 (is_in_list): Uncomment definition.
6003 (scan_prog_file): Don't add AIX shared libraries initializer
6004 to constructor list if linking statically.
6006 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
6008 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
6010 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
6012 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
6014 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
6017 2018-06-19 Martin Liska <mliska@suse.cz>
6019 * config/i386/i386.c (ix86_can_inline_p): Do not use
6020 ipa_fn_summaries::get_create.
6021 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
6023 (devirtualization_time_bonus): Likewise.
6024 (ipcp_propagate_stage): Likewise.
6025 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
6026 (edge_set_predicate): Likewise.
6027 (evaluate_conditions_for_known_args): Likewise.
6028 (evaluate_properties_for_edge): Likewise.
6029 (ipa_call_summary::reset): Tranform to ...
6030 (ipa_call_summary::~ipa_call_summary): ... this.
6031 (ipa_fn_summary::reset): Transform to ...
6032 (ipa_fn_summary::~ipa_fn_summary): ... this.
6033 (ipa_fn_summary_t::remove): Rename to ...
6034 (ipa_fn_summary_t::remove_callees): ... this.
6035 (ipa_fn_summary_t::duplicate): Use placement new
6036 instead of memory copy.
6037 (ipa_call_summary_t::duplicate): Likewise.
6038 (ipa_call_summary_t::remove): Remove.
6039 (dump_ipa_call_summary): Change get_create to get.
6040 (ipa_dump_fn_summary): Dump only when summary exists.
6041 (analyze_function_body): Use symbol_summary::get instead
6043 (compute_fn_summary): Likewise.
6044 (estimate_edge_devirt_benefit): Likewise.
6045 (estimate_edge_size_and_time): Likewise.
6046 (inline_update_callee_summaries): Likewise.
6047 (remap_edge_change_prob): Likewise.
6048 (remap_edge_summaries): Likewise.
6049 (ipa_merge_fn_summary_after_inlining): Likewise.
6050 (write_ipa_call_summary): Likewise.
6051 (ipa_fn_summary_write): Likewise.
6052 (ipa_free_fn_summary): Likewise.
6053 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
6054 (struct ipa_call_summary): Likewise.
6055 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
6057 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
6058 (estimate_size_after_inlining): Likewise.
6059 (estimate_growth): Likewise.
6060 (growth_likely_positive): Likewise.
6061 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
6062 (inline_call): Likewise.
6063 * ipa-inline.c (caller_growth_limits): Likewise.
6064 (can_inline_edge_p): Likewise.
6065 (can_inline_edge_by_limits_p): Likewise.
6066 (compute_uninlined_call_time): Likewise.
6067 (compute_inlined_call_time): Likewise.
6068 (want_inline_small_function_p): Likewise.
6069 (edge_badness): Likewise.
6070 (update_caller_keys): Likewise.
6071 (update_callee_keys): Likewise.
6072 (inline_small_functions): Likewise.
6073 (inline_to_all_callers_1): Likewise.
6074 (dump_overall_stats): Likewise.
6075 (early_inline_small_functions): Likewise.
6076 (early_inliner): Likewise.
6077 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
6078 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
6079 * ipa-pure-const.c (malloc_candidate_p): Likewise.
6080 * ipa-split.c (execute_split_functions): Likewise.
6081 * symbol-summary.h: Likewise.
6082 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
6084 2018-06-19 Richard Biener <rguenther@suse.de>
6086 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
6087 (vectorize_loops): ... here. Fix dbgcnt handling.
6088 (try_vectorize_loop): Wrap try_vectorize_loop_1.
6090 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
6093 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
6094 ieee128 argument takes up only one (vector) register, not two (floating
6097 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
6099 * gimplify.c (gimplify_init_constructor): Really never clear for an
6100 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
6102 2018-06-19 Richard Biener <rguenther@suse.de>
6104 PR tree-optimization/86179
6105 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
6106 after failed recognition.
6108 2018-06-18 Martin Sebor <msebor@redhat.com>
6111 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
6112 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
6113 Handle integer subtraction.
6114 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
6115 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
6117 2018-06-18 David Malcolm <dmalcolm@redhat.com>
6119 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
6120 param from rtx to rtx_insn *.
6121 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
6123 (frv_ifcvt_modify_insn): Likwise.
6124 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
6125 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
6126 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
6127 as_a <rtx_insn *> cast to local "unprotected_region" once
6128 it's been established that it's not NULL or pc_rtx.
6129 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
6130 param "sethi" from rtx to rtx_insn *.
6131 (nds32_group_float_insns): Likewise for param "insn".
6132 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
6134 (vax_output_int_subtract): Likewise.
6135 * config/vax/vax.c (vax_output_int_add): Likewise for param
6137 (vax_output_int_subtract): Likewise.
6138 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
6139 (emit_pattern_after): Likewise for param "after".
6140 (emit_insn_after): Likewise.
6141 (emit_jump_insn_after): Likewise.
6142 (emit_call_insn_after): Likewise.
6143 (emit_debug_insn_after): Likewise.
6144 (emit_pattern_before): Likewise for param "before".
6145 (emit_insn_before): Likewise.
6146 (emit_jump_insn_before): Likewise.
6147 * final.c (get_insn_template): Likewise for param "insn", removing
6149 * output.h (get_insn_template): Likewise for 2nd param.
6150 * rtl.h (emit_insn_before): Likewise.
6151 (emit_jump_insn_before): Likewise.
6152 (emit_debug_insn_before_noloc): Likewise.
6153 (emit_insn_after): Likewise.
6154 (emit_jump_insn_after): Likewise.
6155 (emit_call_insn_after): Likewise.
6156 (emit_debug_insn_after): Likewise.
6157 (set_insn_deleted): Likewise for param.
6159 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
6162 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
6163 floating point modes, so that IFmode is numerically greater than
6164 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
6165 to declare the ordering. This prevents IFmode from being
6166 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
6167 machine. Include rs6000-modes.h to share the fractional values
6168 between genmodes* and the rest of the compiler.
6172 * config/rs6000/rs6000-modes.h: New file.
6173 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
6174 meaning of rs6000_long_double_size so that 126..128 selects an
6175 appropriate 128-bit floating point type.
6176 (rs6000_option_override_internal): Likewise.
6177 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
6178 (TARGET_LONG_DOUBLE_128): Change the meaning of
6179 rs6000_long_double_size so that 126..128 selects an appropriate
6180 128-bit floating point type.
6181 (LONG_DOUBLE_TYPE_SIZE): Update comment.
6182 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
6183 source and destination to match the standard usage.
6184 (truncifkf2): Likewise.
6185 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
6186 ISA 2.07 to use an explicit clobber, instead of passing in a
6188 (copysign<mode>3_soft): Likewise.
6190 2018-06-18 David Malcolm <dmalcolm@redhat.com>
6192 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
6193 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
6194 (vect_slp_analyze_instance_dependence): Likewise.
6195 (vect_enhance_data_refs_alignment): Likewise.
6196 (vect_analyze_data_refs_alignment): Likewise.
6197 (vect_slp_analyze_and_verify_instance_alignment
6198 (vect_analyze_data_ref_accesses): Likewise.
6199 (vect_prune_runtime_alias_test_list): Likewise.
6200 (vect_analyze_data_refs): Likewise.
6201 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
6202 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
6203 (vect_analyze_scalar_cycles_1): Likewise.
6204 (vect_get_loop_niters): Likewise.
6205 (vect_analyze_loop_form_1): Likewise.
6206 (vect_update_vf_for_slp): Likewise.
6207 (vect_analyze_loop_operations): Likewise.
6208 (vect_analyze_loop): Likewise.
6209 (vectorizable_induction): Likewise.
6210 (vect_transform_loop): Likewise.
6211 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
6212 * tree-vect-slp.c (vect_analyze_slp): Likewise.
6213 (vect_make_slp_decision): Likewise.
6214 (vect_detect_hybrid_slp): Likewise.
6215 (vect_slp_analyze_operations): Likewise.
6216 (vect_slp_bb): Likewise.
6217 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
6218 (vectorizable_bswap): Likewise.
6219 (vectorizable_call): Likewise.
6220 (vectorizable_simd_clone_call): Likewise.
6221 (vectorizable_conversion): Likewise.
6222 (vectorizable_assignment): Likewise.
6223 (vectorizable_shift): Likewise.
6224 (vectorizable_operation): Likewise.
6225 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
6227 2018-06-18 Martin Sebor <msebor@redhat.com>
6229 PR tree-optimization/81384
6230 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
6231 * builtins.c (expand_builtin_strnlen): New function.
6232 (expand_builtin): Call it.
6233 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
6234 * builtins.def (BUILT_IN_STRNLEN): New.
6235 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
6236 Warn for bounds in excess of maximum object size.
6237 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
6238 single-value ranges. Handle strnlen.
6239 (handle_builtin_strlen): Handle strnlen.
6240 (strlen_check_and_optimize_stmt): Same.
6241 * doc/extend.texi (Other Builtins): Document strnlen.
6243 2018-06-18 Maya Rashish <coypu@sdf.org>
6245 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
6246 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
6247 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
6249 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
6251 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
6253 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6255 * tree.c (escaped_string::escape): Replace cast to char * by
6256 const_cast<char *> (unescaped).
6258 2018-06-18 Nick Clifton <nickc@redhat.com>
6261 * tree.c (escaped_string): New class. Converts an unescaped
6262 string into its escaped equivalent.
6263 (warn_deprecated_use): Use the new class to convert the
6264 deprecation message, if present.
6265 (test_escaped_strings): New self test.
6266 (test_c_tests): Add test_escaped_strings.
6267 * doc/extend.texi (deprecated): Add a note that the
6268 deprecation message is affected by the -fmessage-length
6269 option, and that control characters will be escaped.
6270 (#pragma GCC error): Document this pragma.
6271 (#pragma GCC warning): Likewise.
6272 * doc/invoke.texi (-fmessage-length): Document this option's
6273 effect on the #warning and #error preprocessor directives and
6274 the deprecated attribute.
6276 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
6278 * tree.c (decl_value_expr_lookup): Revert latest change.
6279 (decl_value_expr_insert): Likewise.
6281 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
6283 * gimplify.c (nonlocal_vlas): Delete.
6284 (nonlocal_vla_vars): Likewise.
6285 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
6287 (gimplify_body): Do not create and destroy nonlocal_vlas.
6288 * tree-nested.c: Include diagnostic.h.
6289 (use_pointer_in_frame): Tweak.
6290 (lookup_field_for_decl): Add assertion and declare the transformation.
6291 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
6292 internal error when the reference is in a wrong context. Do not
6293 create a debug decl by default.
6294 (note_nonlocal_block_vlas): Delete.
6295 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
6296 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
6297 create a debug decl by default.
6298 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
6299 call is in a wrong context.
6300 (fixup_vla_decls): New function.
6301 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
6302 debug variables were created.
6303 * tree.c (decl_value_expr_lookup): Add checking assertion.
6304 (decl_value_expr_insert): Likewise.
6306 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
6309 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
6310 * tree-scalar-evolution.c (interpret_expr): Likewise.
6311 (expression_expensive_p): Likewise.
6312 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
6313 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
6314 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
6315 (ssa_defined_by_minus_one_stmt_p): New.
6317 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
6320 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
6321 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
6322 * dojump.c (do_jump): Likewise.
6323 * expr.c (expand_expr_real_2): Check operand type's sign.
6324 * fold-const.c (const_unop): Handle ABSU_EXPR.
6325 (fold_abs_const): Likewise.
6326 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
6327 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
6328 (strip_sign_op_1): Likesise.
6329 * match.pd: Add new pattern to generate ABSU_EXPR.
6330 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
6331 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
6332 * tree-eh.c (operation_could_trap_helper_p): Likewise.
6333 * tree-inline.c (estimate_operator_cost): Likewise.
6334 * tree-pretty-print.c (dump_generic_node): Likewise.
6335 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
6336 * tree.def (ABSU_EXPR): New.
6338 2018-06-16 Jakub Jelinek <jakub@redhat.com>
6341 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
6342 documented as preserved for backward compatibility only.
6343 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
6345 PR rtl-optimization/86108
6346 * bb-reorder.c (create_forwarder_block): Renamed to ...
6347 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
6348 jump from new landing pad to the second part.
6349 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
6352 2018-06-15 Jakub Jelinek <jakub@redhat.com>
6355 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
6356 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
6357 Only call store_expr for halves if the mode is the same.
6360 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
6361 Fix up comment formatting.
6363 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
6365 * typed-splay-tree.h (typed_splay_tree::remove): New function.
6366 (typed_splay_tree::closure,
6367 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
6368 (typed_splay_tree::typed_splay_tree,
6369 typed_splay_tree::operator =): Declared private.
6370 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
6371 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
6372 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
6373 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
6374 typed_splay_tree::splay_tree_splay,
6375 typed_splay_tree::splay_tree_foreach_helper,
6376 typed_splay_tree::splay_tree_insert,
6377 typed_splay_tree::splay_tree_remove,
6378 typed_splay_tree::splay_tree_lookup,
6379 typed_splay_tree::splay_tree_predecessor,
6380 typed_splay_tree::splay_tree_successor,
6381 typed_splay_tree::splay_tree_min,
6382 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
6383 (typed_splay_tree::root, typed_splay_tree::comp,
6384 typed_splay_tree::delete_key,
6385 typed_splay_tree::delete_value): New data members.
6386 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
6387 typed_splay_tree::remove.
6389 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
6391 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
6392 -mginv and -mno-ginv to the assembler.
6393 * config/mips/mips.opt (-mcrc): New option.
6395 * doc/invoke.text (-mcrc): Document.
6398 2018-06-15 Nick Clifton <nickc@redhat.com>
6401 * tree.c (escaped_string): New class. Converts an unescaped
6402 string into its escaped equivalent.
6403 (warn_deprecated_use): Use the new class to convert the
6404 deprecation message, if present.
6405 (test_escaped_strings): New self test.
6406 (test_c_tests): Add test_escaped_strings.
6407 * doc/extend.texi (deprecated): Add a note that the
6408 deprecation message is affected by the -fmessage-length
6409 option, and that control characters will be escaped.
6410 (#pragma GCC error): Document this pragma.
6411 (#pragma GCC warning): Likewise.
6412 * doc/invoke.texi (-fmessage-length): Document this option's
6413 effect on the #warning and #error preprocessor directives and
6414 the deprecated attribute.
6416 2018-06-15 Richard Biener <rguenther@suse.de>
6418 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
6419 here, also noting vector size used.
6420 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
6421 size used in MSG_OPTIMIZED_LOCATIONS dump.
6422 (pass_slp_vectorize::execute): Adjust.
6424 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
6427 * config/arc/arc.c (arc_return_address_register): Fix
6430 2018-06-15 Richard Biener <rguenther@suse.de>
6433 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
6434 leave useless conversion stripping to force_gimple_operand_gsi.
6435 (gimplify_build2): Likewise.
6436 (gimplify_build1): Likewise.
6438 2018-06-15 Richard Biener <rguenther@suse.de>
6441 * tree-cfg.c (move_stmt_op): unshare invariant addresses
6442 before adjusting their block.
6444 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
6446 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
6447 multilibs for *-*-rtems*.
6448 * config/riscv/t-rtems: New file.
6450 2018-06-14 Jakub Jelinek <jakub@redhat.com>
6453 * match.pd ((A +- CST1) +- CST2): Punt if last resort
6454 unsigned_type_for returns NULL.
6457 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
6458 subregs of multi-word pseudos unless the float mode has word size.
6460 2018-06-14 Richard Biener <rguenther@suse.de>
6463 * tree-vect-generic.c (build_word_mode_vector_type): Remove
6464 duplicate and harmful type_hash_canon.
6465 * tree.c (type_hash_canon): Assert we didn't find ourselves.
6467 2018-06-14 Richard Biener <rguenther@suse.de>
6470 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
6473 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
6475 * config/rtems.h (STDINT_LONG32): Define.
6477 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
6478 Prachi Godbole <prachi.godbole@imgtec.com>
6480 * config/mips/mips-cpus.def: Define P6600.
6481 * config/mips/mips-tables.opt: Regenerate.
6482 * config/mips/mips.c (mips_ucbranch_type): New enum.
6483 (mips_rtx_cost_data): Add support for P6600.
6484 (mips_issue_rate): Likewise.
6485 (mips_multipass_dfa_lookahead): Likewise.
6486 (mips_avoid_hazard): Likewise.
6487 (mips_reorg_process_insns): Likewise.
6488 (mips_classify_branch_p6600): New function.
6489 * config/mips/mips.h (TUNE_P6600): New define.
6490 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
6491 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
6492 * config/mips/mips.md: Include p6600.md.
6493 (processor): Add p6600.
6494 * config/mips/p6600.md: New file.
6495 * doc/invoke.texi: Add p6600 to supported architectures.
6497 2018-06-13 Martin Sebor <msebor@redhat.com>
6499 PR tree-optimization/86114
6500 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
6502 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
6504 2018-06-13 Richard Biener <rguenther@suse.de>
6506 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
6507 Properly set vector type of the intermediate stmt.
6508 * tree-vect-stmts.c (vectorizable_operation): The destination
6509 var always has vectype_out type.
6511 2018-06-13 Jeff Law <law@redhat.com>
6513 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
6514 integer 0 for argument to print_rtl_with_bb.
6515 (rl78_reorg): Likewise.
6517 2018-06-13 David Malcolm <dmalcolm@redhat.com>
6519 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
6520 from rtx to rtx_insn *.
6521 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
6523 (add_sched_insns_for_speculation): Likewise for local "target",
6524 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
6525 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
6526 from rtx_insn ** to rtx_code_label **.
6527 (reorg_emit_nops): Likewise.
6528 (c6x_reorg): Likewise for local "call_labels".
6529 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
6531 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
6532 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
6533 the loops over LABEL_REFS.
6534 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
6536 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
6537 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
6538 (split_branches): Strengthen local "olabel" from rtx to
6539 rtx_insn *, adding a safe_as_a cast.
6540 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
6542 (add_insn_after): Likewise for first two params.
6543 (add_insn_before): Likewise.
6544 (remove_insn): Likewise for param.
6545 (emit_pattern_before_noloc): Likewise for second and third params.
6546 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
6547 (emit_call_insn_before_noloc): Likewise.
6548 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
6550 (emit_barrier_before): Likewise.
6551 (emit_label_before): Strengthen "label" param from "rtx" to
6552 "rtx_code_label *". Strengthen "before" param from "rtx" to
6554 (emit_insn_after_1): Strengthen "after" param from "rtx" to
6556 (emit_pattern_after_noloc): Likewise.
6557 (emit_insn_after_noloc): Likewise.
6558 (emit_jump_insn_after_noloc): Likewise.
6559 (emit_call_insn_after_noloc): Likewise.
6560 (emit_debug_insn_after_noloc): Likewise.
6561 (emit_barrier_after): Likewise.
6562 (emit_label_after): Likewise for both params.
6563 (emit_pattern_after_setloc): Likewise for "after" param. Convert
6564 "loc" param from "int" to "location_t".
6565 (emit_insn_after_setloc): Likewise.
6566 (emit_jump_insn_after_setloc): Likewise.
6567 (emit_call_insn_after_setloc): Likewise.
6568 (emit_debug_insn_after_setloc): Likewise.
6569 (emit_pattern_before_setloc): Likewise for "before" param. Convert
6570 "loc" param from "int" to "location_t".
6571 (emit_pattern_before): Convert NULL_RTX to NULL.
6572 (emit_insn_before_setloc): Convert "loc" param from "int" to
6574 (emit_jump_insn_before_setloc): Likewise.
6575 (emit_call_insn_before_setloc): Likewise.
6576 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
6577 rtx_insn *. Convert "loc" param from "int" to "location_t".
6578 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
6579 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
6580 Convert 3rd param from "int" to "location_t".
6581 (emit_barrier_before, emit_barrier_after, next_real_insn):
6582 Strengthen param from rtx to rtx_insn *.
6583 (emit_label_before): Strengthen 1st param from "rtx" to
6584 "rtx_code_label *". Strengthen 2nd param from "rtx" to
6586 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
6587 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
6588 Strengthen 2nd param from "rtx" to "rtx_insn *".
6589 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
6590 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
6591 Likewise. Convert 3rd param from "int" to "location_t".
6592 (emit_label_after): Strengthen 1st param from "rtx" to
6594 (next_real_insn, remove_insn): Strengthen param from "rtx" to
6596 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
6597 from "rtx" to "rtx_insn *".
6599 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
6601 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
6602 bodies streamed in with -Q.
6603 * dumpfile.c (dump_files): Add lto-stream-out dump file.
6604 * dumpfile.h (tree_dump_index): Add lto_stream_out.
6605 * gimple-streamer-out.c: Include gimple-pretty-print.h
6606 (output_bb): Dump stmts streamed.
6607 * lto-section-out.c: Include print-tree.h
6608 (lto_begin_section): Dump sections created.
6609 (lto_output_decl_index): Dump decl encoded.
6610 * lto-streamer-out.c: Include print-tree.h
6611 (create_output_block): Dump output block created.
6612 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
6613 (output_function): Dump function output.
6614 (output_constructor): Dump constructor streamed.
6615 (write_global_stream): Output indexes encoded.
6616 (produce_asm_for_decls): Dump streams encoded.
6617 * lto-streamer.c (streamer_dump_file): New global var.
6618 * lto-streamer.h (streamer_dump_file): Declare.
6619 * passes.c (ipa_write_summaries): Initialize streamer dump.
6620 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
6623 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
6626 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
6627 offsets for register save directives. Emit a second batch of save
6628 directives, if need be, when the function accesses prior frames.
6630 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
6632 * config/arc/fpu.md (fmasf4): Force operand to register.
6633 (fnmasf4): Likewise.
6635 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
6637 * config/arc/arc-protos.h (arc_pad_return): Remove.
6638 * config/arc/arc.c (machine_function): Remove force_short_suffix
6640 (arc_print_operand): Adjust printing of '&'.
6641 (arc_verify_short): Remove conditional printing of short suffix.
6642 (arc_final_prescan_insn): Remove reference to size_reason.
6643 (pad_return): New function.
6644 (arc_reorg): Call pad_return.
6645 (arc_pad_return): Remove.
6646 (arc_init_machine_status): Remove reference to force_short_suffix.
6647 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
6648 (attr length): When attribute iscompact is true force to 2
6649 regardless; in the case of maybe check if we want to force the
6650 instruction to have 4 bytes length.
6651 (nopv): Change it to generate 4 byte long nop as well.
6652 (blockage): New pattern.
6653 (simple_return): Remove call to arc_pad_return.
6654 (p_return_i): Likewise.
6656 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
6658 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
6660 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
6662 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
6665 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
6667 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
6668 for ARC700 and ARCv2.
6670 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
6673 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
6674 operands[2] instead of operands[1].
6677 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
6679 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
6680 case, check whether the outer register overlaps an unallocatable
6681 register, not just whether it fits the required class.
6683 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
6685 * poly-int.h (can_div_trunc_p): Add new overload in which all values
6687 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
6688 (memrefs_conflict_p): Likewise.
6689 (init_alias_analysis): Likewise.
6690 * cfgexpand.c (expand_debug_expr): Likewise.
6691 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
6692 * cse.c (fold_rtx): Likewise.
6693 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
6694 * expr.c (emit_block_move_hints): Likewise.
6695 (clear_storage_hints, push_block, emit_push_insn): Likewise.
6696 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
6697 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
6698 (emit_group_store): Likewise.
6699 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
6700 to read the PRE/POST_MODIFY increment.
6701 * calls.c (store_one_arg): Use strip_offset.
6702 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
6704 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
6706 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
6707 (simplify_binary_operation_1): Extend CONST_INT handling to
6709 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
6710 than a HOST_WIDE_INT.
6711 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
6713 (adjust_mems, add_stores): Update accodingly.
6714 (vt_canonicalize_addr): Track polynomial offsets.
6715 (emit_note_insn_var_location): Likewise.
6716 (vt_add_function_parameter): Likewise.
6717 (vt_initialize): Likewise.
6719 2018-06-12 Jeff Law <law@redhat.com>
6721 * config.gcc (alpha*-*-freebsd*): Remove.
6722 * config/alpha/freebsd.h: Remove.
6724 2018-06-12 David Malcolm <dmalcolm@redhat.com>
6727 * spellcheck-tree.c (levenshtein_distance): Rename to...
6728 (get_edit_distance): ...this, and update for underlying renaming.
6729 * spellcheck-tree.h (levenshtein_distance): Rename to...
6730 (get_edit_distance): ...this.
6731 * spellcheck.c (levenshtein_distance): Rename to...
6732 (get_edit_distance): ...this. Convert from Levenshtein distance
6733 to Damerau-Levenshtein distance by supporting transpositions of
6734 adjacent characters. Rename "v1" to "v_next" and "v0" to
6736 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
6737 (selftest::test_edit_distance_unit_test_oneway): ...this, and
6738 update for underlying renaming.
6739 (selftest::levenshtein_distance_unit_test): Rename to...
6740 (selftest::test_get_edit_distance_unit): ...this, and update for
6741 underlying renaming.
6742 (selftest::test_find_closest_string): Add example from PR 69968
6743 where transposition helps
6744 (selftest::test_metric_conditions): Update for renaming.
6745 (selftest::test_metric_conditions): Likewise.
6746 (selftest::spellcheck_c_tests): Likewise.
6747 * spellcheck.h (levenshtein_distance): Rename both overloads to...
6748 (get_edit_distance): ...this.
6749 (best_match::consider): Update for renaming.
6751 2018-06-12 Martin Sebor <msebor@redhat.com>
6753 PR tree-optimization/85259
6754 * builtins.c (compute_objsize): Handle constant offsets.
6755 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
6756 true iff a warning has been issued.
6757 * gimple.h (gimple_nonartificial_location): New function.
6758 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
6759 gimple_nonartificial_location and handle -Wno-system-headers.
6760 (handle_builtin_stxncpy): Same.
6762 2018-06-12 Martin Sebor <msebor@redhat.com>
6765 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
6767 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
6769 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
6770 BUILTIN_VEC_XST entries for pointer to double and long long.
6772 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
6775 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
6777 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
6780 2018-06-12 Martin Liska <mliska@suse.cz>
6782 * doc/options.texi: Document IntegerRange.
6784 2018-06-12 Martin Liska <mliska@suse.cz>
6786 * config/i386/i386.opt: Make MPX-related options as Deprecated.
6787 * opt-functions.awk: Handle Deprecated flag.
6788 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
6790 (read_cmdline_option): Report warning for a deprecated option.
6791 * opts.h (struct cl_option): Add new field cl_deprecated.
6792 (CL_ERR_DEPRECATED): New.
6794 2018-06-12 Martin Liska <mliska@suse.cz>
6796 * doc/options.texi: Document Deprecated option flag.
6798 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
6800 * config/arc/arc-arch.h (arc_extras): New enum.
6801 (arc_cpu_t):Add field extra.
6802 (arc_cpu_types): Consider the extras.
6803 * config/arc/arc-cpus.def: Add extras info.
6804 * config/arc/arc-opts.h (processor_type): Consider extra field.
6805 * config/arc/arc.c (arc_override_options): Handle extra field.
6807 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
6809 * config/arc/arc-arch.h: Update ARC_OPTX macro.
6810 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
6812 * config/arc/arc.c (arc_init): Update pic warning.
6813 (irq_range): Update irq range parsing warnings.
6814 (arc_override_options): Update various warning messages.
6815 (arc_handle_aux_attribute): Likewise.
6817 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
6819 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
6821 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
6823 * doc/sourcebuild.texi: Document usage of line number 0 in verify
6824 compiler messages directives.
6826 2018-06-12 Matthew Fortune <mfortune@gmail.com>
6828 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
6829 * config/mips/mips-tables.opt: Regenerate.
6830 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
6832 * doc/invoke.texi: Document -march=i6500.
6834 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
6836 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
6837 (i6400_gpmul): Add cpu_unit.
6838 (i6400_gpdiv): Likewise.
6839 (i6400_msa_add_d): Update reservations.
6840 (i6400_msa_int_add) Likewise.
6841 (i6400_msa_short_logic3) Likewise.
6842 (i6400_msa_short_logic2) Likewise.
6843 (i6400_msa_short_logic) Likewise.
6844 (i6400_msa_move) Likewise.
6845 (i6400_msa_cmp) Likewise.
6846 (i6400_msa_short_float2) Likewise.
6847 (i6400_msa_div_d) Likewise.
6848 (i6400_msa_long_logic1) Likewise.
6849 (i6400_msa_long_logic2) Likewise.
6850 (i6400_msa_mult) Likewise.
6851 (i6400_msa_long_float2) Likewise.
6852 (i6400_msa_long_float4) Likewise.
6853 (i6400_msa_long_float5) Likewise.
6854 (i6400_msa_long_float8) Likewise.
6855 (i6400_fpu_fadd): Include frint type.
6856 (i6400_fpu_store): New define_insn_reservation.
6857 (i6400_fpu_load): Likewise.
6858 (i6400_fpu_move): Likewise.
6859 (i6400_fpu_fcmp): Likewise.
6860 (i6400_fpu_fmadd): Likewise.
6861 (i6400_int_mult): Include imul3nc type and update reservation.
6862 (i6400_int_div): Include idiv3 type and update reservation.
6863 (i6400_int_load): Update to check type not move_type.
6864 (i6400_int_store): Likewise.
6865 (i6400_int_prefetch): Set zero latency.
6867 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
6869 * gcc.c: Document new %@{...} sequence.
6870 (LINK_COMMAND_SPEC): Use it for the -L switches.
6871 (cpp_unique_options): Use it for the -I switches.
6872 (at_file_argbuf): New global variable.
6873 (in_at_file): Likewise.
6874 (alloc_args): Create at_file_argbuf.
6875 (clear_args): Truncate at_file_argbuf.
6876 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
6877 (open_at_file): New function.
6878 (close_at_file): Likewise.
6879 (create_at_file): Delete.
6880 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
6883 (validate_switches_from_spec): Deal with %@{...} sequence.
6884 (validate_switches): Likewise.
6885 (driver::finalize): Call clear_args.
6887 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
6889 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
6891 2018-06-11 Martin Sebor <msebor@redhat.com>
6893 * doc/invoke.texi (-Wall): List -Wc++17-compat.
6894 (Wno-class-memaccess): Add @opindex.
6895 (Wno-templates, Wno-multiple-inheritance): Same.
6896 (Wno-virtual-inheritance, Wno-namespaces): Same.
6897 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
6898 (Wno-format-overflow, Wno-format-truncation): Same.
6899 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
6900 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
6901 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
6902 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
6903 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
6904 (Wno-misspelled-isr): Same.
6906 2018-06-11 Martin Sebor <msebor@redhat.com>
6908 * PR tree-optimization/86083
6909 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
6911 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
6913 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
6915 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
6918 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
6919 on the correct operand.
6920 (*movdi_internal64): Ditto.
6922 2018-06-11 Martin Liska <mliska@suse.cz>
6924 PR tree-optimization/86089
6925 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
6927 2018-06-11 Julia Koval <julia.koval@intel.com>
6929 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
6930 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
6931 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
6933 2018-06-11 Olivier Hainque <hainque@adacore.com>
6935 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
6936 for Ada with strict dwarf2.
6938 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
6941 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
6944 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
6946 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
6948 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
6950 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
6953 2018-06-08 Martin Liska <mliska@suse.cz>
6955 * tree-cfg.h (debug_function): Fix argument type to match
6958 2018-06-08 Martin Liska <mliska@suse.cz>
6960 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
6961 Remove usage of MPX-related (and removed) fields.
6962 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
6964 2018-06-08 David Malcolm <dmalcolm@redhat.com>
6966 * cfg.c (debug): Use TDF_NONE rather than 0.
6967 * cfghooks.c (debug): Likewise.
6968 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
6969 (struct dump_option_value_info): Convert to...
6970 (struct kv_pair): ...this template type.
6971 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
6973 (optinfo_verbosity_options): Likewise.
6974 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
6976 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
6977 than int for "optgroup_flags" param.
6978 (dump_generic_expr_loc): Use dump_flags_t rather than int for
6980 (dump_dec): Likewise.
6981 (dump_finish): Use TDF_NONE rather than 0.
6982 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
6983 rather than int for "optgroup_flags" param. Use TDF_NONE rather
6984 than 0. Update for change to option_ptr.
6985 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
6986 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
6987 0. Update for changes to optinfo_verbosity_options and
6989 (opt_info_switch_p): Convert optgroup_flags from int to
6991 (dump_basic_block): Use dump_flags_t rather than int
6992 for "dump_kind" param.
6993 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
6994 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
6995 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
6996 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
6997 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
6998 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
6999 TDF_NONE): Convert from macros to...
7000 (enum dump_flag): ...this new enum.
7001 (dump_flags_t): Update to use enum.
7002 (operator|, operator&, operator~, operator|=, operator&=):
7003 Implement for dump_flags_t.
7004 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
7005 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
7006 Convert from macros to...
7007 (enum optgroup_flag): ...this new enum.
7008 (optgroup_flags_t): New typedef.
7009 (operator|, operator|=): Implement for optgroup_flags_t.
7010 (struct dump_file_info): Convert field "alt_flags" to
7011 dump_flags_t. Convert field "optgroup_flags" to
7013 (dump_basic_block): Use dump_flags_t rather than int for param.
7014 (dump_generic_expr_loc): Likewise.
7015 (dump_dec): Likewise.
7016 (dump_register): Convert param "optgroup_flags" to
7018 (opt_info_enable_passes): Likewise.
7019 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
7021 * gimple-pretty-print.c (debug): Likewise.
7022 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
7023 (merged_store_group::apply_stores): Likewise.
7024 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
7025 * gimple.c (verify_gimple_pp): Likewise.
7026 * graphite-poly.c (print_pbb_body): Likewise.
7027 * passes.c (pass_manager::register_one_dump_file): Convert
7028 local "optgroup_flags" to optgroup_flags_t.
7029 * print-tree.c (print_node): Use TDF_NONE rather than 0.
7031 (debug_body): Likewise.
7032 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
7033 to optgroup_flags_t.
7034 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
7036 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
7037 (convert_mult_to_fma): Likewise.
7038 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
7039 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
7040 * tree-vect-data-refs.c (dump_lower_bound): Convert param
7041 "dump_kind" to dump_flags_t.
7043 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
7045 * config/rs6000/rs6000.c (min, max): Delete.
7047 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
7049 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
7050 -mabi=spe and -mabi=no-spe.
7052 2018-06-08 Martin Liska <mliska@suse.cz>
7054 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
7055 where we expect an existing summary.
7057 2018-06-08 Martin Liska <mliska@suse.cz>
7059 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
7060 * ipa-inline.h (estimate_edge_growth): Likewise.
7062 2018-06-08 Martin Liska <mliska@suse.cz>
7064 * cgraph.c (function_version_hasher::hash): Use
7065 cgraph_node::get_uid ().
7066 (function_version_hasher::equal):
7067 * cgraph.h (cgraph_node::get_uid): New method.
7068 * ipa-inline.c (update_caller_keys): Use
7069 cgraph_node::get_uid ().
7070 (update_callee_keys): Likewise.
7071 * ipa-utils.c (searchc): Likewise.
7072 (ipa_reduced_postorder): Likewise.
7073 * lto-cgraph.c (input_node): Likewise.
7074 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
7075 * symbol-summary.h (symtab_insertion): Likewise.
7076 (symtab_removal): Likewise.
7077 (symtab_duplication): Likewise.
7078 * tree-pretty-print.c (dump_function_header): Likewise.
7079 * tree-sra.c (convert_callers_for_node): Likewise.
7081 2018-06-08 Martin Liska <mliska@suse.cz>
7083 * cgraph.c (symbol_table::create_edge): Always assign a new
7085 (symbol_table::free_edge): Do not recycle numbers.
7086 * cgraph.h (cgraph_edge::get): New method.
7087 * symbol-summary.h (symtab_removal): Use it.
7088 (symtab_duplication): Likewise.
7089 (call_summary::hashable_uid): Remove.
7091 2018-06-08 Martin Liska <mliska@suse.cz>
7093 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
7094 (initialize_growth_caches): Remove.
7095 (free_growth_caches): Likewise.
7096 (do_estimate_edge_time): Use edge_growth_cache.
7097 (do_estimate_edge_size): Likewise.
7098 (do_estimate_edge_hints): Likewise.
7099 * ipa-inline.c (reset_edge_caches): Likewise.
7100 (recursive_inlining): Likewise.
7101 (inline_small_functions): Likewise.
7102 * ipa-inline.h (initialize_growth_caches): Remove.
7103 (estimate_edge_size): Likewise.
7104 (estimate_edge_time): Likewise.
7105 (estimate_edge_hints): Likewise.
7106 (reset_edge_growth_cache): Likewise.
7107 * symbol-summary.h (call_summary::remove): New method.
7109 2018-06-08 Martin Liska <mliska@suse.cz>
7111 * ipa-cp.c (class edge_clone_summary): New summary.
7112 (grow_edge_clone_vectors): Remove.
7113 (ipcp_edge_duplication_hook): Remove.
7114 (class edge_clone_summary_t): New call_summary class.
7115 (ipcp_edge_removal_hook): Remove.
7116 (edge_clone_summary_t::duplicate): New function.
7117 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
7118 (create_specialized_node): Likewise.
7119 (ipcp_driver): Initialize edge_clone_summaries and do not
7122 2018-06-08 Martin Liska <mliska@suse.cz>
7124 * symbol-summary.h (get): New function.
7125 (call_summary::m_initialize_when_cloning): New class member.
7127 2018-06-08 Martin Liska <mliska@suse.cz>
7129 * cgraph.c (cgraph_node::remove): Do not recycle uid.
7130 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
7131 (symbol_table::allocate_cgraph_symbol): Do not set uid.
7132 * passes.c (uid_hash_t): Record removed_nodes by their uids.
7133 (remove_cgraph_node_from_order): Use the removed_nodes set.
7134 (do_per_function_toporder): Likwise.
7135 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
7136 instead of summary_uid.
7137 (symtab_removal): Likewise.
7138 (symtab_duplication): Likewise.
7140 2018-06-08 Martin Liska <mliska@suse.cz>
7142 * ipa-cp.c (ipcp_store_bits_results): Use
7143 ipcp_transformation_sum.
7144 (ipcp_store_vr_results): Likewise.
7145 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
7147 (ipcp_transformation_initialize): ... this.
7148 (ipa_set_node_agg_value_chain):
7149 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
7150 (write_ipcp_transformation_info): Likewise.
7151 (read_ipcp_transformation_info): Likewise.
7152 (ipcp_update_bits): Likewise.
7153 (ipcp_update_vr): Likewise.
7154 (ipcp_transform_function): Likewise.
7155 * ipa-prop.h: Rename ipcp_transformation_summary to
7156 ipcp_transformation.
7157 (class ipcp_transformation_t): New function summary.
7158 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
7159 (ipa_get_agg_replacements_for_node): Likewise.
7161 2018-06-08 Martin Liska <mliska@suse.cz>
7163 * ipa-pure-const.c (struct funct_state_d): Do it class instead
7165 (class funct_state_summary_t): New function_summary class.
7166 (has_function_state): Remove.
7167 (get_function_state): Likewise.
7168 (set_function_state): Likewise.
7169 (add_new_function): Likewise.
7170 (funct_state_summary_t::insert): New function.
7171 (duplicate_node_data): Remove.
7172 (remove_node_data): Remove.
7173 (funct_state_summary_t::duplicate): New function.
7174 (register_hooks): Create new funct_state_summaries.
7175 (pure_const_generate_summary): Use it.
7176 (pure_const_write_summary): Likewise.
7177 (pure_const_read_summary): Likewise.
7178 (propagate_pure_const): Likewise.
7179 (propagate_nothrow): Likewise.
7180 (dump_malloc_lattice): Likewise.
7181 (propagate_malloc): Likewise.
7182 (execute): Do not register hooks, just remove summary
7184 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
7187 2018-06-08 Martin Liska <mliska@suse.cz>
7189 * ipa-reference.c (remove_node_data): Remove.
7190 (duplicate_node_data): Likewise.
7191 (class ipa_ref_var_info_summary_t): New class.
7192 (class ipa_ref_opt_summary_t): Likewise.
7193 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
7194 (get_reference_optimization_summary): Use
7195 ipa_ref_opt_sum_summaries.
7196 (set_reference_vars_info): Remove.
7197 (set_reference_optimization_summary): Likewise.
7198 (ipa_init): Create summaries.
7199 (init_function_info): Use function summary.
7200 (ipa_ref_opt_summary_t::duplicate): New function.
7201 (ipa_ref_opt_summary_t::remove): New function.
7202 (get_read_write_all_from_node): Fix GNU coding style.
7203 (propagate): Use function summary.
7204 (write_node_summary_p): Fix GNU coding style.
7205 (stream_out_bitmap): Likewise.
7206 (ipa_reference_read_optimization_summary): Use function summary.
7207 (ipa_reference_c_finalize): Do not release hooks.
7209 2018-06-08 Martin Liska <mliska@suse.cz>
7211 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
7212 (analyze_function_body): Extract multiple calls of get_create.
7213 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
7214 * ipa-inline.c (recursive_inlining): Use ::get method.
7215 * ipa-inline.h (estimate_edge_growth): Likewise.
7217 2018-06-08 Martin Liska <mliska@suse.cz>
7219 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
7221 (hsa_function_summary::hsa_function_summary): Use the new enum
7223 (hsa_gpu_implementation_p): Use hsa_summaries::get.
7224 * hsa-gen.c (hsa_get_host_function): Likewise.
7225 (get_brig_function_name): Likewise.
7226 * ipa-hsa.c (process_hsa_functions): Likewise.
7227 (ipa_hsa_write_summary): Likewise.
7228 * symbol-summary.h (symtab_duplication): Use ::get function/
7229 (get): New function.
7231 2018-06-08 Martin Liska <mliska@suse.cz>
7233 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
7235 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
7236 (hsa_register_kernel): Likewise.
7237 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
7238 * hsa-gen.c (hsa_get_host_function): Likewise.
7239 (get_brig_function_name): Likewise.
7240 (generate_hsa): Likewise.
7241 (pass_gen_hsail::execute): Likewise.
7242 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
7243 (devirtualization_time_bonus): Likewise.
7244 (ipcp_propagate_stage): Likewise.
7245 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
7246 (edge_set_predicate): Likewise.
7247 (evaluate_conditions_for_known_args): Likewise.
7248 (evaluate_properties_for_edge): Likewise.
7249 (ipa_fn_summary::reset): Likewise.
7250 (ipa_fn_summary_t::duplicate): Likewise.
7251 (dump_ipa_call_summary): Likewise.
7252 (ipa_dump_fn_summary): Likewise.
7253 (analyze_function_body): Likewise.
7254 (compute_fn_summary): Likewise.
7255 (estimate_edge_devirt_benefit): Likewise.
7256 (estimate_edge_size_and_time): Likewise.
7257 (estimate_calls_size_and_time): Likewise.
7258 (estimate_node_size_and_time): Likewise.
7259 (inline_update_callee_summaries): Likewise.
7260 (remap_edge_change_prob): Likewise.
7261 (remap_edge_summaries): Likewise.
7262 (ipa_merge_fn_summary_after_inlining): Likewise.
7263 (ipa_update_overall_fn_summary): Likewise.
7264 (read_ipa_call_summary): Likewise.
7265 (inline_read_section): Likewise.
7266 (write_ipa_call_summary): Likewise.
7267 (ipa_fn_summary_write): Likewise.
7268 (ipa_free_fn_summary): Likewise.
7269 * ipa-hsa.c (process_hsa_functions): Likewise.
7270 (ipa_hsa_write_summary): Likewise.
7271 (ipa_hsa_read_section): Likewise.
7272 * ipa-icf.c (sem_function::merge): Likewise.
7273 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
7274 (do_estimate_edge_time): Likewise.
7275 (estimate_size_after_inlining): Likewise.
7276 (estimate_growth): Likewise.
7277 (growth_likely_positive): Likewise.
7278 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
7279 (inline_call): Likewise.
7280 * ipa-inline.c (caller_growth_limits): Likewise.
7281 (can_inline_edge_p): Likewise.
7282 (can_inline_edge_by_limits_p): Likewise.
7283 (compute_uninlined_call_time): Likewise.
7284 (compute_inlined_call_time): Likewise.
7285 (want_inline_small_function_p): Likewise.
7286 (edge_badness): Likewise.
7287 (update_caller_keys): Likewise.
7288 (update_callee_keys): Likewise.
7289 (recursive_inlining): Likewise.
7290 (inline_small_functions): Likewise.
7291 (inline_to_all_callers_1): Likewise.
7292 (dump_overall_stats): Likewise.
7293 (early_inline_small_functions): Likewise.
7294 (early_inliner): Likewise.
7295 * ipa-inline.h (estimate_edge_growth): Likewise.
7296 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
7297 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
7298 * ipa-prop.h (IPA_NODE_REF): Likewise.
7299 (IPA_EDGE_REF): Likewise.
7300 * ipa-pure-const.c (malloc_candidate_p): Likewise.
7301 (propagate_malloc): Likewise.
7302 * ipa-split.c (execute_split_functions): Likewise.
7303 * symbol-summary.h: Rename get to get_create.
7305 (get_create): Likewise.
7306 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
7308 2018-06-08 Martin Liska <mliska@suse.cz>
7310 * symbol-summary.h (release): Move definition out of class
7312 (symtab_removal): Likewise.
7313 (symtab_duplication): Likewise.
7315 2018-06-08 Martin Liska <mliska@suse.cz>
7317 * symbol-summary.h (function_summary): Move constructor
7318 implementation out of class declaration.
7319 (release): Likewise.
7320 (symtab_insertion): Likewise.
7321 (symtab_removal): Likewise.
7322 (symtab_duplication): Likewise.
7325 2018-06-08 Martin Liska <mliska@suse.cz>
7327 * Makefile.in: Remove support for MPX (macros, related functions,
7328 fields in cgraph_node, ...).
7329 * builtin-types.def (BT_BND): Likewise.
7330 (BT_FN_BND_CONST_PTR): Likewise.
7331 (BT_FN_CONST_PTR_BND): Likewise.
7332 (BT_FN_VOID_PTR_BND): Likewise.
7333 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
7334 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
7335 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
7336 (expand_builtin_mempcpy_with_bounds): Likewise.
7337 (expand_builtin_memset_with_bounds): Likewise.
7338 (expand_builtin_memset_args): Likewise.
7339 (std_expand_builtin_va_start): Likewise.
7340 (expand_builtin): Likewise.
7341 (expand_builtin_with_bounds): Likewise.
7342 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
7343 (DEF_LIB_BUILTIN_CHKP): Likewise.
7344 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
7345 (DEF_CHKP_BUILTIN): Likewise.
7346 (BUILT_IN_MEMCPY): Likewise.
7347 (BUILT_IN_MEMMOVE): Likewise.
7348 (BUILT_IN_MEMPCPY): Likewise.
7349 (BUILT_IN_MEMSET): Likewise.
7350 (BUILT_IN_STPCPY): Likewise.
7351 (BUILT_IN_STRCAT): Likewise.
7352 (BUILT_IN_STRCHR): Likewise.
7353 (BUILT_IN_STRCPY): Likewise.
7354 (BUILT_IN_STRLEN): Likewise.
7355 (BUILT_IN_MEMCPY_CHK): Likewise.
7356 (BUILT_IN_MEMMOVE_CHK): Likewise.
7357 (BUILT_IN_MEMPCPY_CHK): Likewise.
7358 (BUILT_IN_MEMSET_CHK): Likewise.
7359 (BUILT_IN_STPCPY_CHK): Likewise.
7360 (BUILT_IN_STRCAT_CHK): Likewise.
7361 (BUILT_IN_STRCPY_CHK): Likewise.
7362 * calls.c (store_bounds): Likewise.
7363 (emit_call_1): Likewise.
7364 (special_function_p): Likewise.
7365 (maybe_warn_nonstring_arg): Likewise.
7366 (initialize_argument_information): Likewise.
7367 (finalize_must_preallocate): Likewise.
7368 (compute_argument_addresses): Likewise.
7369 (expand_call): Likewise.
7370 * cfgexpand.c (expand_call_stmt): Likewise.
7371 (expand_return): Likewise.
7372 (expand_gimple_stmt_1): Likewise.
7373 (pass_expand::execute): Likewise.
7374 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
7375 (cgraph_node::remove): Likewise.
7376 (cgraph_node::dump): Likewise.
7377 (cgraph_node::verify_node): Likewise.
7378 * cgraph.h (chkp_function_instrumented_p): Likewise.
7379 (symtab_node::get_alias_target): Likewise.
7380 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
7381 (cgraph_local_p): Likewise.
7382 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
7383 (cgraph_edge::rebuild_references): Likewise.
7384 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
7385 (walk_polymorphic_call_targets): Likewise.
7386 (cgraph_node::expand_thunk): Likewise.
7387 (symbol_table::output_weakrefs): Likewise.
7388 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
7389 (ix86_handle_option): Likewise.
7390 * config/i386/constraints.md: Likewise.
7391 * config/i386/i386-builtin-types.def (BND): Likewise.
7395 * config/i386/i386-builtin.def (BDESC_END): Likewise.
7396 (BDESC_FIRST): Likewise.
7398 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
7399 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
7400 * config/i386/i386.c (enum reg_class): Likewise.
7401 (ix86_target_string): Likewise.
7402 (ix86_option_override_internal): Likewise.
7403 (ix86_conditional_register_usage): Likewise.
7404 (ix86_valid_target_attribute_inner_p): Likewise.
7405 (ix86_set_indirect_branch_type): Likewise.
7406 (ix86_set_current_function): Likewise.
7407 (ix86_function_arg_regno_p): Likewise.
7408 (init_cumulative_args): Likewise.
7409 (ix86_function_arg_advance): Likewise.
7410 (ix86_function_arg): Likewise.
7411 (ix86_pass_by_reference): Likewise.
7412 (ix86_function_value_regno_p): Likewise.
7413 (ix86_function_value_1): Likewise.
7414 (ix86_function_value_bounds): Likewise.
7415 (ix86_return_in_memory): Likewise.
7416 (ix86_setup_incoming_vararg_bounds): Likewise.
7417 (ix86_va_start): Likewise.
7418 (indirect_thunk_need_prefix): Likewise.
7419 (print_reg): Likewise.
7420 (ix86_print_operand): Likewise.
7421 (ix86_expand_call): Likewise.
7422 (ix86_output_function_return): Likewise.
7423 (reg_encoded_number): Likewise.
7424 (BDESC_VERIFYS): Likewise.
7425 (ix86_init_mpx_builtins): Likewise.
7426 (ix86_init_builtins): Likewise.
7427 (ix86_emit_cmove): Likewise.
7428 (ix86_emit_move_max): Likewise.
7429 (ix86_expand_builtin): Likewise.
7430 (ix86_builtin_mpx_function): Likewise.
7431 (ix86_get_arg_address_for_bt): Likewise.
7432 (ix86_load_bounds): Likewise.
7433 (ix86_store_bounds): Likewise.
7434 (ix86_load_returned_bounds): Likewise.
7435 (ix86_store_returned_bounds): Likewise.
7436 (ix86_class_likely_spilled_p): Likewise.
7437 (ix86_hard_regno_mode_ok): Likewise.
7438 (x86_order_regs_for_local_alloc): Likewise.
7439 (ix86_mitigate_rop): Likewise.
7440 (ix86_bnd_prefixed_insn_p): Likewise.
7441 (ix86_mpx_bound_mode): Likewise.
7442 (ix86_make_bounds_constant): Likewise.
7443 (ix86_initialize_bounds): Likewise.
7444 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
7445 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
7446 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
7447 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
7448 (TARGET_CHKP_BOUND_MODE): Likewise.
7449 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
7450 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
7451 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
7452 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
7453 * config/i386/i386.h (TARGET_MPX): Likewise.
7454 (TARGET_MPX_P): Likewise.
7455 (VALID_BND_REG_MODE): Likewise.
7456 (FIRST_BND_REG): Likewise.
7457 (LAST_BND_REG): Likewise.
7458 (enum reg_class): Likewise.
7459 (BND_REG_P): Likewise.
7460 (BND_REGNO_P): Likewise.
7461 (BNDmode): Likewise.
7462 (ADJUST_INSN_LENGTH): Likewise.
7463 * config/i386/i386.md: Likewise.
7464 * config/i386/i386.opt: Likewise.
7465 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
7466 (defined): Likewise.
7467 (LINK_MPX): Likewise.
7468 (MPX_SPEC): Likewise.
7469 (LIBMPX_SPEC): Likewise.
7470 (LIBMPXWRAPPERS_SPEC): Likewise.
7471 (CHKP_SPEC): Likewise.
7472 * config/i386/predicates.md: Likewise.
7473 * dbxout.c (dbxout_type): Likewise.
7474 * doc/extend.texi: Likewise.
7475 * doc/invoke.texi: Likewise.
7476 * doc/md.texi: Likewise.
7477 * doc/tm.texi: Likewise.
7478 * doc/tm.texi.in: Likewise.
7479 * dwarf2out.c (is_base_type): Likewise.
7480 (gen_formal_types_die): Likewise.
7481 (gen_subprogram_die): Likewise.
7482 (gen_type_die_with_usage): Likewise.
7483 (gen_decl_die): Likewise.
7484 (dwarf2out_late_global_decl): Likewise.
7485 * expr.c (expand_assignment): Likewise.
7486 (emit_storent_insn): Likewise.
7487 (store_expr_with_bounds): Likewise.
7488 (store_expr): Likewise.
7489 (expand_expr_real_1): Likewise.
7490 * expr.h (store_expr_with_bounds): Likewise.
7491 * function.c (use_register_for_decl): Likewise.
7492 (struct bounds_parm_data): Likewise.
7493 (assign_parms_augmented_arg_list): Likewise.
7494 (assign_parm_find_entry_rtl): Likewise.
7495 (assign_parm_is_stack_parm): Likewise.
7496 (assign_parm_load_bounds): Likewise.
7497 (assign_bounds): Likewise.
7498 (assign_parms): Likewise.
7499 (expand_function_start): Likewise.
7500 * gcc.c (CHKP_SPEC): Likewise.
7501 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
7502 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
7503 (wrestrict_dom_walker::check_call): Likewise.
7504 * gimple.c (gimple_build_call_from_tree): Likewise.
7505 * gimple.h (enum gf_mask): Likewise.
7506 (gimple_call_with_bounds_p): Likewise.
7507 (gimple_call_set_with_bounds): Likewise.
7508 * gimplify.c (gimplify_init_constructor): Likewise.
7509 * ipa-cp.c (initialize_node_lattices): Likewise.
7510 (propagate_constants_across_call): Likewise.
7511 (find_more_scalar_values_for_callers_subset): Likewise.
7512 * ipa-hsa.c (process_hsa_functions): Likewise.
7513 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
7514 * ipa-icf.c (sem_function::merge): Likewise.
7515 * ipa-inline.c (early_inliner): Likewise.
7516 * ipa-pure-const.c (warn_function_noreturn): Likewise.
7517 (warn_function_cold): Likewise.
7518 (propagate_pure_const): Likewise.
7519 * ipa-ref.h (enum GTY): Likewise.
7520 * ipa-split.c (find_retbnd): Likewise.
7521 (consider_split): Likewise.
7522 (split_function): Likewise.
7523 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
7524 * ipa.c (walk_polymorphic_call_targets): Likewise.
7525 (symbol_table::remove_unreachable_nodes): Likewise.
7526 (process_references): Likewise.
7527 (cgraph_build_static_cdtor_1): Likewise.
7528 * lto-cgraph.c (lto_output_node): Likewise.
7529 (output_refs): Likewise.
7530 (compute_ltrans_boundary): Likewise.
7531 (input_overwrite_node): Likewise.
7532 (input_node): Likewise.
7533 (input_cgraph_1): Likewise.
7534 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
7535 * passes.c (pass_manager::execute_early_local_passes): Likewise.
7536 (class pass_chkp_instrumentation_passes): Likewise.
7537 (make_pass_chkp_instrumentation_passes): Likewise.
7538 * passes.def: Likewise.
7539 * rtl.h (struct GTY): Likewise.
7540 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
7541 * stor-layout.c (layout_type): Likewise.
7542 * symtab.c: Likewise.
7543 * target.def: Likewise.
7544 * targhooks.c (default_chkp_bound_type): Likewise.
7545 (default_chkp_bound_mode): Likewise.
7546 (default_builtin_chkp_function): Likewise.
7547 (default_chkp_function_value_bounds): Likewise.
7548 (default_chkp_make_bounds_constant): Likewise.
7549 (default_chkp_initialize_bounds): Likewise.
7550 * targhooks.h (default_chkp_bound_type): Likewise.
7551 (default_chkp_bound_mode): Likewise.
7552 (default_builtin_chkp_function): Likewise.
7553 (default_chkp_function_value_bounds): Likewise.
7554 (default_chkp_make_bounds_constant): Likewise.
7555 (default_chkp_initialize_bounds): Likewise.
7556 * toplev.c (compile_file): Likewise.
7557 (process_options): Likewise.
7558 * tree-core.h (DEF_BUILTIN): Likewise.
7559 (DEF_BUILTIN_CHKP): Likewise.
7560 * tree-inline.c (declare_return_variable): Likewise.
7561 (remap_gimple_stmt): Likewise.
7562 (copy_bb): Likewise.
7563 (initialize_inlined_parameters): Likewise.
7564 (expand_call_inline): Likewise.
7565 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
7566 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
7567 (make_pass_ipa_chkp_produce_thunks): Likewise.
7568 (make_pass_chkp): Likewise.
7569 (make_pass_chkp_opt): Likewise.
7570 (make_pass_chkp_instrumentation_passes): Likewise.
7571 * tree-pretty-print.c (dump_generic_node): Likewise.
7572 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
7573 * tree-ssa-dce.c (propagate_necessity): Likewise.
7574 (eliminate_unnecessary_stmts): Likewise.
7575 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
7576 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
7577 * tree-ssa-sccvn.h: Likewise.
7578 * tree-ssa-strlen.c (get_string_length): Likewise.
7579 (valid_builtin_call): Likewise.
7580 (adjust_last_stmt): Likewise.
7581 (handle_builtin_strchr): Likewise.
7582 (handle_builtin_strcpy): Likewise.
7583 (handle_builtin_stxncpy): Likewise.
7584 (handle_builtin_memcpy): Likewise.
7585 (handle_builtin_strcat): Likewise.
7586 (strlen_check_and_optimize_stmt): Likewise.
7587 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
7588 * tree-streamer-in.c: Likewise.
7589 * tree-streamer.c (record_common_node): Likewise.
7590 * tree.c (tree_code_size): Likewise.
7591 (wide_int_to_tree_1): Likewise.
7592 (type_contains_placeholder_1): Likewise.
7593 (build_common_tree_nodes): Likewise.
7594 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
7595 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
7596 (POINTER_BOUNDS_P): Likewise.
7597 (BOUNDED_TYPE_P): Likewise.
7598 (BOUNDED_P): Likewise.
7599 (CALL_WITH_BOUNDS_P): Likewise.
7600 (pointer_bounds_type_node): Likewise.
7601 * value-prof.c (gimple_ic): Likewise.
7602 * var-tracking.c (vt_add_function_parameters): Likewise.
7603 * varasm.c (make_decl_rtl): Likewise.
7604 (assemble_start_function): Likewise.
7605 (output_constant): Likewise.
7606 (maybe_assemble_visibility): Likewise.
7607 * varpool.c (ctor_for_folding): Likewise.
7608 * chkp-builtins.def: Remove.
7609 * ipa-chkp.c: Remove.
7610 * ipa-chkp.h: Remove.
7611 * rtl-chkp.c: Remove.
7612 * rtl-chkp.h: Remove.
7613 * tree-chkp-opt.c: Remove.
7614 * tree-chkp.c: Remove.
7615 * tree-chkp.h: Remove.
7617 2018-06-07 Carl Love <cel@us.ibm.com>
7619 * config/rs6000/vsx.md (vextract_fp_from_shorth,
7620 vextract_fp_from_shortl): Add BE support.
7622 2018-06-07 Paul Koning <ni1d@arrl.net>
7624 * compare-elim.c (try_merge_compare): Don't merge compare if
7625 address contains a side effect.
7626 (try_eliminate_compare): Likewise.
7628 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
7630 * config.gcc: Support "tremont".
7631 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
7632 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
7634 * config/i386/i386.c (m_TREMONT): Define.
7635 (processor_target_table): Add "tremont".
7636 (PTA_TREMONT): Define.
7637 (ix86_lea_outperforms): Add TARGET_TREMONT.
7638 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
7639 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
7640 and M_INTEL_GOLDMONT_PLUS.
7641 (fold_builtin_cpu): Add "tremont".
7642 (ix86_add_stmt_cost): Add TARGET_TREMONT.
7643 (ix86_option_override_internal): Add "tremont".
7644 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
7645 (processor_type): Add PROCESSOR_TREMONT.
7646 * config/i386/x86-tune.def: Add m_TREMONT.
7647 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
7649 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
7651 * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
7652 symbol defined for msp430i* devices to be lower case.
7654 2018-06-07 Richard Biener <rguenther@suse.de>
7656 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
7657 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
7658 Properly wrap signed arithmetic if overflow wraps.
7660 2018-06-07 Jakub Jelinek <jakub@redhat.com>
7662 PR tree-optimization/69615
7663 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
7664 of a cast from a same precision integral SSA_NAME in a bb dominated
7665 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
7666 cast to utype if rhs2 has already a compatible type.
7668 2018-06-07 Richard Biener <rguenther@suse.de>
7670 PR tree-optimization/85935
7671 * graphite-scop-detection.c (find_params_in_bb): Analyze
7672 condition operands with respect to the correct loop. Assert
7673 the analysis doesn't fail.
7675 2018-06-04 Carl Love <cel@us.ibm.com>
7677 * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
7678 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
7679 as it is slightly cheaper.
7680 (first_match_or_eos_index_<mode>):
7681 Calculate index using natural element order.
7682 (first_match_index_<mode>):
7683 Calculate index using natural element order.
7684 (first_match_or_eos_index_<mode>):
7685 Calculate index using natural order.
7686 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
7687 for BE and LE modes.
7688 * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
7689 P9V_BUILTIN_VCLZLSBB_V16QI.
7690 * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
7693 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
7695 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
7696 indentation and line wrap for many prototypes. Add missing
7697 @smallexample directives around block of prototypes for vec_xl and
7700 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
7702 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
7703 track if we pass or return IEEE 128-bit floating point.
7704 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
7705 C++ mangling that is compatible with GCC 8.1.
7706 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
7707 (init_cumulative_args): Note if we pass or return IEEE 128-bit
7708 floating point types.
7709 (rs6000_function_arg_advance_1): Likewise.
7710 (rs6000_mangle_type): Optionally generate mangled names that match
7711 what GCC 8.1 generated for IEEE 128-bit floating point types.
7712 (rs6000_globalize_decl_name): If we have an external function that
7713 passes or returns IEEE 128-bit types, generate a weak reference
7714 from the mangled name used in GCC 8.1 to the current mangled
7716 (rs6000_init_builtins): Make __ibm128 use the long double type if
7717 long double is IBM extended double. Make __float128 use the long
7718 double type if long double is IEEE 128-bit.
7721 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
7722 macro for __ibm128 built-in functions.
7723 (PACK_IF): Add __ibm128 pack/unpack functions.
7724 (UNPACK_IF): Likewise.
7725 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
7726 enable long double built-in functions if long double is IEEE
7727 128-bit floating point.
7728 (rs6000_invalid_builtin): Update long double built-in function
7730 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
7731 functions, adjust the built-in function to use the long double
7732 built-in function if __ibm128 and long double are the same type.
7733 * doc/extend.texi (PowerPC builtins): Update documention for
7734 __builtin_{,un}pack_longdouble. Add documentation for
7735 __builtin_{,un}pack_ibm128.
7737 2018-06-06 Jim Wilson <jimw@sifive.com>
7739 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
7740 (struct machine_function): New field interrupt_mode.
7741 (riscv_handle_type_attribute): New function. Add forward declaration.
7742 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
7743 (riscv_expand_epilogue): Check interrupt_mode field.
7744 (riscv_set_current_function): Check interrupt attribute args and
7745 set interrupt_mode field.
7746 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
7747 (riscv_sret, riscv_uret): New.
7748 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
7749 new arguments to interrupt attribute.
7751 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
7754 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
7755 Don't handle -mcpu=power8 if -mpower9-vector is also used.
7757 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
7759 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
7760 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
7761 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
7762 several redundant entries.
7764 2018-06-06 David Malcolm <dmalcolm@redhat.com>
7766 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
7767 type from "rtx" to "rtx_insn *".
7768 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
7769 for local "call_insn", removing cast.
7770 (ix86_expand_call): Likewise, introducing a "call_insn" local.
7772 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
7774 PR tree-optimization/86066
7775 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
7776 for BIT_INSERT_EXPR stores.
7778 2018-06-06 Richard Biener <rguenther@suse.de>
7780 PR tree-optimization/86062
7781 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
7782 component refs ontop
7783 of to be offsetted base.
7785 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
7787 * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
7788 to be static and remove check on interrupt attribute name.
7790 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
7792 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
7793 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
7795 2018-06-05 Steve Ellcey <sellcey@cavium.com>
7798 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
7800 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
7801 Remove second argument, change how error is called.
7802 (aarch64_layout_arg): Remove second argument from
7803 aarch64_err_no_fpadvsimd call.
7804 (aarch64_init_cumulative_args): Ditto.
7805 (aarch64_gimplify_va_arg_expr): Ditto.
7806 * config/aarch64/aarch64.md (mov<mode>): Ditto.
7808 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
7810 * config/i386/i386.md (simple_return_indirect_internal): New expander.
7811 (*simple_return_indirect_internal<mode>): Rename from
7812 simple_return_indirect_internal. Use W mode iterator.
7813 (rstorssp): New expander.
7814 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
7815 (clrssbsy): New expander.
7816 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
7818 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
7820 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
7822 (cmse_check_pointed_object): Likewise.
7824 2018-06-05 Martin Liska <mliska@suse.cz>
7826 PR gcov-profile/47618
7827 * doc/invoke.texi: Document how -fprofile-dir format
7830 2018-06-05 Richard Biener <rguenther@suse.de>
7832 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
7833 removal pretend DOM info isn't available so we do not update
7834 it and only remove edges, not dominated blocks. Actually free
7835 DOM info in case we removed something. Remove unreachable blocks.
7836 (mfb_keep_latches): Work with either DOM info or marked backedges.
7837 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
7838 first. Mark backedges if DOM info isn't available.
7839 (Re-)compute DOM info after cleanup_control_flow_pre.
7841 2018-06-05 Richard Biener <rguenther@suse.de>
7843 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
7844 (locus_discrim_hasher::hash): Adjust.
7845 (locus_discrim_hasher::equal): Likewise.
7846 (next_discriminator_for_locus): Work on line directly.
7847 (same_line_p): Pass in expanded locus1 as well.
7848 (assign_discriminators): Avoid redundant location expansions.
7850 2018-06-05 Richard Biener <rguenther@suse.de>
7852 PR tree-optimization/86046
7853 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
7854 if required after clearing TREE_ADDRESSABLE.
7856 2018-06-05 Richard Biener <rguenther@suse.de>
7858 PR tree-optimization/86047
7859 * tree-ssa-loop.c (for_each_index): Glob handling of all
7860 decls and constants and really handle all of them.
7862 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7865 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
7866 qualifier_void_pointer and qualifier_const_void_pointer.
7867 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
7868 (arm_init_builtins): Handle the above.
7869 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
7870 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
7871 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
7874 2018-06-05 Martin Liska <mliska@suse.cz>
7876 * auto-profile.c (read_autofdo_file): Do not use
7877 gcov_ctr_summary struct.
7878 (afdo_callsite_hot_enough_for_early_inline): Likewise.
7879 * coverage.c (struct counts_entry): Likewise.
7880 (read_counts_file): Read just single summary entry.
7881 (get_coverage_counts): Use gcov_summary struct.
7882 * coverage.h (get_coverage_counts): Likewise.
7883 * gcov-dump.c (dump_working_sets): Likewise.
7884 (tag_summary): Dump just single summary.
7885 * gcov-io.c (gcov_write_summary): Write just histogram
7887 (gcov_read_summary): Read just single summary.
7888 (compute_working_sets): Use gcov_summary struct.
7889 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
7890 of GCOV_COUNTERS_SUMMABLE.
7891 (GCOV_COUNTERS_SUMMABLE): Remove.
7892 (GCOV_FIRST_VALUE_COUNTER): Replace with
7893 GCOV_COUNTER_V_INTERVAL.
7894 (struct gcov_ctr_summary): Remove.
7895 (struct gcov_summary): Directly use fields of former
7897 (compute_working_sets): Use gcov_summary struct.
7898 * gcov.c (read_count_file): Do not use ctrs fields.
7899 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
7901 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
7903 * profile.c: Likewise.
7904 * profile.h: Likewise.
7906 2018-06-05 Martin Liska <mliska@suse.cz>
7908 PR gcov-profile/84846
7909 * gcov.c (output_lines): Print working directory only
7910 in intermediate format.
7912 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
7914 * config/s390/s390-builtin-types.def: Add void function type.
7915 * config/s390/s390-builtins.def: Use the function type for the
7918 2018-06-04 Jim Wilson <jimw@sifive.com>
7920 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
7922 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
7923 and maybe_eh_return. Change regno to unsigned int. Use new args to
7924 handle EH_RETURN_DATA_REGNO registers properly.
7925 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
7926 (riscv_expand_epilogue): Update comment. Change argument name and
7927 type. Update code to use new name and type. Pass new args to
7928 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
7930 * config/riscv/riscv.md (NORMAL_RETURN): New.
7931 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
7932 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
7933 (eh_return): Call gen_eh_return_internal and emit barrier.
7934 (eh_return_internal): Call riscv_expand_epilogue.
7936 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
7938 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
7939 bit_insertion field and declare can_be_merged_into method.
7940 (merged_store_group::can_be_merged_into): New method.
7941 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
7942 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
7943 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
7945 2018-06-04 Richard Biener <rguenther@suse.de>
7947 PR tree-optimization/85955
7948 * builtins.c (fold_builtin_sincos): Convert pointers to
7949 destination to appropriate type before dereferencing.
7951 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
7953 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
7955 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
7957 * expr.c (expand_expr_real_1): Force the operand into memory if
7958 its TYPE_MODE is BLKmode and if there is no integer mode for
7959 the number of bits being extracted.
7961 2018-06-04 Jakub Jelinek <jakub@redhat.com>
7965 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
7966 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
7968 2018-06-04 Richard Biener <rguenther@suse.de>
7970 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
7971 (cleanup_tree_cfg_noloop): ... single caller. Do
7972 start_recording_case_labels later.
7974 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
7976 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
7977 to _IMMINTRIN_H_INCLUDED.
7978 * config/i386/pconfigintrin.h: Ditto.
7979 * config/i386/waitpkgintrin.h: Ditto.
7980 * config/i386/immintrin.h: Add includes for sgxintrin.h,
7981 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
7982 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
7983 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
7984 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
7985 waitpkgintrin.h and cldemoteintrin.h.
7987 2018-06-04 Richard Biener <rguenther@suse.de>
7989 PR tree-optimization/86038
7990 * tracer.c (find_best_successor): Check probability for
7991 being initialized, bail out if not.
7993 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
7996 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
7997 of bits to ignore when comparing architectures.
7999 2018-06-04 Jakub Jelinek <jakub@redhat.com>
8001 PR tree-optimization/69615
8002 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
8003 maximum or minimum of the type, try to merge it also as if
8004 range1 is + [-, x - 1] or + [x + 1, -].
8007 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
8009 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
8011 PR tree-optimization/86034
8012 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
8013 the unsigned bitfield type in a bit insertion sequence if it does not
8014 have a larger precision than the bitfield size.
8015 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
8017 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
8019 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
8021 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
8023 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
8024 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
8025 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
8026 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
8028 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
8030 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
8031 Disable -fdelete-null-pointer-checks for ELF toolchain.
8033 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
8034 Kito Cheng <kito.cheng@gmail.com>
8036 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
8037 (nds32le-*-*, nds32be-*-*): Integrate checking process.
8038 (nds32*-*-*): Add glibc and uclibc conditions.
8039 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
8040 (TARGET_EXCEPT_UNWIND_INFO): Define.
8041 * config/nds32/elf.h: New file.
8042 * config/nds32/linux.h: New file.
8043 * config/nds32/nds32-elf.opt: New file.
8044 * config/nds32/nds32-linux.opt: New file.
8045 * config/nds32/nds32-fp-as-gp.c
8046 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
8047 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
8049 (nds32_asm_file_end): Ditto.
8050 (nds32_print_operand): Ditto.
8051 (nds32_insert_attributes): Ditto.
8052 (nds32_init_libfuncs): New function.
8053 (TARGET_HAVE_TLS): Define.
8054 (TARGET_INIT_LIBFUNCS): Define.
8055 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
8057 (TARGET_ELF): Apply different mcmodel setting.
8058 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
8059 been migrated into elf.h and linux.h files.
8060 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
8061 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
8062 (mcmodel): The content has been migrated into nds32-elf.opt and
8063 nds32-linux.opt files.
8064 * config/nds32/t-elf: New file.
8065 * config/nds32/t-linux: New file.
8067 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
8068 Shiva Chen <shiva0217@gmail.com>
8070 * config/nds32/constants.md (unspec_volatile_element): Add
8071 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
8072 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
8074 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
8075 (make_pass_nds32_fp_as_gp): Declare.
8076 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
8078 (nds32_asm_function_end_prologue): Remove unused asm output.
8079 (nds32_asm_function_begin_epilogue): Remove unused asm output.
8080 (nds32_asm_file_start): Output necessary fp_as_gp information.
8081 (nds32_option_override): Adjust register usage.
8082 (nds32_expand_prologue): Consider fp_as_gp situation.
8083 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
8084 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
8087 (simple_return): Ditto.
8088 (omit_fp_begin): Output special directive for fp_as_gp.
8089 (omit_fp_end): Output special directive for fp_as_gp.
8090 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
8091 mforbid-fp-as-gp): New options.
8093 2018-06-01 Mark Wielaard <mark@klomp.org>
8095 * dwarf2out.c (dwarf2out_finish): Remove generation of
8096 DW_AT_loclists_base.
8098 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
8100 * gimple-ssa-store-merging.c: Include gimple-fold.h.
8101 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
8102 (struct merged_store_group): Add bit_insertion field.
8103 (dump_char_array): Use standard hexadecimal format.
8104 (merged_store_group::merged_store_group): Set bit_insertion to false.
8105 (merged_store_group::apply_stores): Use optimal buffer size. Deal
8106 with BIT_INSERT_EXPR stores. Move up code updating the mask and
8107 also print the mask in the dump file.
8108 (pass_store_merging::gate): Minor tweak.
8109 (imm_store_chain_info::coalesce_immediate): Fix wrong association
8110 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
8111 stores with INTEGER_CST stores.
8112 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
8113 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
8114 and use it throughout. Generate bit insertion sequences if need be.
8115 (pass_store_merging::process_store): Remove redundant condition.
8116 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
8118 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
8120 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
8121 the 128-bit floating point types. Fix function comment.
8123 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8125 * config/aarch64/aarch64-simd.md
8126 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
8128 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
8131 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
8133 PR tree-optimization/85989
8134 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
8136 (backprop::intersect_uses): Check it when deciding whether this
8137 is a backedge reference.
8138 (backprop::process_block): Add each phi to m_visited_phis
8139 after visiting it, then clear it at the end.
8141 2018-06-01 Richard Biener <rguenther@suse.de>
8143 * tree-vectorizer.h (vect_dr_stmt): New function.
8144 (vect_get_load_cost): Adjust.
8145 (vect_get_store_cost): Likewise.
8146 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
8147 Use vect_dr_stmt instead of DR_SMTT.
8148 (vect_record_base_alignments): Likewise.
8149 (vect_calculate_target_alignment): Likewise.
8150 (vect_compute_data_ref_alignment): Likewise and make static.
8151 (vect_update_misalignment_for_peel): Likewise.
8152 (vect_verify_datarefs_alignment): Likewise.
8153 (vector_alignment_reachable_p): Likewise.
8154 (vect_get_data_access_cost): Likewise. Pass down
8155 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
8156 (vect_get_peeling_costs_all_drs): Likewise.
8157 (vect_peeling_hash_get_lowest_cost): Likewise.
8158 (vect_enhance_data_refs_alignment): Likewise.
8159 (vect_find_same_alignment_drs): Likewise.
8160 (vect_analyze_data_refs_alignment): Likewise.
8161 (vect_analyze_group_access_1): Likewise.
8162 (vect_analyze_group_access): Likewise.
8163 (vect_analyze_data_ref_access): Likewise.
8164 (vect_analyze_data_ref_accesses): Likewise.
8165 (vect_vfa_segment_size): Likewise.
8166 (vect_small_gap_p): Likewise.
8167 (vectorizable_with_step_bound_p): Likewise.
8168 (vect_prune_runtime_alias_test_list): Likewise.
8169 (vect_analyze_data_refs): Likewise.
8170 (vect_supportable_dr_alignment): Likewise.
8171 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
8172 (vect_gen_prolog_loop_niters): Likewise.
8173 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
8174 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
8176 (vect_recog_mask_conversion_pattern): Likewise.
8177 (vect_try_gather_scatter_pattern): Likewise.
8178 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
8179 to vect_get_store_cost.
8180 (vect_get_store_cost): Get stmt_info instead of DR.
8181 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
8182 (vect_get_load_cost): Get stmt_info instead of DR.
8184 2018-06-01 Richard Biener <rguenther@suse.de>
8187 * gimple-fold.c (var_decl_component_p): Also allow offsetted
8188 vars wrapped in MEM_REFs.
8190 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
8192 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
8193 Fix subreg tests so that we only return a choice between
8194 GENERAL_REGS and FP_REGS if the original classes included both.
8196 2018-06-01 Richard Biener <rguenther@suse.de>
8199 * tree-ssa-structalias.c (get_function_part_constraint):
8200 Handle NULL fi->decl.
8201 (find_func_aliases_for_call): Properly handle indirect
8202 fi from direct call.
8203 (find_func_clobbers): Likewise.
8204 (ipa_pta_execute): Likewise.
8205 (create_variable_info_for): For functions that are ifunc_resolver
8206 resolve to a varinfo that contains the result of the resolver call.
8207 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
8210 2018-05-31 Michael Collison <michael.collison@arm.com>
8212 * config/aarch64/aarch64.md:
8213 (*fix_to_zero_extenddfdi2): New pattern.
8214 * gcc.target/aarch64/fix_extend1.c: New testcase.
8216 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
8220 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
8221 and BUILT_IN_STRNCMP_EQ.
8222 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
8223 BUILT_IN_STRNCMP_EQ.
8224 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
8225 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
8226 (gimple_fold_builtin): Likewise.
8227 * tree-ssa-strlen.c (compute_string_length): New function.
8228 (determine_min_obsize): New function.
8229 (handle_builtin_string_cmp): New function to handle calls to
8230 string compare functions.
8231 (strlen_optimize_stmt): Add handling to builtin string compare
8233 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
8234 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
8235 * tree.c (build_common_builtin_nodes): Add new defines of
8236 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
8238 2018-05-31 Jakub Jelinek <jakub@redhat.com>
8241 * bb-reorder.c (pass_partition_blocks::gate): Return false for
8242 functions with naked attribute.
8244 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
8246 * config/i386/sse.md (avx_vec_concat<mode>):
8247 Substitute concat_tg_mode mode attribute with xtg_mode.
8248 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
8249 (concat_tg_mode): Remove mode attribute.
8251 2018-05-31 Martin Sebor <msebor@redhat.com>
8254 * calls.c (alloc_max_size): Correct a logic error/typo.
8255 Treat excessive arguments as infinite. Warn for invalid arguments.
8256 * doc/invoke.texi (-Walloc-size-larger-than): Update.
8258 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
8261 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
8262 and movx for Haswell.
8264 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
8265 Cesar Philippidis <cesar@codesourcery.com>
8268 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
8269 when emitting error on private/firstprivate reductions.
8270 * omp-low.c (lower_omp_target): Avoid reference-type processing
8271 on pointers for firstprivate clause.
8273 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
8275 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
8278 * config/aarch64/aarch64-simd.md
8279 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
8280 (aarch64_ld1_x3_<mode>): Likewise
8281 (aarch64_st1x2<VALLDIF:mode>): Likewise
8282 (aarch64_st1_x2_<mode>): Likewise
8283 (aarch64_st1x3<VALLDIF:mode>): Likewise
8284 (aarch64_st1_x3_<mode>): Likewise
8285 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
8286 (vld1_s8_x3): Likewise.
8287 (vld1_u16_x3): Likewise.
8288 (vld1_s16_x3): Likewise.
8289 (vld1_u32_x3): Likewise.
8290 (vld1_s32_x3): Likewise.
8291 (vld1_u64_x3): Likewise.
8292 (vld1_s64_x3): Likewise.
8293 (vld1_f16_x3): Likewise.
8294 (vld1_f32_x3): Likewise.
8295 (vld1_f64_x3): Likewise.
8296 (vld1_p8_x3): Likewise.
8297 (vld1_p16_x3): Likewise.
8298 (vld1_p64_x3): Likewise.
8299 (vld1q_u8_x3): Likewise.
8300 (vld1q_s8_x3): Likewise.
8301 (vld1q_u16_x3): Likewise.
8302 (vld1q_s16_x3): Likewise.
8303 (vld1q_u32_x3): Likewise.
8304 (vld1q_s32_x3): Likewise.
8305 (vld1q_u64_x3): Likewise.
8306 (vld1q_s64_x3): Likewise.
8307 (vld1q_f16_x3): Likewise.
8308 (vld1q_f32_x3): Likewise.
8309 (vld1q_f64_x3): Likewise.
8310 (vld1q_p8_x3): Likewise.
8311 (vld1q_p16_x3): Likewise.
8312 (vld1q_p64_x3): Likewise.
8313 (vst1_s64_x2): Likewise.
8314 (vst1_u64_x2): Likewise.
8315 (vst1_f64_x2): Likewise.
8316 (vst1_s8_x2): Likewise.
8317 (vst1_p8_x2): Likewise.
8318 (vst1_s16_x2): Likewise.
8319 (vst1_p16_x2): Likewise.
8320 (vst1_s32_x2): Likewise.
8321 (vst1_u8_x2): Likewise.
8322 (vst1_u16_x2): Likewise.
8323 (vst1_u32_x2): Likewise.
8324 (vst1_f16_x2): Likewise.
8325 (vst1_f32_x2): Likewise.
8326 (vst1_p64_x2): Likewise.
8327 (vst1q_s8_x2): Likewise.
8328 (vst1q_p8_x2): Likewise.
8329 (vst1q_s16_x2): Likewise.
8330 (vst1q_p16_x2): Likewise.
8331 (vst1q_s32_x2): Likewise.
8332 (vst1q_s64_x2): Likewise.
8333 (vst1q_u8_x2): Likewise.
8334 (vst1q_u16_x2): Likewise.
8335 (vst1q_u32_x2): Likewise.
8336 (vst1q_u64_x2): Likewise.
8337 (vst1q_f16_x2): Likewise.
8338 (vst1q_f32_x2): Likewise.
8339 (vst1q_f64_x2): Likewise.
8340 (vst1q_p64_x2): Likewise.
8341 (vst1_s64_x3): Likewise.
8342 (vst1_u64_x3): Likewise.
8343 (vst1_f64_x3): Likewise.
8344 (vst1_s8_x3): Likewise.
8345 (vst1_p8_x3): Likewise.
8346 (vst1_s16_x3): Likewise.
8347 (vst1_p16_x3): Likewise.
8348 (vst1_s32_x3): Likewise.
8349 (vst1_u8_x3): Likewise.
8350 (vst1_u16_x3): Likewise.
8351 (vst1_u32_x3): Likewise.
8352 (vst1_f16_x3): Likewise.
8353 (vst1_f32_x3): Likewise.
8354 (vst1_p64_x3): Likewise.
8355 (vst1q_s8_x3): Likewise.
8356 (vst1q_p8_x3): Likewise.
8357 (vst1q_s16_x3): Likewise.
8358 (vst1q_p16_x3): Likewise.
8359 (vst1q_s32_x3): Likewise.
8360 (vst1q_s64_x3): Likewise.
8361 (vst1q_u8_x3): Likewise.
8362 (vst1q_u16_x3): Likewise.
8363 (vst1q_u32_x3): Likewise.
8364 (vst1q_u64_x3): Likewise.
8365 (vst1q_f16_x3): Likewise.
8366 (vst1q_f32_x3): Likewise.
8367 (vst1q_f64_x3): Likewise.
8368 (vst1q_p64_x3): Likewise.
8370 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
8372 * config/msp430/msp430.c (msp430_output_labelref): Prepend
8373 user_label_prefix to name.
8375 * tree-core.h: Update comment about the format of NAME string
8376 passed to handler in attribute_spec.
8378 * config/msp430/msp430.md: Remove erroneous subreg expression from
8379 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
8380 zero_extend{q,h}isi2.
8382 2018-05-30 Borislav Petkov <bp@suse.de>
8384 * doc/extend.texi: Document some architecture specific
8385 constraints and sort entries.
8387 2018-05-30 Martin Sebor <msebor@redhat.com>
8390 * builtins.c (expand_builtin_stpcpy_1): New function.
8391 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
8392 only if the former succeeds.
8394 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
8396 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
8399 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
8401 * doc/invoke.texi (-flinker-output): Document
8403 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
8405 * passes.c (ipa_write_summaries): Only modify statements if body
8407 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
8408 incrementally linking.
8409 (ipa_passes): Likewise.
8410 * lto-cgraph.c (lto_output_node): When incrementally linking do not
8411 pass down resolution info.
8412 * common.opt (flag_incremental_link): Update info.
8413 * gcc.c (plugin specs): Turn flinker-output=* to
8414 -plugin-opt=-linker-output-known
8415 * toplev.c (compile_file): Also cut compilation when doing incremental
8417 * flag-types. (enum lto_partition_model): Add
8418 LTO_LINKER_OUTPUT_NOLTOREL.
8419 (invoke.texi): Add -flinker-output docs.
8420 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
8421 link same way as WPA; do not stream in dead initializers.
8423 * dwarf2out.c (dwarf2out_die_ref_for_decl,
8424 darf2out_register_external_decl): Support incremental link.
8426 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
8428 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
8430 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
8432 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
8433 it down to simple_object_copy_lto_debug_sections.
8434 (run_gcc): Determine incremental LTO link time and configure
8435 lto1 into non-wpa mode, disable renaming of debug sections.
8437 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
8439 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
8440 descriptions of various incorrectly documented functions.
8442 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
8445 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
8446 address check not strict.
8448 2018-05-30 Richard Biener <rguenther@suse.de>
8450 PR tree-optimization/85964
8451 * tracer.c (better_p): Drop initialized count check, we only
8452 call the function with initialized counts now.
8453 (find_best_successor): Do find a best edge if one
8454 has uninitialized count.
8455 (find_best_predecessor): Likewise. Do BB frequency check only
8456 if count is initialized.
8458 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
8460 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
8461 (aarch64_ldrstr_offset_compare): New.
8462 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
8463 load/store orderings.
8464 (aarch64_gen_adjusted_ldpstp): Likewise.
8466 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
8468 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
8469 Check for subset of GENERAL_REGS and FP_REGS.
8470 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
8473 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
8475 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
8476 and wi::to_poly_offset. Add the current offset and then check
8477 whether the sum fits, rather than using an unchecked addition of
8478 a checked term. Check for a shwi rather than a uhwi.
8479 * expr.c (get_bit_range): Use tree_to_poly_uint64.
8480 (store_constructor): Use poly_int_tree_p.
8481 (expand_expr_real_1): Likewise.
8482 * function.c (assign_temp): Likewise.
8483 * fold-const.c (const_binop): Use poly_int_tree_p and
8485 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
8487 * ipa-icf-gimple.c (func_checker::compare_operand): Use
8488 to_poly_offset for MEM offsets.
8489 * ipa-icf.c (sem_variable::equals): Likewise.
8490 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
8491 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
8492 wi::to_poly_offset for BIT_FIELD_REF offsets.
8493 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
8495 * var-tracking.c (emit_note_insn_var_location): Use
8496 tree_to_poly_uint64.
8498 2018-05-29 Jim Wilson <jimw@sifive.com>
8500 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
8502 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
8505 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
8506 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
8508 (sse4_1_round<mode>2): Use nonimmediate_operand
8509 for operand 1 predicate.
8511 2018-05-29 Martin Sebor <msebor@redhat.com>
8512 Richard Biener <rguenther@suse.de>
8515 * calls.c (get_size_range): Call determine_value_range instead
8516 of get_value_range..
8517 * tree-vrp.h (determine_value_range): Declared new function.
8518 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
8520 2018-05-29 Richard Biener <rguenther@suse.de>
8522 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
8523 sure to use non-pattern stmts for get_earlier_stmt arguments.
8524 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
8525 called on pattern stmts.
8526 (get_later_stmt): Likewise.
8528 2018-05-29 Martin Liska <mliska@suse.cz>
8530 PR gcov-profile/85759
8531 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
8534 2018-05-29 Jakub Jelinek <jakub@redhat.com>
8536 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
8538 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
8542 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
8543 VEC_PACK_FLOAT_EXPR): New tree codes.
8544 * tree-pretty-print.c (op_code_prio): Handle
8545 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
8546 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
8547 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
8548 * tree-inline.c (estimate_operator_cost): Likewise.
8549 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
8550 * fold-const.c (const_binop): Likewise.
8551 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
8552 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
8553 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
8554 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
8555 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
8556 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
8557 * expr.c (expand_expr_real_2): Likewise.
8558 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
8559 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
8560 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
8562 * optabs.c (expand_widen_pattern_expr): For
8563 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
8564 sign from result type rather than operand's type.
8565 (expand_binop_directly): For vec_packu_float_optab and
8566 vec_packs_float_optab allow result type to be different from operand's
8568 * optabs-tree.c (optab_for_tree_code): Handle
8569 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
8570 VEC_PACK_FLOAT_EXPR. Formatting fixes.
8571 * tree-vect-generic.c (expand_vector_operations_1): Handle
8572 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
8573 VEC_PACK_FLOAT_EXPR.
8574 * tree-vect-stmts.c (supportable_widening_operation): Handle
8576 (supportable_narrowing_operation): Handle FLOAT_EXPR.
8577 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
8578 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
8579 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
8580 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
8582 (vec_pack<floatprefix>_float_<mode>): New expander.
8583 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
8585 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
8586 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
8587 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
8588 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
8589 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
8591 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
8592 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
8593 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
8594 VEC_PACK_FLOAT_EXPR): Document.
8596 2018-05-29 Richard Biener <rguenther@suse.de>
8598 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
8600 (stmt_vec_info_vec): Make pointer.
8601 (init_stmt_vec_info_vec): Remove.
8602 (free_stmt_vec_info_vec): Likewise.
8603 (set_stmt_vec_info_vec): New function.
8604 (free_stmt_vec_infos): Likewise.
8605 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
8606 (set_vinfo_for_stmt): Likewise.
8607 (get_earlier_stmt): Likewise.
8608 (get_later_stmt): Likewise.
8609 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
8610 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
8611 (vec_info::~vec_info): Free stmt_vec_infos.
8612 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
8613 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
8614 (pass_slp_vectorize::execute): Likewise.
8615 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
8616 (free_stmt_vec_info_vec): Likewise.
8617 (set_stmt_vec_info_vec): New function.
8618 (free_stmt_vec_infos): Likewise.
8619 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
8620 the global stmt_vec_info_vec.
8621 * tree-parloops.c (gather_scalar_reductions): Use
8622 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
8625 2018-05-29 Richard Biener <rguenther@suse.de>
8627 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
8629 2018-05-29 Martin Liska <mliska@suse.cz>
8630 David Malcolm <dmalcolm@redhat.com>
8632 * vec.c (test_reverse): New.
8633 (vec_c_tests): Add new test.
8634 * vec.h (vl_ptr>::reverse): New function.
8636 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
8638 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
8640 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
8643 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
8645 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
8647 2018-05-28 Richard Biener <rguenther@suse.de>
8649 PR tree-optimization/85933
8650 * tree-vect-data-refs.c (vect_record_base_alignments): Only
8651 look at stmts marked as vectorizable.
8653 2018-05-28 Richard Biener <rguenther@suse.de>
8655 PR tree-optimization/85934
8656 * tree-vect-generic.c (expand_vector_operations_1): Hoist
8657 vector boolean check before scalar optimization.
8659 2018-05-28 Jakub Jelinek <jakub@redhat.com>
8661 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
8664 2018-05-28 Mark Wielaard <mark@klomp.org>
8666 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
8667 if it is an expression containing a minus sign.
8669 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
8671 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
8673 2018-05-27 Paul Koning <ni1d@arrl.net>
8675 * config/pdp11/pdp11.md (truncsihi2): Remove.
8677 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
8678 Chung-Ju Wu <jasonwucj@gmail.com>
8680 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
8682 (unaligned_store_dw): Ditto.
8683 * config/nds32/nds32-memory-manipulation.c
8684 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
8685 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
8686 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
8687 (emit_setmem_word_loop): Rename to ...
8688 (emit_setmem_doubleword_loop): ... this.
8689 (nds32_gen_dup_4_byte_to_word_value): New function.
8690 (nds32_gen_dup_8_byte_to_double_word_value): New function.
8691 (nds32_expand_setmem_loop): Refine implementation.
8692 (nds32_expand_setmem_loop_v3m): Ditto.
8693 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
8696 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
8698 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
8700 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
8702 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
8703 (nds32_init_machine_status): Initialize machine->attr_naked_p and
8704 machine->attr_no_prologue_p.
8705 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
8706 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
8707 (nds32_expand_epilogue): Consider attr_naked_p.
8708 (nds32_expand_epilogue_v3pop): Likewise.
8709 (nds32_can_use_return_insn): Likewise.
8710 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
8711 attr_no_prologue_p fields.
8712 * config/nds32/nds32.opt (mret-in-naked-func): New option.
8714 2018-05-27 Jakub Jelinek <jakub@redhat.com>
8717 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
8719 * config/i386/sse.md
8720 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
8722 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
8724 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
8726 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
8728 (*<floatsuffix>floatv2div2sf2): Rename to ...
8729 (*float<floatunssuffix>v2div2sf2): ... this.
8730 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
8731 (float<floatunssuffix>v2div2sf2_mask): ... this.
8732 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
8733 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
8734 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
8736 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
8738 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
8740 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
8742 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
8744 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
8746 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
8747 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
8748 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
8749 gen_ufix_truncv8dfv8si2.
8750 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
8751 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
8752 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
8753 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
8754 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
8755 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
8756 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
8757 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
8759 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
8763 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
8765 2018-05-25 Jim Wilson <jimw@sifive.com>
8767 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
8768 * config/riscv/riscv.c (struct machine_function): Add
8769 interrupt_handler_p and attribute_checked_p fields.
8770 (riscv_attribute_table): Add interrupt.
8771 (riscv_interrupt_type_p): New.
8772 (riscv_save_reg_p): Save extra regs for interrupt handler.
8773 (riscv_use_save_libcall): Return false for interrupt handler.
8774 (riscv_first_stack_step): Add forward declaration.
8775 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
8776 for interrupt handler with large frame. Use it for saved reg list.
8777 (riscv_expand_prologue): Move flag_stack_usage_info support to
8778 eliminate duplication.
8779 (riscv_expand_epilogue): Generate mret for interrupt handler.
8780 (riscv_epilogue_uses): New.
8781 (riscv_can_use_return_insn): Return false for interrupt handler.
8782 (riscv_function_ok_for_sibcall): Likewise.
8783 (riscv_set_current_function): Add interrupt handler support.
8784 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
8785 * config/riscv/riscv.md (UNSPECV_MRET): New.
8787 (riscv_frflags, riscv_fsflags): Use tab after opcode.
8789 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
8791 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
8793 PR tree-optimization/85712
8794 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
8795 this candidate has already been replaced in-situ by a copy.
8797 2018-05-25 Jason Merrill <jason@redhat.com>
8799 PR c++/80485 - inline function non-zero address.
8800 * symtab.c (nonzero_address): Check DECL_COMDAT.
8802 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
8805 * config/alpha/alpha.md (ashlsi3): New insn pattern.
8806 (*ashlsi_se): Rename from *ashldi_se. Define as sign
8807 extension of SImode operation. Use const123_operand predicate.
8808 (*saddsi_1): Remove.
8809 (*saddl_se_1): Ditto.
8811 (*ssubl_se_1): Ditto.
8812 * config/alpha/predicates.md (const123_operand): New predicate.
8813 * config/alpha/constraints.md (P): Use IN_RANGE.
8815 2018-05-25 Richard Biener <rguenther@suse.de>
8817 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
8819 (ref_maybe_used_by_stmt_p): Likewise.
8820 (stmt_may_clobber_ref_p): Likewise.
8821 (stmt_may_clobber_ref_p_1): Likewise.
8822 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
8824 (ref_maybe_used_by_stmt_p): Likewise.
8825 (stmt_may_clobber_ref_p): Likewise.
8826 (stmt_may_clobber_ref_p_1): Likewise.
8827 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
8828 the alias oracle to disambiguate DRs with stmts DR analysis
8830 (vect_analyze_data_refs): Do not give up on not analyzable
8831 DRs for BB vectorization. Remove code truncating the dataref
8834 2018-05-25 Jakub Jelinek <jakub@redhat.com>
8837 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
8838 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
8839 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
8841 2018-05-25 Richard Biener <rguenther@suse.de>
8843 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
8844 function, combining stmt data ref gathering and fatal analysis
8846 (vect_analyze_data_refs): Remove now redudnant code and simplify.
8847 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
8848 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
8849 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
8850 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
8852 2018-05-25 Bin Cheng <bin.cheng@arm.com>
8854 PR tree-optimization/85720
8855 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
8856 SCC if all partitions are builtins.
8857 (version_loop_by_alias_check): New parameter. Generate cancelable
8858 runtime alias check if all partitions are builtins.
8859 (distribute_loop): Update call to above function.
8861 2018-05-25 Bin Cheng <bin.cheng@arm.com>
8863 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
8864 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
8865 (parm_default_def_partition_arg): Ditto.
8866 (set_parm_default_def_partition): Ditto.
8867 (get_parm_default_def_partitions): Ditto and make it static.
8868 (get_undefined_value_partitions): Ditto and make it static.
8869 (remove_ssa_form): Refactor call to init_var_map here.
8870 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
8871 computation for loop region.
8872 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
8873 (register_default_def): Delete.
8874 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
8875 (parm_default_def_partition_arg): Ditto.
8876 (set_parm_default_def_partition): Ditto.
8877 (get_parm_default_def_partitions): Ditto and make it static.
8878 (get_undefined_value_partitions): Ditto and make it static.
8879 (coalesce_with_default, coalesce_with_default): Update comment.
8880 (create_coalesce_list_for_region): New func factored out from
8881 create_outofssa_var_map.
8882 (populate_coalesce_list_for_outofssa): New func factored out from
8883 create_outofssa_var_map and coalesce_ssa_name.
8884 (create_outofssa_var_map): Delete.
8885 (coalesce_ssa_name): Refactor to support live range computation.
8886 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
8887 (get_parm_default_def_partitions): Delete.
8888 (get_undefined_value_partitions): Ditto.
8889 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
8890 computation for loop region.
8891 (new_tree_live_info, loe_visit_block): Ditto.
8892 (live_worklist, set_var_live_on_entry): Ditto.
8893 (calculate_live_on_exit, verify_live_on_entry): Ditto.
8894 * tree-ssa-live.h (struct _var_map): New fields.
8895 (init_var_map): Change decl.
8896 (region_contains_p): New.
8898 2018-05-25 Bin Cheng <bin.cheng@arm.com>
8900 * tree-ssa-live.h (live_merge_and_clear): Delete.
8902 2018-05-25 Richard Biener <rguenther@suse.de>
8905 * tree-dump.c (dequeue_and_dump): Remove access to removed
8906 operand 2 of a SWITCH_EXPR.
8908 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
8910 * doc/sourcebuild.texi (vect_double_cond_arith): Include
8911 multiplication and division.
8912 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
8913 (cond_udiv@var{m}, cond_umod@var{m}): Document.
8914 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
8915 (cond_udiv_optab, cond_umod_optab): New optabs.
8916 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
8917 (IFN_COND_RDIV): New internal functions.
8918 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
8919 TRUNC_MOD_EXPR and RDIV_EXPR.
8920 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
8921 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
8923 (SVE_INT_BINARY): Include mult.
8924 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
8925 (optab, sve_int_op): Handle mult.
8926 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
8928 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
8929 for SVE_INT_BINARY_SD.
8931 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
8933 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
8934 (optab, sve_int_op): Handle div and udiv.
8935 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
8936 for SVE_INT_BINARY_SD.
8937 (*<optab><mode>3): New insn for the same.
8939 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
8941 * tree-vect-patterns.c: Include predict.h.
8942 (vect_recog_divmod_pattern): Restrict check for division support
8943 to when optimizing for size.
8945 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
8947 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
8948 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
8949 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
8950 (gimple_match_op::set_op): Likewise.
8951 (gimple_resimplify4): Declare.
8952 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
8953 (expr::gen_transform): Likewise.
8954 (decision_tree::gen): Generate a simplification routine for 4 operands.
8955 * gimple-match-head.c (gimple_simplify): Add an overload for
8956 4 operands. In the top-level function, handle up to 4 call
8957 arguments and call gimple_resimplify4.
8958 (gimple_resimplify4): New function.
8959 (build_call_internal): Pass a fourth operand.
8960 (maybe_push_to_seq): Likewise.
8961 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
8962 Fold VEC_COND_EXPRs of an operation and a default value into
8963 an IFN_COND_* function if possible.
8964 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
8966 (SVE_COND_FP_BINARY): Include them.
8967 (optab, sve_fp_op): Handle them.
8968 (SVE_INT_BINARY_REV): New code iterator.
8969 (SVE_COND_FP_BINARY_REV): New int iterator.
8970 (commutative): New int attribute.
8971 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
8973 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
8975 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
8976 (*cond_<optab><mode>): New patterns for reversed operands.
8978 2018-05-25 Richard Biener <rguenther@suse.de>
8980 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
8981 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
8982 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
8983 (STMT_VINFO_GROUPED_ACCESS): Adjust.
8984 * tree-vect-data-refs.c (everywhere): Adjust users.
8985 * tree-vect-loop.c (everywhere): Likewise.
8986 * tree-vect-slp.c (everywhere): Likewise.
8987 * tree-vect-stmts.c (everywhere): Likewise.
8988 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
8990 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8992 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
8994 (gcc_cv_as_section_exclude): ... this.
8995 Try Solaris as #exclude syntax.
8996 * configure: Regenerate.
8997 * config.in: Regenerate.
8998 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
9000 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
9001 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
9003 * varasm.c (default_elf_asm_named_section): Don't check if
9004 HAVE_GAS_SECTION_EXCLUDE is defined.
9006 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
9008 * doc/md.texi: Update the documentation of the cond_* optabs
9009 to mention the new final operand. Fix GET_MODE_NUNITS call.
9010 Describe the scalar case too.
9011 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
9012 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
9014 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
9015 (get_conditional_internal_fn): Update comment.
9016 * tree-vect-loop.c (vectorizable_reduction): Pass the original
9017 accumulator value as a final argument to conditional functions.
9018 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
9019 a define_expand and add an "else" operand. Assert for now that
9020 the else operand is equal to operand 2. Use SVE_INT_BINARY and
9021 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
9022 (*cond_<optab><mode>): New patterns.
9023 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
9024 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
9025 (UNSPEC_COND_EOR): Delete.
9026 (optab): Remove associated mappings.
9027 (SVE_INT_BINARY): New code iterator.
9028 (sve_int_op): Remove int attribute and add "minus" to the code
9030 (SVE_COND_INT_OP): Delete.
9031 (SVE_COND_FP_OP): Rename to...
9032 (SVE_COND_FP_BINARY): ...this.
9034 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
9036 * optabs.c (can_reuse_operands_p): New function.
9037 (maybe_legitimize_operands): Try to reuse the results for
9040 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
9042 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
9043 Add {q} suffix to insn mnemonic.
9045 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
9047 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
9048 (msp430_warn_func_return): New.
9050 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
9052 * fold-const.c (tree_nonzero_bits): New function.
9053 * fold-const.h (tree_nonzero_bits): Likewise.
9054 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
9055 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
9057 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
9061 * varasm.c (assemble_alias): Check ifunc_resolver only on
9064 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
9067 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
9068 when memory input operand is handled.
9070 2018-05-24 Luis Machado <luis.machado@linaro.org>
9072 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
9074 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
9076 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
9078 * match.pd: Delay FMA folds until after vectorization.
9080 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
9083 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
9084 address check not strict.
9086 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
9088 * gimple-match.h (gimple_match_op): New class.
9089 (mprts_hook): Replace parameters with a gimple_match_op *.
9090 (maybe_build_generic_op): Likewise.
9091 (gimple_simplified_result_is_gimple_val): Replace parameters with
9092 a const gimple_match_op *.
9093 (gimple_simplify): Replace code_helper * and tree * parameters with
9094 a gimple_match_op * parameter.
9095 (gimple_resimplify1): Replace code_helper *, tree and tree *
9096 parameters with a gimple_match_op * parameter.
9097 (gimple_resimplify2): Likewise.
9098 (gimple_resimplify3): Likewise.
9099 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
9100 parameters with a gimple_match_op * parameter.
9101 * gimple-match-head.c (gimple_simplify): Change prototypes of
9102 auto-generated functions to take a gimple_match_op * instead of
9103 separate code_helper * and tree * parameters. Make the same
9104 change in the top-level overload and update calls to the
9105 gimple_resimplify routines. Update calls to the auto-generated
9106 functions and to maybe_push_res_to_seq in the publicly-facing
9107 operation-specific gimple_simplify overloads.
9108 (gimple_match_op::MAX_NUM_OPS): Define.
9109 (gimple_resimplify1): Replace rcode and ops with a single res_op
9110 parameter. Update call to gimple_simplify.
9111 (gimple_resimplify2): Likewise.
9112 (gimple_resimplify3): Likewise.
9113 (mprts_hook): Replace parameters with a gimple_match_op *.
9114 (maybe_build_generic_op): Likewise.
9115 (build_call_internal): Replace type, nargs and ops with
9116 a gimple_match_op *.
9117 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
9118 with a single gimple_match_op *. Update calls to mprts_hook,
9119 build_call_internal and gimple_simplified_result_is_gimple_val.
9120 Factor out code that is common to the tree_code and combined_fn cases.
9121 * genmatch.c (expr::gen_transform): Replace tem_code and
9122 tem_ops with a gimple_match_op called tem_op. Update calls
9123 to the gimple_resimplify functions and maybe_push_res_to_seq.
9124 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
9125 res_ops. Update call to the gimple_resimplify functions.
9126 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
9127 (decision_tree::gen): Make the functions take a gimple_match_op *
9128 called res_op instead of separate res_code and res_ops parameters.
9129 Update call accordingly.
9130 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
9131 and ops with a single res_op parameter. Update calls to
9132 maybe_build_generic_op and maybe_push_res_to_seq.
9133 (fold_stmt_1): Update calls to gimple_simplify and
9134 replace_stmt_with_simplification.
9135 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
9136 and gimple_simplified_result_is_gimple_val.
9137 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
9139 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
9140 with a gimple_match_op *.
9141 (vn_nary_build_or_lookup): Likewise. Update call to
9142 vn_nary_build_or_lookup_1.
9143 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
9144 gimple_match_op *. Update calls to the gimple_resimplify routines
9145 and to gimple_simplified_result_is_gimple_val.
9146 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
9147 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
9148 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
9149 (visit_nary_op): Likewise.
9150 (visit_reference_op_load): Likewise.
9152 2018-05-23 Luis Machado <luis.machado@linaro.org>
9154 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
9155 modifier for printing the step amount.
9157 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
9160 * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
9163 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
9165 * doc/sourcebuild.texi (Endianness): New subsubsection.
9167 2018-05-23 Luis Machado <luis.machado@linaro.org>
9169 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
9170 <prefetch_dynamic_strides>: New const bool field.
9171 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
9172 prefetch_dynamic_strides.
9173 (exynosm1_prefetch_tune): Likewise.
9174 (thunderxt88_prefetch_tune): Likewise.
9175 (thunderx_prefetch_tune): Likewise.
9176 (thunderx2t99_prefetch_tune): Likewise.
9177 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
9179 (aarch64_override_options_internal): Update to set
9180 PARAM_PREFETCH_DYNAMIC_STRIDES.
9181 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
9182 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
9183 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
9184 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
9185 prefetch-dynamic-strides setting.
9187 2018-05-23 Luis Machado <luis.machado@linaro.org>
9189 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
9190 <minimum_stride>: New const int field.
9191 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
9192 minimum_stride field defaulting to -1.
9193 (exynosm1_prefetch_tune): Likewise.
9194 (thunderxt88_prefetch_tune): Likewise.
9195 (thunderx_prefetch_tune): Likewise.
9196 (thunderx2t99_prefetch_tune): Likewise.
9197 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
9198 <default_opt_level>: Set to 3.
9199 (aarch64_override_options_internal): Update to set
9200 PARAM_PREFETCH_MINIMUM_STRIDE.
9201 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
9202 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
9203 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
9204 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
9205 stride is constant and is below the minimum stride threshold.
9207 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9209 * config/arm/arm-cpus.in (mode26): Delete.
9210 (armv4): Delete mode26 reference.
9211 * config/arm/arm.c (arm_configure_build_target): Delete use of
9214 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
9216 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
9218 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
9219 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
9221 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
9222 (floatunsdidf2): Ditto.
9224 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
9226 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
9227 (fixuns_trunc<mode>si2_avx512f): Ditto.
9228 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
9229 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
9230 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
9232 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
9234 PR rtl-optimization/79985
9235 * df-scan.c (df_insn_refs_collect): Remove special case for
9236 global registers and asm statements.
9238 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
9240 * extend.texi (Global Register Variables): Rewrite the bullet list.
9241 Note that the register is available for allocation. Note that access
9242 via inline asm must use constraints. Add note about async-signal
9243 handlers. Remove paragraph about automagic register selection.
9245 2018-05-23 Richard Biener <rguenther@suse.de>
9247 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
9248 of fixed offset from memset VN.
9250 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
9252 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
9254 (alloc_cand_and_find_basis): Initialize first_interp field.
9255 (slsr_process_mul): Modify first_interp field.
9256 (slsr_process_add): Likewise.
9257 (slsr_process_cast): Modify first_interp field for each new
9259 (slsr_process_copy): Likewise.
9260 (dump_candidate): Dump first_interp field.
9261 (replace_mult_candidate): Process all interpretations, not just
9263 (replace_rhs_if_not_dup): Likewise.
9264 (replace_one_candidate): Likewise.
9266 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
9268 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
9270 (aarch64_needs_frame_chain): New function.
9271 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
9273 2018-05-23 Sudakshina Das <sudi.das@arm.com>
9276 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
9277 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
9278 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
9279 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
9280 as true for strict-align.
9281 (aarch64_can_inline_p): Perform checks even when callee has no
9282 attributes to check for strict alignment.
9283 * doc/extend.texi (AArch64 Function Attributes): Document
9285 * doc/invoke.texi: (AArch64 Options): Likewise.
9287 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
9289 PR tree-optimization/85853
9290 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
9291 the handling of the root of the node to...
9292 (vect_slp_analyze_node_operations_1): ...this new function,
9293 and run the whole thing with the child nodes' def types
9294 set according to their SLP node's def type.
9296 2018-05-23 Richard Biener <rguenther@suse.de>
9299 * tree-data-ref.c (create_runtime_alias_checks): Defer
9300 and ignore overflow warnings.
9302 2018-05-23 Yury Gribov <tetra2005@gmail.com>
9304 PR tree-optimization/85822
9305 * tree-vrp.c (is_masked_range_test): Fix handling of negative
9308 2018-05-23 Richard Biener <rguenther@suse.de>
9310 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
9311 memset constants via native_interpret_expr.
9313 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
9316 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
9318 (cgraph_node::create_alias): Likewise.
9319 (cgraph_node::get_availability): Check ifunc_resolver instead
9320 of looking up ifunc attribute.
9321 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
9322 * varasm.c (do_assemble_alias): Likewise.
9323 (assemble_alias): Likewise.
9324 (default_binds_local_p_3): Likewise.
9325 * cgraph.h (cgraph_node): Add ifunc_resolver.
9326 (cgraph_node::only_called_directly_or_aliased_p): Return false
9328 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
9330 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
9331 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
9332 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
9333 instead of looking up ifunc attribute.
9335 2018-05-22 Luis Machado <luis.machado@linaro.org>
9337 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
9339 2018-05-22 Martin Sebor <msebor@redhat.com>
9342 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
9343 only when expasion succeeds.
9344 (expand_builtin_strcmp): Same.
9345 (expand_builtin_strncmp): Same.
9347 2018-05-22 Martin Sebor <msebor@redhat.com>
9349 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
9351 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
9352 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9354 * config/aarch64/aarch64-ldpstp.md: Replace uses of
9355 aarch64_mem_pair_operand with memory_operand and delete operand swapping
9357 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
9358 Add check for legitimate_address.
9359 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
9360 (aarch64_swap_ldrstr_operands): New.
9361 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
9364 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
9365 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9367 * config/aarch64/aarch64.md: New patterns to generate stp
9369 (store_pair_sw, store_pair_dw): New patterns to generate stp for
9370 single words and double words.
9371 (load_pair_sw, load_pair_dw): Likewise.
9372 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
9374 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
9376 * config/aarch64/aarch64-ldpstp.md: Modify peephole
9377 for different mode ldpstp and add peephole for merged zero stores.
9379 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
9381 (aarch64_gen_store_pair): Rename calls to match new patterns.
9382 (aarch64_gen_load_pair): Rename calls to match new patterns.
9383 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
9384 (load_pair<DREG:mode><DREG2:mode>): ... This.
9385 (store_pair<mode>): Rename to...
9386 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
9387 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
9389 (V_INT_EQUIV): Handle SImode.
9390 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
9393 2018-05-22 Martin Sebor <msebor@redhat.com>
9396 * calls.c (maybe_warn_nonstring_arg): Use string length to set
9397 or ajust the presumed bound on an operation to avoid unnecessary
9400 2018-05-22 Martin Sebor <msebor@redhat.com>
9402 PR tree-optimization/85826
9403 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
9404 assuming that a DECL necesarily has a constant size.
9406 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
9409 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
9411 2018-05-22 Richard Biener <rguenther@suse.de>
9413 PR tree-optimization/85834
9414 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
9415 non-constant and non-zero memset arguments.
9417 2018-05-22 Martin Liska <mliska@suse.cz>
9420 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
9422 2018-05-22 Richard Biener <rguenther@suse.de>
9424 PR tree-optimization/85863
9425 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
9426 comparisons when vectype is specified.
9427 (vectorizable_condition): Do not specify vectype for
9428 vect_is_simple_cond when SLP vectorizing.
9430 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
9433 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
9434 define __ibm128 as long double.
9435 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
9436 as a distinct type when IEEE 128-bit support is enabled.
9437 (init_float128_ieee): Fix up conversions between IFmode and IEEE
9438 128-bit types to use the correct functions.
9439 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
9440 convert between 128-bit floating point types that have different
9441 modes but the same representation, instead of using gen_lowpart to
9443 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
9445 (IFKF_reg): New attributes to give the register constraints for
9447 (extend<mode>tf2_internal): New insns to mark an explicit
9448 conversion between 128-bit floating point types that have a
9449 different mode but share the same representation.
9451 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
9453 PR tree-optimization/85814
9454 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
9455 a null return from get_strinfo when unsharing the next
9456 strinfo in the chain.
9458 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
9461 * varasm.c (weak_finish): Clean up weak_decls.
9463 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9465 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
9466 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
9467 UNSPEC_UADALP values.
9468 * config/aarch64/iterators.md (ABAL): New int iterator.
9471 (sur): Add mappings for the above.
9472 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
9474 (aarch64_<sur>abal<mode>_4): Likewise.
9475 (aarch64_<sur>adalp<mode>_3): Likewise.
9476 (<sur>sadv16qi): New define_expand.
9478 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
9480 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
9481 (*movdf_internal): Ditto.
9482 (*rcpsf2_sse): Ditto.
9483 (*rsqrtsf2_sse): Ditto.
9484 (*sqrt<mode>2_sse): Ditto.
9486 2018-05-21 Tamar Christina <tamar.christina@arm.com>
9488 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
9490 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
9491 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
9492 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
9493 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
9495 * config/aarch64/arm_neon.h: Likewise.
9496 * config/aarch64/iterators.md (VQ_I): New.
9498 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
9500 * config.gcc: Add arc/t-multilib-linux to tmake_file for
9502 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
9505 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
9507 * config/nds32/constraints.md (S): New constraint.
9508 * config/nds32/nds32.md (call_internal): Use constraint S.
9509 (call_value_internal): Likewise.
9510 (sibcall_internal): Likewise.
9511 (sibcall_value_internal): Likewise.
9513 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
9514 Chung-Ju Wu <jasonwucj@gmail.com>
9516 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
9519 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
9520 Chung-Ju Wu <jasonwucj@gmail.com>
9522 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
9523 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
9524 (nds32_rtx_costs_impl): Simplify.
9525 (nds32_address_cost_impl): Simplify.
9526 (nds32_init_rtx_costs): New function.
9527 (nds32_rtx_costs_speed_prefer): Likewise.
9528 (nds32_rtx_costs_size_prefer): Likewise.
9529 (nds32_address_cost_speed_prefer): Likewise.
9530 (nds32_address_cost_speed_fwprop): Likewise.
9531 (nds32_address_cost_size_prefer): Likewise.
9532 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
9533 * config/nds32/nds32.c (nds32_option_override): Use
9534 nds32_init_rtx_costs function.
9536 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
9538 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
9539 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
9540 (TARGET_PIPELINE_N8): Likewise.
9541 (TARGET_PIPELINE_N10): Likewise.
9542 (TARGET_PIPELINE_N13): Likewise.
9543 (TARGET_PIPELINE_GRAYWOLF): Likewise.
9545 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
9547 * config/nds32/nds32-fpu.md: Update copyright year.
9549 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
9551 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
9553 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
9555 * config/nds32/nds32.c
9556 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
9557 * config/nds32/nds32.opt (minline-asm-r15): New option.
9559 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
9561 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
9563 * config/nds32/nds32.md (abssi2): New pattern.
9565 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
9567 * config/i386/i386.md (rex64namesuffix): New mode attribute.
9568 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
9569 Merge insn pattern from sse_cvtsi2ss<round_name> and
9570 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
9571 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
9572 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
9573 using SWI48 mode iterator.
9574 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
9575 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
9576 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
9577 pattern from sse_cvttss2si<round_saeonly_name>
9578 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
9579 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
9580 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
9581 using SWI48 mode iterator.
9582 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
9583 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
9584 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
9585 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
9586 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
9587 using SWI48 mode iterator.
9588 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
9589 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
9590 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
9591 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
9592 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
9593 SWI48 mode iterator.
9594 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
9595 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
9596 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
9597 pattern from sse_cvttsd2si<round_saeonly_name>
9598 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
9600 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
9602 * config/nds32/nds32-md-auxiliary.c
9603 (nds32_valid_smw_lwm_base_p): Refine.
9604 (nds32_output_smw_single_word): Refine.
9605 (nds32_output_smw_double_word): New.
9606 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
9608 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
9610 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
9611 (nds32_output_stack_pop): Refine.
9612 (nds32_expand_unaligned_load): Refine.
9613 (nds32_expand_unaligned_store): Refine.
9615 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
9616 Chung-Ju Wu <jasonwucj@gmail.com>
9618 * config/nds32/constants.md: Add TP_REGNUM constant.
9619 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
9620 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
9622 * config/nds32/nds32-doubleword.md: Consider flag_pic.
9623 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
9624 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
9625 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
9626 and PIC code generation.
9627 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
9629 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
9631 * config/nds32/nds32.md: Support TLS and PIC.
9632 * config/nds32/nds32.c: Support TLS and PIC.
9633 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
9634 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
9637 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
9639 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
9640 mode with E_ prefix.
9642 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
9643 Chung-Ju Wu <jasonwucj@gmail.com>
9645 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
9646 * config/nds32/nds32-md-auxiliary.c
9647 (symbolic_reference_mentioned_p): New.
9648 (nds32_legitimize_ict_address): New.
9649 (nds32_expand_ict_move): New.
9650 (nds32_indirect_call_referenced_p): New.
9651 (nds32_symbol_binds_local_p): Delete.
9652 (nds32_long_call_p): Modify.
9653 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
9654 * config/nds32/nds32-protos.h
9655 (symbolic_reference_mentioned_p): Declare.
9656 (nds32_legitimize_ict_address): Declare.
9657 (nds32_expand_ict_move): Declare.
9658 (nds32_indirect_call_referenced_p): Declare.
9659 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
9660 (nds32_relax_group): Use nds32_ict_const_p as condition.
9661 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
9662 (nds32_asm_file_start): Output ict_model directive in asm code.
9663 (nds32_legitimate_address_p): Consider indirect call.
9664 (nds32_print_operand): Consider indirect call.
9665 (nds32_print_operand_address): Consider indirect call.
9666 (nds32_insert_attributes): Handle "indirect_call" attribute.
9667 (TARGET_LEGITIMATE_ADDRESS_P): Define.
9668 (TARGET_LEGITIMATE_CONSTANT_P): Define.
9669 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
9670 (TARGET_DELEGITIMIZE_ADDRESS): Define.
9671 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
9672 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
9673 (TARGET_ICT_MODEL_SMALL): Define.
9674 (TARGET_ICT_MODEL_LARGE): Define.
9675 * config/nds32/nds32.md (movsi): Consider ict model.
9676 (call, call_value): Consider ict model.
9677 (sibcall, sibcall_value): Consider ict model.
9678 * config/nds32/nds32.opt (mict-model): New option.
9679 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
9682 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
9683 Monk Chiang <sh.chiang04@gmail.com>
9684 Jim Wilson <jimw@sifive.com>
9686 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
9687 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
9688 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
9689 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
9690 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
9691 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
9692 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
9693 compute save_libcall_adjustment properly.
9694 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
9695 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
9696 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
9697 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
9698 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
9699 (ABI_SPEC): Handle mabi=ilp32e.
9700 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
9701 (RVE): Add RVE mask.
9702 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
9703 <-march>: Add rv32e as an example.
9705 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
9708 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
9709 (intra_create_variable_infos): Handle C++ constructors.
9711 2018-05-18 Martin Liska <mliska@suse.cz>
9713 * passes.def: Remove a redundant pass.
9715 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
9718 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
9720 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9722 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
9724 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
9726 (armv2, armv2a, armv3, armv3m): Delete architectures.
9727 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
9728 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
9729 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
9731 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
9732 (*mulsidi3adddi): Likewise.
9733 (mulsidi3): Likewise.
9734 (*mulsidi3_nov6): Likewise.
9735 (umulsidi3): Likewise.
9736 (umulsidi3_nov6): Likewise.
9737 (umaddsidi4): Likewise.
9738 (*umulsidi3adddi): Likewise.
9739 (smulsi3_highpart): Likewise.
9740 (*smulsi3_highpart_nov6): Likewise.
9741 (umulsi3_highpart): Likewise.
9742 (*umulsi3_highpart_nov6): Likewise.
9743 * config/arm/arm.h (arm_arch3m): Delete.
9744 * config/arm/arm.c (arm_arch3m): Delete.
9745 (arm_option_override_internal): Update armv3-related comment.
9746 (arm_configure_build_target): Delete use of isa_bit_mode32.
9747 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
9748 (arm_rtx_costs_internal): Delete check of arm_arch3m.
9749 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
9751 (mulusa3): Likewise.
9752 * config/arm/arm-protos.h (arm_arch3m): Delete.
9753 * config/arm/arm-tables.opt: Regenerate.
9754 * config/arm/arm-tune.md: Likewise.
9755 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
9756 deleted architectures.
9758 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9760 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
9761 (armv5t, armv5te): New features.
9762 (ARMv5, ARMv5e): Delete fgroups.
9763 (ARMv5t, ARMv5te): Adjust for above changes.
9765 (armv5, armv5e): Delete arches.
9766 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
9768 (*call_reg_arm): Likewise.
9769 (*call_value_reg_armv5): Likewise.
9770 (*call_value_reg_arm): Likewise.
9771 (*call_symbol): Likewise.
9772 (*call_value_symbol): Likewise.
9773 (*sibcall_insn): Likewise.
9774 (*sibcall_value_insn): Likewise.
9776 (prefetch): Likewise.
9777 (define_split and define_peephole2 dependent on arm_arch5):
9779 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
9781 (TARGET_ARM_QBIT): Likewise.
9782 (TARGET_DSP_MULTIPLY): Likewise.
9783 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
9784 (arm_arch5, arm_arch5e): Delete.
9785 (arm_arch5t, arm_arch5te): Declare.
9786 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
9787 (arm_arch5t): Declare.
9788 (arm_option_reconfigure_globals): Update for the above.
9789 (arm_options_perform_arch_sanity_checks): Update comment, replace
9790 use of arm_arch5 with arm_arch5t.
9791 (use_return_insn): Likewise.
9792 (arm_emit_call_insn): Likewise.
9793 (output_return_instruction): Likewise.
9794 (arm_final_prescan_insn): Likewise.
9795 (arm_coproc_builtin_available): Likewise.
9796 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
9797 arm_arch5e with arm_arch5t and arm_arch5te.
9798 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
9799 (arm_arch5t, arm_arch5te): Declare.
9800 * config/arm/arm-tables.opt: Regenerate.
9801 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
9802 * config/arm/t-multilib: Likewise.
9803 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
9804 instead of arm_arch5.
9805 (*call_reg_thumb1): Likewise.
9806 (*call_value_reg_thumb1_v5): Likewise.
9807 (*call_value_reg_thumb1): Likewise.
9808 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
9810 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
9812 2018-05-18 Martin Liska <mliska@suse.cz>
9814 PR gcov-profile/84846
9815 * doc/gcov.texi: Document -t option of gcov tool.
9817 2018-05-18 Martin Liska <mliska@suse.cz>
9819 PR gcov-profile/84846
9820 * gcov.c (print_usage): Add new -t option.
9821 (process_args): Handle the option.
9822 (generate_results): Use stdout as output when requested by
9825 2018-05-18 Martin Liska <mliska@suse.cz>
9827 PR gcov-profile/84846
9828 * coverage.c (coverage_init): Write PWD to .gcno file.
9829 * doc/gcov.texi: Document how working directory is printed.
9830 * gcov-dump.c (dump_gcov_file): Print PWD.
9831 * gcov.c (output_intermediate_file): Likewise.
9832 (read_graph_file): Read PWD string.
9833 (output_lines): Print PWD.
9835 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9838 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
9839 for retval and return false if all args to phi are zero.
9841 2018-05-18 Richard Biener <rguenther@suse.de>
9843 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
9845 (evrp_dom_walker::before_dom_children): Call it.
9847 2018-05-18 Richard Biener <rguenther@suse.de>
9849 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
9850 results when processing array refs with variable index.
9852 2018-05-18 Toon Moene <toon@moene.org>
9854 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
9855 directly after that of -floop-interchange. Indicate that both
9856 options are enabled by default when specifying -O3.
9858 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9860 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
9861 iterator. Delete separate integer-mode vec_set<mode> expander.
9862 (aarch64_simd_vec_setv2di): Delete.
9863 (vec_setv2di): Delete.
9864 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
9865 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
9866 the "w, r" alternative.
9868 2018-05-18 Martin Liska <mliska@suse.cz>
9870 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
9871 * tree-pass.h (make_pass_lower_switch_O0): New function.
9872 * tree-switch-conversion.c (node_has_low_bound): Remove.
9873 (node_has_high_bound): Likewise.
9874 (node_is_bounded): Likewise.
9875 (class pass_lower_switch): Make it a template type and create
9877 (pass_lower_switch::execute): Add template argument.
9878 (make_pass_lower_switch): New function.
9879 (make_pass_lower_switch_O0): New function.
9880 (do_jump_if_equal): Remove.
9881 (emit_case_nodes): Simplify to just handle all 3 cases and leave
9882 all the hard work to tree optimization passes.
9884 2018-05-18 Martin Liska <mliska@suse.cz>
9886 * dbgcnt.c (limit_low): Renamed from limit.
9887 (limit_high): New variable.
9888 (dbg_cnt_is_enabled): Check for upper limit.
9889 (dbg_cnt): Adjust dumping.
9890 (dbg_cnt_set_limit_by_index): Add new argument for high
9892 (dbg_cnt_set_limit_by_name): Likewise.
9893 (dbg_cnt_process_single_pair): Parse new format.
9894 (dbg_cnt_process_opt): Use strtok.
9895 (dbg_cnt_list_all_counters): Remove 'value' and add
9897 * doc/invoke.texi: Document changes.
9899 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
9901 * doc/sourcebuild.texi (scalar_all_fma): Document.
9902 * tree.def (FMA_EXPR): Delete.
9903 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
9904 * internal-fn.c (ternary_direct): New macro.
9905 (expand_ternary_optab_fn): Likewise.
9906 (direct_ternary_optab_supported_p): Likewise.
9907 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
9908 * builtins.c (fold_builtin_fma): Delete.
9909 (fold_builtin_3): Don't call it.
9910 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
9911 * expr.c (expand_expr_real_2): Likewise.
9912 * fold-const.c (operand_equal_p): Likewise.
9913 (fold_ternary_loc): Likewise.
9914 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
9915 * gimple.c (DEFTREECODE): Likewise.
9916 * gimplify.c (gimplify_expr): Likewise.
9917 * optabs-tree.c (optab_for_tree_code): Likewise.
9918 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
9919 * tree-eh.c (operation_could_trap_p): Likewise.
9920 (stmt_could_throw_1_p): Likewise.
9921 * tree-inline.c (estimate_operator_cost): Likewise.
9922 * tree-pretty-print.c (dump_generic_node): Likewise.
9923 (op_code_prio): Likewise.
9924 * tree-ssa-loop-im.c (stmt_cost): Likewise.
9925 * tree-ssa-operands.c (get_expr_operands): Likewise.
9926 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
9927 * fold-const-call.h (fold_fma): Delete.
9928 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
9929 CFN_FNMA and CFN_FNMS.
9931 * genmatch.c (combined_fn): New enum.
9932 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
9933 (commutative_op): New function.
9934 (commutate): Use it. Handle more than 2 operands.
9935 (dt_operand::gen_gimple_expr): Use commutative_op.
9936 (parser::parse_expr): Allow :c to be used with non-binary
9937 operators if the commutative operand is known.
9938 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
9939 CFN_FMS, CFN_FNMA and CFN_FNMS.
9940 (backprop::process_assign_use): Remove FMA_EXPR handling.
9941 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
9942 (gen_hsa_fma): New function.
9943 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
9944 IFN_FNMA and IFN_FNMS.
9945 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
9946 * gimple-fold.h (follow_all_ssa_edges): Declare.
9947 * gimple-fold.c (follow_all_ssa_edges): New function.
9948 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
9949 gimple_build interface and use follow_all_ssa_edges to fold the result.
9950 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
9951 instead of checking for optabs directly.
9952 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
9953 rather than FMA_EXPRs.
9954 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
9955 call to IFN_FMA instead of an FMA_EXPR.
9957 2018-05-17 Jim Wilson <jimw@sifive.com>
9959 * expr.c (do_tablejump): When converting index to Pmode, if we have a
9960 sign extended promoted subreg, and the range does not have the sign bit
9961 set, then do a sign extend.
9963 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
9964 test, check for sign extended subreg and/or constant operands, and
9965 do a sign extend in that case.
9967 2018-05-17 Steve Ellcey <sellcey@cavium.com>
9969 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
9970 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
9972 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
9973 Change logics_shift_reg to logics_shift_imm.
9974 (thunderx2t99_fp_loadpair_basic): Delete.
9975 (thunderx2t99_fp_storepair_basic): Delete.
9976 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
9977 (thunderx2t99_asimd_polynomial): Delete.
9978 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
9979 and neon_fp_mul_d_scalar_q.
9980 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
9981 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
9982 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
9983 (thunderx2t99_asimd_lut): Add missing tbl types.
9984 (thunderx2t99_asimd_ext): Delete.
9985 (thunderx2t99_asimd_load1_1_mult): Delete.
9986 (thunderx2t99_asimd_load1_2_mult): Delete.
9987 (thunderx2t99_asimd_load1_ldp): New.
9988 (thunderx2t99_asimd_load1): New.
9989 (thunderx2t99_asimd_load2): Add missing *load2* types.
9990 (thunderx2t99_asimd_load3): New.
9991 (thunderx2t99_asimd_load4): New.
9992 (thunderx2t99_asimd_store1_1_mult): Delete.
9993 (thunderx2t99_asimd_store1_2_mult): Delete.
9994 (thunderx2t99_asimd_store2_mult): Delete.
9995 (thunderx2t99_asimd_store2_onelane): Delete.
9996 (thunderx2t99_asimd_store_stp): New.
9997 (thunderx2t99_asimd_store1): New.
9998 (thunderx2t99_asimd_store2): New.
9999 (thunderx2t99_asimd_store3): New.
10000 (thunderx2t99_asimd_store4): New.
10002 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
10004 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
10005 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
10007 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
10008 Segher Boessenkool <segher@kernel.crashing.org>
10011 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
10014 2018-05-17 Richard Biener <rguenther@suse.de>
10016 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
10017 for pruning loop and prune defs feeding only already visited PHIs.
10019 2018-05-17 Richard Biener <rguenther@suse.de>
10021 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
10023 2018-05-17 Bin Cheng <bin.cheng@arm.com>
10024 Richard Biener <rguenther@suse.de>
10026 PR tree-optimization/85793
10027 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
10028 for VMAT_ELEMENTWISE.
10030 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
10032 * internal-fn.h (lookup_internal_fn): Declare
10033 * internal-fn.c (lookup_internal_fn): New function.
10034 * gimple.c (gimple_build_call_from_tree): Handle calls to
10035 internal functions.
10036 * gimple-pretty-print.c (dump_gimple_call): Print "." before
10037 internal function names.
10038 * tree-pretty-print.c (dump_generic_node): Likewise.
10039 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
10041 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
10043 * gimple-fold.h (gimple_build): Make the function forms take
10044 combined_fn rather than built_in_function.
10045 (gimple_simplify): Likewise.
10046 * gimple-match-head.c (gimple_simplify): Likewise.
10047 * gimple-fold.c (gimple_build): Likewise.
10048 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
10049 rather than gimple_build_call_internal.
10050 (get_initial_defs_for_reduction): Likewise.
10051 (vect_create_epilog_for_reduction): Likewise.
10052 (vectorizable_live_operation): Likewise.
10054 2018-05-17 Martin Liska <mliska@suse.cz>
10056 * gimple-ssa-sprintf.c (format_directive): Do not use
10057 space in between 'G_' and '('.
10059 2018-05-17 Jakub Jelinek <jakub@redhat.com>
10062 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
10063 even if the mask is not all ones.
10066 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
10068 (ix86_gimple_fold_builtin): Likewise.
10071 * config/i386/i386.c: Include tree-vector-builder.h.
10072 (ix86_vector_shift_count): New function.
10073 (ix86_fold_builtin): Fold shift builtins by scalar count.
10074 (ix86_gimple_fold_builtin): Likewise.
10076 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
10077 _mm512_setzero): New intrinsics.
10079 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
10080 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10082 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
10083 code generation for cases where splatting a value is not useful.
10084 * simplify-rtx.c (simplify_ternary_operation): Simplify
10085 vec_merge across a vec_duplicate and a paradoxical subreg forming
10086 a vector mode to a vec_concat.
10088 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
10090 * config.gcc: Support "goldmont-plus".
10091 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
10093 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
10094 PROCESSOR_GOLDMONT_PLUS.
10095 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
10096 (processor_target_table): Add "goldmont-plus".
10097 (PTA_GOLDMONT_PLUS): Define.
10098 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
10099 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
10100 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
10101 (fold_builtin_cpu): Add "goldmont-plus".
10102 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
10103 (ix86_option_override_internal): Add "goldmont-plus".
10104 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
10105 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
10106 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
10107 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
10109 2018-05-17 Richard Biener <rguenther@suse.de>
10111 PR tree-optimization/85757
10112 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
10113 remove defs that only feed that PHI from further processing.
10115 2018-05-16 Jim Wilson <jimw@sifive.com>
10117 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
10119 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
10121 2018-05-16 Mark Wielaard <mark@klomp.org>
10123 * dwarf2out.c (count_index_strings): New function.
10124 (output_indirect_strings): Call count_index_strings and generate
10125 header for dwarf_version >= 5.
10127 2018-05-16 Mark Wielaard <mark@klomp.org>
10129 * dwarf2out.c (dwarf_FORM): New function.
10130 (set_indirect_string): Use dwarf_FORM.
10131 (reset_indirect_string): Likewise.
10132 (size_of_die): Likewise.
10133 (value_format): Likewise.
10134 (output_die): Likewise.
10135 (add_skeleton_AT_string): Likewise.
10136 (output_macinfo_op): Likewise.
10137 (index_string): Likewise.
10138 (output_index_string_offset): Likewise.
10139 (output_index_string): Likewise.
10140 (count_index_strings): Likewise.
10142 2018-05-16 Carl Love <cel@us.ibm.com>
10144 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
10145 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
10147 2018-05-16 Martin Jambor <mjambor@suse.cz>
10149 * ipa-prop.c (ipa_free_all_edge_args): Remove.
10150 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
10152 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
10154 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
10155 (fnma<mode>4): Likewise.
10156 (fms<mode>4): Likewise.
10157 (fnms<mode>4): Likewise.
10158 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
10159 (aarch64_fnma<mode>4): Likewise.
10160 (aarch64_fms<mode>4): Likewise.
10161 (aarch64_fnms<mode>4): Likewise.
10162 (aarch64_fnmadd<mode>4): Likewise.
10164 2018-05-16 Jason Merrill <jason@redhat.com>
10166 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
10168 2018-05-16 Richard Biener <rguenther@suse.de>
10170 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
10171 (dump_stmt_cost): Declare.
10172 (add_stmt_cost): Dump cost we add.
10173 (add_stmt_costs): New function.
10174 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
10175 No longer exported.
10176 (vect_analyze_stmt): Adjust prototype.
10177 (vectorizable_condition): Likewise.
10178 (vectorizable_live_operation): Likewise.
10179 (vectorizable_reduction): Likewise.
10180 (vectorizable_induction): Likewise.
10181 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
10182 cost vector to pass to vectorizable_ and record afterwards.
10183 (vect_model_reduction_cost): Take cost vector argument and adjust.
10184 (vect_model_induction_cost): Likewise.
10185 (vectorizable_reduction): Likewise.
10186 (vectorizable_induction): Likewise.
10187 (vectorizable_live_operation): Likewise.
10188 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
10189 SLP_TREE_NUMBER_OF_VEC_STMTS.
10190 (vect_analyze_slp_cost_1): Remove.
10191 (vect_analyze_slp_cost): Likewise.
10192 (vect_slp_analyze_node_operations): Take visited args and
10193 a target cost vector. Avoid processing already visited stmt sets.
10194 (vect_slp_analyze_operations): Use a local cost vector to gather
10195 costs and register those of non-discarded instances.
10196 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
10197 (vect_schedule_slp_instance): Remove copying of
10198 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
10200 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
10201 adding cost. Record cost entry location.
10202 (vect_prologue_cost_for_slp_op): Function to compute cost of
10203 a constant or invariant generated for SLP vect in the prologue,
10204 split out from vect_analyze_slp_cost_1.
10205 (vect_model_simple_cost): Make static. Adjust for SLP costing.
10206 (vect_model_promotion_demotion_cost): Likewise.
10207 (vect_model_store_cost): Likewise, make static.
10208 (vect_model_load_cost): Likewise.
10209 (vectorizable_bswap): Add cost vector arg and adjust.
10210 (vectorizable_call): Likewise.
10211 (vectorizable_simd_clone_call): Likewise.
10212 (vectorizable_conversion): Likewise.
10213 (vectorizable_assignment): Likewise.
10214 (vectorizable_shift): Likewise.
10215 (vectorizable_operation): Likewise.
10216 (vectorizable_store): Likewise.
10217 (vectorizable_load): Likewise.
10218 (vectorizable_condition): Likewise.
10219 (vectorizable_comparison): Likewise.
10220 (can_vectorize_live_stmts): Likewise.
10221 (vect_analyze_stmt): Likewise.
10222 (vect_transform_stmt): Adjust calls to vectorizable_*.
10223 * tree-vectorizer.c: Include gimple-pretty-print.h.
10224 (dump_stmt_cost): New function.
10226 2018-05-16 Richard Biener <rguenther@suse.de>
10228 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
10229 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
10230 * tree-ssa-dse.c: Include tree-ssa-loop.h.
10231 (check_name): New callback.
10232 (dse_classify_store): Track cycles via a visited bitmap of PHI
10233 defs and simplify handling of in-loop and across loop dead stores
10234 and properly fail for loop-variant refs. Handle byte-tracking with
10235 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
10238 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
10240 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
10241 (vect_get_mask_type_for_stmt): Likewise.
10242 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
10244 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
10245 to determine the statement's vector type and the vector type that
10246 should be used for calculating nunits. Deal with cases in which
10247 the type has to be deferred.
10248 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
10249 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
10250 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
10251 (vect_determine_vf_for_stmt): New functions, split out from...
10252 (vect_determine_vectorization_factor): ...here.
10253 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
10254 (vect_get_mask_type_for_stmt): New functions, split out from
10255 vect_determine_vectorization_factor.
10257 2018-05-16 Richard Biener <rguenther@suse.de>
10259 * tree-cfg.c (verify_gimple_assign_ternary): Properly
10260 verify the [VEC_]COND_EXPR embedded comparison.
10262 2018-05-15 Martin Sebor <msebor@redhat.com>
10264 PR tree-optimization/85753
10265 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
10266 RECORD_TYPE in addition to ARRAY_TYPE.
10268 2018-05-15 Martin Sebor <msebor@redhat.com>
10270 PR middle-end/85643
10271 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
10273 2018-05-15 Richard Biener <rguenther@suse.de>
10275 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
10276 add by_clobber_p one. Change algorithm to collect all defs
10277 representing uses we need to walk and try reducing them to
10278 a single one before failing.
10279 (dse_dom_walker::dse_optimize_stmt): Adjust.
10281 2018-05-13 Mark Wielaard <mark@klomp.org>
10283 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
10284 (size_of_loc_descr): Likewise.
10285 (output_loc_operands): Likewise.
10286 (output_loc_operands_raw): Likewise.
10287 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
10288 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
10289 (hash_loc_operands): Likewise.
10290 (compare_loc_operands): Likewise.
10292 2018-05-14 Mark Wielaard <mark@klomp.org>
10294 * dwarf2out.c (count_index_addrs): New function.
10295 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
10297 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10299 PR tree-optimization/83648
10300 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
10301 return value as malloc candidate.
10303 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10306 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
10307 param as true in call to suggest_attribute.
10309 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
10311 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
10312 -mreadonly-in-sdata.
10314 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10316 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
10318 (aarch64_crypto_aesd_fused): Likewise.
10320 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
10322 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
10323 (movsi_aarch64): Likewise.
10324 (load_pairsi): Likewise.
10325 (load_pairdi): Likewise.
10326 (store_pairsi): Likewise.
10327 (store_pairdi): Likewise.
10328 (load_pairsf): Likewise.
10329 (load_pairdf): Likewise.
10330 (store_pairsf): Likewise.
10331 (store_pairdf): Likewise.
10332 (zero_extend): Likewise.
10333 (trunc): Swap alternatives.
10334 (fcvt_target): Add '?' to prefer w over r.
10336 2018-05-14 Jakub Jelinek <jakub@redhat.com>
10339 * config/i386/i386.md: Disallow non-commutative arithmetics in
10340 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
10341 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
10342 in the peephole2 before it.
10344 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
10346 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
10347 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
10348 (ix86_handle_option): Handle -mcldemote.
10349 * config.gcc: New header.
10350 * config/i386/cldemoteintrin.h: New file.
10351 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
10352 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
10354 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
10355 OPTION_MASK_ISA_CLDEMOTE.
10356 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
10357 (ix86_valid_target_attribute_inner_p): Ditto.
10358 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
10359 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
10360 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
10361 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
10362 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
10364 * config/i386/i386.opt: Add -mcldemote.
10365 * config/i386/x86intrin.h: New header.
10366 * doc/invoke.texi: Add -mcldemote.
10368 2018-05-14 Richard Biener <rguenther@suse.de>
10370 * doc/match-and-simplify.texi: Adjust :s documentation.
10372 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
10374 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
10375 intended memcpy size.
10376 (REORDER_45): Likewise.
10378 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
10380 * sort.cc: New file.
10381 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
10382 * vec.c (qsort_chk): Use gcc_qsort.
10383 * Makefile.in (OBJS-libcommon): Add sort.o.
10384 (build/sort.o): New target. Use it...
10385 (BUILD_RTL): ... here, and...
10386 (build/gencfn-macros): ... here, and...
10387 (build/genmatch): ... here.
10389 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
10390 Chung-Ju Wu <jasonwucj@gmail.com>
10392 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
10393 * config/nds32/nds32-graywolf.md: New file.
10394 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
10395 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
10397 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
10398 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
10399 * config/nds32/nds32.md (pipeline_model): Add graywolf.
10400 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
10401 * config/nds32/pipelines.md: Include n15 settings.
10403 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
10404 Chung-Ju Wu <jasonwucj@gmail.com>
10406 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
10407 * config/nds32/nds32-n13.md: New file.
10408 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
10409 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
10411 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
10412 * config/nds32/nds32.md (pipeline_model): Add n13.
10413 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
10414 * config/nds32/pipelines.md: Include n13 settings.
10416 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
10417 Chung-Ju Wu <jasonwucj@gmail.com>
10419 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
10420 * config/nds32/nds32-n10.md: New file.
10421 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
10422 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
10424 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
10425 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
10426 * config/nds32/nds32.md (pipeline_model): Add n10.
10427 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
10428 * config/nds32/pipelines.md: Include n10 settings.
10430 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
10431 Kito Cheng <kito.cheng@gmail.com>
10432 Chung-Ju Wu <jasonwucj@gmail.com>
10434 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
10435 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
10436 Add enum values for DSP extension instructions.
10437 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
10439 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
10440 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
10441 New code iterators.
10442 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
10443 * config/nds32/nds32-dspext.md: New file for DSP implementation.
10444 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
10445 * config/nds32/nds32-intrinsic.md: Likewise.
10446 * config/nds32/nds32_intrinsic.h: Likewise.
10447 * config/nds32/nds32-md-auxiliary.c: Likewise.
10448 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
10449 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
10450 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
10451 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
10452 * config/nds32/nds32-protos.h: New declarations for DSP extension.
10453 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
10454 TYPE_DMAC in switch statement.
10455 * config/nds32/nds32.c: New checking and implementation for DSP
10456 extension instructions.
10457 * config/nds32/nds32.h: Likewise.
10458 * config/nds32/nds32.md: Likewise.
10459 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
10460 * config/nds32/predicates.md: Implement new predicates for DSP
10463 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
10465 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
10466 Reformat alternatives and attributes so it is easier to identify
10467 which constraints/attributes go with which instruction.
10468 (mov<mode>_hardfloat32, FMOVE64): Likewise.
10469 (mov<mode>_softfloat32, FMOVE64): Likewise.
10470 (mov<mode>_hardfloat64, FMOVE64): Likewise.
10471 (mov<mode>_softfloat64, FMOVE64): Likewise.
10473 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
10475 * doc/extend.texi (PowerPC Built-in Functions): Rename this
10477 (Basic PowerPC Built-in Functions): The new name of the
10478 subsection previously known as "PowerPC Built-in Functions".
10479 (Basic PowerPC Built-in Functions Available on all Configurations):
10481 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
10482 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
10483 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
10484 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
10486 2018-05-11 Martin Jambor <mjambor@suse.cz>
10489 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
10492 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
10495 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
10497 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
10499 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
10500 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
10501 (ix86_handle_option): Handle -mwaitpkg.
10502 * config.gcc: New header.
10503 * config/i386/cpuid.h (bit_WAITPKG): New bit.
10504 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
10505 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
10507 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
10508 OPTION_MASK_ISA_WAITPKG.
10509 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
10510 (ix86_option_override_internal): Add PTA_WAITPKG.
10511 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
10512 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
10513 IX86_BUILTIN_TPAUSE.
10514 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
10515 __builtin_ia32_umwait and __builtin_ia32_tpause.
10516 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
10517 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
10518 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
10519 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
10520 UNSPECV_TPAUSE): New.
10521 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
10522 * config/i386/i386.opt: Add -mwaitpkg.
10523 * config/i386/waitpkgintrin.h: New file.
10524 * config/i386/x86intrin.h: New header.
10525 * doc/invoke.texi: Add -mwaitpkg.
10527 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
10530 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
10532 (cortex-m0): Use armv6s-m isa.
10533 (cortex-m0plus): Likewise.
10534 (cortex-m1): Likewise.
10535 (cortex-m0.small-multiply): Likewise.
10536 (cortex-m0plus.small-multiply): Likewise.
10537 (cortex-m1.small-multiply): Likewise.
10539 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
10540 Jakub Jelinek <jakub@redhat.com>
10542 PR tree-optimization/85692
10543 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
10544 source permute as well.
10546 2018-05-11 Martin Liska <mliska@suse.cz>
10549 * doc/extend.texi: Document LLVM style format for no_sanitize
10552 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
10554 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
10555 mode_supports_vsx_dform_quad to mode_supports_dq_form.
10556 (mode_supports_vsx_dform_quad): Likewise.
10557 (mode_supports_vmx_dform): Move these functions to be next to the
10558 other mode_supports functions.
10559 (mode_supports_dq_form): Likewise.
10560 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
10561 mode_supports_dq_form.
10562 (reg_offset_addressing_ok_p): Likewise.
10563 (offsettable_ok_by_alignment): Likewise.
10564 (rs6000_legitimate_offset_address_p): Likewise.
10565 (legitimate_lo_sum_address_p): Likewise.
10566 (rs6000_legitimize_address): Likewise.
10567 (rs6000_legitimize_reload_address): Likewise.
10568 (rs6000_secondary_reload_inner): Likewise.
10569 (rs6000_preferred_reload_class): Likewise.
10570 (rs6000_output_move_128bit): Likewise.
10572 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
10574 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
10575 Generate SImode target register for null target.
10576 <case IX86_BUILTIN_XGETBV>: Ditto.
10577 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
10578 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
10580 2018-05-10 Carl Love <cel@us.ibm.com>
10582 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
10583 dcbtt and dcbtstt if operands[2] is 0.
10585 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
10588 * config/i386/sse.md (usadv64qi): New expander.
10590 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
10592 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
10593 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
10594 -maltivec=be support.
10595 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
10596 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
10597 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
10598 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
10599 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
10600 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
10601 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
10602 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
10603 altivec_vsumsws): Adjust.
10604 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
10605 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
10606 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
10608 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
10609 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
10610 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
10611 (altivec_lve<VI_char>x): Delete expand.
10612 (*altivec_lve<VI_char>x_internal): Rename to...
10613 (altivec_lve<VI_char>x): ... this.
10614 (altivec_lvxl_<mode>): Delete expand.
10615 (*altivec_lvxl_<mode>_internal): Rename to ...
10616 (altivec_lvxl_<mode>): ... this.
10617 (altivec_stvxl_<mode>): Delete expand.
10618 (*altivec_stvxl_<mode>_internal): Rename to ...
10619 (altivec_stvxl_<mode>): ... this.
10620 (altivec_stve<VI_char>x): Delete expand.
10621 (*altivec_stve<VI_char>x_internal): Rename to ...
10622 (altivec_stve<VI_char>x): ... this.
10623 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
10624 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
10625 reduc_plus_scal_<mode>): Adjust.
10626 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
10628 (rs6000_cpu_cpp_builtins): Adjust.
10629 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
10630 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
10631 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
10632 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
10633 -maltivec=be support.
10634 (rs6000_split_vec_extract_var): Adjust.
10635 (rs6000_split_v4si_init): Adjust.
10636 (swap_selector_for_mode): Delete.
10637 (altivec_expand_lvx_be, altivec_expand_stvx_be,
10638 altivec_expand_stvex_be): Delete.
10639 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
10640 -maltivec=be support.
10641 (rs6000_gimple_fold_builtin): Ditto.
10642 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
10644 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
10645 (TARGET_DIRECT_MOVE_64BIT): Adjust.
10646 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
10647 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
10648 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
10649 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
10650 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
10651 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
10652 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
10653 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
10654 anonymous split): Adjust.
10655 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
10656 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
10658 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
10660 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
10661 when --with-gxx-include-dir is also specified.
10662 * configure: Regenerate.
10664 2018-05-09 Jim Wilson <jimw@sifive.com>
10667 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
10668 * config/riscv/t-withmultilib: New.
10669 * config/riscv/withmultilib.h: New.
10670 * doc/install.texi: Document RISC-V --with-multilib-list support.
10672 2018-05-09 Richard Biener <rguenther@suse.de>
10674 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
10676 (vect_bb_vectorization_profitable_p): Adjust. Compute
10677 actual scalar cost using the cost vector and the add_stmt_cost
10680 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
10682 PR rtl-optimization/85645
10683 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
10684 in the REG_CFA_REGISTER note for LR, don't leave it empty.
10686 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
10688 PR rtl-optimization/85645
10689 * shrink-wrap.c (spread_components): Return a boolean saying if
10690 anything was changed.
10691 (try_shrink_wrapping_separate): Iterate spread_components until
10692 nothing changes anymore.
10694 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
10696 PR rtl-optimization/85645
10697 * regrename.c (build_def_use): Also kill the chains that include the
10698 destination of a REG_CFA_REGISTER note.
10700 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
10702 PR rtl-optimization/85645
10703 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
10704 insn that has a REG_CFA_REGISTER note.
10706 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
10708 * cfgexpand.c (expand_clobber): New function.
10709 (expand_gimple_stmt_1): Use it.
10710 * tree-vect-stmts.c (vect_clobber_variable): New function,
10712 (vectorizable_simd_clone_call): ...here.
10713 (vectorizable_store): Emit a clobber either side of an
10714 IFN_STORE_LANES sequence.
10715 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
10717 2018-05-09 Tom de Vries <tom@codesourcery.com>
10720 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
10721 (define_insn "trap_if_false"): Add exit after trap.
10723 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
10725 PR rtl-optimization/85638
10726 * bb-reorder.c: Include common/common-target.h.
10727 (create_forwarder_block): New function extracted from...
10728 (fix_up_crossing_landing_pad): ...here. Rename into...
10729 (dw2_fix_up_crossing_landing_pad): ...this.
10730 (sjlj_fix_up_crossing_landing_pad): New function.
10731 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
10732 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
10733 from both partitions and exit the loop after one iteration.
10735 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
10738 * doc/extend.texi (PowerPC Built-in Functions): Rename this
10740 (Basic PowerPC Built-in Functions): The new name of the
10741 subsection previously known as "PowerPC Built-in Functions".
10742 (Basic PowerPC Built-in Functions Available on all Configurations):
10744 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
10746 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
10748 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
10750 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
10753 2018-05-08 Jim Wilson <jimw@sifive.com>
10755 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
10756 (LD_EMUL_SUFFIX): New.
10757 (LINK_SPEC): Use it.
10759 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
10761 * doc/extend.texi (PowerPC Built-in Functions): Rename this
10763 (Basic PowerPC Built-in Functions): The new name of the
10764 subsection previously known as "PowerPC Built-in Functions".
10765 (Basic PowerPC Built-in Functions Available on all Configurations):
10767 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
10769 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
10771 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
10773 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
10776 2018-05-08 Jakub Jelinek <jakub@redhat.com>
10779 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
10780 after cmpelim optimization.
10782 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
10784 * config.gcc: Support "goldmont".
10785 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
10786 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
10787 PROCESSOR_GOLDMONT.
10788 * config/i386/i386.c (m_GOLDMONT): Define.
10789 (processor_target_table): Add "goldmont".
10790 (PTA_GOLDMONT): Define.
10791 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
10792 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
10793 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
10794 (fold_builtin_cpu): Add "goldmont".
10795 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
10796 (ix86_option_override_internal): Add "goldmont".
10797 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
10798 (processor_type): Add PROCESSOR_GOLDMONT.
10799 * config/i386/i386.md: Add CPU "glm".
10800 * config/i386/glm.md: New file.
10801 * config/i386/x86-tune.def: Add m_GOLDMONT.
10802 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
10804 2018-05-08 Jakub Jelinek <jakub@redhat.com>
10807 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
10809 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
10810 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
10811 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
10814 * config/i386/i386.c (ix86_fold_builtin): Handle
10815 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
10818 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
10819 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
10821 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
10824 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
10825 (check_arch): Likewise.
10826 (check_fpu): Return the result rather than printing it.
10827 (end arch): Fix operator precedence.
10828 (end cpu): Likewise.
10829 (END): Print the result from check_fpu.
10831 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
10832 Alan Hayward <alan.hayward@arm.com>
10833 David Sherwood <david.sherwood@arm.com>
10835 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
10836 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
10837 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
10838 (*fcmuo<mode>_and): New patterns.
10840 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
10842 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
10843 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
10844 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
10845 (cmp_op, sve_imm_con): New code attributes.
10846 (SVE_COND_INT_CMP, imm_con): Delete.
10847 (cmp_op): Remove above unspecs from int attribute.
10848 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
10850 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
10851 comparison-specific unspecs.
10852 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
10853 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
10854 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
10855 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
10856 (*vec_fcm<cmp_op><mode>): Rename to...
10857 (*fcm<cmp_op><mode>): ...this and adjust likewise.
10858 (*vec_fcmuo<mode>): Rename to...
10859 (*fcmuo<mode>): ...this and adjust likewise.
10860 (*pred_fcm<cmp_op><mode>): New pattern.
10861 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
10862 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
10864 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
10866 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
10867 (aarch64_emit_sve_predicated_cond): New function.
10868 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
10869 (aarch64_emit_unspec_cond_or): Replace with...
10870 (aarch64_emit_sve_or_conds): ...this new function. Use
10871 aarch64_emit_sve_ptrue_op for the individual comparisons and
10872 aarch64_emit_binop to OR them together.
10873 (aarch64_emit_inverted_unspec_cond): Replace with...
10874 (aarch64_emit_sve_inverted_cond): ...this new function. Use
10875 aarch64_emit_sve_ptrue_op for the comparison and
10876 aarch64_emit_unop to invert the result.
10877 (aarch64_expand_sve_vec_cmp_float): Update after the above
10878 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
10880 2018-05-07 Nathan Sidwell <nathan@acm.org>
10882 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
10883 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
10884 (Backwards Compatibility): Likewise.
10886 2018-05-07 Luis Machado <luis.machado@linaro.org>
10890 2018-05-07 Luis Machado <luis.machado@linaro.org>
10892 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
10893 <prefetch_dynamic_strides>: New const bool field.
10894 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
10895 prefetch_dynamic_strides.
10896 (exynosm1_prefetch_tune): Likewise.
10897 (thunderxt88_prefetch_tune): Likewise.
10898 (thunderx_prefetch_tune): Likewise.
10899 (thunderx2t99_prefetch_tune): Likewise.
10900 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
10902 (aarch64_override_options_internal): Update to set
10903 PARAM_PREFETCH_DYNAMIC_STRIDES.
10904 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
10905 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
10906 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
10907 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
10908 prefetch-dynamic-strides setting.
10910 2018-05-07 Luis Machado <luis.machado@linaro.org>
10912 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
10913 <minimum_stride>: New const int field.
10914 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
10915 minimum_stride field.
10916 (exynosm1_prefetch_tune): Likewise.
10917 (thunderxt88_prefetch_tune): Likewise.
10918 (thunderx_prefetch_tune): Likewise.
10919 (thunderx2t99_prefetch_tune): Likewise.
10920 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
10921 (aarch64_override_options_internal): Update to set
10922 PARAM_PREFETCH_MINIMUM_STRIDE.
10923 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
10924 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
10925 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
10926 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
10927 stride is constant and is below the minimum stride threshold.
10929 2018-05-07 Luis Machado <luis.machado@linaro.org>
10931 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
10934 2018-05-07 Luis Machado <luis.machado@linaro.org>
10936 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
10937 <prefetch_dynamic_strides>: New const bool field.
10938 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
10939 prefetch_dynamic_strides.
10940 (exynosm1_prefetch_tune): Likewise.
10941 (thunderxt88_prefetch_tune): Likewise.
10942 (thunderx_prefetch_tune): Likewise.
10943 (thunderx2t99_prefetch_tune): Likewise.
10944 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
10946 (aarch64_override_options_internal): Update to set
10947 PARAM_PREFETCH_DYNAMIC_STRIDES.
10948 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
10949 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
10950 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
10951 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
10952 prefetch-dynamic-strides setting.
10954 2018-05-07 Luis Machado <luis.machado@linaro.org>
10956 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
10957 <minimum_stride>: New const int field.
10958 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
10959 minimum_stride field.
10960 (exynosm1_prefetch_tune): Likewise.
10961 (thunderxt88_prefetch_tune): Likewise.
10962 (thunderx_prefetch_tune): Likewise.
10963 (thunderx2t99_prefetch_tune): Likewise.
10964 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
10965 (aarch64_override_options_internal): Update to set
10966 PARAM_PREFETCH_MINIMUM_STRIDE.
10967 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
10968 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
10969 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
10970 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
10971 stride is constant and is below the minimum stride threshold.
10973 2018-05-06 Jakub Jelinek <jakub@redhat.com>
10976 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
10977 the type is addressable. Don't force op into register if it has
10980 2018-05-05 Roland McGrath <mcgrathr@google.com>
10983 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
10984 any section for which we don't know a specific type it should have,
10985 regardless of name. Previously this was done only for the exact
10986 names ".init_array", ".fini_array", and ".preinit_array".
10987 (default_elf_asm_named_section): Add comment about
10988 relationship with default_section_type_flags and SECTION_NOTYPE.
10989 (get_section): Don't consider it a type conflict if one side has
10990 SECTION_NOTYPE and the other doesn't, as long as neither has the
10991 SECTION_BSS et al used in the default_section_type_flags logic.
10993 2018-05-05 Tom de Vries <tom@codesourcery.com>
10996 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
10997 (workaround_barsyncs): New function.
10998 (nvptx_reorg): Use workaround_barsyncs.
10999 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
11000 (define_expand "nvptx_membar_cta"): New define_expand.
11001 (define_insn "*nvptx_membar_cta"): New insn.
11003 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
11005 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
11006 To improve optimization opportunities.
11007 * builtin-types.def: The new needed builtin types for the above.
11009 2018-05-04 Richard Biener <rguenther@suse.de>
11011 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
11012 * gimple-ssa-store-merging.c
11013 (imm_store_chain_info::output_merged_store): Remove redundant create,
11014 release split_store vector contents on failure.
11015 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
11016 scalar stmt vector on cache hit.
11018 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
11020 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
11022 * config.gcc (powerpc-xilinx-eabi*): Remove.
11023 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
11025 (fusion_addis_mem_combo_load): Ditto.
11026 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
11028 (rs6000_cpu_cpp_builtins): Ditto.
11029 * config/rs6000/rs6000-linux.c
11030 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
11031 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
11032 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
11034 (rs6000_setup_reg_addr_masks): Ditto.
11035 (rs6000_init_hard_regno_mode_ok): Ditto.
11036 (rs6000_option_override_internal): Ditto.
11037 (legitimate_lo_sum_address_p): Ditto.
11038 (rs6000_legitimize_address): Ditto.
11039 (rs6000_legitimize_reload_address): Ditto.
11040 (rs6000_legitimate_address_p): Ditto.
11041 (abi_v4_pass_in_fpr): Ditto.
11042 (setup_incoming_varargs): Ditto.
11043 (rs6000_gimplify_va_arg): Ditto.
11044 (rs6000_split_multireg_move): Ditto.
11045 (rs6000_savres_strategy): Ditto.
11046 (rs6000_emit_prologue_components): Ditto.
11047 (rs6000_emit_epilogue_components): Ditto.
11048 (rs6000_emit_prologue): Ditto.
11049 (rs6000_emit_epilogue): Ditto.
11050 (rs6000_elf_file_end): Ditto.
11051 (rs6000_function_value): Ditto.
11052 (rs6000_libcall_value): Ditto.
11053 * config/rs6000/rs6000.h: Ditto.
11054 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
11055 (TARGET_MINMAX): ... this. New.
11056 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
11057 * config/rs6000/rs6000.md: Remove Xilinx FP support.
11058 (*movsi_internal1_single): Delete.
11059 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
11060 mfpu=, mxilinx-fpu): Delete.
11061 * config/rs6000/singlefp.h: Delete.
11062 * config/rs6000/sysv4.h: Remove Xilinx FP support.
11063 * config/rs6000/t-rs6000: Ditto.
11064 * config/rs6000/t-xilinx: Delete.
11065 * config/rs6000/titan.md: Adjust for fp_type removal.
11066 * config/rs6000/vsx.md: Remove Xilinx FP support.
11067 (VStype_simple): Delete.
11068 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
11069 * config/rs6000/xfpu.h: Delete.
11070 * config/rs6000/xfpu.md: Delete.
11071 * config/rs6000/xilinx.h: Delete.
11072 * config/rs6000/xilinx.opt: Delete.
11073 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
11074 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
11076 2018-05-04 Tom de Vries <tom@codesourcery.com>
11079 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
11082 2018-05-04 Richard Biener <rguenther@suse.de>
11084 PR middle-end/85627
11085 * tree-complex.c (update_complex_assignment): We are always in SSA form.
11086 (expand_complex_div_wide): Likewise.
11087 (expand_complex_operations_1): Likewise.
11088 (expand_complex_libcall): Preserve EH info of the original stmt.
11089 (tree_lower_complex): Handle removed blocks.
11090 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
11091 on complex multiplication and division libcall builtins.
11093 2018-05-04 Richard Biener <rguenther@suse.de>
11095 PR middle-end/85574
11096 * fold-const.c (negate_expr_p): Restrict negation of operand
11097 zero of a division to when we know that can happen without
11099 (fold_negate_expr_1): Likewise.
11101 2018-05-04 Jakub Jelinek <jakub@redhat.com>
11104 * real.h (real_nextafter): Declare.
11105 * real.c (real_nextafter): New function.
11106 * fold-const-call.c (fold_const_nextafter): New function.
11107 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
11108 CASE_CFN_NEXTTOWARD.
11109 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
11110 even when arg1_mode is different from arg0_mode.
11112 2018-05-03 Nathan Sidwell <nathan@acm.org>
11114 * doc/extend.texi (Deprecated Features): Remove
11115 -ffriend-injection.
11116 (Backwards Compatibility): Likewise.
11117 * doc/invoke.texi (C++ Language Options): Likewise.
11118 (C++ Dialect Options): Likewise.
11120 2018-05-03 Jakub Jelinek <jakub@redhat.com>
11123 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
11124 _mm512_mask_mullox_epi64): New intrinsics.
11126 2018-05-03 Tom de Vries <tom@codesourcery.com>
11129 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
11130 dump files): Add offload-tree.
11132 2018-05-03 Richard Biener <rguenther@suse.de>
11134 PR tree-optimization/85615
11135 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
11136 to loops not nested in BBs loop father to avoid creating multi-entry
11139 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11141 PR tree-optimization/70291
11142 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
11143 arguments. Change return type to tree. Emit libcall as a new
11144 statement rather than replacing existing one when inplace_p is true.
11145 (expand_complex_multiplication_components): New function.
11146 (expand_complex_multiplication): Expand floating-point complex
11147 multiplication using the above.
11148 (expand_complex_division): Rename inner_type parameter to type.
11149 Update expand_complex_libcall call-site.
11150 (expand_complex_operations_1): Update expand_complex_multiplication
11151 and expand_complex_division call-sites.
11153 2018-05-02 Jakub Jelinek <jakub@redhat.com>
11156 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
11157 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
11158 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
11159 the highest significant bit of the shift count mask is clear. In
11160 check whether and[sq]i3 is needed verify that all significant bits
11161 of the shift count other than the highest are set.
11163 2018-05-02 Tom de Vries <tom@codesourcery.com>
11166 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
11167 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
11168 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
11169 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
11170 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
11171 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
11172 __builtin_goacc_parlevel_size.
11174 2018-05-02 Richard Biener <rguenther@suse.de>
11176 PR tree-optimization/85597
11177 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
11178 do not use split vect_get_vec_defs call but call vect_get_slp_defs
11181 2018-05-02 Tom de Vries <tom@codesourcery.com>
11184 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
11185 dump files): Add ltrans-tree.
11187 2018-05-02 Tom de Vries <tom@codesourcery.com>
11190 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
11191 dump files): Add wpa-ipa.
11193 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
11195 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
11196 powerpc*-*-linux*paired* target.
11197 * config/rs6000/750cl.h: Delete.
11198 * config/rs6000/paired.h: Delete.
11199 * config/rs6000/paired.md: Delete.
11200 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
11202 * config/rs6000/rs6000-builtin.def: Remove paired float support.
11203 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
11204 comment. Remove paired float support.
11205 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
11206 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
11208 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
11209 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
11211 * config/rs6000/rs6000.c: Remove paired float support.
11212 (paired_expand_vector_init, paired_expand_vector_move,
11213 paired_emit_vector_compare, paired_emit_vector_cond_expr,
11214 (paired_expand_lv_builtin, paired_expand_stv_builtin,
11215 paired_expand_builtin, paired_expand_predicate_builtin,
11216 paired_init_builtins): Delete.
11217 * config/rs6000/rs6000.h: Remove paired float support.
11218 * config/rs6000/rs6000.md: Remove paired float support.
11219 (move_from_CR_ov_bit): Delete.
11220 * config/rs6000/rs6000.opt (mpaired): Delete.
11221 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
11222 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
11224 2018-05-02 Richard Biener <rguenther@suse.de>
11226 PR middle-end/85567
11227 * gimplify.c (gimplify_save_expr): When in SSA form allow
11228 SAVE_EXPRs to compute to SSA vars.
11230 2018-05-02 Jakub Jelinek <jakub@redhat.com>
11233 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
11234 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
11235 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
11236 clobber operands[2], instead use a new pseudo. Formatting fixes.
11238 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
11240 PR tree-optimization/85586
11241 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
11242 exit early for statements in the same group if the accesses are
11245 2018-05-02 Tom de Vries <tom@codesourcery.com>
11248 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
11251 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
11253 PR tree-optimization/85143
11254 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
11256 2018-05-01 Tom de Vries <tom@codesourcery.com>
11259 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
11260 not found" error message.
11262 2018-05-01 Tom de Vries <tom@codesourcery.com>
11265 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
11266 * vec.c (test_ordered_remove_if): New function.
11267 (vec_c_tests): Call test_ordered_remove_if.
11268 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
11269 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
11270 * tree-vect-patterns.c (vect_pattern_recog_1): Use
11271 VEC_ORDERED_REMOVE_IF.
11273 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11275 PR tree-optimization/82665
11276 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
11277 pointer subtraction where arguments come from a memchr call.
11279 2018-05-01 Jakub Jelinek <jakub@redhat.com>
11281 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
11282 --push-state --as-needed and --pop-state instead of --as-needed and
11283 --no-as-needed if ld supports it.
11284 * configure: Regenerated.
11287 * doc/install.texi2html: Replace _002d with - and _002a with * in
11288 generated html files using sed.
11290 2018-04-30 David Malcolm <dmalcolm@redhat.com>
11293 * gcc-rich-location.c (blank_line_before_p): New function.
11294 (use_new_line): New function.
11295 (gcc_rich_location::add_fixit_insert_formatted): New function.
11296 * gcc-rich-location.h
11297 (gcc_rich_location::add_fixit_insert_formatted): New function.
11299 2018-04-30 David Malcolm <dmalcolm@redhat.com>
11301 * selftest.c (assert_streq): Rename "expected" and "actual" to
11302 "val1" and "val2". Extend NULL-handling to cover both inputs
11303 symmetrically, while still requiring both to be non-NULL for a pass.
11304 * selftest.h (assert_streq): Rename "expected" and "actual" to
11306 (ASSERT_EQ): Likewise.
11307 (ASSERT_EQ_AT): Likewise.
11308 (ASSERT_KNOWN_EQ): Likewise.
11309 (ASSERT_KNOWN_EQ_AT): Likewise.
11310 (ASSERT_NE): Likewise.
11311 (ASSERT_MAYBE_NE): Likewise.
11312 (ASSERT_MAYBE_NE_AT): Likewise.
11313 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
11314 the assertion to pass.
11315 (ASSERT_STREQ_AT): Likewise.
11317 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
11319 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
11320 interaction with -pie.
11322 2018-04-30 David Malcolm <dmalcolm@redhat.com>
11324 * selftest.h: Fix alphabetization of per-source-file selftest
11327 2018-04-30 Jason Merrill <jason@redhat.com>
11329 PR c++/61982 - dead stores to destroyed objects.
11330 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
11333 2018-04-30 Jason Merrill <jason@redhat.com>
11335 * tree.c (build_clobber): New.
11336 * tree.h: Declare it.
11337 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
11339 2018-04-30 David Malcolm <dmalcolm@redhat.com>
11341 * diagnostic-show-locus.c (layout::layout): Update for
11342 location_get_source_line returning a char_span.
11343 (struct char_span): Move to input.h.
11344 (struct correction): Update for fields in char_span becoming
11346 (struct source_line): Update for location_get_source_line
11347 returning a char_span.
11348 (layout::print_line): Likewise.
11349 * edit-context.c (edited_file::print_content): Likewise.
11350 (edited_file::print_diff_hunk): Likewise.
11351 (edited_file::print_run_of_changed_lines): Likewise.
11352 (edited_file::get_num_lines): Likewise.
11353 (edited_line::edited_line): Likewise.
11354 * final.c (asm_show_source): Likewise.
11355 * input.c (location_get_source_line): Convert return type
11356 from const char * to char_span, losing the final "line_len"
11358 (dump_location_info): Update for the above.
11359 (get_substring_ranges_for_loc): Likewise. Use a char_span
11360 when handling the literal within the line.
11361 (test_reading_source_line): Update for location_get_source_line
11362 returning a char_span.
11363 * input.h (class char_span): Move here from
11364 diagnostic-show-locus.c, converting from a struct to a class.
11365 Make data members private.
11366 (char_span::operator bool): New.
11367 (char_span::length): New.
11368 (char_span::get_buffer): New.
11369 (char_span::operator[]): New.
11370 (char_span::subspan): Make const.
11371 (char_span::xstrdup): New.
11372 (location_get_source_line): Convert return type from const char *
11373 to char_span, losing the final "line_size" param.
11375 2018-04-30 Jan Hubicka <jh@suse.cz>
11377 * lto-wrapper.c (ltrans_priorities): New static var.
11378 (cmp_priority): New.
11379 (run_gcc): Read priorities and if doing parallel build order
11380 the Makefile by them.
11382 2018-04-30 David Malcolm <dmalcolm@redhat.com>
11384 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
11386 2018-04-30 Richard Biener <rguenther@suse.de>
11388 * tree-cfg.c (verify_address): Remove base argument, add
11389 flag whether to check TREE_ADDRESSABLE and do that.
11390 (verify_expr): Remove.
11391 (verify_types_in_gimple_reference): Add pieces from verify_expr.
11392 (verify_gimple_assign_single): Likewise.
11393 (verify_gimple_switch): Likewise.
11394 (verify_expr_location_1): Dereference tp once. Add (disabled)
11395 piece from verify_expr.
11396 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
11398 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
11400 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
11402 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
11404 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
11405 (small_data_pattern): Likewise.
11406 (arc_rewrite_small_data): Likewise.
11407 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
11408 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
11409 (get_symbol_alignment): New function.
11410 (legitimate_small_data_address_p): Likewise.
11411 (legitimate_scaled_address): Update, call
11412 legitimate_small_data_address_p.
11413 (output_sdata): New static variable.
11414 (arc_print_operand): Update how we handle small data operands.
11415 (arc_print_operand_address): Likewise.
11416 (arc_legitimate_address_p): Update, use
11417 legitimate_small_data_address_p.
11418 (arc_rewrite_small_data_p): Remove.
11419 (arc_rewrite_small_data_1): Likewise.
11420 (arc_rewrite_small_data): Likewise.
11421 (small_data_pattern): Likewise.
11422 (compact_sda_memory_operand): Update to use
11423 legitimate_small_data_address_p and get_symbol_alignment.
11424 (prepare_move_operands): Don't rewite sdata pattern.
11425 (prepare_extend_operands): Remove.
11426 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
11428 (zero_extendqisi2): Likewise.
11429 (zero_extendhisi2): Likewise.
11430 (extendqihi2): Likewise.
11431 (extendqisi2): Likewise.
11432 (extendhisi2): Likewise.
11433 (addsi3): Likewise.
11434 (subsi3): Likewise.
11435 (andsi3): Likewise.
11436 * config/arc/constraints.md (Usd): Change it to memory constraint.
11438 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
11440 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
11441 as source of std instructions.
11442 * config/arc/arc.md (movsi_insn): Update pattern predicate to
11443 allow 6-bit constants as source for store instructions.
11444 (movdi_insn): Update instruction pattern to allow 6-bit constants
11445 as source for store instructions.
11447 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
11449 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
11451 2018-04-30 Nathan Sidwell <nathan@acm.org>
11452 Sandra Loosemore <sandra@codesourcery.com>
11454 * dumpfile.c (dump_open): Allow '-' for stdout.
11455 * doc/invoke.texi (Developer Options): Document dump filename
11456 determination early. Document stdin/stdout selection.
11458 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
11460 Microblaze Target: PIC data text relative
11462 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
11463 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
11465 * config/microblaze/microblaze.h (microblaze_constant_address_p):
11466 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
11467 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
11468 New addressing mode for data-text relative position indepenedent code.
11469 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
11470 'ADDRESS_SYMBOLIC_TXT_REL'.
11471 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
11472 (microblaze_legitimate_pic_operand): Exclude function calls from
11473 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
11474 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
11476 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
11477 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
11478 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
11479 for 'address + offset'.
11480 (microblaze_expand_prologue): Add new function prologue call for
11482 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
11483 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
11484 table in case of TARGET_PIC_DATA_TEXT_REL.
11485 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
11486 * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
11487 Add new macros 'UNSPEC_TEXT',
11488 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
11489 + exclude function calls from 'UNSPEC_PLT' in case of data text
11491 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
11492 new target hook for generating address diff vector tables in case of
11494 * doc/tm.texi : Regenerate.
11495 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
11496 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
11497 of addr diff vector generation.
11498 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
11499 target hook definition.
11500 * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
11501 Add default function for generate_pic_addr_diff_vec -> flag_pic.
11502 * doc/invoke.texi (Add new pic option): Add new microblaze pic
11503 option for data text relative.
11505 2018-04-30 Richard Biener <rguenther@suse.de>
11507 * tree-chrec.h (evolution_function_is_constant_p): Remove
11509 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
11511 2018-04-30 Richard Biener <rguenther@suse.de>
11514 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
11516 2018-04-30 Richard Biener <rguenther@suse.de>
11518 PR tree-optimization/28364
11519 PR tree-optimization/85275
11520 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
11521 copying first exit test.
11523 2018-04-28 Mark Wielaard <mark@klomp.org>
11525 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
11526 dwarf_version >= 5.
11527 (dwarf_AT): Handle DW_AT_addr_base.
11528 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
11530 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
11533 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
11534 (*ashl<dwi>3_doubleword_mask_1): Ditto.
11535 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
11536 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
11538 2018-04-28 Richard Biener <rguenther@suse.de>
11540 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
11541 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
11542 to reflect use. Only add interesting stmts.
11544 2018-04-27 Martin Jambor <mjambor@suse.cz>
11547 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
11548 the jump function allows for passing through aggregate values.
11550 2018-04-27 David Malcolm <dmalcolm@redhat.com>
11552 * input.h (in_system_header_at): Convert from macro to inline
11554 (from_macro_expansion_at): Likewise.
11555 (from_macro_definition_at): Likewise.
11557 2018-04-27 Jeff Law <law@redhat.com>
11559 * config.gcc: Mark tile* targets as deprecated/obsolete.
11561 2018-04-27 Richard Biener <rguenther@suse.de>
11563 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
11566 2018-04-27 Richard Biener <rguenther@suse.de>
11568 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
11570 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
11572 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
11573 with Yd constraint. Set "preferred_for_speed" attribute from
11574 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
11575 with Yd constraint.
11576 (*movdi_internal): Ditto.
11577 (movti_interunit splitters): Remove
11578 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
11579 (movdi_interunit splitters): Ditto.
11580 * config/i386/constraints.md (Ye): Remove.
11581 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
11583 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11586 * config/aarch64/constraints.md (Usg): Limit to 31.
11587 (Usj): Limit to 63.
11589 2018-04-27 Jakub Jelinek <jakub@redhat.com>
11591 PR tree-optimization/85529
11592 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
11593 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
11594 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
11595 zero extension or masking of the MSB bit.
11596 (optimize_range_tests): Add FIRST_BB argument, pass it through
11597 to optimize_range_tests_var_bound.
11598 (maybe_optimize_range_tests, reassociate_bb): Adjust
11599 optimize_range_tests callers.
11601 2018-04-26 Richard Biener <rguenther@suse.de>
11602 Jakub Jelinek <jakub@redhat.com>
11604 * cgraph.h (symbol_table): Just declare debug method here.
11605 * symtab.c (symbol_table::debug): Define.
11607 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
11609 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
11611 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
11613 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
11614 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
11615 (*movdi_internal): Substitute Yi and Yj constraint with x
11616 and Ym and Yn constraint with y constraint. Update "isa"
11617 attribute and set "preferred_for_speed" attribute from
11618 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
11619 (*movsi_internal): Ditto.
11620 (*movdf_internal): Ditto.
11621 (*movsf_internal): Ditto.
11622 (*zero_extendsidi2): Ditto.
11623 * config/i386/sse.md (vec_set<mode>_0): Ditto.
11624 (sse2_loadld): Ditto.
11625 (*vec_extract<ssevecmodelower>_0): Ditto.
11626 (*vec_extractv4si_0_zext_sse4): Ditto.
11627 (vec_concatv2di): Ditto.
11628 (*vec_dup<mode>): Ditto.
11629 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
11630 * config/i386/constraints.md (Yi): Remove.
11635 2018-04-26 Nathan Sidwell <nathan@acm.org>
11637 * dumpfile.c (dump_open): New.
11638 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
11639 (dump_finish): Detect stdio/stderr by value not name.
11641 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
11643 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
11645 2018-04-26 Tom de Vries <tom@codesourcery.com>
11648 * config/nvptx/nvptx.c (verify_neutering_jumps)
11649 (verify_neutering_labels): New function
11650 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
11652 2018-04-26 Tom de Vries <tom@codesourcery.com>
11655 * config/nvptx/nvptx.c (needs_neutering_p): New function.
11656 (nvptx_single): Use needs_neutering_p to skip over insns that do not
11659 2018-04-26 Richard Biener <rguenther@suse.de>
11660 Tom de Vries <tom@codesourcery.com>
11663 * lto-streamer-out.c (output_function): Fixup loops if required to match
11664 discovery done in the reader.
11666 2018-04-26 Richard Biener <rguenther@suse.de>
11668 PR tree-optimization/85116
11669 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
11670 have a loop exit from the single latch predecessor. Remove
11671 case of header with just condition.
11672 (ch_base::copy_headers): Exclude infinite loops from any
11674 (pass_ch::execute): Record exits.
11676 2018-04-26 Richard Biener <rguenther@suse.de>
11678 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
11679 prologue cost vector and pass it to vect_get_load_cost.
11680 (vect_get_peeling_costs_all_drs): Likewise.
11681 (vect_peeling_hash_get_lowest_cost): Likewise.
11682 (vect_enhance_data_refs_alignment): Likewise.
11684 2018-04-26 Richard Biener <rguenther@suse.de>
11686 PR middle-end/85450
11687 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
11688 checking of integer<->pointer conversions.
11689 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
11690 sign-/zero-extending pointer types.
11691 (expand_omp_for_static_chunk): Likewise.
11693 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
11694 Jean Lee <xiaoyur347@gmail.com>
11696 * config/mips/mips.c (mips_asan_shadow_offset): New function.
11697 (TARGET_ASAN_SHADOW_OFFSET): Define.
11698 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
11699 true for -fsanitize=address.
11701 2018-04-25 Mark Wielaard <mark@klomp.org>
11703 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
11706 2018-04-25 Jakub Jelinek <jakub@redhat.com>
11708 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
11709 than "alu", remove explicit "memory" and "imm_disp" attributes.
11710 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
11712 PR middle-end/85414
11713 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
11714 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
11715 gen_lowpart_no_emit.
11717 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
11720 * config/i386/i386.c (ix86_expand_builtin): Change memory
11721 operand to XI, extend p0 to Pmode.
11722 * config/i386/i386.md: Change unspec volatile and operand
11723 1 mode to XI, change operand 0 mode to P.
11725 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
11727 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
11728 GET_MODE_MASK before any checking.
11729 (nds32_can_use_bset_p): Likewise.
11730 (nds32_can_use_btgl_p): Likewise.
11732 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
11734 * config/nds32/nds32-doubleword.md: New define_split pattern for
11735 illegal register number.
11737 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
11739 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
11741 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
11743 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
11745 2018-04-25 Richard Biener <rguenther@suse.de>
11747 * lto-streamer.h (LTO_major_version): Bump to 8.
11749 2018-04-25 Jakub Jelinek <jakub@redhat.com>
11751 * BASE-VER: Set to 9.0.0.
11753 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
11755 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
11756 in __abskf2 and __powikf2.
11758 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11761 * config/aarch64/constraints.md (Usg, Usj): New constraints.
11762 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
11763 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
11764 Use the above on operand 2. Reindent.
11765 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
11767 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
11770 * common/config/i386/i386-common.c (ix86_handle_option): Don't
11772 * config/i386/i386.opt (mcet): Removed.
11773 * doc/install.texi: Remove -mcet documentation.
11774 * doc/invoke.texi: Likewise.
11776 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
11779 * doc/install.texi: Remove -mcet from bootstrap-cet.
11781 2018-04-24 Jakub Jelinek <jakub@redhat.com>
11784 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
11785 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
11789 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
11790 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
11791 containing a CONST_VECTOR.
11793 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
11795 * doc/install.texi: Update newlib dependency for nvptx.
11797 2018-04-24 Jakub Jelinek <jakub@redhat.com>
11800 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
11801 instead of INTVAL when shifting x left.
11803 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
11805 PR tree-optimization/85478
11806 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
11807 vect_grouped_store_supported for single element vectors.
11809 2018-04-24 Richard Biener <rguenther@suse.de>
11812 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
11813 load cost increase to the case of non-constant step.
11815 2018-04-24 Jakub Jelinek <jakub@redhat.com>
11818 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
11819 destination if any_malformed_asm.
11821 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
11823 PR middle-end/85496
11824 * expr.c (store_field): In the bitfield case, if the value comes from
11825 a function call and is returned in registers by means of a PARALLEL,
11826 do not change the mode of the temporary unless BLKmode and VOIDmode.
11828 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
11830 PR rtl-optimization/85423
11831 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
11832 dependencies to debug insns when the previous insn is non-debug.
11834 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
11836 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
11837 enums into a single definition.
11838 (fls): Fix predicates and printing.
11841 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
11843 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
11844 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
11845 and short u6 immediate.
11846 (check_if_valid_sleep_operand): Remove.
11847 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
11849 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
11851 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
11852 flag_always_save_lp condition.
11853 * config/nds32/nds32.opt (malways-save-lp): New option.
11855 2018-04-22 Shiva Chen <shiva0217@gmail.com>
11857 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
11858 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
11859 * config/nds32/nds32.h
11860 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
11861 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
11863 2018-04-22 Shiva Chen <shiva0217@gmail.com>
11865 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
11866 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
11868 2018-04-22 Shiva Chen <shiva0217@gmail.com>
11869 Chung-Ju Wu <jasonwucj@gmail.com>
11871 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
11873 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
11874 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
11876 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
11878 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
11880 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
11882 * config/nds32/nds32-protos.h (nds32_data_alignment,
11883 nds32_local_alignment): Declare.
11884 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
11885 nds32_local_alignment): New functions.
11886 (TARGET_CONSTANT_ALIGNMENT): Define.
11887 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
11889 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
11891 * config/nds32/nds32.c
11892 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
11893 (TARGET_MODES_TIEABLE_P): Likewise.
11895 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
11897 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
11898 level Ofast and Og.
11900 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
11901 Chung-Ju Wu <jasonwucj@gmail.com>
11903 * config/nds32/constants.md (unspec_volatile_element): Add enum values
11904 for unaligned access.
11905 * config/nds32/nds32-intrinsic.c: Implementation of expanding
11907 * config/nds32/nds32-intrinsic.md: Likewise.
11908 * config/nds32/nds32_intrinsic.h: Likewise.
11909 * config/nds32/nds32.h (nds32_builtins): Likewise.
11910 * config/nds32/nds32.opt (munaligned-access): New option.
11911 * config/nds32/nds32.c (nds32_asm_file_start): Display
11912 flag_unaligned_access status.
11914 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
11916 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
11917 -mno-relax is present.
11918 * config/riscv/linux.h (LINK_SPEC): Ditto.
11920 2018-04-20 Martin Sebor <msebor@redhat.com>
11923 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
11925 (gimple_fold_builtin_stxcpy_chk): Same.
11926 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
11928 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
11931 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
11932 __powikf2 when long double is IEEE 128-bit.
11934 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
11936 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
11937 step to make sure stack always aligned.
11939 2018-04-20 Carl Love <cel@us.ibm.com>
11942 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
11943 size check for arg0.
11945 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
11946 Tom de Vries <tom@codesourcery.com>
11949 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
11950 Emit insns for calls too.
11951 (nvptx_find_par): Always look for worker-level predecessor insn.
11952 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
11954 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
11955 (nvptx_process_pars): Propagate frames for calls.
11957 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
11960 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
11962 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
11963 (ix86_handle_option): Don't handle OPT_mibt.
11964 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
11966 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
11968 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
11969 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
11970 (ix86_target_macros): Define __CET__ with flag_cf_protection
11971 for -fcf-protection.
11972 * config/i386/i386.c (isa2_opts): Remove -mibt.
11973 * config/i386/i386.h (TARGET_IBT): Removed.
11974 (TARGET_IBT_P): Likewise.
11975 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
11976 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
11977 * config/i386/i386.opt (mcet): Update help message.
11978 (mshstk): Likewise.
11980 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
11981 -mcet as an alias for -mshstk.
11983 2018-04-20 Richard Biener <rguenther@suse.de>
11985 PR middle-end/85475
11986 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
11987 complexity by forcing a single use of the multiply operand.
11989 2018-04-20 Martin Jambor <mjambor@suse.cz>
11992 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
11993 recursion dependency to only apply to non-clones.
11995 2018-04-20 Martin Jambor <mjambor@suse.cz>
11998 * ipa-cp.c (create_specialized_node): Check that clones of
11999 self-recursive edges exist during IPA-CP.
12001 2018-04-19 Toon Moene <toon@moene.org>
12003 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
12006 2018-04-19 Jakub Jelinek <jakub@redhat.com>
12008 PR tree-optimization/85467
12009 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
12010 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
12011 VECTOR_CST element to type.
12013 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
12016 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
12017 * config/i386/i386.md (builtin_setjmp_setup): Removed.
12018 (builtin_longjmp): Likewise.
12019 (save_stack_nonlocal): New pattern.
12020 (restore_stack_nonlocal): Likewise.
12022 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
12025 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
12026 Replace ASM_OUTPUT_LABEL with fprintf.
12028 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
12031 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
12032 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
12033 * config/i386/i386-c.c (ix86_target_macros_internal): Also
12034 define __IBT__ and __SHSTK__ for -fcf-protection.
12035 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
12037 (ix86_trampoline_init): Likewise.
12038 (x86_output_mi_thunk): Likewise.
12039 (ix86_notrack_prefixed_insn_p): Likewise.
12040 (ix86_option_override_internal): Don't disallow -fcf-protection.
12041 * config/i386/i386.md (rdssp<mode>): Also enable for
12043 (incssp<mode>): Likewise.
12044 (nop_endbr): Likewise.
12045 * config/i386/i386.opt (mcet): Change help message to built-in
12048 (mshstk): Likewise.
12049 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
12050 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
12051 enable CET built-in functions.
12053 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
12055 * common/config/i386/i386-common.c
12056 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
12057 OPTION_MASK_ISA_MOVDIRI_UNSET,
12058 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
12059 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
12060 * config.gcc (movdirintrin.h): New header.
12061 * config/i386/cpuid.h (bit_MOVDIRI,
12062 bit_MOVDIR64B): New bits.
12063 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
12065 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
12066 (VOID, PVOID, PCVOID)): New function types.
12067 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
12068 __builtin_ia32_directstoreu_u64,
12069 __builtin_ia32_movdir64b): New builtins.
12070 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
12071 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
12073 (ix86_valid_target_attribute_inner_p): Ditto.
12074 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
12075 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
12076 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
12077 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
12078 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
12079 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
12080 (movdiri<mode>, movdir64b_<mode>): New.
12081 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
12082 * config/i386/immintrin.h: Include movdirintrin.h.
12083 * config/i386/movdirintrin.h: New file.
12084 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
12086 2018-04-19 Richard Biener <rguenther@suse.de>
12088 PR middle-end/85455
12089 * cfg.c (clear_bb_flags): When loop state says we have
12090 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
12092 2018-04-19 Richard Biener <rguenther@suse.de>
12094 PR tree-optimization/84737
12095 * tree-vect-data-refs.c (vect_copy_ref_info): New function
12096 copying restrict info.
12097 (vect_setup_realignment): Use it.
12098 * tree-vectorizer.h (vect_copy_ref_info): Declare.
12099 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
12100 the first DR to all generated stores.
12101 (vectorizable_load): Likewise for loads.
12103 2018-04-19 Jakub Jelinek <jakub@redhat.com>
12105 PR tree-optimization/85446
12106 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
12107 the integral and pointer types to have the same precision.
12109 * doc/install.texi: Document --disable-cet being the default and
12112 2018-04-18 Martin Liska <mliska@suse.cz>
12114 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
12117 2018-04-18 Martin Liska <mliska@suse.cz>
12120 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
12123 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
12124 arguments if they are comparable.
12126 2018-04-18 Martin Liska <mliska@suse.cz>
12129 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
12132 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
12135 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
12138 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
12139 ENDBR after calling __morestack.
12141 2018-04-18 David Malcolm <dmalcolm@redhat.com>
12144 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
12145 by using gcc_base_ver to generate a gcc_driver_version, and use
12146 it when generating GCC_DRIVER_NAME.
12147 * configure: Regenerate.
12149 2018-04-18 Jakub Jelinek <jakub@redhat.com>
12152 * config.gcc: Obsolete powerpc*-*-*spe*.
12154 2018-04-17 Jakub Jelinek <jakub@redhat.com>
12157 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
12158 (stabstr_D): Change type of unum from unsigned int to
12159 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
12162 2018-04-17 Jim Wilson <jimw@sifive.com>
12165 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
12166 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
12167 Set arg_pointer_offset after using pretend_args_size.
12169 2018-04-17 Jakub Jelinek <jakub@redhat.com>
12171 PR rtl-optimization/85431
12172 * dse.c (record_store): Ignore zero width stores.
12175 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
12176 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
12177 __builtin_stack_restore rather than after it.
12178 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
12179 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
12180 argument instead of virtual_dynamic_stack_rtx.
12182 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
12184 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
12186 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
12187 Add note to error message to explain internal mapping of overloaded
12188 built-in function name to non-overloaded built-in function name.
12189 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
12192 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
12195 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
12196 where the inputs overlap with the output.
12198 2018-04-17 Jakub Jelinek <jakub@redhat.com>
12200 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
12201 (=v, v) alternative and explicit "memory" attribute.
12202 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
12203 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
12205 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
12206 "sselog1" type instead of "sselog".
12207 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
12208 "sselog". Remove explicit "memory" attribute.
12209 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
12210 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
12212 (vec_extract_hi_v32hi): Merge all alternatives into one, use
12213 "sselog1" type instead of "sselog". Remove explicit "memory"
12215 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
12216 use "sselog1" type instead of "sselog". Remove explicit "memory"
12218 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
12219 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
12221 (vec_extract_hi_v64qi): Merge all alternatives into one, use
12222 "sselog1" type instead of "sselog". Remove explicit "memory"
12224 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
12225 use "sselog1" type instead of "sselog". Remove explicit "memory"
12229 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
12231 PR middle-end/85414
12232 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
12235 2018-04-17 Martin Jambor <mjambor@suse.cz>
12238 * ipa-cp.c (create_specialized_node): Call
12239 expand_all_artificial_thunks if necessary.
12241 2018-04-17 Martin Liska <mliska@suse.cz>
12244 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
12245 in message, remote space in between '_G' and '('.
12247 2018-04-17 Jakub Jelinek <jakub@redhat.com>
12250 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
12251 avx512f_vmcmp<mode>3<round_saeonly_name>,
12252 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
12253 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
12254 avx512f_rndscale<mode><round_saeonly_name>,
12255 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
12256 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
12257 Use %<iptr>2 instead of %2 for -masm=intel.
12258 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
12259 avx512f_vcvttss2usi<round_saeonly_name>,
12260 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
12262 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
12263 avx512f_vcvttsd2usi<round_saeonly_name>,
12264 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
12265 Use %q1 instead of %1 for -masm=intel.
12266 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
12267 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
12268 of %3 for -masm=intel.
12269 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
12271 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
12273 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
12275 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
12276 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
12278 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
12280 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
12281 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
12282 %g1 and one with %0 and %1.
12283 (avx512er_vmrcp28<mode><round_saeonly_name>,
12284 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
12285 %1 for -masm=intel.
12286 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
12287 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
12288 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
12289 of %0 and %{%4%} for -masm=intel.
12290 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
12291 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
12292 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
12293 order of %0 and %{%5%}%{z%} for -masm=intel.
12295 2018-04-17 Jan Hubicka <jh@suse.cz>
12298 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
12300 2018-04-17 Martin Liska <mliska@suse.cz>
12303 * multiple_target.c (create_dispatcher_calls): Set apostrophes
12304 for target_clone error message. Make default implementation
12305 clone to be a local declaration.
12306 (separate_attrs): Add new argument and check for an empty
12308 (expand_target_clones): Handle it.
12309 (ipa_target_clone): Make redirection just for target_clones
12312 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
12313 Tom de Vries <tom@codesourcery.com>
12315 PR middle-end/84955
12316 * omp-expand.c (expand_oacc_for): Add dummy false branch for
12317 tiled basic blocks without omp continue statements.
12319 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
12322 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
12323 vec_extract expression as having side effects to make sure it gets
12326 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
12329 * config/i386/i386.c (get_builtin_code_for_version): Check
12332 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
12335 * config.gcc: Support "skylake".
12336 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
12338 * config/i386/i386.c (m_SKYLAKE): Define.
12339 (processor_target_table): Add "skylake".
12340 (ix86_option_override_internal): Add "skylake".
12341 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
12342 PROCESSOR_CANNONLAKE.
12343 (get_builtin_code_for_version): Fix priority for
12344 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
12345 PROCESSOR_SKYLAKE-AVX512.
12346 * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
12347 (processor_type): Add PROCESSOR_SKYLAKE.
12349 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
12350 Jason Merrill <jason@redhat.com>
12353 * convert.c (convert_to_integer_1): Use direct recursion for
12354 enumeral types and types with a precision less than the number
12355 of bits in their mode.
12357 2018-04-16 Julia Koval <julia.koval@intel.com>
12360 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
12361 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
12363 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
12366 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
12367 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
12368 and -mno-direct-move.
12370 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
12373 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
12374 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
12375 Ensure negative shifts result in {0}.
12377 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
12379 PR rtl-optimization/79916
12380 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
12381 regs (if any) to define how to gnerate SD moves when LRA is in
12384 2018-04-13 Jakub Jelinek <jakub@redhat.com>
12386 PR rtl-optimization/85393
12387 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
12388 * except.c (expand_dw2_landing_pad_for_region): Make static.
12389 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
12390 a label and unconditional jump to old_bb, rather than
12391 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
12394 PR rtl-optimization/85376
12395 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
12396 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
12397 instead of a specific value.
12399 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
12400 Bin Cheng <bin.cheng@arm.com>
12402 PR tree-optimization/82965
12403 PR tree-optimization/83991
12404 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
12405 by_profile_only parameter.
12406 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
12407 information if the loop was predicted to iterate too many times.
12408 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
12410 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
12413 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
12416 2018-04-13 Martin Liska <mliska@suse.cz>
12417 Jakub Jelinek <jakub@redhat.com>
12419 PR middle-end/81657
12420 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
12421 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
12422 * builtins.c (expand_builtin_memory_copy_args): Use
12423 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
12424 handle dest_addr == pc_rtx.
12426 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
12429 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
12430 asked to not generate direct moves.
12431 (fix_trunc<mode>si2_stfiwx): Similar.
12432 (fix_trunc<mode>si2_internal): Similar.
12434 2018-04-12 Jakub Jelinek <jakub@redhat.com>
12437 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
12438 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
12439 lookup if dest in some wider mode is known to be const0_rtx and
12440 if so, record permanent equivalence for it to be ZERO_EXTEND of
12441 the narrower mode destination.
12443 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
12445 * lto-streamer-out.c (output_function): Revert 259346.
12446 * omp-expand.c (expand_oacc_for): Likewise.
12448 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
12450 PR rtl-optimization/85354
12451 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
12452 * sel-sched.c (sel_global_init): ... here.
12454 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
12457 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
12458 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
12459 mode for PE-COFF targets.
12460 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
12461 (i386_pe_asm_lto_end): Likewise.
12462 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
12463 (TARGET_ASM_LTO_END): Likewise.
12464 * config/i386/winnt.c (saved_debug_info_level): New static variable.
12465 (i386_pe_asm_lto_start): New function.
12466 (i386_pe_asm_lto_end): Likewise.
12468 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
12469 Richard Biener <rguenther@suse.de>
12471 PR middle-end/84955
12472 * lto-streamer-out.c (output_function): Fix CFG loop state before
12474 * omp-expand.c (expand_oacc_for): Handle calls to internal
12475 functions like regular functions.
12477 2018-04-12 Richard Biener <rguenther@suse.de>
12480 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
12481 for the early LTO debug to properly generate references to it
12482 during DIE emission. Do not re-use that for the skeleton for
12484 (dwarf2out_early_finish): Likewise.
12486 2018-04-12 Jakub Jelinek <jakub@redhat.com>
12489 * config/i386/sse.md
12490 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
12491 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
12492 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
12493 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
12494 and output is a reg, avoid creating invalid lowpart subreg, but
12495 instead split into a 512-bit move. Don't split if not AVX512VL,
12496 input is xmm16+ reg and output is a mem.
12497 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
12498 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
12499 xmm16+ reg and output is a mem.
12501 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
12503 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
12504 also for flag_dwarf2_cfi_asm.
12506 2018-04-12 Jakub Jelinek <jakub@redhat.com>
12508 PR rtl-optimization/85342
12509 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
12510 a bool scalar var inside of the loop instead. Don't try to update
12511 recog_data.operand after failed apply_change_group.
12513 2018-04-12 Tom de Vries <tom@codesourcery.com>
12516 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
12517 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
12518 array with flexible array member as array without given dimension.
12519 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
12520 argument for undefined param to true.
12522 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
12525 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
12526 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
12527 from PowerPC section.
12528 * config/rs6000/sysv4.opt (mcall-): Improve help text.
12529 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
12530 help text that is too long.
12531 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
12532 help text that is too long.
12533 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
12534 help text that is too long.
12536 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
12538 * config/alpha/alpha.md (stack_probe_internal): Rename
12539 from "probe_stack". Update all callers.
12541 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
12543 PR rtl-optimization/84566
12544 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
12545 sched_macro_fuse_insns.
12547 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
12550 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
12551 (compute_block_dependences): ... from here.
12553 2018-04-11 Jakub Jelinek <jakub@redhat.com>
12555 PR tree-optimization/85331
12556 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
12557 from int to HOST_WIDE_INT.
12559 2018-04-11 Martin Jambor <mjambor@suse.cz>
12562 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
12563 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
12564 not the same as the source val.
12565 (cgraph_edge_brings_value_p): New parameter.
12566 (gather_edges_for_value): Pass destination value to
12567 cgraph_edge_brings_value_p.
12568 (perhaps_add_new_callers): Likewise.
12569 (get_info_about_necessary_edges): Likewise and exclude values brought
12570 only by self-recursive edges.
12571 (create_specialized_node): Redirect only clones of self-calling edges.
12572 (+self_recursive_pass_through_p): New function.
12573 (find_more_scalar_values_for_callers_subset): Use it.
12574 (find_aggregate_values_for_callers_subset): Likewise.
12575 (known_aggs_to_agg_replacement_list): Removed.
12576 (decide_whether_version_node): Re-calculate known constants for all
12577 remaining context clones.
12579 2018-04-11 Richard Biener <rguenther@suse.de>
12582 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
12583 from early DWARF output.
12584 (dwarf2out_early_finish): Output line info unconditionally into
12585 early DWARF and add reference to it.
12587 2018-04-11 Jakub Jelinek <jakub@redhat.com>
12590 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
12591 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
12592 other than V2DFmode using iptr mode attribute.
12593 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
12595 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
12597 PR rtl-optimization/84659
12598 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
12600 2018-04-11 Jakub Jelinek <jakub@redhat.com>
12603 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
12605 (output_loc_list): Pass address of a dummy size variable even in the
12606 locview handling loop.
12607 (index_location_lists): Add comment on why skip_loc_list_entry can't
12610 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
12613 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
12616 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
12619 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
12620 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
12621 and -mstring-compare-inline-limit.
12623 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
12626 * config/rs6000/rs6000.md (allocate_stack): Put the residual size
12627 for stack clash protection in a register whenever we need it to be in
12630 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
12632 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
12633 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
12635 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
12638 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
12640 (mlong-double-): Ditto.
12641 * config/rs6000/sysv4.opt (msdata=): Ditto.
12642 (mtls-size=): Ditto.
12644 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
12646 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
12647 erroneous entries for
12648 "vector int vec_ldl (int, long int *)", and
12649 "vector unsigned int vec_ldl (int, unsigned long int *)".
12650 Add comments and entries for
12651 "vector bool char vec_ldl (int, bool char *)",
12652 "vector bool short vec_ldl (int, bool short *)",
12653 "vector bool int vec_ldl (int, bool int *)",
12654 "vector bool long long vec_ldl (int, bool long long *)",
12655 "vector pixel vec_ldl (int, pixel *)",
12656 "vector long long vec_ldl (int, long long *)",
12657 "vector unsigned long long vec_ldl (int, unsigned long long *)".
12658 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
12659 type tree bool_long_long_type_node and correct definition of
12660 bool_V2DI_type_node to make reference to this new type tree.
12661 (rs6000_mangle_type): Replace erroneous reference to
12662 bool_long_type_node with bool_long_long_type_node.
12663 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
12664 comments to emphasize sign distinctions for char and int types and
12665 replace RS6000_BTI_bool_long constant with
12666 RS6000_BTI_bool_long_long constant. Also add comment to restrict
12667 use of RS6000_BTI_pixel.
12668 (bool_long_type_node): Remove this macro definition.
12669 (bool_long_long_type_node): New macro definition
12671 2018-04-10 Jakub Jelinek <jakub@redhat.com>
12673 PR rtl-optimization/85300
12674 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
12675 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
12676 simplify_unary_operation fails.
12678 2018-04-10 Martin Liska <mliska@suse.cz>
12680 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
12681 cgraph_edge and ipa_ref.
12683 2018-04-10 Jakub Jelinek <jakub@redhat.com>
12687 * config/i386/sse.md
12688 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
12689 computation of the VEC_MERGE selector from mask.
12690 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
12691 Fix decoding of the VEC_MERGE selector into mask.
12693 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
12695 PR tree-optimization/85286
12696 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
12698 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
12700 * final.c (final_1): Set insn_last_address as well as
12701 insn_current_address.
12703 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12706 * explow.c (emit_stack_probe): Call validize_mem on memory location
12707 before passing it to gen_probe_stack. Create address operand and
12708 legitimize it for the probe_stack_address case.
12710 2018-04-09 Jan Hubicka <jh@suse.cz>
12713 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
12714 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
12715 * tree.c (free_lang_data_in_type): Fix handling of binfos;
12717 (free_lang_data): Rebuild type inheritance graph.
12719 2018-04-09 Martin Sebor <msebor@redhat.com>
12721 * invoke.texi (-finline-small-functions): Mention other optimization
12723 (-findirect-inlining, -fpartial-inlining): Same.
12724 (-finline-functions-called-once): Same.
12725 (-freorder-blocks-and-partition): Same.
12727 2018-04-09 Jan Hubicka <jh@suse.cz>
12730 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
12731 jumps; choose last target that matches the criteria (i.e.
12732 no partition changes for non-crossing jumps).
12733 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
12734 support for redirecting crossing jumps to non-crossing.
12736 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
12738 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
12739 also for naked functions.
12741 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
12743 * config/arc/arc.md (add_shift): New pattern.
12744 (add_shift2): Likewise.
12745 (sub_shift): Likewise.
12746 (sub_shift_cmp0_noout): Likewise.
12747 (compare_si_ashiftsi): Likewise.
12748 (xbfu_cmp0_noout): New combine pattern.
12749 (xbfu_cmp0"): Likewise.
12750 (movsi_set_cc_insn): Place the predicable variant first.
12751 (commutative_binary_cmp0_noout): Remove clobber.
12752 (commutative_binary_cmp0): New pattern.
12753 (noncommutative_binary_cmp0): Likewise.
12754 (noncommutative_binary_cmp0_noout): Likewise.
12755 (noncommutative_binary_comparison_result_used): Removed.
12756 (rsub_cmp0): New pattern.
12757 (rsub_cmp0_noout): Likewise.
12758 (extzvsi): Changed, keep only meaningful variants.
12759 (SQH, SEZ): New iterators.
12760 (SQH_postfix): New mode attribute.
12761 (SEZ_prefix): New code attribute.
12762 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
12763 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
12764 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
12765 of numerical value.
12766 (noncommutative_operator): Check the availability of barrel
12769 2018-04-09 Richard Biener <rguenther@suse.de>
12771 PR tree-optimization/85284
12772 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
12773 Only use the niter constraining form of simple_iv when the exit
12774 is always executed.
12776 2018-04-09 Tom de Vries <tom@codesourcery.com>
12779 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
12780 (define_expand "*memory_barrier"): New define_expand.
12781 (define_insn "memory_barrier"): New insn.
12783 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
12785 PR rtl-optimization/80463
12786 PR rtl-optimization/83972
12787 PR rtl-optimization/83480
12789 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
12790 correct producer for the insn.
12791 (tidy_control_flow): Fixup seqnos in case of debug insns.
12793 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
12795 PR rtl-optimization/83913
12797 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
12798 different sched-times when merging exprs.
12800 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
12802 PR rtl-optimization/83962
12804 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
12805 tidy_fallthru_edge and tidy_control_flow.
12807 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
12809 PR rtl-optimization/83530
12811 * sel-sched.c (force_next_insn): New global variable.
12812 (remove_insn_for_debug): When force_next_insn is true, also leave only
12813 next insn in the ready list.
12814 (sel_sched_region): When the region wasn't scheduled, make another pass
12815 over it with force_next_insn set to 1.
12817 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
12819 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
12821 * config/nds32/constants.md (unspec_volatile_element): Add enum values
12822 for interrupt control.
12823 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
12824 functions for interrupt control.
12825 * config/nds32/nds32-intrinsic.md: Likewise.
12826 * config/nds32/nds32_intrinsic.h: Likewise.
12827 * config/nds32/nds32.h (nds32_builtins): Likewise.
12829 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
12831 * config/nds32/nds32.c (nds32_init_machine_status,
12832 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
12833 strict_aligned_p field.
12834 (nds32_expand_to_rtl_hook): New function.
12835 (TARGET_EXPAND_TO_RTL_HOOK): Define.
12836 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
12838 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
12839 Chung-Ju Wu <jasonwucj@gmail.com>
12841 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
12842 * config/nds32/nds32-n7.md: New file.
12843 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
12844 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
12846 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
12847 * config/nds32/nds32.md (pipeline_model): Add n7.
12848 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
12849 * config/nds32/pipelines.md: Include n7 settings.
12851 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
12852 Chung-Ju Wu <jasonwucj@gmail.com>
12854 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
12855 * config/nds32/nds32-e8.md: New file.
12856 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
12857 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
12859 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
12860 * config/nds32/nds32.md (pipeline_model): Add e8.
12861 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
12862 * config/nds32/pipelines.md: Include e8 settings.
12864 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
12865 Chung-Ju Wu <jasonwucj@gmail.com>
12867 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
12868 * config/nds32/nds32-n8.md: New file.
12869 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
12870 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
12872 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
12873 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
12874 * config/nds32/nds32.md (pipeline_model): Add n8.
12875 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
12876 * config/nds32/pipelines.md: Include n8 settings.
12878 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
12879 Chung-Ju Wu <jasonwucj@gmail.com>
12881 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
12882 * config/nds32/nds32-n9-2r1w.md: New file.
12883 * config/nds32/nds32-n9-3r2w.md: New file.
12884 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
12885 nds32_register_ports): New or modify for cpu n9.
12886 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
12888 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
12889 * config/nds32/nds32-utils.c: New file.
12890 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
12891 TARGET_MUL_SLOW): Define.
12892 * config/nds32/nds32.md (pipeline_model): New attribute.
12893 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
12894 New options that support cpu n9.
12895 * config/nds32/pipelines.md: Include n9 settings.
12896 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
12898 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
12900 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
12901 information if necessary.
12902 (output_cond_branch_compare_zero): Likewise.
12903 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
12904 (nds32_target_alignment): Refine for alignment.
12905 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
12906 (FUNCTION_BOUNDARY): Modify.
12907 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
12909 * config/nds32/nds32.opt (malways-align, malign-functions): New.
12911 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
12913 * config/nds32/constants.md (unspec_volatile_element): Add values for
12914 TLB operation and data prefetch.
12915 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
12916 functions for TLB operation and data prefetch.
12917 * config/nds32/nds32-intrinsic.md: Likewise.
12918 * config/nds32/nds32_intrinsic.h: Likewise.
12919 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
12920 (nds32_print_operand): Likewise.
12921 * config/nds32/nds32.h (nds32_builtins): Likewise.
12923 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
12924 Andrew Pinski <pinsika@gcc.gnu.org>
12926 PR middle-end/82976
12927 * match.pd: Use constant_boolean_node of correct type instead of
12928 boolean_true_node or boolean_false_node for simplifying
12929 pointer comparisons to zero.
12931 2018-04-07 Jakub Jelinek <jakub@redhat.com>
12933 PR tree-optimization/80021
12934 * tree.c (verify_type_variant): Make error call in verify_variant_match
12935 translatable and remove final full stop.
12937 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
12939 * config/nds32/constants.md (unspec_volatile_element): Add
12940 UNSPEC_VOLATILE_EH_RETURN.
12941 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
12942 nds32_output_stack_pop): Support dwarf exception handling process.
12943 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
12944 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
12945 exception handling process.
12946 (nds32_compute_stack_frame): Likewise.
12947 (nds32_return_addr_rtx): Likewise.
12948 (nds32_initial_elimination_offset): Likewise.
12949 (nds32_expand_prologue): Likewise.
12950 (nds32_expand_epilogue): Likewise.
12951 (nds32_dynamic_chain_address): New function.
12952 * config/nds32/nds32.h (machine_function): Add fields for dwarf
12953 exception handling.
12954 (DYNAMIC_CHAIN_ADDRESS): Define.
12955 (EH_RETURN_DATA_REGNO): Define.
12956 (EH_RETURN_STACKADJ_RTX): Define.
12957 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
12958 patterns for dwarf exception handling.
12960 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
12962 * config/nds32/nds32.h: Clean up obsolete macros.
12964 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
12966 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
12967 Add enum values for particular instructions.
12968 * config/nds32/nds32-intrinsic.c: Implementation of expanding
12969 particular intrinsic functions.
12970 * config/nds32/nds32-intrinsic.md: Likewise.
12971 * config/nds32/nds32_intrinsic.h: Likewise.
12972 * config/nds32/nds32.h (nds32_builtins): Likewise.
12973 * config/nds32/nds32.md (type): Add pbsad and pbsada.
12974 (btst, ave): New patterns for particular instructions.
12976 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
12978 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
12979 Add enum values for atomic load/store and memory sync.
12980 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
12982 * config/nds32/nds32-intrinsic.md: Likewise.
12983 * config/nds32/nds32_intrinsic.h: Likewise.
12984 * config/nds32/nds32.h (nds32_builtins): Likewise.
12986 2018-04-07 Jakub Jelinek <jakub@redhat.com>
12988 PR tree-optimization/85257
12989 * fold-const.c (native_encode_vector): If not all elts could fit
12990 and off is -1, return 0 rather than offset.
12991 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
12992 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
12993 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
12994 adjust buffer in native_interpret_expr call.
12996 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
12998 * config/nds32/constants.md (unspec_volatile_element): Add cache
12999 control enum values.
13000 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
13001 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
13002 * config/nds32/nds32.c (nds32_cctl_names): New.
13003 (nds32_print_operand): Handle cache control register names.
13004 * config/nds32/nds32.h (nds32_builtins): New enum values.
13005 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
13007 * config/nds32/nds32.md (type): Add mmu.
13008 * config/nds32/pipelines.md (simple_insn): Add mmu.
13010 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
13012 * config/nds32/nds32.md (type): Remove call.
13013 * config/nds32/pipelines.md (simple_insn): Likewise.
13015 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
13017 * config/nds32/constants.md (unspec_volatile_element): Add
13018 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
13019 UNSPEC_VOLATILE_FMFCFG.
13020 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
13021 description for fmfcfg and fmfcsr.
13022 (bdesc_1arg): Add fmtcsr.
13023 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
13024 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
13025 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
13026 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
13027 unspec_fmfcfg): New patterns.
13028 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
13029 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
13030 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
13031 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
13032 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
13033 __nds32__fmfcfg): Define.
13035 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
13037 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
13038 intrinsic register names.
13039 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
13040 intrinsic register enum values and macros.
13042 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
13044 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
13045 for load/store addressing form.
13046 (nds32_print_operand_address): Likewise.
13048 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
13051 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
13052 based on LABEL_REF. Remove useless assertion.
13053 (pic_address_needs_scratch): Fix formatting.
13054 (sparc_legitimize_pic_address): Minor tweaks.
13055 (sparc_delegitimize_address): Adjust assertion accordingly.
13056 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
13057 into symbolic_operand.
13058 (movsi_high_pic_label_ref): Likewise.
13059 (movsi_lo_sum_pic_label_ref): Likewise.
13060 (movdi_pic_label_ref): Likewise.
13061 (movdi_high_pic_label_ref): Likewise.
13062 (movdi_lo_sum_pic_label_ref): Likewise.
13064 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
13066 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
13067 custom LIB_SPEC setup.
13069 2018-04-06 Ruslan Bukin <br@bsdpad.com>
13070 Kito Cheng <kito.cheng@gmail.com>
13072 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
13073 * config/riscv/freebsd.h: New.
13075 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
13077 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
13078 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
13081 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
13082 Kito Cheng <kito.cheng@gmail.com>
13084 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
13085 nds32_output_call, nds32_symbol_binds_local_p): New functions.
13086 * config/nds32/nds32-protos.h (nds32_output_call,
13087 nds32_output_return): Declare.
13088 * config/nds32/nds32.md: Refine all the call and return patterns.
13090 2018-04-06 Jakub Jelinek <jakub@redhat.com>
13093 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
13094 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
13096 PR rtl-optimization/84872
13097 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
13098 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
13099 EDGE_CROSSING edge.
13101 2018-04-06 Tamar Christina <tamar.christina@arm.com>
13103 * expr.c (copy_blkmode_to_reg): Revert 254862.
13104 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
13106 2018-04-06 Richard Biener <rguenther@suse.de>
13108 PR middle-end/85244
13109 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
13110 after seeing a component reference with an adjacent field. Treat
13111 refs to arrays at struct end of external decls similar to
13112 refs to unconstrained commons.
13114 2018-04-06 Jakub Jelinek <jakub@redhat.com>
13117 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
13118 look through SAVE_EXPRs with non-side-effects argument. Adjust
13120 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
13123 2018-04-06 Richard Biener <rguenther@suse.de>
13125 PR middle-end/85180
13126 * alias.c (find_base_term): New wrapper around find_base_term
13127 unwinding CSELIB_VAL_PTR changes.
13128 (find_base_term): Do not restore CSELIB_VAL_PTR during the
13131 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
13133 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
13135 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
13136 constant definitions.
13137 ("nop"): lr 0,0 -> nopr r0
13138 ("nop_lr0", "nop_lr1"): New insn definitions.
13140 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
13142 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
13143 NDS32_V3PUSH_AVAILABLE_P macro.
13145 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
13146 Chung-Ju Wu <jasonwucj@gmail.com>
13148 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
13149 (nds32*-*-*): Add float and fpu_config into supported_defaults.
13150 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
13151 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
13152 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
13153 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
13154 * config/nds32/constraints.md: New constraints and checking for hard
13155 float configuration.
13156 * config/nds32/iterators.md: New mode iterator and attribute for hard
13157 float configuration.
13158 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
13160 * config/nds32/nds32-fpu.md: New file.
13161 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
13162 deal with hard float code generation.
13163 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
13165 (abi_type, float_reg_number): New enum type.
13166 * config/nds32/nds32-predicates.c: New predicates for hard float.
13167 * config/nds32/nds32-protos.h: Declare functions for hard float.
13168 * config/nds32/nds32.c: Implementation for hard float configuration.
13169 * config/nds32/nds32.h: Definitions for hard float configuration.
13170 * config/nds32/nds32.md: Include hard float machine description and
13171 modify patterns for hard float configuration.
13172 * config/nds32/nds32.opt: New options for hard float configuration.
13173 * config/nds32/predicates.md: New predicates for hard float
13176 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
13178 * common/config/nds32/nds32-common.c
13179 (nds32_option_optimization_table): Enable -mreleax-hint by default.
13181 2018-04-05 Jakub Jelinek <jakub@redhat.com>
13183 PR middle-end/85195
13184 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
13185 CONSTRUCTOR_ELT (ctor, ...)->value.
13187 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
13190 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
13192 2018-04-05 Tom de Vries <tom@codesourcery.com>
13195 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
13198 2018-04-05 Shiva Chen <shiva0217@gmail.com>
13199 Kito Cheng <kito.cheng@gmail.com>
13201 * config/nds32/constraints.md (U33): Fine-tune checking condition.
13202 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
13203 * config/nds32/nds32.h (nds32_16bit_address_type): Add
13204 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
13206 2018-04-05 Shiva Chen <shiva0217@gmail.com>
13207 Kito Cheng <kito.cheng@gmail.com>
13209 * config/nds32/constraints.md (Ufe): New memory constraint.
13210 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
13211 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
13212 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
13214 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
13215 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
13217 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
13219 * config/nds32/nds32.md: Use optimize_size in the condition for
13220 alu-shift instructions.
13222 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
13224 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
13226 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
13228 * config/nds32/nds32.md (negsi2): Refine pattern.
13230 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
13231 Chung-Ju Wu <jasonwucj@gmail.com>
13233 * config/nds32/iterators.md (shift_rotate): New code iterator.
13234 (shift): New code attribute.
13235 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
13236 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
13237 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
13238 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
13239 bit-wise operations.
13240 (andsi3, *andsi3): Ditto.
13241 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
13242 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
13243 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
13244 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
13245 nds32_ior_operand, nds32_xor_operand): New predicates.
13247 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
13249 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
13250 (addsi3, subsi3): ... this.
13252 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
13254 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
13256 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
13258 * config/nds32/nds32.md: Adjust indention.
13260 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
13262 * config/nds32/nds32.md (feature): New attribute.
13264 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
13266 * config/nds32/nds32.md (subtype): New attribute.
13268 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
13271 * config/arm/arm-builtins.c (arm_expand_builtin): Change
13272 expansion to perform a bitwise AND of the argument followed by a
13273 boolean negation of the result.
13275 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
13277 PR rtl-optimization/84878
13278 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
13279 the basic block. Assert the use reference is not artificial and that
13280 it has an associated insn.
13282 2018-04-04 Michael Matz <matz@suse.de>
13284 * builtins.c (compute_objsize): Pass correct operand
13285 to array_at_struct_end_p.
13287 2018-04-04 Richard Biener <rguenther@suse.de>
13290 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
13291 from contexts for DINFO_LEVEL_TERSE and below.
13293 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
13295 * config/nds32/nds32-doubleword.md (move_<mode>): Require
13296 resiter_operand condition.
13297 * config/nds32/nds32.md (*move<mode>): Ditto.
13299 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
13300 Monk Chiang <sh.chiang04@gmail.com>
13302 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
13304 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
13306 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
13308 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
13309 Kito Cheng <kito.cheng@gmail.com>
13311 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
13312 nds32_cond_code_str, output_cond_branch,
13313 output_cond_branch_compare_zero, nds32_expand_cbranch,
13314 nds32_expand_cstore, nds32_expand_movcc,
13315 nds32_output_cbranchsi4_equality_zero,
13316 nds32_output_cbranchsi4_equality_reg,
13317 nds32_output_cbranchsi4_equality_reg_or_const_int,
13318 nds32_output_cbranchsi4_greater_less_zero: New functions.
13319 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
13320 nds32_expand_cstore, nds32_expand_movcc,
13321 nds32_output_cbranchsi4_equality_zero,
13322 nds32_output_cbranchsi4_equality_reg,
13323 nds32_output_cbranchsi4_equality_reg_or_const_int,
13324 nds32_output_cbranchsi4_greater_less_zero): Declare.
13325 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
13326 nds32_rimm11s_operand): New predicates.
13327 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
13328 * config/nds32/nds32.md: Rewrite all the branch and conditional move
13331 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
13333 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
13334 * config/nds32/nds32.md: Ditto.
13335 * config/nds32/pipelines.md: Ditto.
13337 2018-04-04 Richard Biener <rguenther@suse.de>
13339 PR tree-optimization/85168
13340 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
13341 propagating abnormals.
13343 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
13345 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
13347 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
13348 Kito Cheng <kito.cheng@gmail.com>
13350 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
13351 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
13352 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
13353 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
13354 * config/nds32/nds32.md (sibcall_internal): New.
13355 (sibcall_register): Remove.
13356 (sibcall_immediate): Remove.
13357 (sibcall_value_internal): New.
13358 (sibcall_value_register): Remove.
13359 (sibcall_value_immediate): Remove.
13360 * config/nds32/predicates.md (nds32_general_register_operand): New.
13361 (nds32_call_address_operand): New.
13363 2018-04-03 Jakub Jelinek <jakub@redhat.com>
13365 PR rtl-optimization/85167
13366 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
13367 bb_defs if *split_p, instead preinitialize it to NULL.
13369 PR tree-optimization/85156
13370 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
13371 evaluating the argument multiple times.
13373 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
13375 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
13377 (_mm_cvtpd_ps): Likewise.
13378 (_mm_cvttpd_epi32): Likewise.
13379 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
13380 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
13381 vector, pixel, and bool following altivec.h include.
13383 2018-04-03 Martin Sebor <msebor@redhat.com>
13385 * doc/extend.texi (Common Function Attributes): Clarify.
13386 (const attribute): Likewise.
13387 (pure attribute): Likewise.
13389 2018-04-03 Jakub Jelinek <jakub@redhat.com>
13392 * config/i386/i386.c (ix86_expand_vector_set): Use
13393 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
13395 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
13397 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
13398 instructions when changing rounding bits to preserve precision bits
13399 in the x87 control word.
13401 2018-04-03 Martin Liska <mliska@suse.cz>
13403 PR tree-optimization/82491
13404 * rtl.h (strip_offset_and_add): Replace += suboffset with
13405 poly_uint64 () + suboffset.
13407 2018-03-29 Martin Liska <mliska@suse.cz>
13408 Martin Jambor <mjambor@suse.cz>
13411 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
13412 param_type is not an integral or pointer type.
13414 2018-04-03 Richard Biener <rguenther@suse.de>
13416 * sese.h (recompute_all_dominators): Remove.
13418 2018-04-02 Martin Sebor <msebor@redhat.com>
13420 * doc/invoke.texi (-Wrestrict): Fix typos.
13422 2018-04-02 Jim Wilson <jimw@sifive.com>
13424 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
13425 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
13426 (<optab>di3, <optab>si3_extend): Likewise.
13427 (<optab>si3_mask, <optab>si3_mask_1): New.
13428 (<optab>di3_mask, <optab>di3_mask_1): New.
13429 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
13430 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
13431 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
13433 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
13435 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
13438 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
13440 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
13441 (nds32_canonicalize_comparison): New function.
13443 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
13444 Kito Cheng <kito.cheng@gmail.com>
13445 Kuan-Lin Chen <kuanlinchentw@gmail.com>
13447 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
13448 * config/nds32/constants.md (unspec_volatile_element): Add
13449 UNSPEC_VOLATILE_RELAX_GROUP.
13450 * config/nds32/nds32-relax-opt.c: New file.
13451 * config/nds32/nds32-predicates.c
13452 (nds32_symbol_load_store_p): New function.
13453 * config/nds32/nds32-protos.h
13454 (nds32_symbol_load_store_p): Declare function.
13455 (make_pass_nds32_relax_opt): Declare new rtl pass function.
13456 * config/nds32/nds32.c
13457 (nds32_register_pass): New function to register pass.
13458 (nds32_register_passes): New function to register passes.
13459 * config/nds32/nds32.md (relax_group): New pattern.
13460 * config/nds32/nds32.opt (mrelax-hint): New option.
13461 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
13463 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
13465 * config/nds32/t-nds32: Modify files dependency.
13467 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
13469 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
13470 (PROFILE_HOOK): Define its implementation.
13472 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
13474 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
13475 type and 32-bit size.
13477 2018-04-01 Jakub Jelinek <jakub@redhat.com>
13479 PR middle-end/85090
13480 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
13481 (V_128_256): New mode iterator.
13482 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
13483 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
13484 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
13486 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
13487 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
13489 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
13492 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
13493 NaN inputs correctly.
13495 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
13498 * config/rs6000/vsx.md (??r): New mode attribute.
13499 (*vsx_mov<mode>_64bit): Use it.
13500 (*vsx_mov<mode>_32bit): Likewise.
13502 2018-03-30 Martin Sebor <msebor@redhat.com>
13504 PR tree-optimization/84818
13505 * builtins.c (check_access): Use warning_n.
13507 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13510 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
13512 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
13515 2018-03-30 Julia Koval <julia.koval@intel.com>
13518 * x86-tune.def (movx, partial_reg_dependency): Enable for
13521 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
13523 PR inline-asm/84985
13524 * lra-constraints.c (process_alt_operands): Move setting
13525 this_alternative_matches below.
13527 2018-03-29 Martin Liska <mliska@suse.cz>
13530 * doc/invoke.texi: Document how LTO works with debug info.
13531 Describe auto-load support of binutils. Mention 'x86-64'
13532 as valid option value of -march option.
13534 2018-03-29 Jakub Jelinek <jakub@redhat.com>
13536 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
13539 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
13540 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
13541 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
13544 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
13547 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
13548 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
13549 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
13550 for RS6000_BTM_POWERPC64.
13551 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
13552 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
13553 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
13556 (DIVDEU): Likewise.
13558 2018-03-28 Carl Love <cel@us.ibm.com>
13561 2017-09-27 Carl Love <cel@us.ibm.com>
13563 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
13564 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
13565 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
13568 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13570 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
13571 instead of __vector bool.
13572 (_mm_max_pu8): Likewise.
13573 (_mm_min_pi16): Likewise.
13575 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
13578 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
13579 (DIVWEUO): Likewise.
13580 (DIVDEO): Likewise.
13581 (DIVDEUO): Likewise.
13582 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
13583 DIVWEUO and DIVDEUO.
13584 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
13585 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
13586 (div_extend): Likewise.
13587 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
13589 (__builtin_divweuo): Likewise.
13590 (__builtin_divdeo): Likewise.
13591 (__builtin_divdeuo): Likewise.
13593 2018-03-28 Jakub Jelinek <jakub@redhat.com>
13596 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
13597 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
13599 PR tree-optimization/82004
13600 * gimple-match-head.c (optimize_pow_to_exp): New function.
13601 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
13602 Don't fold to exp if optimize_pow_to_exp is false.
13604 2018-03-28 Martin Liska <mliska@suse.cz>
13607 * calls.c (initialize_argument_information): Fix trailing space.
13608 * common.opt: Fix typo and provide better explanation for
13609 -fsanitize-coverage option.
13610 * config/i386/i386.opt: Fix typo.
13612 2018-03-28 Jakub Jelinek <jakub@redhat.com>
13613 Martin Liska <mliska@suse.cz>
13616 * gimplify.c (asan_poison_variable): Don't do the check for
13617 gimplify_omp_ctxp here.
13618 (gimplify_decl_expr): Do it here.
13619 (gimplify_target_expr): Likewise.
13621 2018-03-28 Martin Liska <mliska@suse.cz>
13624 * config/i386/i386.c (ix86_function_arg_advance): Do not call
13625 chkp_type_bounds_count if MPX is not enabled.
13627 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
13629 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
13631 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
13634 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
13635 function to create the function decl for complex long double
13636 multiply and divide for -mabi=ieeelongdouble.
13637 (init_float128_ieee): Call it.
13639 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
13642 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
13643 -fcf-protection=branch -mibt.
13644 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
13646 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
13649 * config/arm/arm.c (arm_valid_symbolic_address): Handle
13650 arm_word_relocations.
13652 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
13655 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
13656 extern array declarations.
13658 2018-03-27 Richard Biener <rguenther@suse.de>
13660 PR middle-end/84067
13661 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
13662 explicit single_use checks.
13664 2018-03-27 Richard Biener <rguenther@suse.de>
13666 PR tree-optimization/85082
13667 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
13670 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
13672 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
13673 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
13674 Turn on fasynchronous-unwind-tables and funwind-tables.
13676 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
13679 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
13680 (*bmi_blsr_<mode>_ccz): Ditto.
13682 2018-03-26 Tom de Vries <tom@codesourcery.com>
13684 PR tree-optimization/85063
13685 * omp-general.c (offloading_function_p): New function. Factor out
13687 * omp-offload.c (pass_omp_target_link::gate): ... here.
13688 * omp-general.h (offloading_function_p): Declare.
13689 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
13690 with attribute omp declare target for offloading functions.
13692 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
13694 PR tree-optimization/84005
13695 * tree-data-ref.h (get_base_for_alignment): Declare.
13696 * tree-data-ref.c (get_base_for_alignment_1): New function.
13697 (get_base_for_alignment): Likewise.
13698 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
13699 get_base_for_alignment to find a suitable base object, instead
13700 of always using drb->base_address.
13702 2018-03-23 Jakub Jelinek <jakub@redhat.com>
13704 PR inline-asm/85022
13705 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
13706 known size by default.
13708 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
13710 PR inline-asm/85030
13711 * lra-constraints.c (process_alt_operands): Don't match BLKmode
13712 and non BLKmode operands.
13714 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13717 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
13718 Clean up attributes.
13720 2018-03-23 Richard Biener <rguenther@suse.de>
13723 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
13724 we are going to emit early debug for LTO.
13726 2018-03-23 Jakub Jelinek <jakub@redhat.com>
13728 PR inline-asm/85034
13729 * function.c (match_asm_constraints_1): Don't optimize if input
13730 doesn't satisfy general_operand predicate for output's mode.
13732 PR inline-asm/85022
13733 * alias.c (write_dependence_p): Don't require for x_canonicalized
13734 non-VOIDmode if x has VOIDmode.
13737 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
13738 just don't try to optimize it rather than assert it never happens.
13740 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
13742 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
13743 macro expansions for definition of ST_INTERNAL_<mode> and
13744 LD_INTERNAL_<mode> builtins.
13745 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
13747 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
13749 (altivec_expand_st_builtin): Likewise.
13750 (altivec_expand_builtin): Remove calls to deleted functions.
13751 (rs6000_address_for_altivec): Delete this function.
13752 * config/rs6000/vector.md: Remove expands for
13753 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
13755 2018-03-22 Sudakshina Das <sudi.das@arm.com>
13758 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
13759 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
13760 re-computing once computed.
13761 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
13762 (arm_init_machine_status): Initialize
13763 machine->static_chain_stack_bytes.
13765 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
13768 * doc/extend.texi: Add four new prototypes for vec_ld.
13769 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
13770 definitions for more logical presentation.
13771 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
13772 entries for V1TI variants of __builtin_altivec_ld builtin.
13773 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
13774 handling of V1TI variant of LVX icode pattern.
13775 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
13776 (rs6000_gimple_fold_builtin): Likewise.
13777 (altivec_init_builtins): Add code to define
13778 __builtin_altivec_lvx_v1ti function.
13780 2018-03-22 Jakub Jelinek <jakub@redhat.com>
13782 PR inline-asm/84941
13783 * function.c (match_asm_constraints_1): Don't do the optimization
13784 if input isn't a REG, SUBREG, MEM or constant.
13786 2018-03-22 Tom de Vries <tom@codesourcery.com>
13788 PR tree-optimization/84956
13789 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
13790 bb_has_abnormal_pred.
13792 2018-03-22 Jakub Jelinek <jakub@redhat.com>
13795 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
13796 DECL_INITIAL (decl) to decl at the end.
13797 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
13798 adjust the comment.
13800 2018-03-21 Joseph Myers <joseph@codesourcery.com>
13802 * doc/extend.texi (__builtin_tgmath): Document when complex
13803 integer types are treated as _Complex _Float64.
13805 2018-03-21 Tom de Vries <tom@codesourcery.com>
13807 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
13809 2018-03-21 Jakub Jelinek <jakub@redhat.com>
13811 PR tree-optimization/84960
13812 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
13813 if it is ENTRY block, move them into single succ of ENTRY in that case.
13815 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
13817 PR tree-optimization/84811
13818 * poly-int.h (poly_span_traits): Remove the T3 parameter and
13819 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
13820 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
13821 (known_subrange_p): Update accordingly. Cast each value involved
13822 in the size comparison, rather than casting the result of the
13825 2018-03-21 Jakub Jelinek <jakub@redhat.com>
13827 PR tree-optimization/84982
13828 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
13829 by flipping the least significant bit rather than all bits from
13830 bitpos to bitpos + bitsize - 1.
13832 2018-03-21 Nathan Sidwell <nathan@acm.org>
13834 * doc/extend.texi (Deprecated Features): Remove mention of
13835 long-deleted deprecations.
13837 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13840 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
13841 * configure: Regenerate.
13843 2018-03-21 Tom de Vries <tom@codesourcery.com>
13845 PR tree-optimization/83126
13846 * tree-parloops.c (num_phis): New function.
13847 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
13849 2018-03-21 Nathan Sidwell <nathan@acm.org>
13851 * doc/extend.texi (Deprecated Features): Update deprecated flags,
13852 mention anon-struct/union members and trailing attributes.
13854 2018-03-21 Bin Cheng <bin.cheng@arm.com>
13856 PR tree-optimization/84969
13857 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
13858 builtin memset partitions if they set different rhs values.
13860 2018-03-21 Jakub Jelinek <jakub@redhat.com>
13862 PR rtl-optimization/84989
13863 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
13864 VEC_DUPLICATE with scalar result mode.
13866 2018-03-21 Martin Liska <mliska@suse.cz>
13869 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
13870 not intended return statement.
13872 2018-03-21 Martin Liska <mliska@suse.cz>
13875 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
13876 (chkp_find_bound_slots_1): Limit number of iterations.
13878 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
13881 * Minor grammar fixes for x86 options.
13883 2018-03-20 Jakub Jelinek <jakub@redhat.com>
13886 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
13887 holding REG_CFA_RESTORE notes, instead turn them into a USE.
13889 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
13892 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
13893 (altivec_lvx_<mode>_1op): Likewise.
13894 (altivec_stvx_<mode>_2op): Likewise.
13895 (altivec_stvx_<mode>_1op): Likewise.
13896 (altivec_lvx_<VM2:mode>): New define_expand.
13897 (altivec_stvx_<VM2:mode>): Likewise.
13898 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
13899 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
13900 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
13901 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
13902 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
13903 (rs6000_gen_lvx): Likewise.
13904 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
13905 (altivec_expand_stv_builtin): Likewise.
13906 (altivec_expand_builtin): Likewise.
13907 * config/rs6000/vector.md: Likewise.
13909 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13912 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
13915 2018-03-20 Richard Biener <rguenther@suse.de>
13918 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
13919 sign-conversions as zero, fall back to standard scalar_stmt
13922 2018-03-20 Martin Liska <mliska@suse.cz>
13925 * predict.c (rebuild_frequencies): Handle case when we have
13926 PROFILE_ABSENT, but flag_guess_branch_prob is false.
13928 2018-03-20 Jakub Jelinek <jakub@redhat.com>
13931 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
13932 flag_section_anchors.
13933 * varasm.c (use_blocks_for_decl_p): Remove hack for
13934 dw2_force_const_mem.
13937 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
13939 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
13940 be created, use lowpart_subreg of operands[0] rather than operands[0]
13942 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
13943 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
13944 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
13945 and n constraint instead of aarch64_shift_imm_di and Usd.
13946 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
13947 (*aarch64_<optab>_reg_minus<mode>3): ... this.
13949 2018-03-20 Sudakshina Das <sudi.das@arm.com>
13952 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
13953 to favor GPR over NEON registers.
13954 (<shift>di3_neon): Likewise.
13956 2018-03-20 Tom de Vries <tom@codesourcery.com>
13959 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
13960 (nvptx_process_pars): Emit bar.sync asap and alap.
13962 2018-03-20 Tom de Vries <tom@codesourcery.com>
13965 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
13966 seen_label if seen_label is already set.
13968 2018-03-20 Jakub Jelinek <jakub@redhat.com>
13971 * config/i386/i386.c (fold_builtin_cpu): For features above 31
13972 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
13973 Use 1U instead of 1. Formatting fixes.
13976 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
13977 instead of TREE_TYPE (s1) for the return value.
13979 2018-03-19 Jakub Jelinek <jakub@redhat.com>
13981 PR tree-optimization/84946
13982 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
13983 bitsize + bitsize in poly_uint64 rather than poly_int64.
13986 * dwarf2asm.c: Include fold-const.c.
13987 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
13988 of decl rather than decl itself.
13990 PR rtl-optimization/84643
13991 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
13993 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
13996 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
13997 calling assemble_variable.
13999 2018-03-19 Sudakshina Das <sudi.das@arm.com>
14002 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
14003 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
14005 2018-03-19 Jim Wilson <jimw@sifive.com>
14008 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
14009 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
14010 (riscv_first_stack_step): Likewise.
14011 (riscv_option_override): Use STACK_BOUNDARY instead of
14012 MIN_STACK_BOUNDARY.
14013 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
14014 MIN_STACK_BOUNDARY.
14015 (BIGGEST_ALIGNMENT): Set to 128.
14016 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
14017 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
14020 2018-03-19 Richard Biener <rguenther@suse.de>
14022 PR tree-optimization/84933
14023 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
14024 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
14026 2018-03-19 Richard Biener <rguenther@suse.de>
14028 PR tree-optimization/84859
14029 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
14030 (cond_if_else_store_replacement): Perform sinking operation on
14031 single-store BBs regardless of MAX_STORES_TO_SINK setting.
14032 Generalize what a BB with a single eligible store is.
14034 2018-03-19 Richard Biener <rguenther@suse.de>
14036 PR tree-optimization/84929
14037 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
14038 chrec_is_positive against non-chrec arg.
14040 2018-03-19 Tamar Christina <tamar.christina@arm.com>
14043 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
14045 2018-03-18 Martin Liska <mliska@suse.cz>
14047 PR rtl-optimization/84635
14048 * regrename.c (build_def_use): Use matches_mode only when
14051 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
14053 PR tree-optimization/84913
14054 * tree-vect-loop.c (vectorizable_reduction): Don't try to
14055 vectorize chains of COND_EXPRs.
14057 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
14059 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
14061 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
14063 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
14065 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
14067 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
14069 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
14070 Kito Cheng <kito.cheng@gmail.com>
14072 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
14073 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
14074 (nds32_adjust_reg_alloc_order): New function.
14075 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
14077 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
14079 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
14080 nds32_print_operand, nds32_print_operand_address): Use
14081 HOST_WIDE_INT_PRINT_DEC instead.
14083 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
14085 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
14087 2018-03-17 Jakub Jelinek <jakub@redhat.com>
14090 * config/i386/i386.c (initial_ix86_tune_features,
14091 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
14092 unsigned long long.
14093 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
14094 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
14095 rather than 1u << ix86_tune. Formatting fix.
14096 (ix86_option_override_internal): Change ix86_arch_mask from
14097 unsigned int to unsigned HOST_WIDE_INT, initialize to
14098 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
14099 (ix86_function_specific_restore): Likewise.
14101 2018-03-16 Jakub Jelinek <jakub@redhat.com>
14104 * postreload.c (reload_combine_recognize_pattern): Perform
14105 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
14106 truncate_int_for_mode the result for the destination's mode.
14109 * hsa-gen.c (mem_type_for_type): Fix comment typo.
14110 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
14112 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
14115 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
14118 * lra-assigns.c (lra_split_hard_reg_for): Don't use
14119 regno_allocno_class_array and sorted_pseudos.
14120 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
14121 insns where regno is used.
14123 2018-03-16 Martin Liska <mliska@suse.cz>
14126 * multiple_target.c (create_dispatcher_calls): Redirect
14127 reference in the symbol table.
14129 2018-03-16 Martin Liska <mliska@suse.cz>
14132 * multiple_target.c (create_dispatcher_calls): Redirect also
14135 2018-03-16 Jakub Jelinek <jakub@redhat.com>
14139 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
14140 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
14141 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
14143 2018-03-16 Julia Koval <julia.koval@intel.com>
14145 * doc/invoke.texi (Skylake Server): Add CLWB.
14146 Cannonlake): Remove CLWB.
14148 2018-03-16 Jakub Jelinek <jakub@redhat.com>
14150 PR tree-optimization/84841
14151 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
14153 (FLOAT_ONE_CONST_TYPE): Define.
14154 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
14155 (sort_by_operand_rank): Put entries with higher constant_type last
14156 rather than first to match comments.
14158 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
14160 * config/nios2/nios2.md (movsi_internal): Fix thinko in
14163 2018-03-15 Jakub Jelinek <jakub@redhat.com>
14166 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
14167 check and use address of target always.
14169 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
14172 * config/i386/i386.c (indirect_thunk_needed): Update comments.
14173 (indirect_thunk_bnd_needed): Likewise.
14174 (indirect_thunks_used): Likewise.
14175 (indirect_thunks_bnd_used): Likewise.
14176 (indirect_return_needed): New.
14177 (indirect_return_bnd_needed): Likewise.
14178 (output_indirect_thunk_function): Add a bool argument for
14180 (output_indirect_thunk_function): Don't generate alias for
14181 function return thunk.
14182 (ix86_code_end): Call output_indirect_thunk_function to generate
14183 function return thunks.
14184 (ix86_output_function_return): Set indirect_return_bnd_needed
14185 and indirect_return_needed instead of indirect_thunk_bnd_needed
14186 and indirect_thunk_needed.
14188 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
14190 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
14191 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
14192 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
14194 2018-03-15 David Malcolm <dmalcolm@redhat.com>
14195 Paul Hua <paul.hua.gm@gmail.com>
14198 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
14200 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
14202 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
14203 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
14204 resp. SFmode cases.
14206 2018-03-15 Tamar Christina <tamar.christina@arm.com>
14209 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
14210 instead of GET_MODE_SIZE when comparing Units.
14212 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
14215 * varasm.c (hash_section): Return an unchangeble hash value
14216 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
14217 Return !aarch64_can_use_per_function_literal_pools_p ().
14219 2018-03-15 Jakub Jelinek <jakub@redhat.com>
14222 * optabs.c (emit_conditional_move): Pass address of cmode's copy
14223 rather than address of cmode as last argument to prepare_cmp_insn.
14225 2018-03-15 Julia Koval <julia.koval@intel.com>
14227 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
14228 F_AVX512VNNI, F_AVX512BITALG): New.
14230 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
14233 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
14234 insn for floating-point loads and stores.
14236 2018-03-14 Carl Love <cel@us.ibm.com>
14238 * config/rs6000/rs6000-c.c: Add macro definitions for
14239 ALTIVEC_BUILTIN_VEC_PERMXOR.
14240 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
14241 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
14242 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
14243 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
14245 * config/doc/extend.texi: Add prototypes for vec_permxor.
14247 2018-03-14 David Malcolm <dmalcolm@redhat.com>
14250 * diagnostic-show-locus.c (class layout_point): Convert m_line
14251 from int to linenum_type.
14252 (line_span::comparator): Use linenum "compare" function when
14253 comparing line numbers.
14254 (test_line_span): New function.
14255 (layout_range::contains_point): Convert param "row" from int to
14257 (layout_range::intersects_line_p): Likewise.
14258 (layout::will_show_line_p): Likewise.
14259 (layout::print_source_line): Likewise.
14260 (layout::should_print_annotation_line_p): Likewise.
14261 (layout::print_annotation_line): Likewise.
14262 (layout::print_leading_fixits): Likewise.
14263 (layout::annotation_line_showed_range_p): Likewise.
14264 (struct line_corrections): Likewise for field m_row.
14265 (line_corrections::line_corrections): Likewise for param "row".
14266 (layout::print_trailing_fixits): Likewise.
14267 (layout::get_state_at_point): Likewise.
14268 (layout::get_x_bound_for_row): Likewise.
14269 (layout::print_line): Likewise.
14270 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
14271 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
14272 * input.c (selftest::test_linenum_comparisons): New function.
14273 (selftest::input_c_tests): Call it.
14274 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
14275 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
14276 * selftest.h (ASSERT_GT): New macro.
14277 (ASSERT_GT_AT): New macro.
14278 (ASSERT_LT): New macro.
14279 (ASSERT_LT_AT): New macro.
14281 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
14283 PR rtl-optimization/84780
14284 * combine.c (distribute_links): Don't make a link based on pc_rtx.
14286 2018-03-14 Martin Liska <mliska@suse.cz>
14288 * tree.c (record_node_allocation_statistics): Use
14289 get_stats_node_kind.
14290 (get_stats_node_kind): New function extracted from
14291 record_node_allocation_statistics.
14292 (free_node): Use get_stats_node_kind.
14294 2018-03-14 Richard Biener <rguenther@suse.de>
14296 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
14297 that the value-set of ANTIC_IN doesn't grow.
14300 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
14302 (BB_VISITED_WITH_VISITED_SUCCS): New define.
14303 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
14305 2018-03-14 Julia Koval <julia.koval@intel.com>
14307 * config.gcc (icelake-client, icelake-server): New.
14309 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
14310 (initial_ix86_arch_features): Ditto.
14311 (PTA_SKYLAKE): Add SGX.
14312 (PTA_ICELAKE): Remove.
14313 (PTA_ICELAKE_CLIENT): New.
14314 (PTA_ICELAKE_SERVER): New.
14315 (ix86_option_override_internal): Split up icelake on icelake client and
14317 (get_builtin_code_for_version): Ditto.
14318 (fold_builtin_cpu): Ditto.
14319 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
14320 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
14321 * config/i386/i386.h (processor_type): Ditto.
14322 * doc/invoke.texi: Ditto.
14324 2018-03-14 Jakub Jelinek <jakub@redhat.com>
14327 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
14328 INTEGER_CST offset, add it together with bitpos / 8 and
14329 sign extend based on POINTER_SIZE.
14333 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
14336 * config/i386/constraints.md (Yc): New register constraint.
14337 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
14338 Use Yc constraint for alternative 2 of operand 0. Remove
14339 preferred_for_speed attribute.
14341 2018-03-14 Richard Biener <rguenther@suse.de>
14343 PR tree-optimization/84830
14344 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
14345 with the old one to avoid oscillations.
14347 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
14350 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
14352 (assign_by_spills): Return a flag of reload assignment failure.
14353 Do not process the reload assignment failures. Do not spill other
14354 reload pseudos if they has the same reg class. Update n if
14356 (lra_assign): Add a return arg. Set up from the result of
14357 assign_by_spills call.
14358 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
14359 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
14360 usage_insns if it is not NULL.
14361 (spill_hard_reg_in_range): New function.
14362 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
14363 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
14364 function prototypes.
14365 (lra_assign): Change prototype.
14366 * lra.c (lra): Add code to deal with fails by splitting hard reg
14369 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
14371 * config/riscv/riscv.opt (mrelax): New option.
14372 * config/riscv/riscv.c (riscv_file_start): Emit ".option
14373 "norelax" when riscv_mrelax is disabled.
14374 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
14376 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14379 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
14380 reassociation for int modes.
14382 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
14384 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
14385 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
14387 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
14388 * config/aarch64/aarch64-sve.md
14389 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
14390 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
14391 (*extend<mode><Vwide>2): Rename to...
14392 (aarch64_sve_extend<mode><Vwide>2): ...this.
14393 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
14394 renaming the old pattern to...
14395 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
14397 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
14398 define_expand, renaming the old pattern to...
14399 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
14400 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
14401 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
14402 account when deciding which SVE instruction the optab should use.
14403 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
14405 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
14407 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
14408 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
14409 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
14410 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
14411 (tlsdesc_small_<mode>): Turn a define_expand and use
14412 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
14413 (tlsdesc_small_advsimd_<mode>): ...this.
14414 (tlsdesc_small_sve_<mode>): New pattern.
14416 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
14418 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
14419 (UNSPEC_UMUL_HIGHPART): New constants.
14420 (MUL_HIGHPART): New int iteraor.
14421 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
14422 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
14424 (*<su>mul<mode>3_highpart): New define_insn.
14426 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
14429 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
14432 2018-03-13 Martin Liska <mliska@suse.cz>
14435 * (sem_item_optimizer::sem_item_optimizer): Initialize new
14437 (sem_item_optimizer::~sem_item_optimizer): Release it.
14438 (sem_item_optimizer::merge_classes): Register variable aliases.
14439 (sem_item_optimizer::fixup_pt_set): New function.
14440 (sem_item_optimizer::fixup_points_to_sets): Likewise.
14441 * ipa-icf.h: Declare new variables and functions.
14443 2018-03-13 Jakub Jelinek <jakub@redhat.com>
14445 PR middle-end/84834
14446 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
14447 integer_pow2p@2 and test integer_pow2p in condition.
14448 (A < 0 ? C : 0): Similarly for @1.
14450 PR middle-end/84831
14451 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
14452 characters starting at p contain '\0' character, don't look beyond
14456 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
14457 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
14460 * reg-stack.c (change_stack): Change update_end var from int to
14461 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
14462 also call set_block_for_insn on the newly added insns and rescan.
14465 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
14466 on the last operand.
14469 * tree.c (stabilize_reference_1): Return save_expr (e) for
14470 STATEMENT_LIST even if it doesn't have side-effects.
14472 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
14474 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
14476 2018-03-12 Renlin Li <renlin.li@arm.com>
14478 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
14479 aarch64_output_scalar_simd_mov_immediate.
14481 2018-03-12 Martin Sebor <msebor@redhat.com>
14483 PR tree-optimization/83456
14484 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
14485 for perfectly overlapping calls to memcpy.
14486 (gimple_fold_builtin_memory_chk): Same.
14487 (gimple_fold_builtin_strcpy): Handle no-warning.
14488 (gimple_fold_builtin_stxcpy_chk): Same.
14489 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
14491 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
14493 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
14494 parameter. Use it for SFmode.
14495 (rs6000_function_arg_advance_1): Adjust.
14496 (rs6000_function_arg): Adjust.
14497 (rs6000_gimplify_va_arg): Pass false for that new parameter.
14499 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
14501 PR rtl-optimization/84169
14502 PR rtl-optimization/84780
14503 * combine.c (can_combine_p): Check for a 2-insn combination whether
14504 the destination register is used between the two insns, too.
14506 2018-03-12 Richard Biener <rguenther@suse.de>
14508 PR tree-optimization/84803
14509 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
14510 for refs DR analysis didn't process.
14512 2018-03-12 Richard Biener <rguenther@suse.de>
14514 PR tree-optimization/84777
14515 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
14516 force-vectorize loops ignore whether we are optimizing for size.
14518 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
14520 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
14521 (TARGET_MD_ASM_ADJUST): Define.
14523 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
14524 Kito Cheng <kito.cheng@gmail.com>
14525 Chung-Ju Wu <jasonwucj@gmail.com>
14527 * config/nds32/nds32.c (nds32_compute_stack_frame,
14528 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
14529 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
14530 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
14531 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
14532 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
14533 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
14534 * config/nds32/nds32.md (prologue, epilogue): Use macro
14535 NDS32_V3PUSH_AVAILABLE_P to do checking.
14537 2018-03-11 Jakub Jelinek <jakub@redhat.com>
14540 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
14541 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
14542 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
14543 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
14544 addition of most attributes on !orig_type_die or the attribute not
14545 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
14547 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
14548 Chung-Ju Wu <jasonwucj@gmail.com>
14550 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
14551 __NDS32_VH__ macro.
14552 * config/nds32/nds32.opt (mvh): New option.
14554 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
14555 Chung-Ju Wu <jasonwucj@gmail.com>
14557 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
14559 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
14560 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
14563 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
14564 Chung-Ju Wu <jasonwucj@gmail.com>
14566 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
14568 * config/nds32/nds32-multiple.md (strlensi): New pattern.
14569 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
14571 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
14572 Kito Cheng <kito.cheng@gmail.com>
14573 Chung-Ju Wu <jasonwucj@gmail.com>
14575 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
14576 UNSPEC_FFMISM and UNSPEC_FLMISM.
14577 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
14578 for ffb, ffmism and flmism.
14579 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
14580 (unspec_ffmism): Ditto.
14581 (unspec_flmism): Ditto.
14582 (nds32_expand_builtin_impl): Check if string extension is available.
14583 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
14584 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
14586 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
14589 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
14592 * lra-assigns.c (assign_by_spills): Return a flag of reload
14593 assignment failure. Do not process the reload assignment
14594 failures. Do not spill other reload pseudos if they has the same
14596 (lra_assign): Add a return arg. Set up from the result of
14597 assign_by_spills call.
14598 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
14599 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
14600 usage_insns if it is not NULL.
14601 (spill_hard_reg_in_range): New function.
14602 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
14603 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
14604 function prototypes.
14605 (lra_assign): Change prototype.
14606 * lra.c (lra): Add code to deal with fails by splitting hard reg
14609 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
14612 * config/i386/i386.opt: Replace Enforcment with Enforcement.
14614 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
14617 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
14618 (dw_val_node): Add val_symbolic_view.
14619 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
14620 (symview_upper_bound): New.
14621 (new_line_info_table): Initialize symviews_since_reset.
14622 (dwarf2out_source_line): Count symviews_since_reset and set
14623 symview_upper_bound.
14624 (dw_val_equal_p): Handle symview.
14625 (add_AT_symview): New.
14626 (print_dw_val): Handle symview.
14627 (attr_checksum, attr_checksum_ordered): Likewise.
14628 (same_dw_val_p, size_of_die): Likewise.
14629 (value_format, output_die): Likewise.
14630 (add_high_low_attributes): Use add_AT_symview for entry_view.
14631 (dwarf2out_finish): Reset symview_upper_bound, clear
14634 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
14637 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
14638 Add strict argument and use it.
14639 (rs6000_split_multireg_move): Update for new strict argument.
14640 (mem_operand_gpr): Disallow all non-offsettable addresses.
14641 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
14643 2018-03-09 Jakub Jelinek <jakub@redhat.com>
14646 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
14647 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
14648 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
14651 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
14652 decl, use remap_type if we want to use the type.
14654 2018-03-09 Martin Sebor <msebor@redhat.com>
14656 PR tree-optimization/84526
14657 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
14659 (builtin_access::generic_overlap): Be prepared to handle non-array
14662 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
14664 PR rtl-optimization/84682
14665 * lra-constraints.c (process_address_1): Check is_address flag
14666 for address constraints.
14667 (process_alt_operands): Likewise.
14668 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
14669 preprocess_constraints.
14670 * recog.h (preprocess_constraints): Add oploc parameter.
14672 * recog.c (preprocess_constraints): Test address_operand for
14673 CT_ADDRESS constraints.
14675 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
14678 * lra-assigns.c (assign_by_spills): Return a flag of reload
14679 assignment failure. Do not process the reload assignment
14680 failures. Do not spill other reload pseudos if they has the same
14682 (lra_assign): Add a return arg. Set up from the result of
14683 assign_by_spills call.
14684 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
14685 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
14686 usage_insns if it is not NULL.
14687 (spill_hard_reg_in_range): New function.
14688 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
14689 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
14690 function prototypes.
14691 (lra_assign): Change prototype.
14692 * lra.c (lra): Add code to deal with fails by splitting hard reg
14695 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14698 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
14699 Accept complain bool parameter. Only emit errors if it is true.
14700 (arm_parse_cpu_option_name): Likewise.
14701 (arm_target_thumb_only): Adjust callers of the above.
14702 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
14703 prototype to take a default true bool parameter.
14704 (arm_parse_arch_option_name): Likewise.
14706 2018-03-09 David Malcolm <dmalcolm@redhat.com>
14707 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
14711 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
14712 * configure: Regenerate.
14713 * configure.ac ("linker --version-script option"): New.
14714 ("linker soname option"): New.
14716 2018-03-09 Richard Biener <rguenther@suse.de>
14718 PR tree-optimization/84775
14719 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
14720 immediate uses of predicate stmts and mark them modified.
14723 PR tree-optimization/84178
14724 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
14726 (version_loop_for_if_conversion): Delay update_ssa call.
14727 (tree_if_conversion): Delay update_ssa until after predicate
14730 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
14733 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
14734 when the function accesses prior frames.
14736 2018-03-08 Jakub Jelinek <jakub@redhat.com>
14739 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
14740 gen_llsym, otherwise call maybe_gen_llsym.
14742 PR inline-asm/84742
14743 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
14744 has ',' character inside of it.
14746 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14749 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
14750 as clobbering CC_REGNUM.
14752 2018-03-08 Richard Biener <rguenther@suse.de>
14754 PR middle-end/84552
14755 * tree-scalar-evolution.c: Include tree-into-ssa.h.
14756 (follow_copies_to_constant): Do not follow SSA names registered
14759 2018-03-08 Richard Biener <rguenther@suse.de>
14761 PR tree-optimization/84178
14762 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
14764 (version_loop_for_if_conversion): Delay update_ssa call.
14765 (tree_if_conversion): Delay update_ssa until after predicate
14768 2018-03-08 David Malcolm <dmalcolm@redhat.com>
14770 PR tree-optimization/84178
14771 * tree-if-conv.c (release_bb_predicate): Remove the
14772 the assertion that the stmts have NULL use_ops.
14773 Discard the statements, asserting that they haven't
14774 yet been added to a BB.
14776 2018-03-08 Richard Biener <rguenther@suse.de>
14778 PR tree-optimization/84746
14779 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
14780 (phi_translate): Pass in destination ANTIC_OUT set.
14781 (phi_translate_1): Likewise. For a simplified result lookup
14782 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
14783 (phi_translate_set): Adjust.
14784 (do_pre_regular_insertion): Likewise.
14785 (do_pre_partial_partial_insertion): Likewise.
14787 2018-03-08 Martin Liska <mliska@suse.cz>
14789 PR gcov-profile/84735
14790 * doc/gcov.texi: Document usage of profile files.
14791 * gcov-io.h: Document changes in the format.
14793 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
14797 * dwarf2out.c (struct dw_line_info_table): Update comments for
14799 (FORCE_RESET_NEXT_VIEW): New.
14800 (FORCE_RESETTING_VIEW_P): New.
14801 (RESETTING_VIEW_P): Check for -1 too.
14802 (ZERO_VIEW_P): Likewise.
14803 (new_line_info_table): Force-reset next view.
14804 (dwarf2out_begin_function): Likewise.
14805 (dwarf2out_source_line): Simplify zero_view_p initialization.
14806 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
14807 view directly. Omit view when omitting .loc at line 0.
14809 2018-03-08 Jakub Jelinek <jakub@redhat.com>
14811 PR tree-optimization/84740
14812 * tree-switch-conversion.c (process_switch): Call build_constructors
14813 only if info.phi_count is non-zero.
14815 PR tree-optimization/84739
14816 * tree-tailcall.c (find_tail_calls): Check call arguments against
14817 DECL_ARGUMENTS (current_function_decl) rather than
14818 DECL_ARGUMENTS (func) when checking for tail recursion.
14820 2018-03-07 Jakub Jelinek <jakub@redhat.com>
14822 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
14823 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
14824 Volker Reichelt's entry and add entries for people that perform
14825 GCC fuzzy testing and report numerous bugs.
14827 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
14830 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
14831 readonly data in sdata, if that is disabled.
14832 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
14833 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
14834 -mreadonly-in-sdata option.
14836 2018-03-07 Martin Sebor <msebor@redhat.com>
14838 PR tree-optimization/84468
14839 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
14840 basic block when looking for nul assignment.
14842 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
14845 * except.h (output_function_exception_table): Adjust prototype.
14846 * except.c (output_function_exception_table): Remove FNNAME parameter
14847 and add SECTION parameter. Ouput one part of the table at a time.
14848 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
14849 the first part of the exception table and emit unwind directives.
14850 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
14851 (i386_pe_seh_cold_init): Likewise.
14852 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
14853 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
14854 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
14855 (ix86_output_call_insn): Emit a nop in one more case for SEH.
14856 * config/i386/winnt.c: Include except.h.
14857 (struct seh_frame_state): Add reg_offset, after_prologue and
14858 in_cold_section fields.
14859 (i386_pe_seh_end_prologue): Set seh->after_prologue.
14860 (i386_pe_seh_cold_init): New function.
14861 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
14862 to seh->in_cold_section.
14863 (seh_emit_push): Record the offset of the push.
14864 (seh_emit_save): Record the offet of the save.
14865 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
14866 Test seh->after_prologue to disregard the epilogue.
14867 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
14868 (i386_pe_end_cold_function): New function.
14870 2018-03-07 Jakub Jelinek <jakub@redhat.com>
14873 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
14874 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
14877 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
14879 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
14880 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
14882 PR middle-end/84723
14883 * multiple_target.c: Include tree-inline.h and intl.h.
14884 (expand_target_clones): Diagnose and fail if node->definition and
14885 !tree_versionable_function_p (node->decl).
14887 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
14889 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
14891 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
14892 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
14893 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
14895 2018-03-06 Jakub Jelinek <jakub@redhat.com>
14898 * combine.c (try_combine): Use reg_or_subregno instead of handling
14899 just paradoxical SUBREGs and REGs.
14901 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
14903 * config/arc/arc.c (arc_finalize_pic): Remove function.
14904 (arc_must_save_register): We use single base PIC register, remove
14905 checks to save/restore the PIC register.
14906 (arc_expand_prologue): Likewise.
14907 * config/arc/arc-protos.h (arc_set_default_type_attributes):
14909 (arc_verify_short): Likewise.
14910 (arc_attr_type): Likewise.
14911 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
14912 (walk_stores): Likewise.
14913 (arc_address_cost): Make it static.
14914 (arc_verify_short): Likewise.
14915 (branch_dest): Likewise.
14916 (arc_attr_type): Likewise.
14917 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
14918 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
14919 (arc_final_prescan_insn): Remove inserting the nops due to
14920 hardware hazards. It is done in reorg step.
14921 (insn_length_variant_t): Remove.
14922 (insn_length_parameters_t): Likewise.
14923 (arc_insn_length_parameters): Likewise.
14924 (arc_get_insn_variants): Likewise.
14925 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
14927 2018-03-06 Jakub Jelinek <jakub@redhat.com>
14929 PR inline-asm/84683
14930 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
14933 PR tree-optimization/84687
14934 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
14936 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
14938 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14940 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
14941 Rename to ppc_speculation_barrier.
14942 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
14943 __builtin_ppc_speculation_barrier.
14945 2018-03-05 Jakub Jelinek <jakub@redhat.com>
14948 * combine.c (combine_simplify_rtx): Don't try to simplify if
14949 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
14952 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
14954 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
14955 to 32 bytes when compiling for POWER9.
14957 2018-03-05 Jakub Jelinek <jakub@redhat.com>
14960 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
14961 regparm >= 3 with no arg reg available also for calls with
14962 flag_force_indirect_call. Pass decl to ix86_function_regparm.
14965 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
14967 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
14969 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
14972 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
14974 2018-03-05 Richard Biener <rguenther@suse.de>
14976 PR tree-optimization/84486
14977 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
14978 When inserting a __builtin_assume_aligned call set the LHS
14979 SSA name alignment info accordingly.
14981 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
14983 PR tree-optimization/84114
14984 * config/aarch64/aarch64.c (aarch64_reassociation_width)
14985 Avoid reassociation of FLOAT_MODE addition.
14987 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
14989 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
14990 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
14991 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
14992 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
14993 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
14994 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
14995 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
14997 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
14998 __builtin_ia32_wbinvd): New builtins.
14999 (SPECIAL_ARGS2): New.
15000 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
15001 (SPECIAL_ARGS2): New.
15002 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
15003 (ix86_valid_target_attribute_inner_p): Ditto.
15004 (ix86_init_mmx_sse_builtins): Add special_args2.
15005 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
15006 TARGET_WBNOINVD_P): New.
15007 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
15008 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
15009 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
15010 * config/i386/immintrin.h (_wbinvd): New intrinsic.
15011 * config/i386/pconfigintrin.h: New file.
15012 * config/i386/wbnoinvdintrin.h: Ditto.
15013 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
15015 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
15017 2018-03-05 Richard Biener <rguenther@suse.de>
15019 PR tree-optimization/84670
15020 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
15022 (BB_VISITED_WITH_VISITED_SUCCS): New define.
15023 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
15024 (compute_antic_aux): Only assert the number of values in ANTIC_IN
15025 doesn't grow if all successors (recursively) were visited at least
15028 2018-03-05 Richard Biener <rguenther@suse.de>
15030 PR tree-optimization/84650
15031 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
15032 if executed in the loop pipeline.
15034 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
15036 * doc/configfiles.texi (Configuration Files): Move info about
15037 conditionalizing $target-protos.h to...
15038 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
15039 differs from $target-protos.h.
15041 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
15042 Chung-Ju Wu <jasonwucj@gmail.com>
15044 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
15045 * config/nds32/nds32-multiple.md (setmemsi): Define.
15046 * config/nds32/nds32-memory-manipulation.c
15047 (nds32_gen_dup_4_byte_to_word_value): New.
15048 (emit_setmem_word_loop): New.
15049 (emit_setmem_byte_loop): New.
15050 (nds32_expand_setmem_loop): New.
15051 (nds32_expand_setmem_loop_v3m): New.
15052 (nds32_expand_setmem_unroll): New.
15053 (nds32_expand_setmem): New.
15055 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
15056 Chung-Ju Wu <jasonwucj@gmail.com>
15058 * config/nds32/nds32-memory-manipulation.c
15059 (nds32_emit_load_store): New.
15060 (nds32_emit_post_inc_load_store): New.
15061 (nds32_emit_mem_move): New.
15062 (nds32_emit_mem_move_block): New.
15063 (nds32_expand_movmemsi_loop_unknown_size): New.
15064 (nds32_expand_movmemsi_loop_known_size): New.
15065 (nds32_expand_movmemsi_loop): New.
15066 (nds32_expand_movmemsi_unroll): New.
15067 (nds32_expand_movmemqi): Rename ...
15068 (nds32_expand_movmemsi): ... to this.
15069 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
15070 (movmemsi): ... to this.
15071 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
15072 (nds32_expand_movmemsi): ... to this.
15074 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
15075 Monk Chiang <sh.chiang04@gmail.com>
15076 Chung-Ju Wu <jasonwucj@gmail.com>
15078 * config/nds32/nds32-protos.h
15079 (nds32_expand_load_multiple): New arguments.
15080 (nds32_expand_store_multiple): Ditto.
15081 (nds32_valid_multiple_load_store): Rename ...
15082 (nds32_valid_multiple_load_store_p): ... to this.
15083 * config/nds32/nds32-memory-manipulation.c
15084 (nds32_expand_load_multiple): Refine implementation.
15085 (nds32_expand_store_multiple): Ditto.
15086 * config/nds32/nds32-multiple.md
15087 (load_multiple): Update nds32_expand_load_multiple interface.
15088 (store_multiple): Update nds32_expand_store_multiple interface.
15089 * config/nds32/nds32-predicates.c
15090 (nds32_valid_multiple_load_store): Rename ...
15091 (nds32_valid_multiple_load_store_p): ... to this and refine
15093 * config/nds32/predicates.md
15094 (nds32_load_multiple_and_update_address_operation): New predicate.
15095 (nds32_store_multiple_and_update_address_operation): New predicate.
15097 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
15098 Chung-Ju Wu <jasonwucj@gmail.com>
15100 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
15101 (combo): New attribute.
15102 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
15104 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
15106 * config/nds32/nds32.opt: Change -mcmodel= default value.
15108 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
15109 Monk Chiang <sh.chiang04@gmail.com>
15110 Chung-Ju Wu <jasonwucj@gmail.com>
15112 * config/nds32/constants.md (unspec_element): New enum.
15113 * config/nds32/constraints.md (Umw): New constraint.
15114 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
15115 * config/nds32/nds32-intrinsic.md: Likewise.
15116 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
15117 (nds32_valid_smw_lwm_base_p): New.
15118 (nds32_output_smw_single_word): New.
15119 (nds32_output_lmw_single_word): New.
15120 (nds32_expand_unaligned_load): New.
15121 (nds32_expand_unaligned_store): New.
15122 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
15123 (nds32_output_smw_single_word): Declare.
15124 (nds32_output_lmw_single_word): Declare.
15125 (nds32_expand_unaligned_load): Declare.
15126 (nds32_expand_unaligned_store): Declare.
15127 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
15128 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
15129 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
15130 NDS32_BUILTIN_UASTORE_DW.
15131 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
15134 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
15135 Kito Cheng <kito.cheng@gmail.com>
15136 Chung-Ju Wu <jasonwucj@gmail.com>
15138 * config/nds32/nds32-intrinsic.c
15139 (nds32_expand_builtin_null_ftype_reg): Delete.
15140 (nds32_expand_builtin_reg_ftype_imm): Ditto.
15141 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
15142 (nds32_read_argument): New.
15143 (nds32_legitimize_target): Ditto.
15144 (nds32_legitimize_argument): Ditto.
15145 (nds32_check_constant_argument): Ditto.
15146 (nds32_expand_unop_builtin): Ditto.
15147 (nds32_expand_unopimm_builtin): Ditto.
15148 (nds32_expand_binop_builtin): Ditto.
15149 (nds32_builtin_decl_impl): Ditto.
15150 (builtin_description): Ditto.
15151 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
15152 (nds32_init_builtins_impl): Ditto.
15153 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
15154 (nds32_builtin_decl): New.
15155 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
15156 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
15158 2018-03-02 Jeff Law <law@redhat.com>
15160 * reorg.c (stop_search_p): Handle DEBUG_INSN.
15161 (redundant_insn, fill_simple_delay_slots): Likewise.
15162 (fill_slots_from_thread): Likewise.
15163 * resource.c (mark_referenced_resources): Likewise.
15164 (mark_set_resources, find_dead_or_set_registers): Likewise.
15166 2018-03-02 Jakub Jelinek <jakub@redhat.com>
15168 * substring-locations.h (format_warning_va): Formatting fix for
15169 ATTRIBUTE_GCC_DIAG.
15170 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
15172 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
15173 * substring-locations.c: Include intl.h.
15174 (format_warning_va): Turned into small wrapper around
15175 format_warning_n_va, renamed to ...
15176 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
15177 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
15179 (format_warning_at_substring_n): New function.
15180 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
15181 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
15182 format_warning_at_substring with just a shorter name instead of
15183 const function pointer.
15184 (fmtwarn_n): New function.
15185 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
15186 appropriate, get rid of all the fmtstr temporaries, move conditionals
15187 with G_() wrapped string literals directly into fmtwarn arguments,
15188 cast dir.len to (int), formatting fixes.
15190 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
15192 * doc/invoke.texi: Remove "Cilk Plus" references.
15194 2018-03-02 Jakub Jelinek <jakub@redhat.com>
15195 Richard Biener <rguenther@suse.de>
15198 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
15199 for error or warning attributes if CALL_FROM_THUNK_P is set.
15202 2018-03-02 Jakub Jelinek <jakub@redhat.com>
15205 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
15206 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
15209 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
15210 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
15212 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
15213 instead of -1U in last predictors element's probability member.
15215 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
15218 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
15219 arguments if they are comparable.
15221 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
15223 PR tree-optimization/84634
15224 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
15225 masks and masked_loop_p with a single loop_masks, making sure it's
15226 null for bb vectorization.
15228 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
15230 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
15231 (vect_analyze_data_ref_access): Use loop->safe_len rather than
15232 loop->force_vectorize to check whether there is no alias.
15234 2018-03-02 Jakub Jelinek <jakub@redhat.com>
15237 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
15239 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
15241 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
15242 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
15243 instead of a loop around prev_real_insn.
15244 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
15247 PR inline-asm/84625
15248 * config/i386/i386.c (ix86_print_operand): Use conditional
15249 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
15252 2018-03-02 Richard Biener <rguenther@suse.de>
15254 PR tree-optimization/84427
15255 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
15256 (bitmap_set_subtract_values): Rewrite to handle multiple
15259 (prune_clobbered_mems): Likewise.
15260 (phi_translate): Take edge instead of pred/phiblock.
15261 (phi_translate_1): Likewise.
15262 (phi_translate_set): Likewise. Insert all translated
15263 exprs for a value into the set, keeping possibly multiple
15264 expressions per value.
15265 (compute_antic_aux): Adjust for phi_translate changes.
15266 When intersecting union the expressions and prune those
15267 not in the final value set, keeping possibly multiple
15268 expressions per value. Do not use value-insertion
15269 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
15270 all expressions. Add verification that the value-sets
15271 only shrink during iteration.
15272 (compute_partial_antic_aux): Adjust for the phi_translate changes.
15273 (do_pre_regular_insertion): Likewise.
15274 (do_pre_partial_partial_insertion): Likewise.
15276 2018-03-02 Richard Biener <rguenther@suse.de>
15279 * config/darwin.c (saved_debug_info_level): New static global.
15280 (darwin_asm_lto_start): Disable debug info generation for LTO out.
15281 (darwin_asm_lto_end): Restore debug info generation settings.
15283 2018-03-01 Martin Liska <mliska@suse.cz>
15286 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
15287 volatile arguments.
15289 2018-03-01 Richard Biener <rguenther@suse.de>
15292 * dwarf2out.c (gen_variable_die): Properly handle late VLA
15293 type annotation with LTO when debug was disabled at compile-time.
15295 2018-03-01 Matthew Fortune <mfortune@gmail.com>
15297 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
15299 (mips_final_postscan_insn): Likewise.
15301 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
15303 PR rtl-optimization/84528
15304 * alias.c (init_alias_target): Add commentary.
15305 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
15306 a unique base value if the frame pointer is not eliminated
15307 to the stack pointer.
15309 2018-03-01 Tom de Vries <tom@codesourcery.com>
15311 PR rtl-optimization/83327
15312 * lra-int.h (hard_regs_spilled_into): Declare.
15313 * lra.c (hard_regs_spilled_into): Define.
15314 (init_reg_info): Init hard_regs_spilled_into.
15315 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
15316 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
15317 (process_bb_lives): Handle hard_regs_spilled_into.
15318 (lra_create_live_ranges_1): Before doing liveness propagation, clear
15319 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
15321 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
15323 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
15324 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
15325 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
15326 * config/rs6000/aix72.h: New file.
15328 2018-02-28 Jakub Jelinek <jakub@redhat.com>
15330 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
15331 instead of warning_at with conditional singular and plural messages
15335 * stor-layout.c (update_alignment_for_field): For
15336 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
15337 && !DECL_PACKED (field), do the alignment update, just use
15338 only desired_align instead of MAX (type_align, desired_align)
15340 (place_field): Don't do known_align < desired_align handling
15341 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
15342 is non-NULL, instead do it after rli->prev_field handling and
15343 only if not within a bitfield word. For DECL_PACKED (field)
15344 use type_align of BITS_PER_UNIT.
15346 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
15348 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
15349 superfluous parentheses and trailing spaces.
15351 2018-02-28 Richard Biener <rguenther@suse.de>
15353 PR tree-optimization/84584
15354 * graphite-scop-detection.c (scop_detection::add_scop): Discard
15355 SCoPs with fake exit edge.
15357 2018-02-28 Martin Liska <mliska@suse.cz>
15360 * timevar.c (timer::print): Fix format to properly print 100%
15363 2018-02-28 Richard Biener <rguenther@suse.de>
15365 PR middle-end/84607
15366 * genmatch.c (capture_info::walk_match): Do not mark
15367 captured expressions without operands as expr_p given
15368 they act more like predicates and should be subject to
15369 "lost tail" side-effect preserving.
15371 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
15373 PR rtl-optimization/81611
15374 * auto-inc-dec.c (attempt_change): Move dead note from
15375 mem_insn if it's the next use of regno
15376 (find_address): Take address use of reg holding
15377 non-incremented value. Add parm to limit search to the named
15379 (merge_in_block): Attempt to use a mem insn that is the next
15380 use of the original regno.
15382 2018-02-27 Martin Sebor <msebor@redhat.com>
15385 * doc/invoke.texi (-Wmissing-attributes): New option.
15386 * print-tree.c (print_node): Handle DECL_UNINLINABLE.
15388 2018-02-27 Martin Sebor <msebor@redhat.com>
15390 PR translation/84207
15391 * diagnostic-core.h (warning_n, error_n, inform_n): Change
15392 n argument to unsigned HOST_WIDE_INT.
15393 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
15394 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
15395 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
15396 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
15398 2018-02-27 Richard Biener <rguenther@suse.de>
15400 PR tree-optimization/84512
15401 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
15402 Do not use the estimate returned from record_stmt_cost for
15403 the scalar iteration cost but sum properly using add_stmt_cost.
15405 2018-02-27 Richard Biener <rguenther@suse.de>
15407 PR tree-optimization/84466
15408 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
15409 Adjust last change to less strictly validate use operands.
15411 2018-02-27 Martin Liska <mliska@suse.cz>
15413 PR gcov-profile/84548
15414 * gcov.c (process_file): Allow partial overlap and consider it
15415 also as group functions.
15416 (output_lines): Properly calculate range of lines for a group.
15418 2018-02-27 Martin Liska <mliska@suse.cz>
15420 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
15421 'ggc' suffixes. Change first column width.
15422 (timer::print): Fix formatting of the column.
15424 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
15426 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
15427 preserve inline entry blocks for the sake of debug inline
15428 entry point markers alone.
15429 (remove_unused_locals): Suggest in comments a better place to
15430 force the preservation of inline entry blocks that are
15431 otherwise unused, but do not preserve them.
15433 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
15435 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
15437 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
15440 * config/i386/constraints.md (Bs): Replace
15441 ix86_indirect_branch_register with
15442 TARGET_INDIRECT_BRANCH_REGISTER.
15444 * config/i386/i386.md (indirect_jump): Likewise.
15445 (tablejump): Likewise.
15446 (*sibcall_memory): Likewise.
15447 (*sibcall_value_memory): Likewise.
15448 Peepholes of indirect call and jump via memory: Likewise.
15449 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
15450 (*sibcall_value_GOT_32): Likewise.
15451 * config/i386/predicates.md (indirect_branch_operand): Likewise.
15452 (GOT_memory_operand): Likewise.
15453 (call_insn_operand): Likewise.
15454 (sibcall_insn_operand): Likewise.
15455 (GOT32_symbol_operand): Likewise.
15456 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
15458 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
15460 PR rtl-optimization/83496
15461 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
15462 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
15463 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
15464 redundant insn, if any.
15465 (relax_delay_slots): Likewise.
15466 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
15468 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
15470 PR tree-optimization/83965
15471 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
15472 that grouped statements are part of a reduction chain. Return
15473 true if the statement is not marked as a reduction itself but
15474 is part of a group.
15475 (vect_recog_dot_prod_pattern): Don't check whether the statement
15476 is part of a group here.
15477 (vect_recog_sad_pattern): Likewise.
15478 (vect_recog_widen_sum_pattern): Likewise.
15480 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
15483 * final.c (rest_of_clean_state): Also look for calls inside sequences.
15485 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
15488 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
15490 (ix86_output_indirect_function_return): New prototype.
15491 (ix86_split_simple_return_pop_internal): Likewise.
15492 * config/i386/i386.c (indirect_return_via_cx): New.
15493 (indirect_return_via_cx_bnd): Likewise.
15494 (indirect_thunk_name): Handle return va CX_REG.
15495 (output_indirect_thunk_function): Create alias for
15496 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
15497 (ix86_output_indirect_jmp): Remove the bool argument.
15498 (ix86_output_indirect_function_return): New function.
15499 (ix86_split_simple_return_pop_internal): Likewise.
15500 * config/i386/i386.md (*indirect_jump): Don't pass false
15501 to ix86_output_indirect_jmp.
15502 (*tablejump_1): Likewise.
15503 (simple_return_pop_internal): Change it to define_insn_and_split.
15504 Call ix86_split_simple_return_pop_internal to split it for
15505 -mfunction-return=.
15506 (simple_return_indirect_internal): Call
15507 ix86_output_indirect_function_return instead of
15508 ix86_output_indirect_jmp.
15510 2018-02-26 Jakub Jelinek <jakub@redhat.com>
15513 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
15514 memset and value initialization afterwards.
15516 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
15518 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
15520 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
15523 * common/config/aarch64/aarch64-common.c
15524 (aarch_option_optimization_table[]): Switch
15525 off fomit-frame-pointer
15527 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
15528 Chung-Ju Wu <jasonwucj@gmail.com>
15530 * config/nds32/nds32-multiple.md (load_multiple): Disallow
15532 (store_multiple): Ditto.
15534 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
15536 * config.gcc: Add --with-cpu support for nds32 target.
15537 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
15538 * config/nds32/nds32.opt: Add -mcpu= option.
15540 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
15542 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
15543 isel=yes): Warn for these deprecated options.
15545 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
15547 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
15548 ISA_2_5_MASKS_EMBEDDED.
15550 2018-02-23 Jakub Jelinek <jakub@redhat.com>
15552 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
15553 p->max as pointers rather than using iterative_hash_expr.
15555 2018-02-23 Carl Love <cel@us.ibm.com>
15557 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
15558 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
15560 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
15561 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
15562 P8V_BUILTIN_VEC_VUNSIGNED2.
15564 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
15567 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
15568 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
15570 * lra-constraints.c (curr_insn_transform): Set up
15571 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
15573 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
15574 * lra-eliminations.c (spill_pseudos): Ditto.
15575 (process_insn_for_elimination): Ditto.
15576 * lra-lives.c (reg_early_clobber_p): Use the new macros.
15577 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
15578 LRA_NON_CLOBBERED_ALT.
15580 2018-02-22 Martin Sebor <msebor@redhat.com>
15582 PR tree-optimization/84480
15583 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
15584 to maybe_diag_stxncpy_trunc. Call it.
15585 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
15586 from gimple_fold_builtin_strcpy. Print inlining stack.
15587 (handle_builtin_stxncpy): Print inlining stack.
15588 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
15590 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
15593 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
15594 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
15595 and -fcheck-pointer-bounds are used together.
15596 (indirect_thunk_prefix): New enum.
15597 (indirect_thunk_need_prefix): New function.
15598 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
15599 "_nt" instead of "_bnd" for NOTRACK prefix.
15600 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
15601 (output_indirect_thunk_function): Likewise.
15603 (ix86_code_end): Update output_indirect_thunk_function calls.
15604 (ix86_output_indirect_branch_via_reg): Replace
15605 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
15606 (ix86_output_indirect_branch_via_push): Likewise.
15607 (ix86_output_function_return): Likewise.
15608 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
15609 incompatible with -fcf-protection=branch and
15610 -fcheck-pointer-bounds.
15612 2018-02-22 Steve Ellcey <sellcey@cavium.com>
15615 * config/aarch64/aarch64.c (aarch64_print_address_internal):
15616 Change gcc_assert call to output_operand_lossage.
15618 2018-02-22 Steve Ellcey <sellcey@cavium.com>
15620 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
15622 2018-02-22 DJ Delorie <dj@redhat.com>
15623 Sebastian Perta <sebastian.perta@renesas.com>
15624 Oleg Endo <olegendo@gcc.gnu.org>
15626 * config/rx/rx.c (rx_rtx_costs): New function.
15627 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
15629 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
15631 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
15633 2018-02-22 Martin Liska <mliska@suse.cz>
15636 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
15637 Add "native" as a possible value.
15639 2018-02-22 Martin Liska <mliska@suse.cz>
15642 * config/i386/i386.c (ix86_option_override_internal):
15643 Add "native" as a possible value for -march and -mtune.
15645 2018-02-22 Jakub Jelinek <jakub@redhat.com>
15648 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
15649 to all type variants.
15651 PR tree-optimization/84503
15652 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
15653 width as info->bitpos + info->bitsize - start.
15654 (merged_store_group::merge_overlapping): Simplify width computation.
15655 (check_no_overlap): New function.
15656 (imm_store_chain_info::try_coalesce_bswap): Compute expected
15657 start + width and last_order of the group, fail if check_no_overlap
15659 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
15660 to group if check_no_overlap fails.
15662 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
15664 * config/rs6000/altivec.md: Delete contraint arguments to
15665 define_expand, define_split, and define_peephole2, and in
15666 define_insn_and_split if always unused.
15667 * config/rs6000/darwin.md: Ditto.
15668 * config/rs6000/dfp.md: Ditto.
15669 * config/rs6000/rs6000.md: Ditto.
15670 * config/rs6000/sync.md: Ditto.
15671 * config/rs6000/vector.md: Ditto.
15672 * config/rs6000/vsx.md: Ditto.
15674 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
15676 * config/rs6000/altivec.md: Write output control strings as braced
15677 blocks instead of double-quoted strings.
15678 * config/rs6000/darwin.md: Ditto.
15679 * config/rs6000/rs6000.md: Ditto.
15680 * config/rs6000/vector.md: Ditto.
15681 * config/rs6000/vsx.md: Ditto.
15683 2018-02-21 Jason Merrill <jason@redhat.com>
15685 PR c++/84314 - ICE with templates and fastcall attribute.
15686 * attribs.c (build_type_attribute_qual_variant): Remove assert.
15688 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
15690 * ipa-cp.c (determine_versionability): Fix comment typos.
15692 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
15695 * ipa-cp.c (determine_versionability): Do not version functions caling
15698 2018-02-21 Martin Liska <mliska@suse.cz>
15701 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
15702 Add "native" as a possible value.
15703 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
15704 the macro when native cpu detection is available.
15706 2018-02-21 Martin Liska <mliska@suse.cz>
15709 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
15710 Add "native" as a possible value.
15711 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
15712 when native cpu detection is available.
15714 2018-02-21 Jakub Jelinek <jakub@redhat.com>
15715 Martin Sebor <msebor@redhat.com>
15717 PR tree-optimization/84478
15718 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
15720 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
15721 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
15722 support which is conservatively correct, for 2 only stay conservative
15723 for maxlen. Formatting and comment capitalization fixes. Add STRICT
15724 argument to the 2 argument get_range_strlen, adjust 6 arg
15725 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
15727 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
15728 (gimple_fold_builtin_strlen): Pass true as last argument to
15731 2018-02-20 Martin Sebor <msebor@redhat.com>
15733 PR middle-end/84095
15734 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
15735 (builtin_memref::set_base_and_offset): Same. Handle inner references.
15736 (builtin_memref::builtin_memref): Factor out parts into
15737 set_base_and_offset and call it.
15739 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
15741 PR middle-end/84406
15742 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
15743 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
15744 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
15745 search at the associated MODE_INT.
15747 2018-02-20 Jeff Law <law@redhat.com>
15749 PR middle-end/82123
15750 PR tree-optimization/81592
15751 PR middle-end/79257
15752 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
15753 for range data rather than using global data.
15754 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
15755 range data rather than using global data.
15756 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
15757 pass it to children as needed.
15758 (struct directive::fmtresult): Similarly.
15759 (struct directive::set_width): Similarly.
15760 (struct directive::set_precision): Similarly.
15761 (format_integer, format_directive, parse_directive): Similarly.
15762 (format_none): Accept unnamed vr_values parameter.
15763 (format_percent, format_floating, format_character): Similarly.
15764 (format_string, format_plain): Similarly.
15765 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
15766 the EVRP range analyzer for range data rather than using global data.
15767 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
15768 gimple-ssa-evrp-analyze.h
15769 (class sprintf_dom_walker): Add after_dom_children member function.
15770 Add evrp_range_analyzer member.
15771 (sprintf_dom_walker::before_dom_children): Call into the EVRP
15772 range analyzer as needed.
15773 (sprintf_dom_walker::after_dom_children): New member function.
15774 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
15776 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
15777 (evrp_range_analyzer::pop_to_marker): Likewise.
15779 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
15781 PR tree-optimization/84419
15782 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
15783 with the required type if its current type is compatible but
15786 2018-02-20 Jakub Jelinek <jakub@redhat.com>
15788 PR middle-end/82004
15789 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
15790 after vectorization.
15792 2018-02-20 Martin Liska <mliska@suse.cz>
15795 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
15796 possible values if we don't have a hint.
15798 2018-02-20 Martin Liska <mliska@suse.cz>
15802 * final.c (shorten_branches): Build align_tab array with one
15804 * opts.c (finish_options): Add alignment option limit check.
15805 (MAX_CODE_ALIGN): Likewise.
15806 (MAX_CODE_ALIGN_VALUE): Likewise.
15807 * doc/invoke.texi: Document maximum allowed option value for
15808 all -falign-* options.
15810 2018-02-19 Jakub Jelinek <jakub@redhat.com>
15813 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
15814 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
15815 * var-tracking.c (emit_note_insn_var_location): Remove all references
15816 to NOTE_INSN_CALL_ARG_LOCATION.
15817 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
15818 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
15819 Use copy_rtx_if_shared.
15820 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
15821 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
15822 (dwarf2out_var_location): Remove handling of
15823 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
15825 * final.c (final_scan_insn): Remove all references to
15826 NOTE_INSN_CALL_ARG_LOCATION.
15827 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
15828 before dumping final insns.
15829 * except.c (emit_note_eh_region_end): Remove all references to
15830 NOTE_INSN_CALL_ARG_LOCATION.
15831 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
15832 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
15833 * config/arc/arc.c (hwloop_optimize): Likewise.
15834 * config/arm/arm.c (create_fix_barrier): Likewise.
15835 * config/s390/s390.c (s390_chunkify_start): Likewise.
15836 * config/sh/sh.c (find_barrier): Likewise.
15837 * config/i386/i386.c (rest_of_insert_endbranch,
15838 ix86_seh_fixup_eh_fallthru): Likewise.
15839 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
15840 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
15841 * config/frv/frv.c (frv_function_prologue): Likewise.
15842 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
15844 (note_outside_basic_block_p): Remove all references to
15845 NOTE_INSN_CALL_ARG_LOCATION.
15846 * gengtype.c (adjust_field_rtx_def): Likewise.
15847 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
15849 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
15850 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
15853 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
15856 PR tree-optimization/84452
15857 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
15858 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
15861 2018-02-19 Martin Liska <mliska@suse.cz>
15864 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
15866 2018-02-19 Martin Liska <mliska@suse.cz>
15867 Richard Sandiford <richard.sandiford@linaro.org>
15869 PR tree-optimization/82491
15870 * gimple-fold.c (get_base_constructor): Make earlier bail out
15873 2018-02-19 Carl Love <cel@us.ibm.com>
15875 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
15876 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
15878 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
15879 P8V_BUILTIN_VEC_NEG.
15881 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
15883 * config/rl78/rl78.md (movdf): New define expand.
15885 2018-02-19 Martin Liska <mliska@suse.cz>
15888 * doc/invoke.texi: Fix typo.
15889 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
15891 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
15893 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
15894 handle rs6000_single_float and rs6000_double_float specially for
15897 2018-02-16 Jeff Law <law@redhat.com>
15899 * config/rx/rx.c (add_pop_cfi_notes): New function.;
15900 (pop_regs): Use it.
15902 2018-02-16 Jakub Jelinek <jakub@redhat.com>
15905 * ipa-inline.c (inline_small_functions): Fix a typo.
15907 2018-02-16 Nathan Sidwell <nathan@acm.org>
15909 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
15911 2018-02-16 Carl Love <cel@us.ibm.com>
15913 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
15914 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
15915 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
15916 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
15917 expansion to P8V_BUILTIN_VEC_FLOAT2.
15919 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
15921 PR rtl-optimization/70023
15922 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
15923 src_regno into account.
15925 2018-02-16 Carl Love <cel@us.ibm.com>
15927 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
15928 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
15929 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
15930 * config/rs6000/rs6000.c: Remove case statements for
15931 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
15932 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
15933 and P9V_BUILTIN_VEC_VINSERT4B.
15934 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
15935 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
15936 * config/rs6000/vsx.md:
15937 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
15940 2018-02-16 Carl Love <cel@us.ibm.com>
15942 * config/rs6000/altivec.h: Add builtin names vec_extract4b
15944 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
15946 * config/rs6000/rs6000-c.c: Add the definitions for
15947 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
15948 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
15949 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
15950 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
15951 definition for insert4b and define insn *insert3b_internal.
15952 * doc/extend.texi: Add documentation for vec_extract4b.
15954 2018-02-16 Nathan Sidwell <nathan@acm.org>
15956 * doc/extend.texi (Backwards Compatibility): Mention friend
15957 injection. Note for-scope is deprecated.
15958 * doc/invoke.texi (-ffriend-injection): Deprecate.
15960 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
15962 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
15963 that moved to I2, also allow destinations that are a paradoxical
15964 subreg (instead of a normal reg).
15966 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
15969 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
15972 2018-02-16 Richard Biener <rguenther@suse.de>
15974 PR tree-optimization/84037
15975 PR tree-optimization/84016
15977 * config/i386/i386.c (ix86_builtin_vectorization_cost):
15978 Adjust vec_construct for the fact we need additional higher latency
15979 128bit inserts for AVX256 and AVX512 vector builds.
15980 (ix86_add_stmt_cost): Scale vector construction cost for
15983 2018-02-16 Richard Biener <rguenther@suse.de>
15985 PR tree-optimization/84417
15986 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
15987 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
15988 (non_rewritable_lvalue_p): Likewise, use poly-ints.
15990 2018-02-16 Martin Liska <mliska@suse.cz>
15993 * internal-fn.def (ASAN_CHECK): Set proper flags.
15994 (ASAN_MARK): Likewise.
15996 2018-02-16 Julia Koval <julia.koval@intel.com>
15998 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
15999 from PTA_CANNONLAKE.
16001 2018-02-16 Jakub Jelinek <jakub@redhat.com>
16004 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
16005 Use ++iter rather than iter++ for std::list iterators.
16006 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
16007 defer deleting them until all nodes in the forest are processed. Do
16008 free even leaf nodes. Change to_process into auto_vec.
16011 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
16012 * vec.h (vec_default_construct): Use memset instead of placement new
16013 if BROKEN_VALUE_INITIALIZATION is defined.
16014 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
16015 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
16018 PR rtl-optimization/83723
16019 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
16020 * lra.c (lra_substitute_pseudo): Likewise. If true, use
16021 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
16023 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
16025 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
16027 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
16029 PR rtl-optimization/81443
16030 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
16031 from inner REGs to paradoxical SUBREGs.
16033 2018-02-16 Richard Biener <rguenther@suse.de>
16035 PR tree-optimization/84399
16036 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
16037 For operands we can analyze at their definition make sure we can
16038 analyze them at each use as well.
16040 2018-02-16 Richard Biener <rguenther@suse.de>
16042 PR tree-optimization/84190
16043 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
16044 volatile accesses if the decl isn't volatile.
16046 2018-02-15 Jason Merrill <jason@redhat.com>
16048 PR c++/84314 - ICE with templates and fastcall attribute.
16049 * attribs.c (build_type_attribute_qual_variant): Don't clobber
16050 TYPE_CANONICAL on an existing type.
16052 2018-02-15 Jakub Jelinek <jakub@redhat.com>
16054 PR tree-optimization/84383
16055 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
16056 dstoff nor call operand_equal_p if dstbase is NULL.
16058 PR tree-optimization/84334
16059 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
16060 also a CONSTANT_CLASS_P, punt.
16062 2018-02-14 Jim Wilson <jimw@sifive.com>
16064 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
16065 first SMALL_OPERAND check. New local min_second_step. Move assert
16066 to where locals are set. Add TARGET_RVC support.
16067 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
16069 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
16071 * doc/invoke.texi: Correct -Wformat-overflow code sample.
16073 2018-02-14 Martin Sebor <msebor@redhat.com>
16075 PR tree-optimization/83698
16076 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
16077 arrays constrain the offset range to their bounds.
16078 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
16079 (builtin_access::overlap): Avoid setting the size of overlap if it's
16081 (maybe_diag_overlap): Also consider arrays when deciding what values
16082 of offsets to include in diagnostics.
16084 2018-02-14 Martin Sebor <msebor@redhat.com>
16087 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
16088 that correspond to the kind of a declaration.
16090 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
16093 * config/pa/pa.md: Load address of PIC label using the linkage table
16094 if the label is nonlocal.
16096 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
16098 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
16099 warning message if user requests -maltivec=be.
16100 * doc/invoke.texi: Document deprecation of -maltivec=be.
16102 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
16105 * config/rs6000/rs6000-c.c: Update definitions for
16106 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
16107 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
16109 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
16112 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
16113 add _get_ssp intrinsics. Remove argument from
16114 __builtin_ia32_rdssp[d|q].
16115 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
16116 * config/i386/i386-builtin.def: Remove argument from
16117 __builtin_ia32_rdssp[d|q].
16118 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
16119 ix86_expand_special_args_builtin for _rdssp[d|q].
16120 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
16121 Clear register before usage.
16122 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
16123 Add documentation for new _get_ssp and _inc_ssp intrinsics.
16125 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
16127 PR tree-optimization/84357
16128 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
16129 operand 1 of an ARRAY_REF too.
16131 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
16134 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
16135 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
16137 (set_of_reg): New struct.
16138 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
16139 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
16140 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
16142 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
16143 Split into bitclr, bitset, bitinvert patterns if appropriate.
16144 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
16145 use rx_fuse_in_memory_bitop.
16146 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
16147 to named insn, correct maximum insn length.
16149 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
16152 * machmode.def: Define a complex mode for PARTIAL_INT.
16153 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
16155 * doc/rtl.texi: Document CSPImode.
16156 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
16158 (msp430_hard_regno_nregs_with_padding): Likewise.
16160 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
16163 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
16165 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
16167 PR rtl-optimization/84169
16168 * combine.c (try_combine): New variable split_i2i3. Set it to true if
16169 we generated a parallel as new i3 and we split that to new i2 and i3
16170 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
16171 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
16172 those to i2, not i1. Partially rewrite this scan code.
16174 2018-02-13 Jakub Jelinek <jakub@redhat.com>
16177 * stor-layout.c (place_field): For variable length fields, adjust
16178 offset_align afterwards not just based on the field's alignment,
16179 but also on the size.
16181 PR middle-end/84309
16182 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
16183 of exps and logs in the use_exp2 case.
16185 2018-02-13 Jeff Law <law@redhat.com>
16187 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
16188 entry for "vector".
16190 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
16193 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
16197 * common.opt (gas-loc-support, gas-locview-support): New.
16198 (ginline-points, ginternal-reset-location-views): New.
16199 * doc/invoke.texi: Document them. Use @itemx where intended.
16200 (gvariable-location-views): Adjust.
16201 * target.def (reset_location_view): New.
16202 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
16203 (TARGET_RESET_LOCATION_VIEW): New.
16204 * doc/tm.texi: Rebuilt.
16205 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
16206 (dwarf2out_default_as_locview_support): New.
16207 (output_asm_line_debug_info): Use option variables.
16208 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
16209 (output_loc_list): Likewise.
16210 (add_high_low_attributes): Check option variables.
16211 Don't output entry view attribute in strict mode.
16212 (gen_inlined_subroutine_die): Check option variables.
16213 (dwarf2out_inline_entry): Likewise.
16214 (init_sections_and_labels): Likewise.
16215 (dwarf2out_early_finish): Likewise.
16216 (maybe_reset_location_view): New, from...
16217 (dwarf2out_var_location): ... here. Call it.
16218 * debug.h (dwarf2out_default_as_loc_support): Declare.
16219 (dwarf2out_default_as_locview_support): Declare.
16220 * hooks.c (hook_int_rtx_insn_0): New.
16221 * hooks.h (hook_int_rtx_insn_0): Declare.
16222 * toplev.c (process_options): Take -gas-loc-support and
16223 -gas-locview-support from dwarf2out. Enable
16224 -gvariable-location-views by default only with locview
16225 assembler support. Enable -ginternal-reset-location-views by
16226 default only if the target defines the corresponding hook.
16227 Enable -ginline-points by default if location views are
16228 enabled; force it disabled if statement frontiers are
16230 * tree-inline.c (expand_call_inline): Check option variables.
16231 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
16233 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
16235 PR tree-optimization/84321
16236 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
16237 handling. Also check whether the anti-range contains any values
16238 that satisfy the mask; switch to a VR_RANGE if not.
16240 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
16243 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
16245 2018-02-13 Martin Jambor <mjambor@suse.cz>
16248 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
16249 of call statements, also set location of a load to a temporary.
16251 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
16253 * config/rl78/rl78.c (add_vector_labels): New function.
16254 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
16255 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
16256 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
16257 which checks that no arguments are passed.
16258 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
16259 * doc/extend.texi: Documentation for the new attribute.
16261 2018-02-13 Andreas Schwab <schwab@suse.de>
16263 * config/riscv/linux.h (CPP_SPEC): Define.
16265 2018-02-13 Jakub Jelinek <jakub@redhat.com>
16268 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
16269 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
16270 OPTION_MASK_ISA_AES as first argument to def_builtin_const
16271 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
16272 instead of OPTION_MASK_ISA_PCLMUL as first argument to
16273 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
16274 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
16275 temporarily for AES and PCLMUL builtins.
16277 PR tree-optimization/84339
16278 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
16279 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
16282 PR middle-end/84309
16283 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
16284 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
16285 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
16287 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
16289 * omp-simd-clone.h: New file.
16290 * omp-simd-clone.c: Include omp-simd-clone.h.
16291 (expand_simd_clones): No longer static.
16292 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
16293 cgraph.h and omp-simd-clone.h.
16294 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
16295 (vect_recog_widen_shift_pattern): Formatting fix.
16296 (vect_pattern_recog_1): Don't check optab for calls.
16299 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
16300 operands[2] into a REG before using gen_lowpart on it.
16302 2018-02-12 Jeff Law <law@redhat.com>
16305 * config/sh/sh.c (find_barrier): Consider a sibling call
16308 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
16309 successfully back substituting a reg.
16311 2018-02-12 Richard Biener <rguenther@suse.de>
16313 PR tree-optimization/84037
16314 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
16315 parameter, move visited init to caller.
16316 (vect_slp_analyze_operations): Separate cost from validity
16317 check, initialize visited once for all instances.
16318 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
16320 * tree-vect-stmts.c (vect_model_simple_cost): Make early
16322 (vect_model_promotion_demotion_cost): Likewise.
16323 (vectorizable_bswap): Guard cost modeling with !slp_node
16324 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
16326 (vectorizable_call): Likewise.
16327 (vectorizable_conversion): Likewise.
16328 (vectorizable_assignment): Likewise.
16329 (vectorizable_shift): Likewise.
16330 (vectorizable_operation): Likewise.
16331 (vectorizable_store): Likewise.
16332 (vectorizable_load): Likewise.
16333 (vectorizable_condition): Likewise.
16334 (vectorizable_comparison): Likewise.
16336 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
16339 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
16340 (ASAN_MARK): Fix fnspec to account for return value, change pointer
16341 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
16343 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
16345 PR middle-end/83665
16346 * params.def (inline-min-speedup): Increase from 8 to 15.
16347 (max-inline-insns-auto): Decrease from 40 to 30.
16348 * ipa-split.c (consider_split): Add some buffer for function to
16349 be considered inlining candidate.
16350 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
16353 2018-02-12 Richard Biener <rguenther@suse.de>
16355 PR tree-optimization/84037
16356 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
16357 matched stmts if we cannot swap the non-matched ones.
16359 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
16361 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
16362 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
16363 _mm_maskz_scalef_round_ss): New intrinsics.
16364 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
16365 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
16366 __builtin_ia32_scalefss_round): Remove.
16367 (__builtin_ia32_scalefsd_mask_round,
16368 __builtin_ia32_scalefss_mask_round): New intrinsics.
16369 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
16370 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
16371 ((match_operand:VF_128 2 "<round_nimm_predicate>"
16372 "<round_constraint>")): Changed to ...
16373 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
16374 "<round_scalar_constraint>")): ... this.
16375 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
16376 %0, %1, %2<round_op3>}"): Changed to ...
16377 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
16378 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
16379 %2<round_scalar_mask_op3>}"): ... this.
16380 * config/i386/subst.md (round_scalar_nimm_predicate): New.
16382 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
16384 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
16385 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
16386 (_mm_maskz_sqrt_round_ss): New intrinsics.
16387 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
16388 (__builtin_ia32_sqrtsd_mask_round)
16389 (__builtin_ia32_sqrtss_mask_round): New builtins.
16390 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
16391 (__builtin_ia32_sqrtss_round): Remove.
16392 (__builtin_ia32_sqrtsd_mask_round)
16393 (__builtin_ia32_sqrtss_mask_round): New builtins.
16394 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
16395 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
16396 ((match_operand:VF_128 1 "vector_operand"
16397 "xBm,<round_constraint>")): Changed to ...
16398 ((match_operand:VF_128 1 "vector_operand"
16399 "xBm,<round_scalar_constraint>")): ... this.
16400 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
16401 %0, %2, %<iptr>1<round_op3>}): Changed to ...
16402 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
16403 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
16404 %<iptr>1<round_scalar_mask_op3>}): ... this.
16405 ((set_attr "prefix" "<round_prefix>")): Changed to ...
16406 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
16408 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
16411 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
16412 Cast vec_cmpeq result to correct type.
16413 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
16414 Cast vec_cmpgt result to correct type.
16416 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
16418 * final.c (final_scan_insn_1): Renamed from...
16419 (final_scan_insn): ... this. New wrapper, to recover
16420 seen from the outermost call in recursive ones.
16421 * config/sparc/sparc.c (output_return): Drop seen from call.
16422 (output_sibcall): Likewise.
16423 * config/visium/visium.c (output_branch): Likewise.
16425 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
16427 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
16430 2018-02-10 Alan Modra <amodra@gmail.com>
16433 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
16434 Specify LR as an input.
16436 2018-02-10 Jakub Jelinek <jakub@redhat.com>
16439 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
16440 remove_member_access_dummy_vars): New functions.
16441 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
16442 lower_omp_1, execute_lower_omp): Use them.
16444 PR rtl-optimization/84308
16445 * shrink-wrap.c (spread_components): Release todo vector.
16447 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
16449 PR rtl-optimization/57193
16450 * ira-color.c (struct allocno_color_data): Add member
16451 conflict_allocno_hard_prefs.
16452 (update_conflict_allocno_hard_prefs): New.
16453 (bucket_allocno_compare_func): Add a preference based on
16454 conflict_allocno_hard_prefs.
16455 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
16456 (color_allocnos): Remove a dead code. Initiate
16457 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
16459 2018-02-09 Jakub Jelinek <jakub@redhat.com>
16462 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
16463 constraint from =wa to wa. Avoid a subreg on the output operand,
16464 instead use a pseudo and subreg it in a move.
16465 (p9_xxbrd_<mode>): Changed to ...
16466 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
16467 (p9_xxbrd_v2df): New expander.
16468 (p9_xxbrw_<mode>): Changed to ...
16469 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
16470 (p9_xxbrw_v4sf): New expander.
16472 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
16474 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
16476 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
16479 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
16480 multiply in 32-bit mode.
16481 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
16482 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
16485 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
16487 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
16488 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
16489 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
16490 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
16492 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
16495 * dwarf2out.c (is_trivial_indirect_ref): New function.
16496 (dwarf2out_late_global_decl): Do not generate a location
16497 attribute for variables that have a non-trivial DECL_VALUE_EXPR
16498 and that are not defined in the current unit.
16500 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
16502 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
16503 instead of a libcall for UNORDERED.
16505 2018-02-09 Tamar Christina <tamar.christina@arm.com>
16508 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
16509 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
16511 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
16514 * config/s390/s390.c (s390_set_current_function): Invoke
16515 s390_indirect_branch_settings also if fndecl didn't change.
16517 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
16519 * config/rs6000/rs6000.md (blockage): Set length to zero.
16521 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
16523 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
16525 2018-02-09 Jakub Jelinek <jakub@redhat.com>
16528 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
16529 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
16533 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
16534 PARALLEL incoming that failed vt_get_decl_and_offset check.
16536 PR middle-end/84237
16537 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
16538 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
16540 (get_variable_section): For decls in named .bss* sections pass true as
16541 second argument to bss_initializer_p.
16543 2018-02-09 Marek Polacek <polacek@redhat.com>
16544 Jakub Jelinek <jakub@redhat.com>
16547 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
16548 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
16549 Sync some changes from cxx_fold_indirect_ref.
16551 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
16553 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
16555 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
16556 (BLOCK_INLINE_ENTRY_LABEL): New.
16557 (dwarf2out_var_location): Disregard inline entry markers.
16558 (inline_entry_data): New struct.
16559 (inline_entry_data_hasher): New hashtable type.
16560 (inline_entry_data_hasher::hash): New.
16561 (inline_entry_data_hasher::equal): New.
16562 (inline_entry_data_table): New variable.
16563 (add_high_low_attributes): Add DW_AT_entry_pc and
16564 DW_AT_GNU_entry_view attributes if a pending entry is found
16565 in inline_entry_data_table. Add old entry_pc attribute only
16566 if debug nonbinding markers are disabled.
16567 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
16568 markers are enabled.
16569 (block_within_block_p, dwarf2out_inline_entry): New.
16570 (dwarf2out_finish): Check that no entries remained in
16571 inline_entry_data_table.
16572 * final.c (reemit_insn_block_notes): Handle inline entry notes.
16573 (final_scan_insn, notice_source_line): Likewise.
16574 (rest_of_clean_state): Skip inline entry markers.
16575 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
16577 * gimple.c (gimple_build_debug_inline_entry): New.
16578 * gimple.h (enum gimple_debug_subcode): Add
16579 GIMPLE_DEBUG_INLINE_ENTRY.
16580 (gimple_build_debug_inline_entry): Declare.
16581 (gimple_debug_inline_entry_p): New.
16582 (gimple_debug_nonbind_marker_p): Adjust.
16583 * insn-notes.def (INLINE_ENTRY): New.
16584 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
16585 inline entry marker notes.
16586 (print_insn): Likewise.
16587 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
16588 (INSN_DEBUG_MARKER_KIND): Likewise.
16589 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
16590 * tree-inline.c (expand_call_inline): Build and insert
16591 debug_inline_entry stmt.
16592 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
16593 inline entry blocks early, if nonbind markers are enabled.
16594 (dump_scope_block): Dump fragment info.
16595 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
16596 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
16597 (gimple_build_debug_inline_entry): New.
16598 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
16599 Enable/disable inline entry points too.
16600 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
16601 (DEBUG_INSN): Describe inline entry markers.
16603 * common.opt (gvariable-location-views): New.
16604 (gvariable-location-views=incompat5): New.
16605 * config.in: Rebuilt.
16606 * configure: Rebuilt.
16607 * configure.ac: Test assembler for view support.
16608 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
16609 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
16610 * dwarf2out.c (var_loc_view): New typedef.
16611 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
16612 (dwarf2out_locviews_in_attribute): New.
16613 (dwarf2out_locviews_in_loclist): New.
16614 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
16615 (enum dw_line_info_opcode): Add LI_adv_address.
16616 (struct dw_line_info_table): Add view.
16617 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
16618 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
16619 (zero_view_p): New variable.
16620 (ZERO_VIEW_P): New macro.
16621 (output_asm_line_debug_info): New.
16622 (struct var_loc_node): Add view.
16623 (add_AT_view_list, AT_loc_list): New.
16624 (add_var_loc_to_decl): Add view param. Test it against last.
16625 (new_loc_list): Add view params. Record them.
16626 (AT_loc_list_ptr): Handle loc and view lists.
16627 (view_list_to_loc_list_val_node): New.
16628 (print_dw_val): Handle dw_val_class_view_list.
16629 (size_of_die): Likewise.
16630 (value_format): Likewise.
16631 (loc_list_has_views): New.
16632 (gen_llsym): Set vl_symbol too.
16633 (maybe_gen_llsym, skip_loc_list_entry): New.
16634 (dwarf2out_maybe_output_loclist_view_pair): New.
16635 (output_loc_list): Output view list or entries too.
16636 (output_view_list_offset): New.
16637 (output_die): Handle dw_val_class_view_list.
16638 (output_dwarf_version): New.
16639 (output_compilation_unit_header): Use it.
16640 (output_skeleton_debug_sections): Likewise.
16641 (output_rnglists, output_line_info): Likewise.
16642 (output_pubnames, output_aranges): Update version comments.
16643 (output_one_line_info_table): Output view numbers in asm comments.
16644 (dw_loc_list): Determine current endview, pass it to new_loc_list.
16645 Call maybe_gen_llsym.
16646 (loc_list_from_tree_1): Adjust.
16647 (add_AT_location_description): Create view list attribute if
16648 needed, check it's absent otherwise.
16649 (convert_cfa_to_fb_loc_list): Adjust.
16650 (maybe_emit_file): Call output_asm_line_debug_info for test.
16651 (dwarf2out_var_location): Reset views as needed. Precompute
16652 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
16653 attribute. Set view.
16654 (new_line_info_table): Reset next view.
16655 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
16656 (dwarf2out_source_line): Likewise. Output view resets and labels to
16657 the assembler, or select appropriate line info opcodes.
16658 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
16659 (optimize_string_length): Catch it. Adjust.
16660 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
16661 dw_val_class_view_list, and remove it if no longer needed.
16662 (hash_loc_list): Hash view numbers.
16663 (loc_list_hasher::equal): Compare them.
16664 (optimize_location_lists): Check whether a view list symbol is
16665 needed, and whether the locview attribute is present, and
16666 whether they match. Remove the locview attribute if no longer
16668 (index_location_lists): Call skip_loc_list_entry for test.
16669 (dwarf2out_finish): Call output_asm_line_debug_info for test.
16670 Use output_dwarf_version.
16671 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
16672 (struct dw_val_node): Add val_view_list.
16673 * final.c (SEEN_NEXT_VIEW): New.
16674 (set_next_view_needed): New.
16675 (clear_next_view_needed): New.
16676 (maybe_output_next_view): New.
16677 (final_start_function): Rename to...
16678 (final_start_function_1): ... this. Take pointer to FIRST,
16679 add SEEN parameter. Emit param bindings in the initial view.
16680 (final_start_function): Reintroduce SEEN-less interface.
16681 (final): Rename to...
16682 (final_1): ... this. Take SEEN parameter. Output final pending
16683 next view at the end.
16684 (final): Reintroduce seen-less interface.
16685 (final_scan_insn): Output pending next view before switching
16686 sections or ending a block. Mark the next view as needed when
16687 outputting variable locations. Notify debug backend of section
16688 changes, and of location view changes.
16689 (rest_of_handle_final): Adjust.
16690 * toplev.c (process_options): Autodetect value for debug variable
16691 location views option. Warn on incompat5 without -gdwarf-5.
16692 * doc/invoke.texi (gvariable-location-views): New.
16693 (gvariable-location-views=incompat5): New.
16694 (gno-variable-location-views): New.
16696 2018-02-08 David Malcolm <dmalcolm@redhat.com>
16698 PR tree-optimization/84136
16699 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
16700 that the result of find_edge is non-NULL.
16702 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
16705 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
16706 storing integer register in SImode. Fix cost of 256 and 512
16707 byte aligned SSE register store.
16709 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
16711 * config/i386/i386.c (ix86_multiplication_cost): Fix
16712 multiplication cost for TARGET_AVX512DQ.
16714 2018-02-08 Marek Polacek <polacek@redhat.com>
16716 PR tree-optimization/84238
16717 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
16720 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
16722 PR tree-optimization/84265
16723 * tree-vect-stmts.c (vectorizable_store): Don't treat
16724 VMAT_CONTIGUOUS accesses as grouped.
16725 (vectorizable_load): Likewise.
16727 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
16729 PR tree-optimization/81635
16730 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
16731 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
16732 (test_round_for_mask): New functions.
16733 (wide_int_cc_tests): Call test_round_for_mask.
16734 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
16735 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
16736 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
16737 range returned by get_range_info.
16739 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
16742 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
16743 * symtab.c: Include builtins.h
16744 (symtab_node::output_to_lto_symbol_table_p): Move here
16745 from lto-streamer-out.c:output_symbol_p.
16746 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
16747 (output_symbol_p): Move all logic to symtab.c
16748 (produce_symtab): Update.
16750 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
16752 * config/s390/s390-opts.h (enum indirect_branch): Define.
16753 * config/s390/s390-protos.h (s390_return_addr_from_memory)
16754 (s390_indirect_branch_via_thunk)
16755 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
16756 (enum s390_indirect_branch_type): Define.
16757 * config/s390/s390.c (struct s390_frame_layout, struct
16758 machine_function): Remove.
16759 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
16760 (indirect_branch_table_label_no, indirect_branch_table_name):
16762 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
16763 (enum s390_indirect_branch_option): Define.
16764 (s390_return_addr_from_memory): New function.
16765 (s390_handle_string_attribute): New function.
16766 (s390_attribute_table): Add new attribute handler.
16767 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
16768 (s390_indirect_branch_via_thunk): New function.
16769 (s390_indirect_branch_via_inline_thunk): New function.
16770 (s390_function_ok_for_sibcall): When jumping via thunk disallow
16771 sibling call optimization for non z10 compiles.
16772 (s390_emit_call): Force indirect branch target to be a single
16773 register. Add r1 clobber for non-z10 compiles.
16774 (s390_emit_epilogue): Emit return jump via return_use expander.
16775 (s390_reorg): Handle JUMP_INSNs as execute targets.
16776 (s390_option_override_internal): Perform validity checks for the
16777 new command line options.
16778 (s390_indirect_branch_attrvalue): New function.
16779 (s390_indirect_branch_settings): New function.
16780 (s390_set_current_function): Invoke s390_indirect_branch_settings.
16781 (s390_output_indirect_thunk_function): New function.
16782 (s390_code_end): Implement target hook.
16783 (s390_case_values_threshold): Implement target hook.
16784 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
16786 * config/s390/s390.h (struct s390_frame_layout)
16787 (struct machine_function): Move here from s390.c.
16788 (TARGET_INDIRECT_BRANCH_NOBP_RET)
16789 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
16790 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
16791 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
16792 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
16793 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
16794 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
16795 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
16796 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
16797 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
16798 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
16799 (mnemonic attribute): Add values which aren't recognized
16801 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
16802 pattern for branch conversion. Fix mnemonic attribute.
16803 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
16804 indirect branch via thunk if requested.
16805 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
16806 ("*indirect_jump"): Disable for branch conversion using out of
16808 ("indirect_jump_via_thunk<mode>_z10")
16809 ("indirect_jump_via_thunk<mode>")
16810 ("indirect_jump_via_inlinethunk<mode>_z10")
16811 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
16812 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
16813 ("casesi_jump_via_inlinethunk<mode>_z10")
16814 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
16815 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
16816 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
16817 ("*indirect2_jump"): Disable for branch conversion.
16818 ("casesi_jump"): Turn into expander and expand patterns for branch
16820 ("return_use"): New expander.
16821 ("*return"): Emit return via thunk and rename it to ...
16822 ("*return<mode>"): ... this one.
16823 * config/s390/s390.opt: Add new options and and enum for the
16826 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
16828 * lra-constraints.c (match_reload): Unconditionally use
16829 gen_lowpart_SUBREG, rather than selecting between that
16830 and equivalent gen_rtx_SUBREG code.
16832 2018-02-08 Richard Biener <rguenther@suse.de>
16834 PR tree-optimization/84233
16835 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
16836 changed flag instead of boguously re-using phi_inserted.
16838 2018-02-08 Martin Jambor <mjambor@suse.cz>
16840 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
16841 static local variables.
16843 2018-02-08 Richard Biener <rguenther@suse.de>
16845 PR tree-optimization/84278
16846 * tree-vect-stmts.c (vectorizable_store): When looking for
16847 smaller vector types to perform grouped strided loads/stores
16848 make sure the mode is supported by the target.
16849 (vectorizable_load): Likewise.
16851 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
16853 * config/aarch64/aarch64.c (aarch64_components_for_bb):
16854 Increase LDP/STP opportunities by adding adjacent callee-saves.
16856 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
16858 PR rtl-optimization/84068
16859 PR rtl-optimization/83459
16860 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
16862 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
16864 PR tree-optimization/84224
16865 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
16866 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
16867 non-zero arguments.
16869 2018-02-07 Iain Sandoe <iain@codesourcery.com>
16872 * config/rs6000/altivec.md (*restore_world): Remove LR use.
16873 * config/rs6000/predicates.md (restore_world_operation): Adjust op
16874 count, remove one USE.
16876 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
16878 * doc/install.texi (Configuration): Document the
16879 --with-long-double-format={ibm,ieee} PowerPC configuration
16883 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
16884 Convert from define_expand to be define_insn_and_split. Rework
16885 float/double/_Float128 conversions to QI/HI/SImode to work with
16886 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
16887 conversions to QI/HImode types did a store and then a load to
16888 truncate the value. For conversions to VSX registers, don't split
16889 the insn, instead emit the code directly. Use the code iterator
16890 any_fix to combine signed and unsigned conversions.
16891 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
16892 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
16893 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
16894 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
16895 (fix_<mode>di2_hw): Likewise.
16896 (fixuns_<mode>di2_hw): Likewise.
16897 (fix_<mode>si2_hw): Likewise.
16898 (fixuns_<mode>si2_hw): Likewise.
16899 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
16900 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
16901 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
16902 fix<uns>_trunc<SFDF:mode>si2_p8.
16903 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
16904 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
16905 (fix<uns>_<mode>_mem): Likewise.
16906 (fctiw<u>z_<mode>_mem): Likewise.
16907 (fix<uns>_<mode>_mem): Likewise.
16908 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
16909 the register allocator from doing a direct move to the GPRs to do
16910 a store, and instead use the ISA 3.0 store byte/half-word from
16911 vector register instruction. For IEEE 128-bit floating point,
16912 also optimize stores of 32-bit ints.
16913 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
16915 2018-02-07 Alan Hayward <alan.hayward@arm.com>
16917 * genextract.c (push_pathstr_operand): New function to support
16919 (walk_rtx): Call push_pathstr_operand.
16920 (print_path): Support [a-zA-Z].
16922 2018-02-07 Richard Biener <rguenther@suse.de>
16924 PR tree-optimization/84037
16925 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
16926 (cse_and_gimplify_to_preheader): Declare.
16927 (vect_get_place_in_interleaving_chain): Likewise.
16928 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
16930 (_loop_vec_info::~_loop_vec_info): Delete it.
16931 (cse_and_gimplify_to_preheader): New function.
16932 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
16933 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
16934 (vectorizable_load): Likewise. For grouped stores always base
16935 the IV on the first element.
16936 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
16937 condition before gimplifying.
16939 2018-02-07 Jakub Jelinek <jakub@redhat.com>
16941 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
16942 *DIV_EXPR and *MOD_EXPR.
16944 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
16947 * config/i386/i386.c (ix86_option_override_internal): Mask out
16948 the CF_SET bit when checking -fcf-protection.
16950 2018-02-07 Tom de Vries <tom@codesourcery.com>
16953 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
16956 2018-02-07 Richard Biener <rguenther@suse.de>
16958 PR tree-optimization/84204
16959 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
16962 PR tree-optimization/84205
16963 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
16964 special-case isl_ast_op_zdiv_r.
16966 PR tree-optimization/84223
16967 * graphite-scop-detection.c (gather_bbs::before_dom_children):
16968 Only add conditions from within the region.
16969 (gather_bbs::after_dom_children): Adjust.
16971 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
16974 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
16975 * config/avr/avr.md: Only post-reload split REG-REG moves if
16976 either register is GENERAL_REG_P.
16978 2018-02-07 Jakub Jelinek <jakub@redhat.com>
16980 PR tree-optimization/84235
16981 * tree-ssa-scopedtables.c
16982 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
16983 if the subtraction is performed in floating point type where NaNs are
16984 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
16985 build 1. Formatting fix.
16987 2018-02-06 Jakub Jelinek <jakub@redhat.com>
16990 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
16991 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
16992 and skip it regardless of bb boundaries. Use CALL_P macro,
16993 don't test INSN_P (insn) together with CALL_P or JUMP_P check
16994 unnecessarily, formatting fix.
16996 2018-02-06 Michael Collison <michael.collison@arm.com>
16998 * config/arm/thumb2.md:
16999 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
17000 (*thumb_mov_notscc): Ditto.
17002 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
17005 * config/rs6000/rs6000.md (su code attribute): Use "u" for
17006 unsigned_fix, not "s".
17008 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
17010 * configure.ac (gcc_fn_eh_frame_ro): New function.
17011 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
17012 correct .eh_frame permissions.
17013 * configure: Regenerate.
17015 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
17017 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
17018 irrelevant options.
17020 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
17022 * config/rs6000/rs6000.c (rs6000_option_override_internal):
17023 Display warning message for -mno-speculate-indirect-jumps.
17025 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
17027 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
17029 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
17031 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
17033 PR tree-optimization/84225
17034 * tree-eh.c (find_trapping_overflow): Only call
17035 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
17037 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
17040 * config/i386/i386.c: Reimplement the check of possible options
17041 -mibt/-mshstk conbination. Change error messages.
17042 * doc/invoke.texi: Fix a typo: remove extra '='.
17044 2018-02-06 Marek Polacek <polacek@redhat.com>
17046 PR tree-optimization/84228
17047 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
17049 2018-02-06 Tamar Christina <tamar.christina@arm.com>
17052 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
17053 emitted arch directives.
17054 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
17055 __ARM_FEATURE_COPROC before changing architectures.
17057 2018-02-06 Richard Biener <rguenther@suse.de>
17059 * config/i386/i386.c (print_reg): Fix typo.
17060 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
17062 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
17064 * configure: Regenerate.
17066 2018-02-05 Martin Sebor <msebor@redhat.com>
17068 PR tree-optimization/83369
17069 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
17072 2018-02-05 Martin Liska <mliska@suse.cz>
17074 * doc/invoke.texi: Cherry-pick upstream r323995.
17076 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
17078 * ira.c (ira_init_register_move_cost): Adjust comment.
17080 2018-02-05 Martin Liska <mliska@suse.cz>
17082 PR gcov-profile/84137
17083 * doc/gcov.texi: Fix typo in documentation.
17085 2018-02-05 Martin Liska <mliska@suse.cz>
17087 PR gcov-profile/83879
17088 * doc/gcov.texi: Document necessity of --dynamic-list-data when
17089 using dlopen functionality.
17091 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
17093 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
17094 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
17095 _mm_maskz_range_ss, _mm_mask_range_round_ss,
17096 _mm_maskz_range_round_ss): New intrinsics.
17097 (__builtin_ia32_rangesd128_round)
17098 (__builtin_ia32_rangess128_round): Remove.
17099 (__builtin_ia32_rangesd128_mask_round,
17100 __builtin_ia32_rangess128_mask_round): New builtins.
17101 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
17102 __builtin_ia32_rangess128_round): Remove.
17103 (__builtin_ia32_rangesd128_mask_round,
17104 __builtin_ia32_rangess128_mask_round): New builtins.
17105 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
17106 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
17107 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
17108 "<round_saeonly_constraint>")): Changed to ...
17109 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
17110 "<round_saeonly_scalar_constraint>")): ... this.
17111 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
17112 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
17113 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
17114 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
17115 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
17117 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
17119 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
17121 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
17122 Remove all values except native, 8540 and 8548.
17124 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
17126 * config/i386/i386.c (ix86_output_function_return): Pass
17127 INVALID_REGNUM, instead of -1, as invalid register number to
17128 indirect_thunk_name and output_indirect_thunk.
17130 2018-02-02 Julia Koval <julia.koval@intel.com>
17132 * config.gcc: Add -march=icelake.
17133 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
17134 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
17135 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
17136 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
17137 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
17138 (processor_target_table): Add icelake.
17139 (ix86_option_override_internal): Handle new PTAs.
17140 (get_builtin_code_for_version): Handle icelake.
17141 (M_INTEL_COREI7_ICELAKE): New.
17142 (fold_builtin_cpu): Handle icelake.
17143 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
17144 * doc/invoke.texi: Add -march=icelake.
17146 2018-02-02 Julia Koval <julia.koval@intel.com>
17148 * config/i386/i386.c (ix86_option_override_internal): Change flags type
17149 to wide_int_bitmask.
17150 * wide-int-bitmask.h: New.
17152 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
17155 * config/i386/i386.md: Replace Pmode with word_mode in
17156 builtin_setjmp_setup and builtin_longjmp to support x32.
17158 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
17162 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
17164 (rs6000_supported_cpu_names): New static variable.
17165 (linux_cpu_translation_table): Likewise.
17166 (elf_platform) <cpu>: Define new static variable and use it.
17167 Translate kernel AT_PLATFORM name to canonical name if needed.
17168 Error if platform name is unknown.
17170 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
17173 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
17175 2018-02-01 Jeff Law <law@redhat.com>
17178 * config/i386/i386.c (release_scratch_register_on_entry): Add new
17179 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
17180 the scratch if RELEASE_VIA_POP is false.
17181 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
17182 If we have to save a temporary register, decrement SIZE appropriately.
17183 Pass new arguments to release_scratch_register_on_entry.
17184 (ix86_adjust_stack_and_probe): Likewise.
17185 (ix86_emit_probe_stack_range): Pass new arguments to
17186 release_scratch_register_on_entry.
17188 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
17190 PR rtl-optimization/84157
17191 * combine.c (change_zero_ext): Use REG_P predicate in
17192 front of HARD_REGISTER_P predicate.
17194 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
17196 * config/avr/avr.c (avr_option_override): Move disabling of
17197 -fdelete-null-pointer-checks to...
17198 * common/config/avr/avr-common.c (avr_option_optimization_table):
17201 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
17203 PR tree-optimization/81635
17204 * tree-data-ref.c (split_constant_offset_1): For types that
17205 wrap on overflow, try to use range info to prove that wrapping
17208 2018-02-01 Renlin Li <renlin.li@arm.com>
17211 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
17212 TAILCALL_ADDR_REGS.
17213 (aarch64_register_move_cost): Likewise.
17214 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
17215 TAILCALL_ADDR_REGS.
17216 (REG_CLASS_NAMES): Likewise.
17217 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
17218 TAILCALL_ADDR_REGS. Remove IP registers.
17219 * config/aarch64/aarch64.md (Ucs): Update register constraint.
17221 2018-02-01 Richard Biener <rguenther@suse.de>
17223 * domwalk.h (dom_walker::dom_walker): Add additional constructor
17224 for specifying RPO order and allow NULL for that.
17225 * domwalk.c (dom_walker::dom_walker): Likewise.
17226 (dom_walker::walk): Handle NULL RPO order.
17227 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
17229 (rewrite_update_dom_walker): Likewise.
17230 (mark_def_dom_walker): Likewise.
17232 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
17234 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
17235 (aarch64_maybe_expand_sve_subreg_move): Declare.
17236 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
17237 * config/aarch64/predicates.md (aarch64_any_register_operand): New
17239 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
17240 that are semantically a reverse operation.
17241 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
17242 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
17243 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
17245 (aarch64_can_change_mode_class): For big-endian, forbid changes
17246 between two SVE modes if they have different element sizes.
17248 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
17250 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
17251 the TImode handling for big-endian targets.
17253 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
17255 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
17256 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
17258 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
17259 Remove BSWAP handing for big-endian targets and use the form of
17260 LD1RQ appropariate for the mode.
17262 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
17264 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
17265 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
17266 duplicated element.
17268 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
17271 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
17272 check for operands that need to go through aarch64_sve_reload_be.
17274 2018-02-01 Jakub Jelinek <jakub@redhat.com>
17276 PR tree-optimization/81661
17277 PR tree-optimization/84117
17278 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
17279 * tree-eh.c: Include gimplify.h.
17280 (find_trapping_overflow, replace_trapping_overflow,
17281 rewrite_to_non_trapping_overflow): New functions.
17282 * tree-vect-loop.c: Include tree-eh.h.
17283 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
17284 * tree-data-ref.c: Include tree-eh.h.
17285 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
17287 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
17289 PR rtl-optimization/84123
17290 * combine.c (change_zero_ext): Check if hard register satisfies
17291 can_change_dest_mode before calling gen_lowpart_SUBREG.
17293 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
17296 * ira.c (ira_init_register_move_cost): Remove assert.
17298 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
17300 PR rtl-optimization/84071
17301 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
17302 * doc/tm.texi: Regenerate.
17304 2018-01-31 Richard Biener <rguenther@suse.de>
17306 PR tree-optimization/84132
17307 * tree-data-ref.c (analyze_miv_subscript): Properly
17308 check whether evolution_function_is_affine_multivariate_p
17309 before calling gcd_of_steps_may_divide_p.
17311 2018-01-31 Julia Koval <julia.koval@intel.com>
17314 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
17315 * config/i386/i386.md (rdpid_rex64) New.
17316 (rdpid): Make 32bit only.
17318 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
17321 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
17322 an IDENTIFIER_NODE for FUNCTION_TYPE's.
17324 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
17327 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
17329 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
17331 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
17333 PR rtl-optimization/84071
17334 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
17335 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
17337 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
17339 * config/arc/arc.c (arc_handle_aux_attribute): New function.
17340 (arc_attribute_table): Add 'aux' attribute.
17341 (arc_in_small_data_p): Consider aux like variables.
17342 (arc_is_aux_reg_p): New function.
17343 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
17344 (arc_get_aux_arg): New function.
17345 (prepare_move_operands): Handle aux-register access.
17346 (arc_handle_aux_attribute): New function.
17347 * doc/extend.texi (ARC Variable attributes): Add subsection.
17349 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
17351 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
17352 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
17353 (arc_attribute_table): Add 'uncached' attribute.
17354 (arc_print_operand): Print '.di' flag for uncached memory
17356 (arc_in_small_data_p): Do not consider for small data the uncached
17358 (arc_is_uncached_mem_p): New function.
17359 * config/arc/predicates.md (compact_store_memory_operand): Check
17360 for uncached memory accesses.
17361 (nonvol_nonimm_operand): Likewise.
17362 * doc/extend.texi (ARC Type Attribute): New subsection.
17364 2018-01-31 Jakub Jelinek <jakub@redhat.com>
17367 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
17368 falign-loops=): Add Optimization flag.
17370 2018-01-30 Jeff Law <law@redhat.com>
17373 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
17374 INT_REGISTERS_SAVED. Check it prior to calling
17375 get_scratch_register_on_entry.
17376 (ix86_adjust_stack_and_probe): Similarly.
17377 (ix86_emit_probe_stack_range): Similarly.
17378 (ix86_expand_prologue): Corresponding changes.
17380 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
17383 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
17384 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
17386 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
17389 * lra-constraints.c (curr_insn_transform): Process AND in the
17392 2018-01-30 Jakub Jelinek <jakub@redhat.com>
17394 PR rtl-optimization/83986
17395 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
17396 dependence against last_pending_memory_flush in addition to
17397 pending_jump_insns.
17399 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
17401 PR tree-optimization/81611
17402 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
17405 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17408 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
17411 2018-01-30 Richard Biener <rguenther@suse.de>
17412 Jakub Jelinek <jakub@redhat.com>
17414 PR tree-optimization/84111
17415 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
17416 inner loops added during recursion, as they don't have up-to-date
17419 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
17422 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
17423 (can_inline_edge_by_limits_p): ... here.
17424 (can_early_inline_edge_p, check_callers,
17425 update_caller_keys, update_callee_keys, recursive_inlining,
17426 add_new_edges_to_heap, speculation_useful_p,
17427 inline_small_functions,
17428 inline_small_functions, flatten_function,
17429 inline_to_all_callers_1): Update.
17431 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
17433 * profile-count.c (profile_count::combine_with_ipa_count): Handle
17436 2018-01-30 Richard Biener <rguenther@suse.de>
17438 PR tree-optimization/83008
17439 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
17440 invariant and constant vector uses in stmts when they need
17441 more than one stmt.
17443 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
17446 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
17447 * configure: Regenerate.
17449 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
17451 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
17453 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
17454 Use gen_rtx_REG rather than gen_lowpart.
17456 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
17458 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
17459 rather than 0 when creating partial subregs.
17461 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
17463 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
17466 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
17469 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
17470 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
17471 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
17472 flags. This restores the settings used before the 2017-07-24.
17473 Turning off pre increment/decrement/modify allows IVOPTS to
17474 optimize DF/SF loops where the index is an int.
17476 2018-01-29 Richard Biener <rguenther@suse.de>
17477 Kelvin Nilsen <kelvin@gcc.gnu.org>
17480 * tree-vect-stmts.c (vectorizable_call): Don't call
17481 targetm.vectorize_builtin_md_vectorized_function if callee is
17484 2018-01-22 Carl Love <cel@us.ibm.com>
17486 * doc/extend.tex: Fix typo in second arg in
17487 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
17489 2018-01-29 Richard Biener <rguenther@suse.de>
17491 PR tree-optimization/84086
17492 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
17493 (flush_ssaname_freelist): When SSA names were released reset
17494 the SCEV hash table.
17496 2018-01-29 Richard Biener <rguenther@suse.de>
17498 PR tree-optimization/84057
17499 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
17500 removed paths when removing edges.
17502 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
17504 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
17505 -mfunction-return=@var{choice}.
17507 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
17509 PR diagnostic/84034
17510 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
17511 Handle CR like TAB.
17512 (layout::print_source_line): Likewise.
17513 (test_get_line_width_without_trailing_whitespace): Add test cases.
17515 2018-01-27 Jakub Jelinek <jakub@redhat.com>
17517 PR middle-end/84040
17518 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
17521 2018-01-26 Jim Wilson <jimw@sifive.com>
17523 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
17525 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
17528 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17530 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
17531 and CMP + SUB-immediate -> SUBS.
17533 2018-01-26 Martin Sebor <msebor@redhat.com>
17535 PR tree-optimization/83896
17536 * tree-ssa-strlen.c (get_string_len): Rename...
17537 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
17538 Avoid assuming length is constant.
17539 (handle_char_store): Use HOST_WIDE_INT for string length.
17541 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
17544 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
17545 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
17547 2018-01-26 Richard Biener <rguenther@suse.de>
17549 PR rtl-optimization/84003
17550 * dse.c (record_store): Only record redundant stores when
17551 the earlier store aliases at least all accesses the later one does.
17553 2018-01-26 Jakub Jelinek <jakub@redhat.com>
17555 PR rtl-optimization/83985
17556 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
17557 REG_CFA_RESTORE insns.
17558 (delete_unmarked_insns): Don't ignore separate shrink wrapping
17559 REG_CFA_RESTORE insns here.
17562 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
17563 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
17565 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
17567 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
17568 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
17569 (arc_init): Likewise.
17570 (arc_override_options): Likewise.
17571 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
17573 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
17575 * config/arc/arc.h (TARGET_DBNZ): Define.
17576 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
17577 properly set the tune attribute.
17578 (dbnz): Use TARGET_DBNZ guard.
17579 * config/arc/arc.opt (mtune): Add core3 option.
17581 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
17583 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
17584 recognize new pic like addresses.
17585 (arc_delegitimize_address): Clean up.
17587 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
17589 * config/arc/arc-arches.def: Option mrf16 valid for all
17591 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
17592 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
17593 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
17594 * config/arc/arc-tables.opt: Regenerate.
17595 * config/arc/arc.c (arc_conditional_register_usage): Handle
17596 reduced register file case.
17597 (arc_file_start): Set must have build attributes.
17598 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
17599 mrf16 option value.
17600 * config/arc/arc.opt (mrf16): Add new option.
17601 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
17602 * config/arc/genmultilib.awk: Handle new mrf16 option.
17603 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
17604 * config/arc/t-multilib: Regenerate.
17605 * doc/invoke.texi (ARC Options): Document mrf16 option.
17607 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
17609 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
17610 * config/arc/arc.c (arc_handle_secure_attribute): New function.
17611 (arc_attribute_table): Add 'secure_call' attribute.
17612 (arc_print_operand): Print secure call operand.
17613 (arc_function_ok_for_sibcall): Don't optimize tail calls when
17615 (arc_is_secure_call_p): New function. * config/arc/arc.md
17616 (call_i): Add support for sjli instruction.
17617 (call_value_i): Likewise.
17618 * config/arc/constraints.md (Csc): New constraint.
17620 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
17621 John Eric Martin <John.Martin@emmicro-us.com>
17623 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
17624 * config/arc/arc.c (_arc_jli_section): New struct.
17625 (arc_jli_section): New type.
17626 (rc_jli_sections): New static variable.
17627 (arc_handle_jli_attribute): New function.
17628 (arc_attribute_table): Add jli_always and jli_fixed attribute.
17629 (arc_file_end): New function.
17630 (TARGET_ASM_FILE_END): Define.
17631 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
17632 (arc_add_jli_section): New function.
17633 (jli_call_scan): Likewise.
17634 (arc_reorg): Call jli_call_scan.
17635 (arc_output_addsi): Remove 'S' from printing asm operand.
17636 (arc_is_jli_call_p): New function.
17637 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
17639 (movhi_insn): Likewise.
17640 (movsi_insn): Likewise.
17641 (movsi_set_cc_insn): Likewise.
17642 (loadqi_update): Likewise.
17643 (load_zeroextendqisi_update): Likewise.
17644 (load_signextendqisi_update): Likewise.
17645 (loadhi_update): Likewise.
17646 (load_zeroextendhisi_update): Likewise.
17647 (load_signextendhisi_update): Likewise.
17648 (loadsi_update): Likewise.
17649 (loadsf_update): Likewise.
17650 (movsicc_insn): Likewise.
17651 (bset_insn): Likewise.
17652 (bxor_insn): Likewise.
17653 (bclr_insn): Likewise.
17654 (bmsk_insn): Likewise.
17655 (bicsi3_insn): Likewise.
17656 (cmpsi_cc_c_insn): Likewise.
17657 (movsi_ne): Likewise.
17658 (movsi_cond_exec): Likewise.
17659 (clrsbsi2): Likewise.
17660 (norm_f): Likewise.
17670 (call_i): Remove 'S' asm letter, add jli instruction.
17671 (call_value_i): Likewise.
17672 * config/arc/arc.op (mjli-always): New option.
17673 * config/arc/constraints.md (Cji): New constraint.
17674 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
17676 (subsf3_fpx): Likewise.
17677 (mulsf3_fpx): Likewise.
17678 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
17680 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
17681 function attrbutes.
17682 * doc/invoke.texi (ARC): Document mjli-always option.
17684 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
17686 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
17687 avoid addition with 0 and use incw and decw where possible.
17689 2018-01-26 Richard Biener <rguenther@suse.de>
17691 PR tree-optimization/81082
17692 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
17693 association if it requires casting to unsigned.
17694 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
17695 from fold_plusminus_mult_expr to catch important cases late when
17696 range info is available.
17698 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
17700 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
17701 * configure.ac (hidden_linkonce): New test.
17702 * configure: Regenerate.
17703 * config.in: Regenerate.
17705 2018-01-26 Julia Koval <julia.koval@intel.com>
17707 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
17708 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
17709 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
17710 _mm_mask_bitshuffle_epi64_mask): Fix type.
17711 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
17712 USI_FTYPE_V4DI_V4DI_USI): Remove.
17713 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
17714 __builtin_ia32_vpshufbitqmb256_mask,
17715 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
17716 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
17717 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
17719 2018-01-26 Alan Modra <amodra@gmail.com>
17722 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
17723 UNSPEC_VBPERMQ. Sort other unspecs.
17725 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
17727 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
17729 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
17731 PR middle-end/83055
17732 * predict.c (drop_profile): Do not push/pop cfun; update also
17734 (handle_missing_profiles): Fix logic looking for zero profiles.
17736 2018-01-25 Jakub Jelinek <jakub@redhat.com>
17738 PR middle-end/83977
17739 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
17740 on functions with #pragma omp declare simd or functions with simd
17742 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
17743 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
17744 Remove trailing \n from warning_at calls.
17746 2018-01-25 Tom de Vries <tom@codesourcery.com>
17749 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
17750 for neutered workers.
17752 2018-01-24 Joseph Myers <joseph@codesourcery.com>
17755 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
17756 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
17758 2018-01-24 Jeff Law <law@redhat.com>
17761 * i386.c (get_probe_interval): Move to earlier point.
17762 (ix86_compute_frame_layout): If -fstack-clash-protection and
17763 the frame is larger than the probe interval, then use pushes
17764 to save registers rather than reg->mem moves.
17765 (ix86_expand_prologue): Remove conditional for int_registers_saved
17768 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
17771 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
17772 min/max for never referenced object.
17774 2018-01-24 Jakub Jelinek <jakub@redhat.com>
17776 PR middle-end/83977
17777 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
17779 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
17780 attributes from DECL_ATTRIBUTES (decl) without affecting
17781 DECL_ATTRIBUTES (current_function_decl).
17782 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
17783 functions with non-NULL DECL_ABSTRACT_ORIGIN.
17785 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
17787 PR tree-optimization/83979
17788 * fold-const.c (fold_comparison): Use constant_boolean_node
17789 instead of boolean_{true,false}_node.
17791 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
17793 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
17796 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
17798 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
17799 Simplify the clause that sets the length attribute.
17800 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
17801 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
17802 clause that sets the length attribute.
17803 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
17805 2018-01-24 Tom de Vries <tom@codesourcery.com>
17808 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
17809 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
17810 Add strict parameter.
17811 (prevent_branch_around_nothing): Insert dummy insn between branch to
17812 label and label with no ptx insn inbetween.
17813 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
17815 2018-01-24 Tom de Vries <tom@codesourcery.com>
17818 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
17819 for neutered threads in warp.
17820 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
17822 2018-01-24 Richard Biener <rguenther@suse.de>
17824 PR tree-optimization/83176
17825 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
17828 2018-01-24 Richard Biener <rguenther@suse.de>
17830 PR tree-optimization/82819
17831 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
17832 code generating pluses that are no-ops in the target precision.
17834 2018-01-24 Richard Biener <rguenther@suse.de>
17836 PR middle-end/84000
17837 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
17839 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
17841 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
17842 to merge probabilities.
17843 * predict.c (probably_never_executed): Also mark as cold functions
17844 with global 0 profile and guessed local profile.
17845 * profile-count.c (profile_probability::combine_with_count): New
17847 * profile-count.h (profile_probability::operator*,
17848 profile_probability::operator*=, profile_probability::operator/,
17849 profile_probability::operator/=): Reduce precision to adjusted
17850 and set value to guessed on contradictory divisions.
17851 (profile_probability::combine_with_freq): Remove.
17852 (profile_probability::combine_wiht_count): Declare.
17853 (profile_count::force_nonzero):: Set to adjusted.
17854 (profile_count::probability_in):: Set quality to adjusted.
17855 * tree-ssa-tail-merge.c (replace_block_by): Use
17856 combine_with_count.
17858 2018-01-23 Andrew Waterman <andrew@sifive.com>
17859 Jim Wilson <jimw@sifive.com>
17861 * config/riscv/riscv.c (riscv_stack_boundary): New.
17862 (riscv_option_override): Set riscv_stack_boundary. Handle
17863 riscv_preferred_stack_boundary_arg.
17864 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
17865 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
17866 (STACK_BOUNDARY): Set to riscv_stack_boundary.
17867 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
17868 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
17869 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
17871 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
17874 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
17875 of struct ix86_frame.
17876 (ix86_expand_epilogue): Likewise. Add a local variable for
17877 the reg_save_offset field in struct ix86_frame.
17879 2018-01-23 Bin Cheng <bin.cheng@arm.com>
17881 PR tree-optimization/82604
17882 * tree-loop-distribution.c (enum partition_kind): New enum item
17883 PKIND_PARTIAL_MEMSET.
17884 (partition_builtin_p): Support above new enum item.
17885 (generate_code_for_partition): Ditto.
17886 (compute_access_range): Differentiate cases that equality can be
17887 proven at all loops, the innermost loops or no loops.
17888 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
17889 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
17890 (finalize_partitions, distribute_loop): Don't fuse partition of
17891 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
17892 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
17893 parloop is enabled.
17895 2018-01-23 Martin Liska <mliska@suse.cz>
17897 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
17898 order to ignore the predictor.
17899 (PRED_POLYMORPHIC_CALL): Likewise.
17900 (PRED_RECURSIVE_CALL): Likewise.
17902 2018-01-23 Martin Liska <mliska@suse.cz>
17904 * tree-profile.c (tree_profiling): Print function header to
17905 aware reader which function we are working on.
17906 * value-prof.c (gimple_find_values_to_profile): Do not print
17907 not interesting value histograms.
17909 2018-01-23 Martin Liska <mliska@suse.cz>
17911 * profile-count.h (enum profile_quality): Add
17912 profile_uninitialized as the first value. Do not number values
17913 as they are zero based.
17914 (profile_count::verify): Update sanity check.
17915 (profile_probability::verify): Likewise.
17917 2018-01-23 Nathan Sidwell <nathan@acm.org>
17919 * doc/invoke.texi (ffor-scope): Deprecate.
17921 2018-01-23 David Malcolm <dmalcolm@redhat.com>
17923 PR tree-optimization/83510
17924 * domwalk.c (set_all_edges_as_executable): New function.
17925 (dom_walker::dom_walker): Convert bool param
17926 "skip_unreachable_blocks" to enum reachability. Move setup of
17927 edge flags to set_all_edges_as_executable and only do it when
17928 reachability is REACHABLE_BLOCKS.
17929 * domwalk.h (enum dom_walker::reachability): New enum.
17930 (dom_walker::dom_walker): Convert bool param
17931 "skip_unreachable_blocks" to enum reachability.
17932 (set_all_edges_as_executable): New decl.
17933 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
17934 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
17936 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
17937 but converting true to REACHABLE_BLOCKS.
17938 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
17940 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
17941 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
17942 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
17944 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
17945 if check_all_array_refs will be called.
17947 2018-01-23 David Malcolm <dmalcolm@redhat.com>
17949 * tree.c (selftest::test_location_wrappers): Add more test
17952 2018-01-23 David Malcolm <dmalcolm@redhat.com>
17954 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
17955 (selftest::test_bit_in_range): Likewise.
17957 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
17960 * doc/sourcebuild.texi (vect_float): Say that the selector
17961 only describes the situation when -funsafe-math-optimizations is on.
17962 (vect_float_strict): Document.
17964 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
17966 PR tree-optimization/83965
17967 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
17968 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
17969 instead of checking only for a reduction.
17970 (vect_recog_widen_sum_pattern): Likewise.
17972 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
17974 * predict.c (probably_never_executed): Only use precise profile info.
17975 (compute_function_frequency): Skip after inlining hack since we now
17976 have quality checking.
17978 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
17980 * profile-count.h (profile_probability::very_unlikely,
17981 profile_probability::unlikely, profile_probability::even): Set
17982 precision to guessed.
17984 2018-01-23 Richard Biener <rguenther@suse.de>
17986 PR tree-optimization/83963
17987 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
17988 Properly terminate dominator walk when crossing the exit edge not
17989 when visiting its source block.
17991 2018-01-23 Jakub Jelinek <jakub@redhat.com>
17994 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
17995 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
17997 2018-01-22 Jakub Jelinek <jakub@redhat.com>
17999 PR tree-optimization/83957
18000 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
18001 semicolon after for body surrounded by braces.
18003 PR tree-optimization/83081
18004 * profile-count.h (profile_probability::split): New method.
18005 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
18006 Use profile_probability::split.
18007 (do_compare_rtx_and_jump): Fix adjustment of probabilities
18008 when splitting a single conditional jump into 2.
18010 2018-01-22 David Malcolm <dmalcolm@redhat.com>
18012 PR tree-optimization/69452
18013 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
18016 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
18018 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
18019 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
18020 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
18022 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
18024 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
18026 * config/rl78/rl78.md (movdi): New define_expand.
18027 * config/rl78/rl78.c (rl78_split_movdi): New function.
18029 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
18032 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
18034 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
18035 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
18036 128-bit to produce an UNSPEC move to get the double word with the
18037 signbit and then a shift directly to do signbit.
18038 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
18039 implementation with a new version that just does either a direct
18040 move or a regular move. Move memory interface to separate insns.
18041 Move insns so they are next to the expander.
18042 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
18043 with signbit move. Split big and little endian case.
18044 (signbit<mode>2_dm_mem_le): Likewise.
18045 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
18046 (signbit<mode>2_dm2): Likewise.
18048 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
18050 * config/rl78/rl78.md (anddi3): New define_expand.
18052 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
18054 * config/rl78/rl78.md (umindi3): New define_expand.
18056 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
18058 * config/rl78/rl78.md (smindi3): New define_expand.
18060 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
18062 * config/rl78/rl78.md (smaxdi3): New define_expand.
18064 2018-01-22 Carl Love <cel@us.ibm.com>
18066 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
18067 LVX_V1TI): Add macro expansion.
18068 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
18069 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
18070 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
18071 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
18072 Change check to determine if the instruction is a byte reversing
18073 entry. Fix typo in comment.
18074 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
18075 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
18076 Add def_builtin calls for new builtins.
18077 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
18078 Add define_insn expansion.
18080 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
18082 * config/rl78/rl78.md (umaxdi3): New define_expand.
18084 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
18086 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
18087 for non-QImode registers.
18089 2018-01-22 Richard Biener <rguenther@suse.de>
18091 PR tree-optimization/83963
18092 * graphite-scop-detection.c (scop_detection::get_sese): Delay
18093 including the loop exit block.
18094 (scop_detection::merge_sese): Likewise.
18095 (scop_detection::add_scop): Do it here instead.
18097 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18099 * doc/sourcebuild.texi (arm_softfloat): Document.
18101 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
18104 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
18105 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
18106 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
18108 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
18109 David Edelsohn <dje.gcc@gmail.com>
18112 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
18113 Change "crset eq" to "crset 2".
18114 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
18115 (*call_indirect_aix<mode>_nospec): Likewise.
18116 (*call_value_indirect_aix<mode>_nospec): Likewise.
18117 (*call_indirect_elfv2<mode>_nospec): Likewise.
18118 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
18119 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
18120 change assembly output from . to $.
18121 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
18122 (indirect_jump<mode>_nospec): Change assembly output from . to $.
18123 (*tablejump<mode>_internal1_nospec): Likewise.
18125 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
18128 * config/sh/sh_optimize_sett_clrt.cc:
18129 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
18131 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
18133 PR tree-optimization/83940
18134 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
18135 offset_dt to vect_constant_def rather than vect_unknown_def_type.
18136 (vect_check_load_store_mask): Add a mask_dt_out parameter and
18137 use it to pass back the definition type.
18138 (vect_check_store_rhs): Likewise rhs_dt_out.
18139 (vect_build_gather_load_calls): Add a mask_dt argument and use
18140 it instead of a call to vect_is_simple_use.
18141 (vectorizable_store): Update calls to vect_check_load_store_mask
18142 and vect_check_store_rhs. Use the dt returned by the latter instead
18143 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
18144 instead of calls to vect_is_simple_use. Pass the scalar rather
18145 than the vector operand to vect_is_simple_use when handling
18146 second and subsequent copies of an rhs value.
18147 (vectorizable_load): Update calls to vect_check_load_store_mask
18148 and vect_build_gather_load_calls. Use the cached mask_dt and
18149 gs_info.offset_dt instead of calls to vect_is_simple_use.
18151 2018-01-20 Jakub Jelinek <jakub@redhat.com>
18153 PR middle-end/83945
18154 * tree-emutls.c: Include gimplify.h.
18155 (lower_emutls_2): New function.
18156 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
18157 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
18158 it before further processing.
18161 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
18162 UINTVAL (trueop1) instead of INTVAL (op1).
18164 2018-01-19 Jakub Jelinek <jakub@redhat.com>
18168 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
18169 INCOMING_FRAME_SP_OFFSET if not defined.
18170 (scan_trace): Add ENTRY argument. If true and
18171 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
18172 emit a note to adjust the CFA offset.
18173 (create_cfi_notes): Adjust scan_trace callers.
18174 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
18175 INCOMING_FRAME_SP_OFFSET in the CIE.
18176 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
18177 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
18179 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
18180 * doc/tm.texi: Regenerated.
18182 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
18184 PR rtl-optimization/83147
18185 * lra-constraints.c (remove_inheritance_pseudos): Use
18186 lra_substitute_pseudo_within_insn.
18188 2018-01-19 Tom de Vries <tom@codesourcery.com>
18189 Cesar Philippidis <cesar@codesourcery.com>
18192 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
18194 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
18197 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
18198 spaces for function labels.
18200 2018-01-19 Martin Liska <mliska@suse.cz>
18202 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
18203 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
18204 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
18205 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
18206 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
18207 (PRED_CONST_RETURN): Change from 69 to 65.
18208 (PRED_NULL_RETURN): Change from 91 to 71.
18209 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
18210 (PRED_LOOP_GUARD): Change from 66 to 73.
18212 2018-01-19 Martin Liska <mliska@suse.cz>
18214 * predict.c (predict_insn_def): Add new assert.
18215 (struct branch_predictor): Change type to signed integer.
18216 (test_prediction_value_range): Amend test to cover
18217 PROB_UNINITIALIZED.
18218 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
18219 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
18220 (PRED_LOOP_ITERATIONS_MAX): Likewise.
18221 (PRED_LOOP_IV_COMPARE): Likewise.
18222 * predict.h (PROB_UNINITIALIZED): Define new constant.
18224 2018-01-19 Martin Liska <mliska@suse.cz>
18226 * predict.c (dump_prediction): Add new format for
18227 analyze_brprob.py script which is enabled with -details
18229 * profile-count.h (precise_p): New function.
18231 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
18233 PR tree-optimization/83922
18234 * tree-vect-loop.c (vect_verify_full_masking): Return false if
18235 there are no statements that need masking.
18236 (vect_active_double_reduction_p): New function.
18237 (vect_analyze_loop_operations): Use it when handling phis that
18238 are not in the loop header.
18240 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
18242 PR tree-optimization/83914
18243 * tree-vect-loop.c (vectorizable_induction): Don't convert
18244 init_expr or apply the peeling adjustment for inductions
18245 that are nested within the vectorized loop.
18247 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18249 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
18252 2018-01-18 Jakub Jelinek <jakub@redhat.com>
18256 * function.h (gimplify_parameters): Add gimple_seq * argument.
18257 * function.c: Include gimple.h and options.h.
18258 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
18259 for the added local temporaries if needed.
18260 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
18261 if there are any parameter cleanups, wrap whole body into a
18262 try/finally with the cleanups.
18264 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
18267 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
18268 Use GET_MODE_CLASS for scalar floating point.
18270 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
18274 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
18275 Fix call of call_cgraph_insertion_hooks.
18277 2018-01-18 Martin Sebor <msebor@redhat.com>
18279 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
18281 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
18284 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
18287 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
18290 * common.opt: (-ffile-prefix-map): New option.
18291 * opts.c (common_handle_option): Defer it.
18292 * opts-global.c (handle_common_deferred_options): Handle it.
18293 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
18294 * file-prefix-map.h: New file.
18295 (remap_debug_filename, add_debug_prefix_map): ...here.
18296 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
18297 * final.c (debug_prefix_map, add_debug_prefix_map
18298 remap_debug_filename): Move to...
18299 * file-prefix-map.c: New file.
18300 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
18301 generalize, get rid of alloca(), use strrchr() instead of strchr().
18302 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
18303 Implement in terms of add_prefix_map().
18304 (remap_macro_filename, remap_debug_filename): Implement in term of
18306 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
18307 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
18308 * dbxout.c: Include file-prefix-map.h.
18309 * varasm.c: Likewise.
18310 * vmsdbgout.c: Likewise.
18311 * xcoffout.c: Likewise.
18312 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
18313 * doc/cppopts.texi (-fmacro-prefix-map): Document.
18314 * doc/invoke.texi (-ffile-prefix-map): Document.
18315 (-fdebug-prefix-map): Update description.
18317 2018-01-18 Martin Liska <mliska@suse.cz>
18319 * config/i386/i386.c (indirect_thunk_name): Document that also
18321 (output_indirect_thunk): Document why both instructions
18322 (pause and lfence) are generated.
18324 2018-01-18 Richard Biener <rguenther@suse.de>
18326 PR tree-optimization/83887
18327 * graphite-scop-detection.c
18328 (scop_detection::get_nearest_dom_with_single_entry): Remove.
18329 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
18330 (scop_detection::merge_sese): Re-implement with a flood-fill
18331 algorithm that properly finds a SESE region if it exists.
18333 2018-01-18 Jakub Jelinek <jakub@redhat.com>
18336 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
18337 pointer_diff optimizations use view_convert instead of convert.
18339 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
18341 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
18342 Generate different code for -mno-speculate-indirect-jumps.
18343 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
18344 (*call_indirect_aix<mode>): Disable for
18345 -mno-speculate-indirect-jumps.
18346 (*call_indirect_aix<mode>_nospec): New define_insn.
18347 (*call_value_indirect_aix<mode>): Disable for
18348 -mno-speculate-indirect-jumps.
18349 (*call_value_indirect_aix<mode>_nospec): New define_insn.
18350 (*sibcall_nonlocal_sysv<mode>): Generate different code for
18351 -mno-speculate-indirect-jumps.
18352 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
18354 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
18356 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
18357 long double type, set the flags for noting the default long double
18358 type, even if we don't pass or return a long double type.
18360 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
18363 * ipa-inline.c (flatten_function): Do not overwrite final inlining
18366 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
18368 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
18369 support for merge[hl].
18370 (fold_mergehl_helper): New helper function.
18371 (tree-vector-builder.h): New #include for tree_vector_builder usage.
18372 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
18373 (altivec_vmrglw_direct): Add xxmrglw insn.
18375 2018-01-17 Andrew Waterman <andrew@sifive.com>
18377 * config/riscv/riscv.c (riscv_conditional_register_usage): If
18378 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
18380 2018-01-17 David Malcolm <dmalcolm@redhat.com>
18383 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
18384 call the lto_location_cache before reading the
18385 DECL_SOURCE_LOCATION of the types.
18387 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
18388 Richard Sandiford <richard.sandiford@linaro.org>
18390 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
18391 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
18392 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
18393 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
18394 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
18396 * config/aarch64/constraints.md (aarch64_movti_operand):
18398 * config/aarch64/predicates.md (Uti): Add new constraint.
18400 2018-01-17 Carl Love <cel@us.ibm.com>
18402 * config/rs6000/vsx.md (define_expand xl_len_r,
18403 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
18404 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
18406 (define_expand, define_insn): Move the shift left from the
18407 define_insn to the define_expand for lxvl and stxvl instructions.
18408 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
18409 and XL_LEN_R definitions to PURE.
18411 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
18413 * config/i386/i386.c (indirect_thunk_name): Declare regno
18414 as unsigned int. Compare regno with INVALID_REGNUM.
18415 (output_indirect_thunk): Ditto.
18416 (output_indirect_thunk_function): Ditto.
18417 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
18418 in the call to output_indirect_thunk_function.
18420 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
18422 PR middle-end/83884
18423 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
18424 rather than the size of inner_type to determine the stack slot size
18425 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
18427 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
18430 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
18433 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
18435 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
18436 endian Linux systems to optionally enable multilibs for selecting
18437 the long double type if the user configured an explicit type.
18438 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
18439 have no long double multilibs if not defined.
18440 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
18441 warn if the user used -mabi={ieee,ibm}longdouble and we built
18442 multilibs for long double.
18443 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
18444 appropriate multilib option.
18445 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
18447 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
18448 for building long double multilibs.
18449 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
18451 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
18453 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
18456 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
18458 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
18461 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
18464 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
18467 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
18469 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
18470 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
18472 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
18474 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
18475 different rtl trees depending on TARGET_64BIT.
18476 (rs6000_gen_lvx): Likewise.
18478 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
18480 * config/visium/visium.md (nop): Tweak comment.
18481 (hazard_nop): Likewise.
18483 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
18485 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
18486 -mspeculate-indirect-jumps.
18487 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
18488 for -mno-speculate-indirect-jumps.
18489 (*call_indirect_elfv2<mode>_nospec): New define_insn.
18490 (*call_value_indirect_elfv2<mode>): Disable for
18491 -mno-speculate-indirect-jumps.
18492 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
18493 (indirect_jump): Emit different RTL for
18494 -mno-speculate-indirect-jumps.
18495 (*indirect_jump<mode>): Disable for
18496 -mno-speculate-indirect-jumps.
18497 (*indirect_jump<mode>_nospec): New define_insn.
18498 (tablejump): Emit different RTL for
18499 -mno-speculate-indirect-jumps.
18500 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
18501 (tablejumpsi_nospec): New define_expand.
18502 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
18503 (tablejumpdi_nospec): New define_expand.
18504 (*tablejump<mode>_internal1): Disable for
18505 -mno-speculate-indirect-jumps.
18506 (*tablejump<mode>_internal1_nospec): New define_insn.
18507 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
18510 2018-01-16 Artyom Skrobov tyomitch@gmail.com
18512 * caller-save.c (insert_save): Drop unnecessary parameter. All
18515 2018-01-16 Jakub Jelinek <jakub@redhat.com>
18516 Richard Biener <rguenth@suse.de>
18519 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
18520 return early, inline manually is_gimple_sizepos. Make sure if we
18521 call gimplify_expr we don't end up with a gimple constant.
18522 * tree.c (variably_modified_type_p): Don't return true for
18523 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
18524 * gimplify.h (is_gimple_sizepos): Remove.
18526 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
18528 PR tree-optimization/83857
18529 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
18530 vectorizable_live_operation for pure SLP statements.
18531 (vectorizable_live_operation): Handle PHIs.
18533 2018-01-16 Richard Biener <rguenther@suse.de>
18535 PR tree-optimization/83867
18536 * tree-vect-stmts.c (vect_transform_stmt): Precompute
18537 nested_in_vect_loop_p since the scalar stmt may get invalidated.
18539 2018-01-16 Jakub Jelinek <jakub@redhat.com>
18542 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
18543 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
18544 If off is not INTEGER_CST, issue a may not be aligned warning
18545 rather than isn't aligned. Use isn%'t rather than isn't.
18546 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
18548 <case MULT_EXPR>: Improve the case when bottom and one of the
18549 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
18550 operand, in that case check if the other operand is multiple of
18551 bottom divided by the INTEGER_CST operand.
18553 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
18556 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
18557 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
18558 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
18559 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
18560 * config/pa/pa.c (pa_function_arg_advance): Likewise.
18561 (pa_function_arg, pa_arg_partial_bytes): Likewise.
18562 (pa_function_arg_size): New function.
18564 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
18566 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
18567 in a separate statement.
18569 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
18571 PR tree-optimization/83847
18572 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
18573 group gathers and scatters.
18575 2018-01-16 Jakub Jelinek <jakub@redhat.com>
18577 PR rtl-optimization/86620
18578 * params.def (max-sched-ready-insns): Bump minimum value to 1.
18580 PR rtl-optimization/83213
18581 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
18582 to last if both are JUMP_INSNs.
18584 PR tree-optimization/83843
18585 * gimple-ssa-store-merging.c
18586 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
18587 store_immediate_info for bswap/nop orig_stores.
18589 2018-01-15 Andrew Waterman <andrew@sifive.com>
18591 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
18593 <UDIV>: Increase cost if !TARGET_DIV.
18595 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
18597 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
18598 (define_attr "cr_logical_3op"): New.
18599 (cceq_ior_compare): Adjust.
18600 (cceq_ior_compare_complement): Adjust.
18601 (*cceq_rev_compare): Adjust.
18602 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
18603 (is_cracked_insn): Adjust.
18604 (insn_must_be_first_in_group): Adjust.
18605 * config/rs6000/40x.md: Adjust.
18606 * config/rs6000/440.md: Adjust.
18607 * config/rs6000/476.md: Adjust.
18608 * config/rs6000/601.md: Adjust.
18609 * config/rs6000/603.md: Adjust.
18610 * config/rs6000/6xx.md: Adjust.
18611 * config/rs6000/7450.md: Adjust.
18612 * config/rs6000/7xx.md: Adjust.
18613 * config/rs6000/8540.md: Adjust.
18614 * config/rs6000/cell.md: Adjust.
18615 * config/rs6000/e300c2c3.md: Adjust.
18616 * config/rs6000/e500mc.md: Adjust.
18617 * config/rs6000/e500mc64.md: Adjust.
18618 * config/rs6000/e5500.md: Adjust.
18619 * config/rs6000/e6500.md: Adjust.
18620 * config/rs6000/mpc.md: Adjust.
18621 * config/rs6000/power4.md: Adjust.
18622 * config/rs6000/power5.md: Adjust.
18623 * config/rs6000/power6.md: Adjust.
18624 * config/rs6000/power7.md: Adjust.
18625 * config/rs6000/power8.md: Adjust.
18626 * config/rs6000/power9.md: Adjust.
18627 * config/rs6000/rs64.md: Adjust.
18628 * config/rs6000/titan.md: Adjust.
18630 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
18632 * config/i386/predicates.md (indirect_branch_operand): Rewrite
18633 ix86_indirect_branch_register logic.
18635 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
18637 * config/i386/constraints.md (Bs): Update
18638 ix86_indirect_branch_register check. Don't check
18639 ix86_indirect_branch_register with GOT_memory_operand.
18641 * config/i386/predicates.md (GOT_memory_operand): Don't check
18642 ix86_indirect_branch_register here.
18643 (GOT32_symbol_operand): Likewise.
18645 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
18647 * config/i386/predicates.md (constant_call_address_operand):
18648 Rewrite ix86_indirect_branch_register logic.
18649 (sibcall_insn_operand): Likewise.
18651 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
18653 * config/i386/constraints.md (Bs): Replace
18654 ix86_indirect_branch_thunk_register with
18655 ix86_indirect_branch_register.
18657 * config/i386/i386.md (indirect_jump): Likewise.
18658 (tablejump): Likewise.
18659 (*sibcall_memory): Likewise.
18660 (*sibcall_value_memory): Likewise.
18661 Peepholes of indirect call and jump via memory: Likewise.
18662 * config/i386/i386.opt: Likewise.
18663 * config/i386/predicates.md (indirect_branch_operand): Likewise.
18664 (GOT_memory_operand): Likewise.
18665 (call_insn_operand): Likewise.
18666 (sibcall_insn_operand): Likewise.
18667 (GOT32_symbol_operand): Likewise.
18669 2018-01-15 Jakub Jelinek <jakub@redhat.com>
18671 PR middle-end/83837
18672 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
18673 type rather than type addr's type points to.
18674 (expand_omp_atomic_mutex): Likewise.
18675 (expand_omp_atomic): Likewise.
18677 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
18680 * config/i386/i386.c (output_indirect_thunk_function): Use
18681 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
18682 for __x86_return_thunk.
18684 2018-01-15 Richard Biener <rguenther@suse.de>
18686 PR middle-end/83850
18687 * expmed.c (extract_bit_field_1): Fix typo.
18689 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18692 * config/arm/iterators.md (VF): New mode iterator.
18693 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
18694 Remove integer-related logic from pattern.
18695 (neon_vabd<mode>_3): Likewise.
18697 2018-01-15 Jakub Jelinek <jakub@redhat.com>
18699 PR middle-end/82694
18700 * common.opt (fstrict-overflow): No longer an alias.
18701 (fwrapv-pointer): New option.
18702 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
18703 also for pointer types based on flag_wrapv_pointer.
18704 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
18705 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
18706 opts->x_flag_wrapv got set.
18707 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
18708 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
18709 POINTER_TYPE_OVERFLOW_UNDEFINED.
18710 * match.pd: Likewise in address comparison pattern.
18711 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
18713 2018-01-15 Richard Biener <rguenther@suse.de>
18716 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
18717 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
18718 Reset type names to their identifier if their TYPE_DECL doesn't
18719 have linkage (and thus is used for ODR and devirt).
18720 (save_debug_info_for_decl): Remove.
18721 (save_debug_info_for_type): Likewise.
18722 (add_tree_to_fld_list): Adjust.
18723 * tree-pretty-print.c (dump_generic_node): Make dumping of
18724 type names more robust.
18726 2018-01-15 Richard Biener <rguenther@suse.de>
18728 * BASE-VER: Bump to 8.0.1.
18730 2018-01-14 Martin Sebor <msebor@redhat.com>
18733 * builtins.c (check_access): Avoid warning when the no-warning bit
18736 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
18738 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
18739 * ira-color (allocno_hard_regs_compare): Likewise.
18741 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
18744 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
18745 Use .pushsection/.popsection.
18747 2018-01-14 Martin Sebor <msebor@redhat.com>
18750 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
18752 2018-01-14 Jakub Jelinek <jakub@redhat.com>
18754 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
18755 entry from extra_headers.
18756 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
18757 extra_headers, make the list bitwise identical to the i?86-*-* one.
18759 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
18761 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
18762 -mcmodel=large with -mindirect-branch=thunk,
18763 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
18764 -mfunction-return=thunk-extern.
18765 * doc/invoke.texi: Document -mcmodel=large is incompatible with
18766 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
18767 -mfunction-return=thunk and -mfunction-return=thunk-extern.
18769 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
18771 * config/i386/i386.c (print_reg): Print the name of the full
18772 integer register without '%'.
18773 (ix86_print_operand): Handle 'V'.
18774 * doc/extend.texi: Document 'V' modifier.
18776 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
18778 * config/i386/constraints.md (Bs): Disallow memory operand for
18779 -mindirect-branch-register.
18781 * config/i386/predicates.md (indirect_branch_operand): Likewise.
18782 (GOT_memory_operand): Likewise.
18783 (call_insn_operand): Likewise.
18784 (sibcall_insn_operand): Likewise.
18785 (GOT32_symbol_operand): Likewise.
18786 * config/i386/i386.md (indirect_jump): Call convert_memory_address
18787 for -mindirect-branch-register.
18788 (tablejump): Likewise.
18789 (*sibcall_memory): Likewise.
18790 (*sibcall_value_memory): Likewise.
18791 Disallow peepholes of indirect call and jump via memory for
18792 -mindirect-branch-register.
18793 (*call_pop): Replace m with Bw.
18794 (*call_value_pop): Likewise.
18795 (*sibcall_pop_memory): Replace m with Bs.
18796 * config/i386/i386.opt (mindirect-branch-register): New option.
18797 * doc/invoke.texi: Document -mindirect-branch-register option.
18799 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
18801 * config/i386/i386-protos.h (ix86_output_function_return): New.
18802 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
18803 set function_return_type.
18804 (indirect_thunk_name): Add ret_p to indicate thunk for function
18806 (output_indirect_thunk_function): Pass false to
18807 indirect_thunk_name.
18808 (ix86_output_indirect_branch_via_reg): Likewise.
18809 (ix86_output_indirect_branch_via_push): Likewise.
18810 (output_indirect_thunk_function): Create alias for function
18811 return thunk if regno < 0.
18812 (ix86_output_function_return): New function.
18813 (ix86_handle_fndecl_attribute): Handle function_return.
18814 (ix86_attribute_table): Add function_return.
18815 * config/i386/i386.h (machine_function): Add
18816 function_return_type.
18817 * config/i386/i386.md (simple_return_internal): Use
18818 ix86_output_function_return.
18819 (simple_return_internal_long): Likewise.
18820 * config/i386/i386.opt (mfunction-return=): New option.
18821 (indirect_branch): Mention -mfunction-return=.
18822 * doc/extend.texi: Document function_return function attribute.
18823 * doc/invoke.texi: Document -mfunction-return= option.
18825 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
18827 * config/i386/i386-opts.h (indirect_branch): New.
18828 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
18829 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
18830 with local indirect jump when converting indirect call and jump.
18831 (ix86_set_indirect_branch_type): New.
18832 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
18833 (indirectlabelno): New.
18834 (indirect_thunk_needed): Likewise.
18835 (indirect_thunk_bnd_needed): Likewise.
18836 (indirect_thunks_used): Likewise.
18837 (indirect_thunks_bnd_used): Likewise.
18838 (INDIRECT_LABEL): Likewise.
18839 (indirect_thunk_name): Likewise.
18840 (output_indirect_thunk): Likewise.
18841 (output_indirect_thunk_function): Likewise.
18842 (ix86_output_indirect_branch_via_reg): Likewise.
18843 (ix86_output_indirect_branch_via_push): Likewise.
18844 (ix86_output_indirect_branch): Likewise.
18845 (ix86_output_indirect_jmp): Likewise.
18846 (ix86_code_end): Call output_indirect_thunk_function if needed.
18847 (ix86_output_call_insn): Call ix86_output_indirect_branch if
18849 (ix86_handle_fndecl_attribute): Handle indirect_branch.
18850 (ix86_attribute_table): Add indirect_branch.
18851 * config/i386/i386.h (machine_function): Add indirect_branch_type
18852 and has_local_indirect_jump.
18853 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
18855 (tablejump): Likewise.
18856 (*indirect_jump): Use ix86_output_indirect_jmp.
18857 (*tablejump_1): Likewise.
18858 (simple_return_indirect_internal): Likewise.
18859 * config/i386/i386.opt (mindirect-branch=): New option.
18860 (indirect_branch): New.
18863 (thunk-inline): Likewise.
18864 (thunk-extern): Likewise.
18865 * doc/extend.texi: Document indirect_branch function attribute.
18866 * doc/invoke.texi: Document -mindirect-branch= option.
18868 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
18871 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
18873 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
18875 * ipa-inline.c (want_inline_small_function_p): Return false if
18876 inlining has already failed with CIF_FINAL_ERROR.
18877 (update_caller_keys): Call want_inline_small_function_p before
18879 (update_callee_keys): Likewise.
18881 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
18883 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
18885 (rs6000_quadword_masked_address_p): Likewise.
18886 (quad_aligned_load_p): Likewise.
18887 (quad_aligned_store_p): Likewise.
18888 (const_load_sequence_p): Add comment to describe the outer-most loop.
18889 (mimic_memory_attributes_and_flags): New function.
18890 (rs6000_gen_stvx): Likewise.
18891 (replace_swapped_aligned_store): Likewise.
18892 (rs6000_gen_lvx): Likewise.
18893 (replace_swapped_aligned_load): Likewise.
18894 (replace_swapped_load_constant): Capitalize argument name in
18895 comment describing this function.
18896 (rs6000_analyze_swaps): Add a third pass to search for vector loads
18897 and stores that access quad-word aligned addresses and replace
18898 with stvx or lvx instructions when appropriate.
18899 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
18900 New function prototype.
18901 (rs6000_quadword_masked_address_p): Likewise.
18902 (rs6000_gen_lvx): Likewise.
18903 (rs6000_gen_stvx): Likewise.
18904 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
18905 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
18906 when memory address is aligned.
18907 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
18908 this split to select lvx instruction when memory address is aligned.
18909 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
18910 instruction when memory address is aligned.
18911 (*vsx_le_perm_load_v16qi): Likewise.
18912 (four unnamed splitters): Modify to select the stvx instruction
18913 when memory is aligned.
18915 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
18917 * predict.c (determine_unlikely_bbs): Handle correctly BBs
18918 which appears in the queue multiple times.
18920 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18921 Alan Hayward <alan.hayward@arm.com>
18922 David Sherwood <david.sherwood@arm.com>
18924 * tree-vectorizer.h (vec_lower_bound): New structure.
18925 (_loop_vec_info): Add check_nonzero and lower_bounds.
18926 (LOOP_VINFO_CHECK_NONZERO): New macro.
18927 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
18928 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
18929 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
18930 fields. Make seg_len the distance travelled, not including the
18932 (dr_direction_indicator): Declare.
18933 (dr_zero_step_indicator): Likewise.
18934 (dr_known_forward_stride_p): Likewise.
18935 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
18937 (runtime_alias_check_p): Allow runtime alias checks with
18939 (operator ==): Compare access_size and align.
18940 (prune_runtime_alias_test_list): Rework for new distinction between
18941 the access_size and seg_len.
18942 (create_intersect_range_checks_index): Likewise. Cope with polynomial
18944 (get_segment_min_max): New function.
18945 (create_intersect_range_checks): Use it.
18946 (dr_step_indicator): New function.
18947 (dr_direction_indicator): Likewise.
18948 (dr_zero_step_indicator): Likewise.
18949 (dr_known_forward_stride_p): Likewise.
18950 * tree-loop-distribution.c (data_ref_segment_size): Return
18951 DR_STEP * (niters - 1).
18952 (compute_alias_check_pairs): Update call to the dr_with_seg_len
18954 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
18955 (vect_preserves_scalar_order_p): New function, split out from...
18956 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
18957 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
18958 (vect_vfa_access_size): New function.
18959 (vect_vfa_align): Likewise.
18960 (vect_compile_time_alias): Take access_size_a and access_b arguments.
18961 (dump_lower_bound): New function.
18962 (vect_check_lower_bound): Likewise.
18963 (vect_small_gap_p): Likewise.
18964 (vectorizable_with_step_bound_p): Likewise.
18965 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
18966 depencies if the vectorization factor is 1. Convert the checks
18967 for nonzero steps into checks on the bounds of DR_STEP. Try using
18968 a bunds check for variable steps if the minimum required step is
18969 relatively small. Update calls to the dr_with_seg_len
18970 constructor and to vect_compile_time_alias.
18971 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
18973 (vect_loop_versioning): Call it.
18974 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
18976 (vect_estimate_min_profitable_iters): Account for any bounds checks.
18978 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18979 Alan Hayward <alan.hayward@arm.com>
18980 David Sherwood <david.sherwood@arm.com>
18982 * doc/sourcebuild.texi (vect_scatter_store): Document.
18983 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
18985 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
18987 * genopinit.c (main): Add supports_vec_scatter_store and
18988 supports_vec_scatter_store_cached to target_optabs.
18989 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
18990 IFN_MASK_SCATTER_STORE.
18991 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
18993 * internal-fn.h (internal_store_fn_p): Declare.
18994 (internal_fn_stored_value_index): Likewise.
18995 * internal-fn.c (scatter_store_direct): New macro.
18996 (expand_scatter_store_optab_fn): New function.
18997 (direct_scatter_store_optab_supported_p): New macro.
18998 (internal_store_fn_p): New function.
18999 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
19000 IFN_MASK_SCATTER_STORE.
19001 (internal_fn_mask_index): Likewise.
19002 (internal_fn_stored_value_index): New function.
19003 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
19004 for scatter stores.
19005 * optabs-query.h (supports_vec_scatter_store_p): Declare.
19006 * optabs-query.c (supports_vec_scatter_store_p): New function.
19007 * tree-vectorizer.h (vect_get_store_rhs): Declare.
19008 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
19009 true for scatter stores.
19010 (vect_gather_scatter_fn_p): Handle scatter stores too.
19011 (vect_check_gather_scatter): Consider using scatter stores if
19012 supports_vec_scatter_store_p.
19013 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
19014 scatter stores too.
19015 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
19016 internal_fn_stored_value_index.
19017 (check_load_store_masking): Handle scatter stores too.
19018 (vect_get_store_rhs): Make public.
19019 (vectorizable_call): Use internal_store_fn_p.
19020 (vectorizable_store): Handle scatter store internal functions.
19021 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
19022 when deciding whether the end of the group has been reached.
19023 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
19024 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
19025 (mask_scatter_store<mode>): New insns.
19027 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19028 Alan Hayward <alan.hayward@arm.com>
19029 David Sherwood <david.sherwood@arm.com>
19031 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
19032 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
19033 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
19035 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
19036 Use vect_truncate_gather_scatter_offset if we can't treat the
19037 operation as a normal gather load or scatter store.
19038 (get_group_load_store_type): Take the gather_scatter_info
19039 as argument. Try using a gather load or scatter store for
19040 single-element groups.
19041 (get_load_store_type): Update calls to get_group_load_store_type
19042 and vect_use_strided_gather_scatters_p.
19044 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19045 Alan Hayward <alan.hayward@arm.com>
19046 David Sherwood <david.sherwood@arm.com>
19048 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
19049 optional tree argument.
19050 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
19052 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
19053 but continue to use the current value as a fallback.
19054 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
19055 to compare the updates.
19056 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
19057 (get_load_store_type): Use it when handling a strided access.
19058 (vect_get_strided_load_store_ops): New function.
19059 (vect_get_data_ptr_increment): Likewise.
19060 (vectorizable_load): Handle strided gather loads. Always pass
19061 a step to vect_create_data_ref_ptr and bump_vector_ptr.
19063 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19064 Alan Hayward <alan.hayward@arm.com>
19065 David Sherwood <david.sherwood@arm.com>
19067 * doc/md.texi (gather_load@var{m}): Document.
19068 (mask_gather_load@var{m}): Likewise.
19069 * genopinit.c (main): Add supports_vec_gather_load and
19070 supports_vec_gather_load_cached to target_optabs.
19071 * optabs-tree.c (init_tree_optimization_optabs): Use
19072 ggc_cleared_alloc to allocate target_optabs.
19073 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
19074 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
19076 * internal-fn.h (internal_load_fn_p): Declare.
19077 (internal_gather_scatter_fn_p): Likewise.
19078 (internal_fn_mask_index): Likewise.
19079 (internal_gather_scatter_fn_supported_p): Likewise.
19080 * internal-fn.c (gather_load_direct): New macro.
19081 (expand_gather_load_optab_fn): New function.
19082 (direct_gather_load_optab_supported_p): New macro.
19083 (direct_internal_fn_optab): New function.
19084 (internal_load_fn_p): Likewise.
19085 (internal_gather_scatter_fn_p): Likewise.
19086 (internal_fn_mask_index): Likewise.
19087 (internal_gather_scatter_fn_supported_p): Likewise.
19088 * optabs-query.c (supports_at_least_one_mode_p): New function.
19089 (supports_vec_gather_load_p): Likewise.
19090 * optabs-query.h (supports_vec_gather_load_p): Declare.
19091 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
19092 and memory_type field.
19093 (NUM_PATTERNS): Bump to 15.
19094 * tree-vect-data-refs.c: Include internal-fn.h.
19095 (vect_gather_scatter_fn_p): New function.
19096 (vect_describe_gather_scatter_call): Likewise.
19097 (vect_check_gather_scatter): Try using internal functions for
19098 gather loads. Recognize existing calls to a gather load function.
19099 (vect_analyze_data_refs): Consider using gather loads if
19100 supports_vec_gather_load_p.
19101 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
19102 (vect_get_gather_scatter_offset_type): Likewise.
19103 (vect_convert_mask_for_vectype): Likewise.
19104 (vect_add_conversion_to_patterm): Likewise.
19105 (vect_try_gather_scatter_pattern): Likewise.
19106 (vect_recog_gather_scatter_pattern): New pattern recognizer.
19107 (vect_vect_recog_func_ptrs): Add it.
19108 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
19109 internal_fn_mask_index and internal_gather_scatter_fn_p.
19110 (check_load_store_masking): Take the gather_scatter_info as an
19111 argument and handle gather loads.
19112 (vect_get_gather_scatter_ops): New function.
19113 (vectorizable_call): Check internal_load_fn_p.
19114 (vectorizable_load): Likewise. Handle gather load internal
19116 (vectorizable_store): Update call to check_load_store_masking.
19117 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
19118 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
19119 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
19120 (aarch64_gather_scale_operand_d): New predicates.
19121 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
19122 (mask_gather_load<mode>): New insns.
19124 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19125 Alan Hayward <alan.hayward@arm.com>
19126 David Sherwood <david.sherwood@arm.com>
19128 * optabs.def (fold_left_plus_optab): New optab.
19129 * doc/md.texi (fold_left_plus_@var{m}): Document.
19130 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
19131 * internal-fn.c (fold_left_direct): Define.
19132 (expand_fold_left_optab_fn): Likewise.
19133 (direct_fold_left_optab_supported_p): Likewise.
19134 * fold-const-call.c (fold_const_fold_left): New function.
19135 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
19136 * tree-parloops.c (valid_reduction_p): New function.
19137 (gather_scalar_reductions): Use it.
19138 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
19139 (vect_finish_replace_stmt): Declare.
19140 * tree-vect-loop.c (fold_left_reduction_fn): New function.
19141 (needs_fold_left_reduction_p): New function, split out from...
19142 (vect_is_simple_reduction): ...here. Accept reductions that
19143 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
19144 (vect_force_simple_reduction): Also store the reduction type in
19145 the assignment's STMT_VINFO_REDUC_TYPE.
19146 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
19147 (merge_with_identity): New function.
19148 (vect_expand_fold_left): Likewise.
19149 (vectorize_fold_left_reduction): Likewise.
19150 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
19151 scalar phi in place for it. Check for target support and reject
19152 cases that would reassociate the operation. Defer the transform
19153 phase to vectorize_fold_left_reduction.
19154 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
19155 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
19156 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
19158 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19160 * tree-if-conv.c (predicate_mem_writes): Remove redundant
19161 call to ifc_temp_var.
19163 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19164 Alan Hayward <alan.hayward@arm.com>
19165 David Sherwood <david.sherwood@arm.com>
19167 * target.def (legitimize_address_displacement): Take the original
19168 offset as a poly_int.
19169 * targhooks.h (default_legitimize_address_displacement): Update
19171 * targhooks.c (default_legitimize_address_displacement): Likewise.
19172 * doc/tm.texi: Regenerate.
19173 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
19174 as an argument, moving assert of ad->disp == ad->disp_term to...
19175 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
19176 Try calling targetm.legitimize_address_displacement before expanding
19177 the address rather than afterwards, and adjust for the new interface.
19178 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
19179 Match the new hook interface. Handle SVE addresses.
19180 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
19181 new hook interface.
19183 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19185 * Makefile.in (OBJS): Add early-remat.o.
19186 * target.def (select_early_remat_modes): New hook.
19187 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
19188 * doc/tm.texi: Regenerate.
19189 * targhooks.h (default_select_early_remat_modes): Declare.
19190 * targhooks.c (default_select_early_remat_modes): New function.
19191 * timevar.def (TV_EARLY_REMAT): New timevar.
19192 * passes.def (pass_early_remat): New pass.
19193 * tree-pass.h (make_pass_early_remat): Declare.
19194 * early-remat.c: New file.
19195 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
19197 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
19199 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19200 Alan Hayward <alan.hayward@arm.com>
19201 David Sherwood <david.sherwood@arm.com>
19203 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
19204 vfm1 with a bound_epilog parameter.
19205 (vect_do_peeling): Update calls accordingly, and move the prologue
19206 call earlier in the function. Treat the base bound_epilog as 0 for
19207 fully-masked loops and retain vf - 1 for other loops. Add 1 to
19208 this base when peeling for gaps.
19209 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
19210 with fully-masked loops.
19211 (vect_estimate_min_profitable_iters): Handle the single peeled
19212 iteration in that case.
19214 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19215 Alan Hayward <alan.hayward@arm.com>
19216 David Sherwood <david.sherwood@arm.com>
19218 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
19219 single-element interleaving even if the size is not a power of 2.
19220 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
19221 accesses for single-element interleaving if the group size is
19224 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19225 Alan Hayward <alan.hayward@arm.com>
19226 David Sherwood <david.sherwood@arm.com>
19228 * doc/md.texi (fold_extract_last_@var{m}): Document.
19229 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
19230 * optabs.def (fold_extract_last_optab): New optab.
19231 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
19232 * internal-fn.c (fold_extract_direct): New macro.
19233 (expand_fold_extract_optab_fn): Likewise.
19234 (direct_fold_extract_optab_supported_p): Likewise.
19235 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
19236 * tree-vect-loop.c (vect_model_reduction_cost): Handle
19237 EXTRACT_LAST_REDUCTION.
19238 (get_initial_def_for_reduction): Do not create an initial vector
19239 for EXTRACT_LAST_REDUCTION reductions.
19240 (vectorizable_reduction): Leave the scalar phi in place for
19241 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
19242 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
19243 epilogue code for EXTRACT_LAST_REDUCTION and defer the
19244 transform phase to vectorizable_condition.
19245 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
19247 (vect_finish_stmt_generation): ...here.
19248 (vect_finish_replace_stmt): New function.
19249 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
19250 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
19252 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
19254 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19255 Alan Hayward <alan.hayward@arm.com>
19256 David Sherwood <david.sherwood@arm.com>
19258 * doc/md.texi (extract_last_@var{m}): Document.
19259 * optabs.def (extract_last_optab): New optab.
19260 * internal-fn.def (EXTRACT_LAST): New internal function.
19261 * internal-fn.c (cond_unary_direct): New macro.
19262 (expand_cond_unary_optab_fn): Likewise.
19263 (direct_cond_unary_optab_supported_p): Likewise.
19264 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
19265 loops using EXTRACT_LAST.
19266 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
19267 (extract_last_<mode>): ...this optab.
19268 (vec_extract<mode><Vel>): Update accordingly.
19270 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19271 Alan Hayward <alan.hayward@arm.com>
19272 David Sherwood <david.sherwood@arm.com>
19274 * target.def (empty_mask_is_expensive): New hook.
19275 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
19276 * doc/tm.texi: Regenerate.
19277 * targhooks.h (default_empty_mask_is_expensive): Declare.
19278 * targhooks.c (default_empty_mask_is_expensive): New function.
19279 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
19280 if the target says that empty masks are expensive.
19281 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
19283 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
19285 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19286 Alan Hayward <alan.hayward@arm.com>
19287 David Sherwood <david.sherwood@arm.com>
19289 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
19290 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
19291 (vect_use_loop_mask_for_alignment_p): New function.
19292 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
19293 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
19294 niters_skip argument. Make sure that the first niters_skip elements
19295 of the first iteration are inactive.
19296 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
19297 Update call to vect_set_loop_masks_directly.
19298 (get_misalign_in_elems): New function, split out from...
19299 (vect_gen_prolog_loop_niters): ...here.
19300 (vect_update_init_of_dr): Take a code argument that specifies whether
19301 the adjustment should be added or subtracted.
19302 (vect_update_init_of_drs): Likewise.
19303 (vect_prepare_for_masked_peels): New function.
19304 (vect_do_peeling): Skip prologue peeling if we're using a mask
19305 instead. Update call to vect_update_inits_of_drs.
19306 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
19308 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
19309 alignment. Do not include the number of peeled iterations in
19310 the minimum threshold in that case.
19311 (vectorizable_induction): Adjust the start value down by
19312 LOOP_VINFO_MASK_SKIP_NITERS iterations.
19313 (vect_transform_loop): Call vect_prepare_for_masked_peels.
19314 Take the number of skipped iterations into account when calculating
19316 * tree-vect-stmts.c (vect_gen_while_not): New function.
19318 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19319 Alan Hayward <alan.hayward@arm.com>
19320 David Sherwood <david.sherwood@arm.com>
19322 * doc/sourcebuild.texi (vect_fully_masked): Document.
19323 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
19324 default value to 0.
19325 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
19327 (vect_analyze_loop_2): ...here. Don't check the vectorization
19328 factor against the number of loop iterations if the loop is
19331 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19332 Alan Hayward <alan.hayward@arm.com>
19333 David Sherwood <david.sherwood@arm.com>
19335 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
19336 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
19337 (dump_groups): Update accordingly.
19338 (iv_use::mem_type): New member variable.
19339 (address_p): New function.
19340 (record_use): Add a mem_type argument and initialize the new
19342 (record_group_use): Add a mem_type argument. Use address_p.
19343 Remove obsolete null checks of base_object. Update call to record_use.
19344 (find_interesting_uses_op): Update call to record_group_use.
19345 (find_interesting_uses_cond): Likewise.
19346 (find_interesting_uses_address): Likewise.
19347 (get_mem_type_for_internal_fn): New function.
19348 (find_address_like_use): Likewise.
19349 (find_interesting_uses_stmt): Try find_address_like_use before
19350 calling find_interesting_uses_op.
19351 (addr_offset_valid_p): Use the iv mem_type field as the type
19352 of the addressed memory.
19353 (add_autoinc_candidates): Likewise.
19354 (get_address_cost): Likewise.
19355 (split_small_address_groups_p): Use address_p.
19356 (split_address_groups): Likewise.
19357 (add_iv_candidate_for_use): Likewise.
19358 (autoinc_possible_for_pair): Likewise.
19359 (rewrite_groups): Likewise.
19360 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
19361 (determine_group_iv_cost): Update after split of USE_ADDRESS.
19362 (get_alias_ptr_type_for_ptr_address): New function.
19363 (rewrite_use_address): Rewrite address uses in calls that were
19364 identified by find_address_like_use.
19366 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19367 Alan Hayward <alan.hayward@arm.com>
19368 David Sherwood <david.sherwood@arm.com>
19370 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
19372 * gimple-expr.h (is_gimple_addressable: Likewise.
19373 * gimple-expr.c (is_gimple_address): Likewise.
19374 * internal-fn.c (expand_call_mem_ref): New function.
19375 (expand_mask_load_optab_fn): Use it.
19376 (expand_mask_store_optab_fn): Likewise.
19378 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19379 Alan Hayward <alan.hayward@arm.com>
19380 David Sherwood <david.sherwood@arm.com>
19382 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
19383 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
19384 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
19385 (cond_umax@var{mode}): Document.
19386 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
19387 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
19388 (cond_umin_optab, cond_umax_optab): New optabs.
19389 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
19390 (COND_IOR, COND_XOR): New internal functions.
19391 * internal-fn.h (get_conditional_internal_fn): Declare.
19392 * internal-fn.c (cond_binary_direct): New macro.
19393 (expand_cond_binary_optab_fn): Likewise.
19394 (direct_cond_binary_optab_supported_p): Likewise.
19395 (get_conditional_internal_fn): New function.
19396 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
19397 Cope with reduction statements that are vectorized as calls rather
19399 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
19400 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
19401 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
19402 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
19403 (UNSPEC_COND_EOR): New unspecs.
19404 (optab): Add mappings for them.
19405 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
19406 (sve_int_op, sve_fp_op): New int attributes.
19408 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19409 Alan Hayward <alan.hayward@arm.com>
19410 David Sherwood <david.sherwood@arm.com>
19412 * optabs.def (while_ult_optab): New optab.
19413 * doc/md.texi (while_ult@var{m}@var{n}): Document.
19414 * internal-fn.def (WHILE_ULT): New internal function.
19415 * internal-fn.h (direct_internal_fn_supported_p): New override
19416 that takes two types as argument.
19417 * internal-fn.c (while_direct): New macro.
19418 (expand_while_optab_fn): New function.
19419 (convert_optab_supported_p): Likewise.
19420 (direct_while_optab_supported_p): New macro.
19421 * wide-int.h (wi::udiv_ceil): New function.
19422 * tree-vectorizer.h (rgroup_masks): New structure.
19423 (vec_loop_masks): New typedef.
19424 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
19425 and fully_masked_p.
19426 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
19427 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
19428 (vect_max_vf): New function.
19429 (slpeel_make_loop_iterate_ntimes): Delete.
19430 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
19431 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
19432 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
19433 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
19434 internal-fn.h, stor-layout.h and optabs-query.h.
19435 (vect_set_loop_mask): New function.
19436 (add_preheader_seq): Likewise.
19437 (add_header_seq): Likewise.
19438 (interleave_supported_p): Likewise.
19439 (vect_maybe_permute_loop_masks): Likewise.
19440 (vect_set_loop_masks_directly): Likewise.
19441 (vect_set_loop_condition_masked): Likewise.
19442 (vect_set_loop_condition_unmasked): New function, split out from
19443 slpeel_make_loop_iterate_ntimes.
19444 (slpeel_make_loop_iterate_ntimes): Rename to..
19445 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
19446 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
19447 (vect_do_peeling): Update call accordingly.
19448 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
19450 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
19451 mask_compare_type, can_fully_mask_p and fully_masked_p.
19452 (release_vec_loop_masks): New function.
19453 (_loop_vec_info): Use it to free the loop masks.
19454 (can_produce_all_loop_masks_p): New function.
19455 (vect_get_max_nscalars_per_iter): Likewise.
19456 (vect_verify_full_masking): Likewise.
19457 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
19458 retries, and free the mask rgroups before retrying. Check loop-wide
19459 reasons for disallowing fully-masked loops. Make the final decision
19460 about whether use a fully-masked loop or not.
19461 (vect_estimate_min_profitable_iters): Do not assume that peeling
19462 for the number of iterations will be needed for fully-masked loops.
19463 (vectorizable_reduction): Disable fully-masked loops.
19464 (vectorizable_live_operation): Likewise.
19465 (vect_halve_mask_nunits): New function.
19466 (vect_double_mask_nunits): Likewise.
19467 (vect_record_loop_mask): Likewise.
19468 (vect_get_loop_mask): Likewise.
19469 (vect_transform_loop): Handle the case in which the final loop
19470 iteration might handle a partial vector. Call vect_set_loop_condition
19471 instead of slpeel_make_loop_iterate_ntimes.
19472 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
19473 (check_load_store_masking): New function.
19474 (prepare_load_store_mask): Likewise.
19475 (vectorizable_store): Handle fully-masked loops.
19476 (vectorizable_load): Likewise.
19477 (supportable_widening_operation): Use vect_halve_mask_nunits for
19479 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
19480 (vect_gen_while): New function.
19481 * config/aarch64/aarch64.md (umax<mode>3): New expander.
19482 (aarch64_uqdec<mode>): New insn.
19484 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19485 Alan Hayward <alan.hayward@arm.com>
19486 David Sherwood <david.sherwood@arm.com>
19488 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
19489 (reduc_xor_scal_optab): New optabs.
19490 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
19491 (reduc_xor_scal_@var{m}): Document.
19492 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
19493 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
19494 internal functions.
19495 * fold-const-call.c (fold_const_call): Handle them.
19496 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
19497 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
19498 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
19499 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
19500 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
19501 (UNSPEC_XORV): New unspecs.
19502 (optab): Add entries for them.
19503 (BITWISEV): New int iterator.
19504 (bit_reduc_op): New int attributes.
19506 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19507 Alan Hayward <alan.hayward@arm.com>
19508 David Sherwood <david.sherwood@arm.com>
19510 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
19511 * internal-fn.def (VEC_SHL_INSERT): New internal function.
19512 * optabs.def (vec_shl_insert_optab): New optab.
19513 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
19514 (duplicate_and_interleave): Likewise.
19515 * tree-vect-loop.c: Include internal-fn.h.
19516 (neutral_op_for_slp_reduction): New function, split out from
19517 get_initial_defs_for_reduction.
19518 (get_initial_def_for_reduction): Handle option 2 for variable-length
19519 vectors by loading the neutral value into a vector and then shifting
19520 the initial value into element 0.
19521 (get_initial_defs_for_reduction): Replace the code argument with
19522 the neutral value calculated by neutral_op_for_slp_reduction.
19523 Use gimple_build_vector for constant-length vectors.
19524 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
19525 but the first group_size elements have a neutral value.
19526 Use duplicate_and_interleave otherwise.
19527 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
19528 Update call to get_initial_defs_for_reduction. Handle SLP
19529 reductions for variable-length vectors by creating one vector
19530 result for each scalar result, with the elements associated
19531 with other scalar results stubbed out with the neutral value.
19532 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
19533 Require IFN_VEC_SHL_INSERT for double reductions on
19534 variable-length vectors, or SLP reductions that have
19535 a neutral value. Require can_duplicate_and_interleave_p
19536 support for variable-length unchained SLP reductions if there
19537 is no neutral value, such as for MIN/MAX reductions. Also require
19538 the number of vector elements to be a multiple of the number of
19539 SLP statements when doing variable-length unchained SLP reductions.
19540 Update call to vect_create_epilog_for_reduction.
19541 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
19542 and remove initial values.
19543 (duplicate_and_interleave): Make public.
19544 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
19545 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
19547 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19548 Alan Hayward <alan.hayward@arm.com>
19549 David Sherwood <david.sherwood@arm.com>
19551 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
19552 (can_duplicate_and_interleave_p): New function.
19553 (vect_get_and_check_slp_defs): Take the vector of statements
19554 rather than just the current one. Remove excess parentheses.
19555 Restriction rejectinon of vect_constant_def and vect_external_def
19556 for variable-length vectors to boolean types, or types for which
19557 can_duplicate_and_interleave_p is false.
19558 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
19559 (duplicate_and_interleave): New function.
19560 (vect_get_constant_vectors): Use gimple_build_vector for
19561 constant-length vectors and suitable variable-length constant
19562 vectors. Use duplicate_and_interleave for other variable-length
19563 vectors. Don't defer the update when inserting new statements.
19565 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19566 Alan Hayward <alan.hayward@arm.com>
19567 David Sherwood <david.sherwood@arm.com>
19569 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
19570 min_profitable_iters doesn't go negative.
19572 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19573 Alan Hayward <alan.hayward@arm.com>
19574 David Sherwood <david.sherwood@arm.com>
19576 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
19577 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
19578 * optabs.def (vec_mask_load_lanes_optab): New optab.
19579 (vec_mask_store_lanes_optab): Likewise.
19580 * internal-fn.def (MASK_LOAD_LANES): New internal function.
19581 (MASK_STORE_LANES): Likewise.
19582 * internal-fn.c (mask_load_lanes_direct): New macro.
19583 (mask_store_lanes_direct): Likewise.
19584 (expand_mask_load_optab_fn): Handle masked operations.
19585 (expand_mask_load_lanes_optab_fn): New macro.
19586 (expand_mask_store_optab_fn): Handle masked operations.
19587 (expand_mask_store_lanes_optab_fn): New macro.
19588 (direct_mask_load_lanes_optab_supported_p): Likewise.
19589 (direct_mask_store_lanes_optab_supported_p): Likewise.
19590 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
19592 (vect_load_lanes_supported): Likewise.
19593 * tree-vect-data-refs.c (strip_conversion): New function.
19594 (can_group_stmts_p): Likewise.
19595 (vect_analyze_data_ref_accesses): Use it instead of checking
19596 for a pair of assignments.
19597 (vect_store_lanes_supported): Take a masked_p parameter.
19598 (vect_load_lanes_supported): Likewise.
19599 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
19600 vect_store_lanes_supported and vect_load_lanes_supported.
19601 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
19602 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
19603 parameter. Don't allow gaps for masked accesses.
19604 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
19605 and vect_load_lanes_supported.
19606 (get_load_store_type): Take a masked_p parameter and update
19607 call to get_group_load_store_type.
19608 (vectorizable_store): Update call to get_load_store_type.
19609 Handle IFN_MASK_STORE_LANES.
19610 (vectorizable_load): Update call to get_load_store_type.
19611 Handle IFN_MASK_LOAD_LANES.
19613 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19614 Alan Hayward <alan.hayward@arm.com>
19615 David Sherwood <david.sherwood@arm.com>
19617 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
19619 * config/aarch64/aarch64-protos.h
19620 (aarch64_sve_struct_memory_operand_p): Declare.
19621 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
19622 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
19623 (VPRED, vpred): Handle SVE structure modes.
19624 * config/aarch64/constraints.md (Utx): New constraint.
19625 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
19626 (aarch64_sve_struct_nonimmediate_operand): New predicates.
19627 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
19628 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
19629 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
19630 structure modes. Split into pieces after RA.
19631 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
19632 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
19634 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
19635 SVE structure modes.
19636 (aarch64_classify_address): Likewise.
19637 (sizetochar): Move earlier in file.
19638 (aarch64_print_operand): Handle SVE register lists.
19639 (aarch64_array_mode): New function.
19640 (aarch64_sve_struct_memory_operand_p): Likewise.
19641 (TARGET_ARRAY_MODE): Redefine.
19643 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19644 Alan Hayward <alan.hayward@arm.com>
19645 David Sherwood <david.sherwood@arm.com>
19647 * target.def (array_mode): New target hook.
19648 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
19649 * doc/tm.texi: Regenerate.
19650 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
19651 * hooks.c (hook_optmode_mode_uhwi_none): New function.
19652 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
19653 targetm.array_mode.
19654 * stor-layout.c (mode_for_array): Likewise. Support polynomial
19657 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19658 Alan Hayward <alan.hayward@arm.com>
19659 David Sherwood <david.sherwood@arm.com>
19661 * fold-const.c (fold_binary_loc): Check the argument types
19662 rather than the result type when testing for a vector operation.
19664 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19666 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
19667 * doc/tm.texi: Regenerate.
19669 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19670 Alan Hayward <alan.hayward@arm.com>
19671 David Sherwood <david.sherwood@arm.com>
19673 * doc/invoke.texi (-msve-vector-bits=): Document new option.
19674 (sve): Document new AArch64 extension.
19675 * doc/md.texi (w): Extend the description of the AArch64
19676 constraint to include SVE vectors.
19677 (Upl, Upa): Document new AArch64 predicate constraints.
19678 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
19680 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
19681 (msve-vector-bits=): New option.
19682 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
19683 SVE when these are disabled.
19684 (sve): New extension.
19685 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
19686 modes. Adjust their number of units based on aarch64_sve_vg.
19687 (MAX_BITSIZE_MODE_ANY_MODE): Define.
19688 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
19689 aarch64_addr_query_type.
19690 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
19691 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
19692 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
19693 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
19694 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
19695 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
19696 (aarch64_simd_imm_zero_p): Delete.
19697 (aarch64_check_zero_based_sve_index_immediate): Declare.
19698 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
19699 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
19700 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
19701 (aarch64_sve_float_mul_immediate_p): Likewise.
19702 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
19703 rather than an rtx.
19704 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
19705 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
19706 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
19707 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
19708 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
19709 (aarch64_regmode_natural_size): Likewise.
19710 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
19711 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
19713 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
19714 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
19715 for VG and the SVE predicate registers.
19716 (V_ALIASES): Add a "z"-prefixed alias.
19717 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
19718 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
19719 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
19720 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
19721 (REG_CLASS_NAMES): Add entries for them.
19722 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
19723 and the predicate registers.
19724 (aarch64_sve_vg): Declare.
19725 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
19726 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
19727 (REGMODE_NATURAL_SIZE): Define.
19728 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
19730 * config/aarch64/aarch64.c: Include cfgrtl.h.
19731 (simd_immediate_info): Add a constructor for series vectors,
19732 and an associated step field.
19733 (aarch64_sve_vg): New variable.
19734 (aarch64_dbx_register_number): Handle VG and the predicate registers.
19735 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
19736 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
19737 (VEC_ANY_DATA, VEC_STRUCT): New constants.
19738 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
19739 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
19740 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
19741 (aarch64_get_mask_mode): New functions.
19742 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
19743 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
19744 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
19745 predicate modes and predicate registers. Explicitly restrict
19746 GPRs to modes of 16 bytes or smaller. Only allow FP registers
19747 to store a vector mode if it is recognized by
19748 aarch64_classify_vector_mode.
19749 (aarch64_regmode_natural_size): New function.
19750 (aarch64_hard_regno_caller_save_mode): Return the original mode
19752 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
19753 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
19754 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
19755 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
19757 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
19758 does not overlap dest if the function is frame-related. Handle
19760 (aarch64_split_add_offset): New function.
19761 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
19762 them aarch64_add_offset.
19763 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
19764 and update call to aarch64_sub_sp.
19765 (aarch64_add_cfa_expression): New function.
19766 (aarch64_expand_prologue): Pass extra temporary registers to the
19767 functions above. Handle the case in which we need to emit new
19768 DW_CFA_expressions for registers that were originally saved
19769 relative to the stack pointer, but now have to be expressed
19770 relative to the frame pointer.
19771 (aarch64_output_mi_thunk): Pass extra temporary registers to the
19773 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
19774 IP0 and IP1 values for SVE frames.
19775 (aarch64_expand_vec_series): New function.
19776 (aarch64_expand_sve_widened_duplicate): Likewise.
19777 (aarch64_expand_sve_const_vector): Likewise.
19778 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
19779 Handle SVE constants. Use emit_move_insn to move a force_const_mem
19780 into the register, rather than emitting a SET directly.
19781 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
19782 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
19783 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
19784 (offset_9bit_signed_scaled_p): New functions.
19785 (aarch64_replicate_bitmask_imm): New function.
19786 (aarch64_bitmask_imm): Use it.
19787 (aarch64_cannot_force_const_mem): Reject expressions involving
19788 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
19789 (aarch64_classify_index): Handle SVE indices, by requiring
19790 a plain register index with a scale that matches the element size.
19791 (aarch64_classify_address): Handle SVE addresses. Assert that
19792 the mode of the address is VOIDmode or an integer mode.
19793 Update call to aarch64_classify_symbol.
19794 (aarch64_classify_symbolic_expression): Update call to
19795 aarch64_classify_symbol.
19796 (aarch64_const_vec_all_in_range_p): New function.
19797 (aarch64_print_vector_float_operand): Likewise.
19798 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
19799 "vN" for FP registers with SVE modes. Handle (const ...) vectors
19800 and the FP immediates 1.0 and 0.5.
19801 (aarch64_print_address_internal): Handle SVE addresses.
19802 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
19803 (aarch64_regno_regclass): Handle predicate registers.
19804 (aarch64_secondary_reload): Handle big-endian reloads of SVE
19806 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
19807 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
19808 (aarch64_convert_sve_vector_bits): New function.
19809 (aarch64_override_options): Use it to handle -msve-vector-bits=.
19810 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
19811 rather than an rtx.
19812 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
19813 Handle SVE vector and predicate modes. Accept VL-based constants
19814 that need only one temporary register, and VL offsets that require
19815 no temporary registers.
19816 (aarch64_conditional_register_usage): Mark the predicate registers
19817 as fixed if SVE isn't available.
19818 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
19819 Return true for SVE vector and predicate modes.
19820 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
19821 rather than an unsigned int. Handle SVE modes.
19822 (aarch64_preferred_simd_mode): Update call accordingly. Handle
19824 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
19826 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
19827 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
19828 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
19829 (aarch64_sve_float_mul_immediate_p): New functions.
19830 (aarch64_sve_valid_immediate): New function.
19831 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
19832 Explicitly reject structure modes. Check for INDEX constants.
19833 Handle PTRUE and PFALSE constants.
19834 (aarch64_check_zero_based_sve_index_immediate): New function.
19835 (aarch64_simd_imm_zero_p): Delete.
19836 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
19837 vector modes. Accept constants in the range of CNT[BHWD].
19838 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
19839 ask for an Advanced SIMD mode.
19840 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
19841 (aarch64_simd_vector_alignment): Handle SVE predicates.
19842 (aarch64_vectorize_preferred_vector_alignment): New function.
19843 (aarch64_simd_vector_alignment_reachable): Use it instead of
19845 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
19846 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
19848 (MAX_VECT_LEN): Delete.
19849 (expand_vec_perm_d): Add a vec_flags field.
19850 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
19851 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
19852 (aarch64_evpc_ext): Don't apply a big-endian lane correction
19854 (aarch64_evpc_rev): Rename to...
19855 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
19856 (aarch64_evpc_rev_global): New function.
19857 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
19858 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
19860 (aarch64_evpc_sve_tbl): New function.
19861 (aarch64_expand_vec_perm_const_1): Update after rename of
19862 aarch64_evpc_rev. Handle SVE permutes too, trying
19863 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
19864 than aarch64_evpc_tbl.
19865 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
19866 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
19867 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
19868 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
19869 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
19870 (aarch64_expand_sve_vcond): New functions.
19871 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
19872 of aarch64_vector_mode_p.
19873 (aarch64_dwarf_poly_indeterminate_value): New function.
19874 (aarch64_compute_pressure_classes): Likewise.
19875 (aarch64_can_change_mode_class): Likewise.
19876 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
19877 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
19878 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
19879 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
19880 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
19881 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
19882 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
19883 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
19885 (Dn, Dl, Dr): Accept const as well as const_vector.
19886 (Dz): Likewise. Compare against CONST0_RTX.
19887 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
19888 of "vector" where appropriate.
19889 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
19890 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
19891 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
19892 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
19893 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
19894 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
19895 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
19896 (v_int_equiv): Extend to SVE modes.
19897 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
19899 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
19900 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
19901 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
19902 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
19903 (SVE_COND_FP_CMP): New int iterators.
19904 (perm_hilo): Handle the new unpack unspecs.
19905 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
19907 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
19908 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
19909 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
19910 (aarch64_equality_operator, aarch64_constant_vector_operand)
19911 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
19912 (aarch64_sve_nonimmediate_operand): Likewise.
19913 (aarch64_sve_general_operand): Likewise.
19914 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
19915 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
19916 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
19917 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
19918 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
19919 (aarch64_sve_float_arith_immediate): Likewise.
19920 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
19921 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
19922 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
19923 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
19924 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
19925 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
19926 (aarch64_sve_float_arith_operand): Likewise.
19927 (aarch64_sve_float_arith_with_sub_operand): Likewise.
19928 (aarch64_sve_float_mul_operand): Likewise.
19929 (aarch64_sve_vec_perm_operand): Likewise.
19930 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
19931 (aarch64_mov_operand): Accept const_poly_int and const_vector.
19932 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
19933 as well as const_vector.
19934 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
19935 in file. Use CONST0_RTX and CONSTM1_RTX.
19936 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
19937 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
19938 Use aarch64_simd_imm_zero.
19939 * config/aarch64/aarch64-sve.md: New file.
19940 * config/aarch64/aarch64.md: Include it.
19941 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
19942 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
19943 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
19944 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
19945 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
19946 (sve): New attribute.
19947 (enabled): Disable instructions with the sve attribute unless
19949 (movqi, movhi): Pass CONST_POLY_INT operaneds through
19950 aarch64_expand_mov_immediate.
19951 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
19952 CNT[BHSD] immediates.
19953 (movti): Split CONST_POLY_INT moves into two halves.
19954 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
19955 Split additions that need a temporary here if the destination
19956 is the stack pointer.
19957 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
19958 (*add<mode>3_poly_1): New instruction.
19959 (set_clobber_cc): New expander.
19961 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19963 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
19964 parameter and use it instead of GET_MODE_SIZE (innermode). Use
19965 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
19966 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
19967 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
19968 Change innermode from fixed_mode_size to machine_mode.
19969 (simplify_subreg): Update call accordingly. Handle a constant-sized
19970 subreg of a variable-length CONST_VECTOR.
19972 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19973 Alan Hayward <alan.hayward@arm.com>
19974 David Sherwood <david.sherwood@arm.com>
19976 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
19977 (add_offset_to_base): New function, split out from...
19978 (create_mem_ref): ...here. When handling a scale other than 1,
19979 check first whether the address is valid without the offset.
19980 Add it into the base if so, leaving the index and scale as-is.
19982 2018-01-12 Jakub Jelinek <jakub@redhat.com>
19985 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
19986 fold_for_warn before checking if arg2 is INTEGER_CST.
19988 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
19990 * config/rs6000/predicates.md (load_multiple_operation): Delete.
19991 (store_multiple_operation): Delete.
19992 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
19993 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
19994 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
19995 guarded by TARGET_STRING.
19996 (rs6000_output_load_multiple): Delete.
19997 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
19998 OPTION_MASK_STRING / TARGET_STRING handling.
19999 (print_operand) <'N', 'O'>: Add comment that these are unused now.
20000 (const rs6000_opt_masks) <"string">: Change mask to 0.
20001 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
20002 (MASK_STRING): Delete.
20003 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
20005 (load_multiple): Delete.
20012 (store_multiple): Delete.
20019 (movmemsi_8reg): Delete.
20020 (corresponding unnamed define_insn): Delete.
20021 (movmemsi_6reg): Delete.
20022 (corresponding unnamed define_insn): Delete.
20023 (movmemsi_4reg): Delete.
20024 (corresponding unnamed define_insn): Delete.
20025 (movmemsi_2reg): Delete.
20026 (corresponding unnamed define_insn): Delete.
20027 (movmemsi_1reg): Delete.
20028 (corresponding unnamed define_insn): Delete.
20029 * config/rs6000/rs6000.opt (mno-string): New.
20030 (mstring): Replace by deprecation warning stub.
20031 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
20033 2018-01-12 Jakub Jelinek <jakub@redhat.com>
20035 * regrename.c (regrename_do_replace): If replacing the same
20036 reg multiple times, try to reuse last created gen_raw_REG.
20039 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
20040 main to workaround a bug in GDB.
20042 2018-01-12 Tom de Vries <tom@codesourcery.com>
20045 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
20047 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
20049 PR rtl-optimization/80481
20050 * ira-color.c (get_cap_member): New function.
20051 (allocnos_conflict_by_live_ranges_p): Use it.
20052 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
20053 (setup_slot_coalesced_allocno_live_ranges): Ditto.
20055 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
20058 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
20059 (*saddl_se_1): Ditto.
20060 (*ssubsi_1): Ditto.
20061 (*ssubl_se_1): Ditto.
20063 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
20065 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
20066 rather than wi::to_widest for DR_INITs.
20067 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
20068 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
20069 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
20071 (vect_analyze_group_access_1): Note that here.
20073 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
20075 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
20076 polynomial type sizes.
20078 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
20080 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
20081 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
20082 (gimple_add_tmp_var): Likewise.
20084 2018-01-12 Martin Liska <mliska@suse.cz>
20086 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
20087 (gimple_alloc_sizes): Likewise.
20088 (dump_gimple_statistics): Use PRIu64 in printf format.
20089 * gimple.h: Change uint64_t to int.
20091 2018-01-12 Martin Liska <mliska@suse.cz>
20093 * tree-core.h: Use uint64_t instead of int.
20094 * tree.c (tree_node_counts): Likewise.
20095 (tree_node_sizes): Likewise.
20096 (dump_tree_statistics): Use PRIu64 in printf format.
20098 2018-01-12 Martin Liska <mliska@suse.cz>
20100 * Makefile.in: As qsort_chk is implemented in vec.c, add
20101 vec.o to linkage of gencfn-macros.
20102 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
20103 passing the info to record_node_allocation_statistics.
20104 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
20106 * ggc-common.c (struct ggc_usage): Add operator== and use
20107 it in operator< and compare function.
20108 * mem-stats.h (struct mem_usage): Likewise.
20109 * vec.c (struct vec_usage): Remove operator< and compare
20110 function. Can be simply inherited.
20112 2018-01-12 Martin Jambor <mjambor@suse.cz>
20115 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
20116 * tree-ssa-math-opts.c: Include domwalk.h.
20117 (convert_mult_to_fma_1): New function.
20118 (fma_transformation_info): New type.
20119 (fma_deferring_state): Likewise.
20120 (cancel_fma_deferring): New function.
20121 (result_of_phi): Likewise.
20122 (last_fma_candidate_feeds_initial_phi): Likewise.
20123 (convert_mult_to_fma): Added deferring logic, split actual
20124 transformation to convert_mult_to_fma_1.
20125 (math_opts_dom_walker): New type.
20126 (math_opts_dom_walker::after_dom_children): New method, body moved
20127 here from pass_optimize_widening_mul::execute, added deferring logic
20129 (pass_optimize_widening_mul::execute): Moved most of code to
20130 math_opts_dom_walker::after_dom_children.
20131 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
20132 * config/i386/i386.c (ix86_option_override_internal): Added
20133 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
20135 2018-01-12 Richard Biener <rguenther@suse.de>
20138 * dwarf2out.c (gen_variable_die): Do not reset old_die for
20139 inline instance vars.
20141 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
20144 * config/rx/rx.c (rx_is_restricted_memory_address):
20145 Handle SUBREG case.
20147 2018-01-12 Richard Biener <rguenther@suse.de>
20149 PR tree-optimization/80846
20150 * target.def (split_reduction): New target hook.
20151 * targhooks.c (default_split_reduction): New function.
20152 * targhooks.h (default_split_reduction): Declare.
20153 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
20154 target requests first reduce vectors by combining low and high
20156 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
20157 (get_vectype_for_scalar_type_and_size): Export.
20158 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
20159 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
20160 * doc/tm.texi: Regenerate.
20161 * config/i386/i386.c (ix86_split_reduction): Implement
20162 TARGET_VECTORIZE_SPLIT_REDUCTION.
20164 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
20167 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
20168 in PIC mode except for TARGET_VXWORKS_RTP.
20169 * config/sparc/sparc.c: Include cfgrtl.h.
20170 (TARGET_INIT_PIC_REG): Define.
20171 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
20172 (sparc_pic_register_p): New predicate.
20173 (sparc_legitimate_address_p): Use it.
20174 (sparc_legitimize_pic_address): Likewise.
20175 (sparc_delegitimize_address): Likewise.
20176 (sparc_mode_dependent_address_p): Likewise.
20177 (gen_load_pcrel_sym): Remove 4th parameter.
20178 (load_got_register): Adjust call to above. Remove obsolete stuff.
20179 (sparc_expand_prologue): Do not call load_got_register here.
20180 (sparc_flat_expand_prologue): Likewise.
20181 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
20182 (sparc_use_pseudo_pic_reg): New function.
20183 (sparc_init_pic_reg): Likewise.
20184 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
20185 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
20187 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
20189 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
20190 Add item for branch_cost.
20192 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
20194 PR rtl-optimization/83565
20195 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
20196 not extend the result to a larger mode for rotate operations.
20197 (num_sign_bit_copies1): Likewise.
20199 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
20202 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
20204 Use values-Xc.o for -pedantic.
20205 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
20207 2018-01-12 Martin Liska <mliska@suse.cz>
20210 * ipa-devirt.c (final_warning_record::grow_type_warnings):
20212 (possible_polymorphic_call_targets): Use it.
20213 (ipa_devirt): Likewise.
20215 2018-01-12 Martin Liska <mliska@suse.cz>
20217 * profile-count.h (enum profile_quality): Use 0 as invalid
20218 enum value of profile_quality.
20220 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
20222 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
20223 -mext-string options.
20225 2018-01-12 Richard Biener <rguenther@suse.de>
20227 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
20228 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
20229 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
20231 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
20233 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
20235 * configure.ac (--with-long-double-format): Add support for the
20236 configuration option to change the default long double format on
20238 * config.gcc (powerpc*-linux*-*): Likewise.
20239 * configure: Regenerate.
20240 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
20241 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
20242 used without modification.
20244 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
20246 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
20247 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
20248 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
20249 MISC_BUILTIN_SPEC_BARRIER.
20250 (rs6000_init_builtins): Likewise.
20251 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
20253 (speculation_barrier): New define_insn.
20254 * doc/extend.texi: Document __builtin_speculation_barrier.
20256 2018-01-11 Jakub Jelinek <jakub@redhat.com>
20259 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
20260 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
20261 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
20263 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
20264 integral modes instead of "ss" and "sd".
20265 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
20266 vectors with 32-bit and 64-bit elements.
20267 (vecdupssescalarmodesuffix): New mode attribute.
20268 (vec_dup<mode>): Use it.
20270 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
20273 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
20274 frame if argument is passed on stack.
20276 2018-01-11 Jakub Jelinek <jakub@redhat.com>
20279 * ree.c (combine_reaching_defs): Optimize also
20280 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
20281 reg2=any_extend(exp); reg1=reg2;, formatting fix.
20283 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
20285 PR middle-end/83189
20286 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
20288 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
20290 PR middle-end/83718
20291 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
20292 after they are computed.
20294 2018-01-11 Bin Cheng <bin.cheng@arm.com>
20296 PR tree-optimization/83695
20297 * gimple-loop-linterchange.cc
20298 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
20299 reset cached scev information after interchange.
20300 (pass_linterchange::execute): Remove call to scev_reset_htab.
20302 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20304 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
20305 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
20306 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
20307 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
20308 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
20309 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
20310 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
20311 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
20312 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
20313 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
20314 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
20315 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
20316 (V_lane_reg): Likewise.
20317 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
20319 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
20320 (vfmal_lane_low<mode>_intrinsic,
20321 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
20322 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
20323 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
20324 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
20325 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
20326 vfmsl_lane_high<mode>_intrinsic): New define_insns.
20328 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20330 * config/arm/arm-cpus.in (fp16fml): New feature.
20331 (ALL_SIMD): Add fp16fml.
20332 (armv8.2-a): Add fp16fml as an option.
20333 (armv8.3-a): Likewise.
20334 (armv8.4-a): Add fp16fml as part of fp16.
20335 * config/arm/arm.h (TARGET_FP16FML): Define.
20336 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
20338 * config/arm/arm-modes.def (V2HF): Define.
20339 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
20340 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
20341 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
20342 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
20343 vfmsl_low, vfmsl_high): New set of builtins.
20344 * config/arm/iterators.md (PLUSMINUS): New code iterator.
20345 (vfml_op): New code attribute.
20346 (VFMLHALVES): New int iterator.
20347 (VFML, VFMLSEL): New mode attributes.
20348 (V_reg): Define mapping for V2HF.
20349 (V_hi, V_lo): New mode attributes.
20350 (VF_constraint): Likewise.
20351 (vfml_half, vfml_half_selector): New int attributes.
20352 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
20354 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
20355 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
20357 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
20358 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
20359 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
20360 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
20362 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
20363 Document new effective target and option set.
20365 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20367 * config/arm/arm-cpus.in (armv8_4): New feature.
20368 (ARMv8_4a): New fgroup.
20369 (armv8.4-a): New arch.
20370 * config/arm/arm-tables.opt: Regenerate.
20371 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
20372 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
20373 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
20374 Add matching rules for -march=armv8.4-a and extensions.
20375 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
20377 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
20380 * config/rx/rx.md (BW): New mode attribute.
20381 (sync_lock_test_and_setsi): Add mode suffix to insn output.
20383 2018-01-11 Richard Biener <rguenther@suse.de>
20385 PR tree-optimization/83435
20386 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
20387 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
20388 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
20390 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
20391 Alan Hayward <alan.hayward@arm.com>
20392 David Sherwood <david.sherwood@arm.com>
20394 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
20396 (aarch64_classify_address): Initialize it. Track polynomial offsets.
20397 (aarch64_print_address_internal): Use it to check for a zero offset.
20399 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
20400 Alan Hayward <alan.hayward@arm.com>
20401 David Sherwood <david.sherwood@arm.com>
20403 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
20404 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
20405 Return a poly_int64 rather than a HOST_WIDE_INT.
20406 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
20407 rather than a HOST_WIDE_INT.
20408 * config/aarch64/aarch64.h (aarch64_frame): Protect with
20409 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
20410 hard_fp_offset, frame_size, initial_adjust, callee_offset and
20411 final_offset from HOST_WIDE_INT to poly_int64.
20412 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
20413 to_constant when getting the number of units in an Advanced SIMD
20415 (aarch64_builtin_vectorized_function): Check for a constant number
20417 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
20419 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
20420 attribute instead of GET_MODE_NUNITS.
20421 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
20422 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
20423 GET_MODE_SIZE for fixed-size registers.
20424 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
20425 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
20426 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
20427 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
20428 (aarch64_print_operand, aarch64_print_address_internal)
20429 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
20430 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
20431 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
20432 Handle polynomial GET_MODE_SIZE.
20433 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
20434 wider than SImode without modification.
20435 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
20436 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
20437 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
20438 passing and returning SVE modes.
20439 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
20440 rather than GEN_INT.
20441 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
20442 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
20443 (aarch64_allocate_and_probe_stack_space): Likewise.
20444 (aarch64_layout_frame): Cope with polynomial offsets.
20445 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
20446 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
20447 polynomial offsets.
20448 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
20449 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
20450 poly_int64 rather than a HOST_WIDE_INT.
20451 (aarch64_get_separate_components, aarch64_process_components)
20452 (aarch64_expand_prologue, aarch64_expand_epilogue)
20453 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
20454 (aarch64_anchor_offset): New function, split out from...
20455 (aarch64_legitimize_address): ...here.
20456 (aarch64_builtin_vectorization_cost): Handle polynomial
20457 TYPE_VECTOR_SUBPARTS.
20458 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
20460 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
20461 number of elements from the PARALLEL rather than the mode.
20462 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
20463 rather than GET_MODE_BITSIZE.
20464 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
20465 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
20466 (aarch64_expand_vec_perm_const_1): Handle polynomial
20467 d->perm.length () and d->perm elements.
20468 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
20469 Apply to_constant to d->perm elements.
20470 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
20471 polynomial CONST_VECTOR_NUNITS.
20472 (aarch64_move_pointer): Take amount as a poly_int64 rather
20474 (aarch64_progress_pointer): Avoid temporary variable.
20475 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
20476 the mode attribute instead of GET_MODE.
20478 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
20479 Alan Hayward <alan.hayward@arm.com>
20480 David Sherwood <david.sherwood@arm.com>
20482 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
20483 x exists before using it.
20484 (aarch64_add_constant_internal): Rename to...
20485 (aarch64_add_offset_1): ...this. Replace regnum with separate
20486 src and dest rtxes. Handle the case in which they're different,
20487 including when the offset is zero. Replace scratchreg with an rtx.
20488 Use 2 additions if there is no spare register into which we can
20489 move a 16-bit constant.
20490 (aarch64_add_constant): Delete.
20491 (aarch64_add_offset): Replace reg with separate src and dest
20492 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
20493 Use aarch64_add_offset_1.
20494 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
20495 an rtx rather than an int. Take the delta as a poly_int64
20496 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
20497 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
20498 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
20499 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
20500 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
20501 and aarch64_add_sp.
20502 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
20503 aarch64_add_constant.
20505 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
20507 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
20508 Use scalar_float_mode.
20510 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
20512 * config/aarch64/aarch64-simd.md
20513 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
20514 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
20515 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
20516 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
20517 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
20518 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
20519 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
20520 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
20521 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
20522 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
20524 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
20527 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
20528 targ_options->x_arm_arch_string is non NULL.
20530 2018-01-11 Tamar Christina <tamar.christina@arm.com>
20532 * config/aarch64/aarch64.h
20533 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
20535 2018-01-11 Sudakshina Das <sudi.das@arm.com>
20538 * expmed.c (emit_store_flag_force): Swap if const op0
20539 and change VOIDmode to mode of op0.
20541 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
20543 PR rtl-optimization/83761
20544 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
20545 than bytes to mode_for_size.
20547 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
20549 PR middle-end/83189
20550 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
20551 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
20554 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
20556 PR middle-end/83575
20557 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
20558 when in layout mode.
20559 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
20560 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
20563 2018-01-10 Michael Collison <michael.collison@arm.com>
20565 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
20566 * config/aarch64/aarch64-option-extension.def: Add
20567 AARCH64_OPT_EXTENSION of 'fp16fml'.
20568 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
20569 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
20570 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
20571 * config/aarch64/constraints.md (Ui7): New constraint.
20572 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
20573 (VFMLA_SEL_W): Ditto.
20576 (VFMLA16_LOW): New int iterator.
20577 (VFMLA16_HIGH): Ditto.
20578 (UNSPEC_FMLAL): New unspec.
20579 (UNSPEC_FMLSL): Ditto.
20580 (UNSPEC_FMLAL2): Ditto.
20581 (UNSPEC_FMLSL2): Ditto.
20582 (f16mac): New code attribute.
20583 * config/aarch64/aarch64-simd-builtins.def
20584 (aarch64_fmlal_lowv2sf): Ditto.
20585 (aarch64_fmlsl_lowv2sf): Ditto.
20586 (aarch64_fmlalq_lowv4sf): Ditto.
20587 (aarch64_fmlslq_lowv4sf): Ditto.
20588 (aarch64_fmlal_highv2sf): Ditto.
20589 (aarch64_fmlsl_highv2sf): Ditto.
20590 (aarch64_fmlalq_highv4sf): Ditto.
20591 (aarch64_fmlslq_highv4sf): Ditto.
20592 (aarch64_fmlal_lane_lowv2sf): Ditto.
20593 (aarch64_fmlsl_lane_lowv2sf): Ditto.
20594 (aarch64_fmlal_laneq_lowv2sf): Ditto.
20595 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
20596 (aarch64_fmlalq_lane_lowv4sf): Ditto.
20597 (aarch64_fmlsl_lane_lowv4sf): Ditto.
20598 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
20599 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
20600 (aarch64_fmlal_lane_highv2sf): Ditto.
20601 (aarch64_fmlsl_lane_highv2sf): Ditto.
20602 (aarch64_fmlal_laneq_highv2sf): Ditto.
20603 (aarch64_fmlsl_laneq_highv2sf): Ditto.
20604 (aarch64_fmlalq_lane_highv4sf): Ditto.
20605 (aarch64_fmlsl_lane_highv4sf): Ditto.
20606 (aarch64_fmlalq_laneq_highv4sf): Ditto.
20607 (aarch64_fmlsl_laneq_highv4sf): Ditto.
20608 * config/aarch64/aarch64-simd.md:
20609 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
20610 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
20611 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
20612 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
20613 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
20614 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
20615 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
20616 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
20617 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
20618 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
20619 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
20620 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
20621 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
20622 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
20623 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
20624 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
20625 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
20626 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
20627 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
20628 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
20629 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
20630 (vfmlsl_low_u32): Ditto.
20631 (vfmlalq_low_u32): Ditto.
20632 (vfmlslq_low_u32): Ditto.
20633 (vfmlal_high_u32): Ditto.
20634 (vfmlsl_high_u32): Ditto.
20635 (vfmlalq_high_u32): Ditto.
20636 (vfmlslq_high_u32): Ditto.
20637 (vfmlal_lane_low_u32): Ditto.
20638 (vfmlsl_lane_low_u32): Ditto.
20639 (vfmlal_laneq_low_u32): Ditto.
20640 (vfmlsl_laneq_low_u32): Ditto.
20641 (vfmlalq_lane_low_u32): Ditto.
20642 (vfmlslq_lane_low_u32): Ditto.
20643 (vfmlalq_laneq_low_u32): Ditto.
20644 (vfmlslq_laneq_low_u32): Ditto.
20645 (vfmlal_lane_high_u32): Ditto.
20646 (vfmlsl_lane_high_u32): Ditto.
20647 (vfmlal_laneq_high_u32): Ditto.
20648 (vfmlsl_laneq_high_u32): Ditto.
20649 (vfmlalq_lane_high_u32): Ditto.
20650 (vfmlslq_lane_high_u32): Ditto.
20651 (vfmlalq_laneq_high_u32): Ditto.
20652 (vfmlslq_laneq_high_u32): Ditto.
20653 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
20654 (AARCH64_FL_FOR_ARCH8_4): New.
20655 (AARCH64_ISA_F16FML): New ISA flag.
20656 (TARGET_F16FML): New feature flag for fp16fml.
20657 (doc/invoke.texi): Document new fp16fml option.
20659 2018-01-10 Michael Collison <michael.collison@arm.com>
20661 * config/aarch64/aarch64-builtins.c:
20662 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
20663 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
20664 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
20665 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
20666 (AARCH64_ISA_SHA3): New ISA flag.
20667 (TARGET_SHA3): New feature flag for sha3.
20668 * config/aarch64/iterators.md (sha512_op): New int attribute.
20669 (CRYPTO_SHA512): New int iterator.
20670 (UNSPEC_SHA512H): New unspec.
20671 (UNSPEC_SHA512H2): Ditto.
20672 (UNSPEC_SHA512SU0): Ditto.
20673 (UNSPEC_SHA512SU1): Ditto.
20674 * config/aarch64/aarch64-simd-builtins.def
20675 (aarch64_crypto_sha512hqv2di): New builtin.
20676 (aarch64_crypto_sha512h2qv2di): Ditto.
20677 (aarch64_crypto_sha512su0qv2di): Ditto.
20678 (aarch64_crypto_sha512su1qv2di): Ditto.
20679 (aarch64_eor3qv8hi): Ditto.
20680 (aarch64_rax1qv2di): Ditto.
20681 (aarch64_xarqv2di): Ditto.
20682 (aarch64_bcaxqv8hi): Ditto.
20683 * config/aarch64/aarch64-simd.md:
20684 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
20685 (aarch64_crypto_sha512su0qv2di): Ditto.
20686 (aarch64_crypto_sha512su1qv2di): Ditto.
20687 (aarch64_eor3qv8hi): Ditto.
20688 (aarch64_rax1qv2di): Ditto.
20689 (aarch64_xarqv2di): Ditto.
20690 (aarch64_bcaxqv8hi): Ditto.
20691 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
20692 (vsha512h2q_u64): Ditto.
20693 (vsha512su0q_u64): Ditto.
20694 (vsha512su1q_u64): Ditto.
20695 (veor3q_u16): Ditto.
20696 (vrax1q_u64): Ditto.
20697 (vxarq_u64): Ditto.
20698 (vbcaxq_u16): Ditto.
20699 * config/arm/types.md (crypto_sha512): New type attribute.
20700 (crypto_sha3): Ditto.
20701 (doc/invoke.texi): Document new sha3 option.
20703 2018-01-10 Michael Collison <michael.collison@arm.com>
20705 * config/aarch64/aarch64-builtins.c:
20706 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
20707 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
20708 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
20709 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
20710 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
20711 (AARCH64_ISA_SM4): New ISA flag.
20712 (TARGET_SM4): New feature flag for sm4.
20713 * config/aarch64/aarch64-simd-builtins.def
20714 (aarch64_sm3ss1qv4si): Ditto.
20715 (aarch64_sm3tt1aq4si): Ditto.
20716 (aarch64_sm3tt1bq4si): Ditto.
20717 (aarch64_sm3tt2aq4si): Ditto.
20718 (aarch64_sm3tt2bq4si): Ditto.
20719 (aarch64_sm3partw1qv4si): Ditto.
20720 (aarch64_sm3partw2qv4si): Ditto.
20721 (aarch64_sm4eqv4si): Ditto.
20722 (aarch64_sm4ekeyqv4si): Ditto.
20723 * config/aarch64/aarch64-simd.md:
20724 (aarch64_sm3ss1qv4si): Ditto.
20725 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
20726 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
20727 (aarch64_sm4eqv4si): Ditto.
20728 (aarch64_sm4ekeyqv4si): Ditto.
20729 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
20730 (sm3part_op): Ditto.
20731 (CRYPTO_SM3TT): Ditto.
20732 (CRYPTO_SM3PART): Ditto.
20733 (UNSPEC_SM3SS1): New unspec.
20734 (UNSPEC_SM3TT1A): Ditto.
20735 (UNSPEC_SM3TT1B): Ditto.
20736 (UNSPEC_SM3TT2A): Ditto.
20737 (UNSPEC_SM3TT2B): Ditto.
20738 (UNSPEC_SM3PARTW1): Ditto.
20739 (UNSPEC_SM3PARTW2): Ditto.
20740 (UNSPEC_SM4E): Ditto.
20741 (UNSPEC_SM4EKEY): Ditto.
20742 * config/aarch64/constraints.md (Ui2): New constraint.
20743 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
20744 * config/arm/types.md (crypto_sm3): New type attribute.
20745 (crypto_sm4): Ditto.
20746 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
20747 (vsm3tt1aq_u32): Ditto.
20748 (vsm3tt1bq_u32): Ditto.
20749 (vsm3tt2aq_u32): Ditto.
20750 (vsm3tt2bq_u32): Ditto.
20751 (vsm3partw1q_u32): Ditto.
20752 (vsm3partw2q_u32): Ditto.
20753 (vsm4eq_u32): Ditto.
20754 (vsm4ekeyq_u32): Ditto.
20755 (doc/invoke.texi): Document new sm4 option.
20757 2018-01-10 Michael Collison <michael.collison@arm.com>
20759 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
20760 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
20761 (AARCH64_FL_FOR_ARCH8_4): New.
20762 (AARCH64_FL_V8_4): New flag.
20763 (doc/invoke.texi): Document new armv8.4-a option.
20765 2018-01-10 Michael Collison <michael.collison@arm.com>
20767 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
20768 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
20769 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
20770 * config/aarch64/aarch64-option-extension.def: Add
20771 AARCH64_OPT_EXTENSION of 'sha2'.
20772 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
20773 (crypto): Disable sha2 and aes if crypto disabled.
20774 (crypto): Enable aes and sha2 if enabled.
20775 (simd): Disable sha2 and aes if simd disabled.
20776 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
20778 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
20779 (TARGET_SHA2): New feature flag for sha2.
20780 (TARGET_AES): New feature flag for aes.
20781 * config/aarch64/aarch64-simd.md:
20782 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
20783 conditional on TARGET_AES.
20784 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
20785 (aarch64_crypto_sha1hsi): Make pattern conditional
20787 (aarch64_crypto_sha1hv4si): Ditto.
20788 (aarch64_be_crypto_sha1hv4si): Ditto.
20789 (aarch64_crypto_sha1su1v4si): Ditto.
20790 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
20791 (aarch64_crypto_sha1su0v4si): Ditto.
20792 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
20793 (aarch64_crypto_sha256su0v4si): Ditto.
20794 (aarch64_crypto_sha256su1v4si): Ditto.
20795 (doc/invoke.texi): Document new aes and sha2 options.
20797 2018-01-10 Martin Sebor <msebor@redhat.com>
20799 PR tree-optimization/83781
20800 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
20803 2018-01-11 Martin Sebor <msebor@gmail.com>
20804 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
20806 PR tree-optimization/83501
20807 PR tree-optimization/81703
20809 * tree-ssa-strlen.c (get_string_cst): Rename...
20810 (get_string_len): ...to this. Handle global constants.
20811 (handle_char_store): Adjust.
20813 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
20814 Jim Wilson <jimw@sifive.com>
20816 * config/riscv/riscv-protos.h (riscv_output_return): New.
20817 * config/riscv/riscv.c (struct machine_function): New naked_p field.
20818 (riscv_attribute_table, riscv_output_return),
20819 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
20820 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
20821 (riscv_compute_frame_info): Only compute frame->mask if not a naked
20823 (riscv_expand_prologue): Add early return for naked function.
20824 (riscv_expand_epilogue): Likewise.
20825 (riscv_function_ok_for_sibcall): Return false for naked function.
20826 (riscv_set_current_function): New.
20827 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
20828 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
20829 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
20830 * doc/extend.texi (RISC-V Function Attributes): New.
20832 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
20834 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
20835 check for 128-bit long double before checking TCmode.
20836 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
20837 128-bit long doubles before checking TFmode or TCmode.
20838 (FLOAT128_IBM_P): Likewise.
20840 2018-01-10 Martin Sebor <msebor@redhat.com>
20842 PR tree-optimization/83671
20843 * builtins.c (c_strlen): Unconditionally return zero for the empty
20845 Use -Warray-bounds for warnings.
20846 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
20847 for non-constant array indices with COMPONENT_REF, arrays of
20848 arrays, and pointers to arrays.
20849 (gimple_fold_builtin_strlen): Determine and set length range for
20850 non-constant character arrays.
20852 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
20854 PR middle-end/81897
20855 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
20858 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
20860 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
20862 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
20865 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
20866 VECTOR_MEM_ALTIVEC_OR_VSX_P.
20867 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
20868 indexed_or_indirect_operand predicate.
20869 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
20870 (*vsx_le_perm_load_v8hi): Likewise.
20871 (*vsx_le_perm_load_v16qi): Likewise.
20872 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
20873 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
20874 (*vsx_le_perm_store_v8hi): Likewise.
20875 (*vsx_le_perm_store_v16qi): Likewise.
20876 (eight unnamed splitters): Likewise.
20878 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
20880 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
20881 * config/rs6000/emmintrin.h: Likewise.
20882 * config/rs6000/mmintrin.h: Likewise.
20883 * config/rs6000/xmmintrin.h: Likewise.
20885 2018-01-10 David Malcolm <dmalcolm@redhat.com>
20888 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
20890 * tree.c (tree_nop_conversion): Return true for location wrapper
20892 (maybe_wrap_with_location): New function.
20893 (selftest::check_strip_nops): New function.
20894 (selftest::test_location_wrappers): New function.
20895 (selftest::tree_c_tests): Call it.
20896 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
20897 (maybe_wrap_with_location): New decl.
20898 (EXPR_LOCATION_WRAPPER_P): New macro.
20899 (location_wrapper_p): New inline function.
20900 (tree_strip_any_location_wrapper): New inline function.
20902 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
20905 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
20906 stack_realign_offset for the largest alignment of stack slot
20908 (ix86_find_max_used_stack_alignment): New function.
20909 (ix86_finalize_stack_frame_flags): Use it. Set
20910 max_used_stack_alignment if we don't realign stack.
20911 * config/i386/i386.h (machine_function): Add
20912 max_used_stack_alignment.
20914 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
20916 * config/arm/arm.opt (-mbranch-cost): New option.
20917 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
20920 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
20923 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
20924 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
20926 2018-01-10 Richard Biener <rguenther@suse.de>
20929 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
20930 early out so it also covers the case where we have a non-NULL
20933 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
20935 PR tree-optimization/83753
20936 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
20937 for non-strided grouped accesses if the number of elements is 1.
20939 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
20942 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
20943 * i386.h (TARGET_USE_GATHER): Define.
20944 * x86-tune.def (X86_TUNE_USE_GATHER): New.
20946 2018-01-10 Martin Liska <mliska@suse.cz>
20949 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
20950 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
20952 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
20953 CLEANUP_NO_PARTITIONING is not set.
20955 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
20957 * doc/rtl.texi: Remove documentation of (const ...) wrappers
20958 for vectors, as a partial revert of r254296.
20959 * rtl.h (const_vec_p): Delete.
20960 (const_vec_duplicate_p): Don't test for vector CONSTs.
20961 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
20962 * expmed.c (make_tree): Likewise.
20965 * common.md (E, F): Use CONSTANT_P instead of checking for
20967 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
20968 checking for CONST_VECTOR.
20970 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
20972 PR middle-end/83575
20973 * predict.c (force_edge_cold): Handle in more sane way edges
20974 with no prediction.
20976 2018-01-09 Carl Love <cel@us.ibm.com>
20978 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
20980 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
20981 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
20982 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
20983 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
20984 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
20985 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
20986 * config/rs6000/rs6000-protos.h: Add extern defition for
20987 rs6000_generate_float2_double_code.
20988 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
20990 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
20991 (float2_v2df): Add define_expand.
20993 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
20996 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
20997 op_mode in the force_to_mode call.
20999 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
21001 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
21002 instead of checking each element individually.
21003 (aarch64_evpc_uzp): Likewise.
21004 (aarch64_evpc_zip): Likewise.
21005 (aarch64_evpc_ext): Likewise.
21006 (aarch64_evpc_rev): Likewise.
21007 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
21008 instead of checking each element individually. Return true without
21010 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
21011 whether all selected elements come from the same input, instead of
21012 checking each element individually. Remove calls to gen_rtx_REG,
21013 start_sequence and end_sequence and instead assert that no rtl is
21016 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
21018 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
21019 order of HIGH and CONST checks.
21021 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
21023 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
21024 if the destination isn't an SSA_NAME.
21026 2018-01-09 Richard Biener <rguenther@suse.de>
21028 PR tree-optimization/83668
21029 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
21031 (canonicalize_loop_form): ... here, renamed from ...
21032 (canonicalize_loop_closed_ssa_form): ... this and amended to
21033 swap successor edges for loop exit blocks to make us use
21034 the RPO order we need for initial schedule generation.
21036 2018-01-09 Joseph Myers <joseph@codesourcery.com>
21038 PR tree-optimization/64811
21039 * match.pd: When optimizing comparisons with Inf, avoid
21040 introducing or losing exceptions from comparisons with NaN.
21042 2018-01-09 Martin Liska <mliska@suse.cz>
21045 * asan.c (shadow_mem_size): Add gcc_assert.
21047 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
21049 Don't save registers in main().
21052 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
21053 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
21054 * config/avr/avr.c (avr_set_current_function): Don't error if
21055 naked, OS_task or OS_main are specified at the same time.
21056 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
21058 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
21060 * common/config/avr/avr-common.c (avr_option_optimization_table):
21061 Switch on -mmain-is-OS_task for optimizing compilations.
21063 2018-01-09 Richard Biener <rguenther@suse.de>
21065 PR tree-optimization/83572
21066 * graphite.c: Include cfganal.h.
21067 (graphite_transform_loops): Connect infinite loops to exit
21068 and remove fake edges at the end.
21070 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
21072 * ipa-inline.c (edge_badness): Revert accidental checkin.
21074 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
21077 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
21078 symbols; not inline clones.
21080 2018-01-09 Jakub Jelinek <jakub@redhat.com>
21083 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
21084 hard registers. Formatting fixes.
21086 PR preprocessor/83722
21087 * gcc.c (try_generate_repro): Pass
21088 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
21089 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
21092 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
21093 Kito Cheng <kito.cheng@gmail.com>
21095 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
21096 (riscv_leaf_function_p): Delete.
21097 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
21099 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
21101 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
21103 (do_ifelse): New function.
21104 (do_isel): New function.
21105 (do_sub3): New function.
21106 (do_add3): New function.
21107 (do_load_mask_compare): New function.
21108 (do_overlap_load_compare): New function.
21109 (expand_compare_loop): New function.
21110 (expand_block_compare): Call expand_compare_loop() when appropriate.
21111 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
21112 option description.
21113 (-mblock-compare-inline-loop-limit): New option.
21115 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
21118 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
21119 Reverse order of second and third operands in first alternative.
21120 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
21121 of first and second elements in UNSPEC_VPERMR vector.
21122 (altivec_expand_vec_perm_le): Likewise.
21124 2018-01-08 Jeff Law <law@redhat.com>
21126 PR rtl-optimizatin/81308
21127 * tree-switch-conversion.c (cfg_altered): New file scoped static.
21128 (process_switch): If group_case_labels makes a change, then set
21130 (pass_convert_switch::execute): If a switch is converted, then
21131 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
21133 PR rtl-optimization/81308
21134 * recog.c (split_all_insns): Conditionally cleanup the CFG after
21137 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
21139 PR target/83663 - Revert r255946
21140 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
21141 generation for cases where splatting a value is not useful.
21142 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
21143 across a vec_duplicate and a paradoxical subreg forming a vector
21144 mode to a vec_concat.
21146 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21148 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
21149 -march=armv8.3-a variants.
21150 * config/arm/t-multilib: Likewise.
21151 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
21153 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
21155 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
21157 (cceq_ior_compare_complement): Give it a name so I can use it, and
21158 change boolean_or_operator predicate to boolean_operator so it can
21159 be used to generate a crand.
21160 (eqne): New code iterator.
21161 (bd/bd_neg): New code_attrs.
21162 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
21163 a single define_insn.
21164 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
21165 decrement (bdnzt/bdnzf/bdzt/bdzf).
21166 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
21167 with the new names of the branch decrement patterns, and added the
21168 names of the branch decrement conditional patterns.
21170 2018-01-08 Richard Biener <rguenther@suse.de>
21172 PR tree-optimization/83563
21173 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
21176 2018-01-08 Richard Biener <rguenther@suse.de>
21178 PR middle-end/83713
21179 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
21181 2018-01-08 Richard Biener <rguenther@suse.de>
21183 PR tree-optimization/83685
21184 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
21185 references to abnormals.
21187 2018-01-08 Richard Biener <rguenther@suse.de>
21190 * dwarf2out.c (output_indirect_strings): Handle empty
21191 skeleton_debug_str_hash.
21192 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
21194 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
21196 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
21197 (emit_store_direct): Likewise.
21198 (arc_trampoline_adjust_address): Likewise.
21199 (arc_asm_trampoline_template): New function.
21200 (arc_initialize_trampoline): Use asm_trampoline_template.
21201 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
21202 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
21203 * config/arc/arc.md (flush_icache): Delete pattern.
21205 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
21207 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
21208 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
21211 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
21214 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
21215 by not USED_FOR_TARGET.
21216 (make_pass_resolve_sw_modes): Likewise.
21218 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
21220 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
21223 2018-01-08 Richard Biener <rguenther@suse.de>
21225 PR middle-end/83580
21226 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
21228 2018-01-08 Richard Biener <rguenther@suse.de>
21230 PR middle-end/83517
21231 * match.pd ((t * 2) / 2) -> t): Add missing :c.
21233 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
21235 PR middle-end/81897
21236 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
21237 basic blocks with a small number of successors.
21238 (convert_control_dep_chain_into_preds): Improve handling of
21240 (dump_predicates): Split apart into...
21241 (dump_pred_chain): ...here...
21242 (dump_pred_info): ...and here.
21243 (can_one_predicate_be_invalidated_p): Add debugging printfs.
21244 (can_chain_union_be_invalidated_p): Improve check for invalidation
21246 (uninit_uses_cannot_happen): Avoid unnecessary if
21247 convert_control_dep_chain_into_preds yielded nothing.
21249 2018-01-06 Martin Sebor <msebor@redhat.com>
21251 PR tree-optimization/83640
21252 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
21253 subtracting negative offset from size.
21254 (builtin_access::overlap): Adjust offset bounds of the access to fall
21255 within the size of the object if possible.
21257 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
21259 PR rtl-optimization/83699
21260 * expmed.c (extract_bit_field_1): Restrict the vector usage of
21261 extract_bit_field_as_subreg to cases in which the extracted
21262 value is also a vector.
21264 * lra-constraints.c (process_alt_operands): Test for the equivalence
21265 substitutions when detecting a possible reload cycle.
21267 2018-01-06 Jakub Jelinek <jakub@redhat.com>
21270 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
21271 by default if flag_selective_schedling{,2}. Formatting fixes.
21273 PR rtl-optimization/83682
21274 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
21275 if it has non-VECTOR_MODE element mode.
21276 (vec_duplicate_p): Likewise.
21278 PR middle-end/83694
21279 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
21280 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
21282 2018-01-05 Jakub Jelinek <jakub@redhat.com>
21285 * config/i386/i386-builtin.def
21286 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
21287 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
21288 Require also OPTION_MASK_ISA_AVX512F in addition to
21289 OPTION_MASK_ISA_GFNI.
21290 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
21291 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
21292 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
21293 to OPTION_MASK_ISA_GFNI.
21294 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
21295 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
21296 OPTION_MASK_ISA_AVX512BW.
21297 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
21298 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
21299 addition to OPTION_MASK_ISA_GFNI.
21300 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
21301 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
21302 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
21303 to OPTION_MASK_ISA_GFNI.
21304 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
21305 a requirement for all ISAs rather than any of them with a few
21307 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
21309 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
21310 bitmasks to be enabled with 3 exceptions, instead of requiring any
21311 enabled ISA with lots of exceptions.
21312 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
21313 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
21314 Change avx512bw in isa attribute to avx512f.
21315 * config/i386/sgxintrin.h: Add license boilerplate.
21316 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
21317 to __AVX512F__ and __AVX512VL to __AVX512VL__.
21318 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
21319 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
21321 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
21322 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
21323 temporarily sse2 rather than sse if not enabled already.
21326 * config/i386/sse.md (VI248_VLBW): Rename to ...
21327 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
21328 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
21329 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
21330 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
21331 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
21332 mode iterator instead of VI248_VLBW.
21334 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
21336 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
21337 (record_modified): Skip clobbers; add debug output.
21338 (param_change_prob): Use sreal frequencies.
21340 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
21342 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
21343 punt for user-aligned variables.
21345 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
21347 * tree-chrec.c (chrec_contains_symbols): Return true for
21350 2018-01-05 Sudakshina Das <sudi.das@arm.com>
21353 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
21354 of (x|y) == x for BICS pattern.
21356 2018-01-05 Jakub Jelinek <jakub@redhat.com>
21358 PR tree-optimization/83605
21359 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
21360 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
21363 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
21365 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
21366 * config/epiphany/rtems.h: New file.
21368 2018-01-04 Jakub Jelinek <jakub@redhat.com>
21369 Uros Bizjak <ubizjak@gmail.com>
21372 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
21373 QIreg_operand instead of register_operand predicate.
21374 * config/i386/i386.c (ix86_rop_should_change_byte_p,
21375 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
21376 comments instead of -fmitigate[-_]rop.
21378 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
21381 * cgraphunit.c (symbol_table::compile): Switch to text_section
21382 before calling assembly_start debug hook.
21383 * run-rtl-passes.c (run_rtl_passes): Likewise.
21386 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
21388 * tree-vrp.c (extract_range_from_binary_expr_1): Check
21389 range_int_cst_p rather than !symbolic_range_p before calling
21390 extract_range_from_multiplicative_op_1.
21392 2018-01-04 Jeff Law <law@redhat.com>
21394 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
21395 redundant test in assertion.
21397 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
21399 * doc/rtl.texi: Document machine_mode wrapper classes.
21401 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
21403 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
21404 using tree_to_uhwi.
21406 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
21408 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
21409 the VEC_PERM_EXPR fold to fail.
21411 2018-01-04 Jakub Jelinek <jakub@redhat.com>
21414 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
21415 to switched_sections.
21417 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
21420 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
21421 test for d.testing.
21423 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
21426 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
21427 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
21429 2018-01-04 Jakub Jelinek <jakub@redhat.com>
21432 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
21433 is BLKmode and bitpos not zero or mode change is needed.
21435 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
21438 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
21441 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
21444 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
21445 instead of MULT rtx. Update all corresponding splitters.
21446 (*saddl_se): Ditto.
21447 (*ssub<modesuffix>): Ditto.
21448 (*ssubl_se): Ditto.
21449 (*cmp_sadd_di): Update split patterns.
21450 (*cmp_sadd_si): Ditto.
21451 (*cmp_sadd_sidi): Ditto.
21452 (*cmp_ssub_di): Ditto.
21453 (*cmp_ssub_si): Ditto.
21454 (*cmp_ssub_sidi): Ditto.
21455 * config/alpha/predicates.md (const23_operand): New predicate.
21456 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
21457 Look for ASHIFT, not MULT inner operand.
21458 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
21460 2018-01-04 Martin Liska <mliska@suse.cz>
21462 PR gcov-profile/83669
21463 * gcov.c (output_intermediate_file): Add version to intermediate
21465 * doc/gcov.texi: Document new field 'version' in intermediate
21466 file format. Fix location of '-k' option of gcov command.
21468 2018-01-04 Martin Liska <mliska@suse.cz>
21471 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
21473 2018-01-04 Jakub Jelinek <jakub@redhat.com>
21475 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
21477 2018-01-03 Martin Sebor <msebor@redhat.com>
21479 PR tree-optimization/83655
21480 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
21481 checking calls with invalid arguments.
21483 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21485 * tree-vect-stmts.c (vect_get_store_rhs): New function.
21486 (vectorizable_mask_load_store): Delete.
21487 (vectorizable_call): Return false for masked loads and stores.
21488 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
21489 instead of gimple_assign_rhs1.
21490 (vectorizable_load): Handle IFN_MASK_LOAD.
21491 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
21493 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21495 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
21497 (vectorizable_mask_load_store): ...here.
21498 (vectorizable_load): ...and here.
21500 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21502 * tree-vect-stmts.c (vect_build_all_ones_mask)
21503 (vect_build_zero_merge_argument): New functions, split out from...
21504 (vectorizable_load): ...here.
21506 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21508 * tree-vect-stmts.c (vect_check_store_rhs): New function,
21510 (vectorizable_mask_load_store): ...here.
21511 (vectorizable_store): ...and here.
21513 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21515 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
21517 (vectorizable_mask_load_store): ...here.
21519 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21521 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
21522 (vect_model_store_cost): Take a vec_load_store_type instead of a
21524 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
21525 (vect_model_store_cost): Take a vec_load_store_type instead of a
21527 (vectorizable_mask_load_store): Update accordingly.
21528 (vectorizable_store): Likewise.
21529 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
21531 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21533 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
21534 IFN_MASK_LOAD calls here rather than...
21535 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
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 * expmed.c (extract_bit_field_1): For vector extracts,
21542 fall back to extract_bit_field_as_subreg if vec_extract
21545 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21546 Alan Hayward <alan.hayward@arm.com>
21547 David Sherwood <david.sherwood@arm.com>
21549 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
21550 they are variable or constant sized.
21551 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
21552 slots for constant-sized data.
21554 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21555 Alan Hayward <alan.hayward@arm.com>
21556 David Sherwood <david.sherwood@arm.com>
21558 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
21559 handling COND_EXPRs with boolean comparisons, try to find a better
21560 basis for the mask type than the boolean itself.
21562 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21564 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
21565 is calculated and how it can be overridden.
21566 * genmodes.c (max_bitsize_mode_any_mode): New variable.
21567 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
21569 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
21572 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21573 Alan Hayward <alan.hayward@arm.com>
21574 David Sherwood <david.sherwood@arm.com>
21576 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
21577 Remove the mode argument.
21578 (aarch64_simd_valid_immediate): Remove the mode and inverse
21580 * config/aarch64/iterators.md (bitsize): New iterator.
21581 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
21582 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
21583 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
21584 aarch64_simd_valid_immediate.
21585 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
21586 (aarch64_reg_or_bic_imm): Likewise.
21587 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
21588 with an insn_type enum and msl with a modifier_type enum.
21589 Replace element_width with a scalar_mode. Change the shift
21590 to unsigned int. Add constructors for scalar_float_mode and
21591 scalar_int_mode elements.
21592 (aarch64_vect_float_const_representable_p): Delete.
21593 (aarch64_can_const_movi_rtx_p)
21594 (aarch64_simd_scalar_immediate_valid_for_move)
21595 (aarch64_simd_make_constant): Update call to
21596 aarch64_simd_valid_immediate.
21597 (aarch64_advsimd_valid_immediate_hs): New function.
21598 (aarch64_advsimd_valid_immediate): Likewise.
21599 (aarch64_simd_valid_immediate): Remove mode and inverse
21600 arguments. Rewrite to use the above. Use const_vec_duplicate_p
21601 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
21602 and aarch64_float_const_representable_p on the result.
21603 (aarch64_output_simd_mov_immediate): Remove mode argument.
21604 Update call to aarch64_simd_valid_immediate and use of
21605 simd_immediate_info.
21606 (aarch64_output_scalar_simd_mov_immediate): Update call
21609 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21610 Alan Hayward <alan.hayward@arm.com>
21611 David Sherwood <david.sherwood@arm.com>
21613 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
21614 (mode_nunits): Likewise CONST_MODE_NUNITS.
21615 * machmode.def (ADJUST_NUNITS): Document.
21616 * genmodes.c (mode_data::need_nunits_adj): New field.
21617 (blank_mode): Update accordingly.
21618 (adj_nunits): New variable.
21619 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
21621 (emit_mode_size_inline): Set need_bytesize_adj for all modes
21622 listed in adj_nunits.
21623 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
21624 listed in adj_nunits. Don't emit case statements for such modes.
21625 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
21626 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
21627 nothing if adj_nunits is nonnull.
21628 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
21629 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
21630 (emit_mode_fbit): Update use of print_maybe_const_decl.
21631 (emit_move_size): Likewise. Treat the array as non-const
21633 (emit_mode_adjustments): Handle adj_nunits.
21635 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21637 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
21638 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
21639 (VECTOR_MODES): Use it.
21640 (make_vector_modes): Take the prefix as an argument.
21642 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21643 Alan Hayward <alan.hayward@arm.com>
21644 David Sherwood <david.sherwood@arm.com>
21646 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
21647 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
21648 for MODE_VECTOR_BOOL.
21649 * machmode.def (VECTOR_BOOL_MODE): Document.
21650 * genmodes.c (VECTOR_BOOL_MODE): New macro.
21651 (make_vector_bool_mode): New function.
21652 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
21654 * lto-streamer-in.c (lto_input_mode_table): Likewise.
21655 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
21657 * stor-layout.c (int_mode_for_mode): Likewise.
21658 * tree.c (build_vector_type_for_mode): Likewise.
21659 * varasm.c (output_constant_pool_2): Likewise.
21660 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
21661 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
21662 for MODE_VECTOR_BOOL.
21663 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
21664 of mode class checks.
21665 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
21666 instead of a list of mode class checks.
21667 (expand_vector_scalar_condition): Likewise.
21668 (type_for_widest_vector_mode): Handle BImode as an inner mode.
21670 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21671 Alan Hayward <alan.hayward@arm.com>
21672 David Sherwood <david.sherwood@arm.com>
21674 * machmode.h (mode_size): Change from unsigned short to
21676 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
21677 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
21678 or if measurement_type is not polynomial.
21679 (fixed_size_mode::includes_p): Check for constant-sized modes.
21680 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
21681 return a poly_uint16 rather than an unsigned short.
21682 (emit_mode_size): Change the type of mode_size from unsigned short
21683 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
21684 (emit_mode_adjustments): Cope with polynomial vector sizes.
21685 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
21687 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
21689 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
21690 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
21691 * caller-save.c (setup_save_areas): Likewise.
21692 (replace_reg_with_saved_mem): Likewise.
21693 * calls.c (emit_library_call_value_1): Likewise.
21694 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
21695 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
21696 (gen_lowpart_for_combine): Likewise.
21697 * convert.c (convert_to_integer_1): Likewise.
21698 * cse.c (equiv_constant, cse_insn): Likewise.
21699 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
21700 (cselib_subst_to_values): Likewise.
21701 * dce.c (word_dce_process_block): Likewise.
21702 * df-problems.c (df_word_lr_mark_ref): Likewise.
21703 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
21704 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
21705 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
21706 (rtl_for_decl_location): Likewise.
21707 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
21708 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
21709 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
21710 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
21711 (expand_expr_real_1): Likewise.
21712 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
21713 (pad_below): Likewise.
21714 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
21715 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
21716 * ira.c (get_subreg_tracking_sizes): Likewise.
21717 * ira-build.c (ira_create_allocno_objects): Likewise.
21718 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
21719 (ira_sort_regnos_for_alter_reg): Likewise.
21720 * ira-costs.c (record_operand_costs): Likewise.
21721 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
21722 (resolve_simple_move): Likewise.
21723 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
21724 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
21725 (lra_constraints): Likewise.
21726 (CONST_POOL_OK_P): Reject variable-sized modes.
21727 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
21728 (add_pseudo_to_slot, lra_spill): Likewise.
21729 * omp-low.c (omp_clause_aligned_alignment): Likewise.
21730 * optabs-query.c (get_best_extraction_insn): Likewise.
21731 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
21732 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
21733 (expand_mult_highpart, valid_multiword_target_p): Likewise.
21734 * recog.c (offsettable_address_addr_space_p): Likewise.
21735 * regcprop.c (maybe_mode_change): Likewise.
21736 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
21737 * regrename.c (build_def_use): Likewise.
21738 * regstat.c (dump_reg_info): Likewise.
21739 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
21740 (find_reloads, find_reloads_subreg_address): Likewise.
21741 * reload1.c (eliminate_regs_1): Likewise.
21742 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
21743 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
21744 (simplify_binary_operation_1, simplify_subreg): Likewise.
21745 * targhooks.c (default_function_arg_padding): Likewise.
21746 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
21747 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
21748 (verify_gimple_assign_ternary): Likewise.
21749 * tree-inline.c (estimate_move_cost): Likewise.
21750 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
21751 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
21752 (get_address_cost_ainc): Likewise.
21753 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
21754 (vect_supportable_dr_alignment): Likewise.
21755 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
21756 (vectorizable_reduction): Likewise.
21757 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
21758 (vectorizable_operation, vectorizable_load): Likewise.
21759 * tree.c (build_same_sized_truth_vector_type): Likewise.
21760 * valtrack.c (cleanup_auto_inc_dec): Likewise.
21761 * var-tracking.c (emit_note_insn_var_location): Likewise.
21762 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
21763 (ADDR_VEC_ALIGN): Likewise.
21765 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21766 Alan Hayward <alan.hayward@arm.com>
21767 David Sherwood <david.sherwood@arm.com>
21769 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
21771 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
21772 or if measurement_type is polynomial.
21773 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
21774 * combine.c (make_extraction): Likewise.
21775 * dse.c (find_shift_sequence): Likewise.
21776 * dwarf2out.c (mem_loc_descriptor): Likewise.
21777 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
21778 (extract_bit_field, extract_low_bits): Likewise.
21779 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
21780 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
21781 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
21782 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
21783 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
21784 * reload.c (find_reloads): Likewise.
21785 * reload1.c (alter_reg): Likewise.
21786 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
21787 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
21788 * tree-if-conv.c (predicate_mem_writes): Likewise.
21789 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
21790 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
21791 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
21792 * valtrack.c (dead_debug_insert_temp): Likewise.
21793 * varasm.c (mergeable_constant_section): Likewise.
21794 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
21796 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21797 Alan Hayward <alan.hayward@arm.com>
21798 David Sherwood <david.sherwood@arm.com>
21800 * expr.c (expand_assignment): Cope with polynomial mode sizes
21801 when assigning to a CONCAT.
21803 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21804 Alan Hayward <alan.hayward@arm.com>
21805 David Sherwood <david.sherwood@arm.com>
21807 * machmode.h (mode_precision): Change from unsigned short to
21809 (mode_to_precision): Return a poly_uint16 rather than an unsigned
21811 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
21812 or if measurement_type is not polynomial.
21813 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
21814 in which the mode is already known to be a scalar_int_mode.
21815 * genmodes.c (emit_mode_precision): Change the type of mode_precision
21816 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
21818 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
21819 for GET_MODE_PRECISION.
21820 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
21821 for GET_MODE_PRECISION.
21822 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
21824 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
21825 (expand_field_assignment, make_extraction): Likewise.
21826 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
21827 (get_last_value): Likewise.
21828 * convert.c (convert_to_integer_1): Likewise.
21829 * cse.c (cse_insn): Likewise.
21830 * expr.c (expand_expr_real_1): Likewise.
21831 * lra-constraints.c (simplify_operand_subreg): Likewise.
21832 * optabs-query.c (can_atomic_load_p): Likewise.
21833 * optabs.c (expand_atomic_load): Likewise.
21834 (expand_atomic_store): Likewise.
21835 * ree.c (combine_reaching_defs): Likewise.
21836 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
21837 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
21838 * tree.h (type_has_mode_precision_p): Likewise.
21839 * ubsan.c (instrument_si_overflow): Likewise.
21841 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21842 Alan Hayward <alan.hayward@arm.com>
21843 David Sherwood <david.sherwood@arm.com>
21845 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
21846 polynomial numbers of units.
21847 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
21848 (valid_vector_subparts_p): New function.
21849 (build_vector_type): Remove temporary shim and take the number
21850 of units as a poly_uint64 rather than an int.
21851 (build_opaque_vector_type): Take the number of units as a
21852 poly_uint64 rather than an int.
21853 * tree.c (build_vector_from_ctor): Handle polynomial
21854 TYPE_VECTOR_SUBPARTS.
21855 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
21856 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
21857 (build_vector_from_val): If the number of units is variable,
21858 use build_vec_duplicate_cst for constant operands and
21859 VEC_DUPLICATE_EXPR otherwise.
21860 (make_vector_type): Remove temporary is_constant ().
21861 (build_vector_type, build_opaque_vector_type): Take the number of
21862 units as a poly_uint64 rather than an int.
21863 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
21865 * cfgexpand.c (expand_debug_expr): Likewise.
21866 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
21867 (store_constructor, expand_expr_real_1): Likewise.
21868 (const_scalar_mask_from_tree): Likewise.
21869 * fold-const-call.c (fold_const_reduction): Likewise.
21870 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
21871 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
21872 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
21873 (fold_relational_const): Likewise.
21874 (native_interpret_vector): Likewise. Change the size from an
21875 int to an unsigned int.
21876 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
21877 TYPE_VECTOR_SUBPARTS.
21878 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
21879 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
21880 duplicating a non-constant operand into a variable-length vector.
21881 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
21882 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
21883 * ipa-icf.c (sem_variable::equals): Likewise.
21884 * match.pd: Likewise.
21885 * omp-simd-clone.c (simd_clone_subparts): Likewise.
21886 * print-tree.c (print_node): Likewise.
21887 * stor-layout.c (layout_type): Likewise.
21888 * targhooks.c (default_builtin_vectorization_cost): Likewise.
21889 * tree-cfg.c (verify_gimple_comparison): Likewise.
21890 (verify_gimple_assign_binary): Likewise.
21891 (verify_gimple_assign_ternary): Likewise.
21892 (verify_gimple_assign_single): Likewise.
21893 * tree-pretty-print.c (dump_generic_node): Likewise.
21894 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
21895 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
21896 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
21897 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
21898 (vect_shift_permute_load_chain): Likewise.
21899 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
21900 (expand_vector_condition, optimize_vector_constructor): Likewise.
21901 (lower_vec_perm, get_compute_type): Likewise.
21902 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
21903 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
21904 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
21905 (vect_recog_mask_conversion_pattern): Likewise.
21906 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
21907 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
21908 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
21909 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
21910 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
21911 (vectorizable_shift, vectorizable_operation, vectorizable_store)
21912 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
21913 (supportable_widening_operation): Likewise.
21914 (supportable_narrowing_operation): Likewise.
21915 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
21917 * varasm.c (output_constant): Likewise.
21919 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21920 Alan Hayward <alan.hayward@arm.com>
21921 David Sherwood <david.sherwood@arm.com>
21923 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
21924 so that both the length == 3 and length != 3 cases set up their
21925 own permute vectors. Add comments explaining why we know the
21926 number of elements is constant.
21927 (vect_permute_load_chain): Likewise.
21929 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21930 Alan Hayward <alan.hayward@arm.com>
21931 David Sherwood <david.sherwood@arm.com>
21933 * machmode.h (mode_nunits): Change from unsigned char to
21935 (ONLY_FIXED_SIZE_MODES): New macro.
21936 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
21937 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
21938 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
21940 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
21941 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
21942 or if measurement_type is not polynomial.
21943 * genmodes.c (ZERO_COEFFS): New macro.
21944 (emit_mode_nunits_inline): Make mode_nunits_inline return a
21946 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
21947 Use ZERO_COEFFS when emitting initializers.
21948 * data-streamer.h (bp_pack_poly_value): New function.
21949 (bp_unpack_poly_value): Likewise.
21950 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
21951 for GET_MODE_NUNITS.
21952 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
21953 for GET_MODE_NUNITS.
21954 * tree.c (make_vector_type): Remove temporary shim and make
21955 the real function take the number of units as a poly_uint64
21956 rather than an int.
21957 (build_vector_type_for_mode): Handle polynomial nunits.
21958 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
21959 * emit-rtl.c (const_vec_series_p_1): Likewise.
21960 (gen_rtx_CONST_VECTOR): Likewise.
21961 * fold-const.c (test_vec_duplicate_folding): Likewise.
21962 * genrecog.c (validate_pattern): Likewise.
21963 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
21964 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
21965 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
21966 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
21967 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
21968 * rtlanal.c (subreg_get_info): Likewise.
21969 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
21970 (vect_grouped_load_supported): Likewise.
21971 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
21972 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
21973 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
21974 (simplify_const_unary_operation, simplify_binary_operation_1)
21975 (simplify_const_binary_operation, simplify_ternary_operation)
21976 (test_vector_ops_duplicate, test_vector_ops): Likewise.
21977 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
21978 instead of CONST_VECTOR_NUNITS.
21979 * varasm.c (output_constant_pool_2): Likewise.
21980 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
21981 explicit-encoded elements in the XVEC for variable-length vectors.
21983 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21985 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
21987 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21988 Alan Hayward <alan.hayward@arm.com>
21989 David Sherwood <david.sherwood@arm.com>
21991 * coretypes.h (fixed_size_mode): Declare.
21992 (fixed_size_mode_pod): New typedef.
21993 * builtins.h (target_builtins::x_apply_args_mode)
21994 (target_builtins::x_apply_result_mode): Change type to
21995 fixed_size_mode_pod.
21996 * builtins.c (apply_args_size, apply_result_size, result_vector)
21997 (expand_builtin_apply_args_1, expand_builtin_apply)
21998 (expand_builtin_return): Update accordingly.
22000 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22002 * cse.c (hash_rtx_cb): Hash only the encoded elements.
22003 * cselib.c (cselib_hash_rtx): Likewise.
22004 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
22005 CONST_VECTOR encoding.
22007 2018-01-03 Jakub Jelinek <jakub@redhat.com>
22008 Jeff Law <law@redhat.com>
22011 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
22012 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
22013 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
22014 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
22017 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
22018 explicitly probe *sp in a noreturn function if there were any callee
22019 register saves or frame pointer is needed.
22021 2018-01-03 Jakub Jelinek <jakub@redhat.com>
22024 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
22025 BLKmode for ternary, binary or unary expressions.
22028 * var-tracking.c (delete_vta_debug_insn): New inline function.
22029 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
22030 insns from get_insns () to NULL instead of each bb separately.
22031 Use delete_vta_debug_insn. No longer static.
22032 (vt_debug_insns_local, variable_tracking_main_1): Adjust
22033 delete_vta_debug_insns callers.
22034 * rtl.h (delete_vta_debug_insns): Declare.
22035 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
22036 instead of variable_tracking_main.
22038 2018-01-03 Martin Sebor <msebor@redhat.com>
22040 PR tree-optimization/83603
22041 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
22042 arguments past the endof the argument list in functions declared
22043 without a prototype.
22044 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
22045 Avoid checking when arguments are null.
22047 2018-01-03 Martin Sebor <msebor@redhat.com>
22050 * doc/extend.texi (attribute const): Fix a typo.
22051 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
22052 issuing -Wsuggest-attribute for void functions.
22054 2018-01-03 Martin Sebor <msebor@redhat.com>
22056 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
22057 offset_int::from instead of wide_int::to_shwi.
22058 (maybe_diag_overlap): Remove assertion.
22059 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
22060 * gimple-ssa-sprintf.c (format_directive): Same.
22061 (parse_directive): Same.
22062 (sprintf_dom_walker::compute_format_length): Same.
22063 (try_substitute_return_value): Same.
22065 2018-01-03 Jeff Law <law@redhat.com>
22067 PR middle-end/83654
22068 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
22069 non-constant residual for zero at runtime and avoid probing in
22070 that case. Reorganize code for trailing problem to mirror handling
22073 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
22075 PR tree-optimization/83501
22076 * tree-ssa-strlen.c (get_string_cst): New.
22077 (handle_char_store): Call get_string_cst.
22079 2018-01-03 Martin Liska <mliska@suse.cz>
22081 PR tree-optimization/83593
22082 * tree-ssa-strlen.c: Include tree-cfg.h.
22083 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
22084 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
22085 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
22087 (strlen_dom_walker::before_dom_children): Call
22088 gimple_purge_dead_eh_edges. Dump tranformation with details
22090 (strlen_dom_walker::before_dom_children): Update call by adding
22091 new argument cleanup_eh.
22092 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
22094 2018-01-03 Martin Liska <mliska@suse.cz>
22097 * cif-code.def (VARIADIC_THUNK): New enum value.
22098 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
22101 2018-01-03 Jan Beulich <jbeulich@suse.com>
22103 * sse.md (mov<mode>_internal): Tighten condition for when to use
22104 vmovdqu<ssescalarsize> for TI and OI modes.
22106 2018-01-03 Jakub Jelinek <jakub@redhat.com>
22108 Update copyright years.
22110 2018-01-03 Martin Liska <mliska@suse.cz>
22113 * ipa-visibility.c (function_and_variable_visibility): Skip
22114 functions with noipa attribure.
22116 2018-01-03 Jakub Jelinek <jakub@redhat.com>
22118 * gcc.c (process_command): Update copyright notice dates.
22119 * gcov-dump.c (print_version): Ditto.
22120 * gcov.c (print_version): Ditto.
22121 * gcov-tool.c (print_version): Ditto.
22122 * gengtype.c (create_file): Ditto.
22123 * doc/cpp.texi: Bump @copying's copyright year.
22124 * doc/cppinternals.texi: Ditto.
22125 * doc/gcc.texi: Ditto.
22126 * doc/gccint.texi: Ditto.
22127 * doc/gcov.texi: Ditto.
22128 * doc/install.texi: Ditto.
22129 * doc/invoke.texi: Ditto.
22131 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22133 * vector-builder.h (vector_builder::m_full_nelts): Change from
22134 unsigned int to poly_uint64.
22135 (vector_builder::full_nelts): Update prototype accordingly.
22136 (vector_builder::new_vector): Likewise.
22137 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
22138 (vector_builder::operator ==): Likewise.
22139 (vector_builder::finalize): Likewise.
22140 * int-vector-builder.h (int_vector_builder::int_vector_builder):
22141 Take the number of elements as a poly_uint64 rather than an
22143 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
22144 from unsigned int to poly_uint64.
22145 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
22146 (vec_perm_indices::new_vector): Likewise.
22147 (vec_perm_indices::length): Likewise.
22148 (vec_perm_indices::nelts_per_input): Likewise.
22149 (vec_perm_indices::input_nelts): Likewise.
22150 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
22151 number of elements per input as a poly_uint64 rather than an
22152 unsigned int. Use the original encoding for variable-length
22153 vectors, rather than clamping each individual element.
22154 For the second and subsequent elements in each pattern,
22155 clamp the step and base before clamping their sum.
22156 (vec_perm_indices::series_p): Handle polynomial element counts.
22157 (vec_perm_indices::all_in_range_p): Likewise.
22158 (vec_perm_indices_to_tree): Likewise.
22159 (vec_perm_indices_to_rtx): Likewise.
22160 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
22161 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
22162 (tree_vector_builder::new_binary_operation): Handle polynomial
22163 element counts. Return false if we need to know the number
22164 of elements at compile time.
22165 * fold-const.c (fold_vec_perm): Punt if the number of elements
22166 isn't known at compile time.
22168 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22170 * vec-perm-indices.h (vec_perm_builder): Change element type
22171 from HOST_WIDE_INT to poly_int64.
22172 (vec_perm_indices::element_type): Update accordingly.
22173 (vec_perm_indices::clamp): Handle polynomial element_types.
22174 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
22175 (vec_perm_indices::all_in_range_p): Likewise.
22176 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
22178 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
22179 polynomial vec_perm_indices element types.
22180 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
22181 * fold-const.c (fold_vec_perm): Likewise.
22182 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
22183 * tree-vect-generic.c (lower_vec_perm): Likewise.
22184 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
22185 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
22186 element type to HOST_WIDE_INT.
22188 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22189 Alan Hayward <alan.hayward@arm.com>
22190 David Sherwood <david.sherwood@arm.com>
22192 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
22193 rather than an int. Use plus_constant.
22194 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
22195 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
22197 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22198 Alan Hayward <alan.hayward@arm.com>
22199 David Sherwood <david.sherwood@arm.com>
22201 * calls.c (emit_call_1, expand_call): Change struct_value_size from
22202 a HOST_WIDE_INT to a poly_int64.
22204 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22205 Alan Hayward <alan.hayward@arm.com>
22206 David Sherwood <david.sherwood@arm.com>
22208 * calls.c (load_register_parameters): Cope with polynomial
22209 mode sizes. Require a constant size for BLKmode parameters
22210 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
22211 forces a parameter to be padded at the lsb end in order to
22212 fill a complete number of words, require the parameter size
22213 to be ordered wrt UNITS_PER_WORD.
22215 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22216 Alan Hayward <alan.hayward@arm.com>
22217 David Sherwood <david.sherwood@arm.com>
22219 * reload1.c (spill_stack_slot_width): Change element type
22220 from unsigned int to poly_uint64_pod.
22221 (alter_reg): Treat mode sizes as polynomial.
22223 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22224 Alan Hayward <alan.hayward@arm.com>
22225 David Sherwood <david.sherwood@arm.com>
22227 * reload.c (complex_word_subreg_p): New function.
22228 (reload_inner_reg_of_subreg, push_reload): Use it.
22230 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22231 Alan Hayward <alan.hayward@arm.com>
22232 David Sherwood <david.sherwood@arm.com>
22234 * lra-constraints.c (process_alt_operands): Reject matched
22235 operands whose sizes aren't ordered.
22236 (match_reload): Refer to this check here.
22238 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22239 Alan Hayward <alan.hayward@arm.com>
22240 David Sherwood <david.sherwood@arm.com>
22242 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
22243 that the mode size is in the set {1, 2, 4, 8, 16}.
22245 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22246 Alan Hayward <alan.hayward@arm.com>
22247 David Sherwood <david.sherwood@arm.com>
22249 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
22250 Use plus_constant instead of gen_rtx_PLUS.
22252 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22253 Alan Hayward <alan.hayward@arm.com>
22254 David Sherwood <david.sherwood@arm.com>
22256 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
22257 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
22258 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
22259 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
22260 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
22261 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
22262 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
22263 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
22264 * config/i386/i386.c (ix86_push_rounding): ...this new function.
22265 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
22267 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
22268 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
22269 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
22270 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
22271 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
22272 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
22273 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
22274 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
22275 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
22276 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
22278 * expr.c (emit_move_resolve_push): Treat the input and result
22279 of PUSH_ROUNDING as a poly_int64.
22280 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
22281 (emit_push_insn): Likewise.
22282 * lra-eliminations.c (mark_not_eliminable): Likewise.
22283 * recog.c (push_operand): Likewise.
22284 * reload1.c (elimination_effects): Likewise.
22285 * rtlanal.c (nonzero_bits1): Likewise.
22286 * calls.c (store_one_arg): Likewise. Require the padding to be
22287 known at compile time.
22289 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22290 Alan Hayward <alan.hayward@arm.com>
22291 David Sherwood <david.sherwood@arm.com>
22293 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
22294 Use plus_constant instead of gen_rtx_PLUS.
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 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
22301 rather than an int.
22303 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22304 Alan Hayward <alan.hayward@arm.com>
22305 David Sherwood <david.sherwood@arm.com>
22307 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
22308 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
22309 via stack temporaries. Treat the mode size as polynomial too.
22311 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22312 Alan Hayward <alan.hayward@arm.com>
22313 David Sherwood <david.sherwood@arm.com>
22315 * expr.c (expand_expr_real_2): When handling conversions involving
22316 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
22317 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
22318 as a poly_uint64 too.
22320 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22321 Alan Hayward <alan.hayward@arm.com>
22322 David Sherwood <david.sherwood@arm.com>
22324 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
22326 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22327 Alan Hayward <alan.hayward@arm.com>
22328 David Sherwood <david.sherwood@arm.com>
22330 * combine.c (can_change_dest_mode): Handle polynomial
22331 REGMODE_NATURAL_SIZE.
22332 * expmed.c (store_bit_field_1): Likewise.
22333 * expr.c (store_constructor): Likewise.
22334 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
22335 and polynomial REGMODE_NATURAL_SIZE.
22336 (gen_lowpart_common): Likewise.
22337 * reginfo.c (record_subregs_of_mode): Likewise.
22338 * rtlanal.c (read_modify_subreg_p): Likewise.
22340 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22341 Alan Hayward <alan.hayward@arm.com>
22342 David Sherwood <david.sherwood@arm.com>
22344 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
22345 numbers of elements.
22347 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22348 Alan Hayward <alan.hayward@arm.com>
22349 David Sherwood <david.sherwood@arm.com>
22351 * match.pd: Cope with polynomial numbers of vector elements.
22353 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22354 Alan Hayward <alan.hayward@arm.com>
22355 David Sherwood <david.sherwood@arm.com>
22357 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
22358 in a POINTER_PLUS_EXPR.
22360 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22361 Alan Hayward <alan.hayward@arm.com>
22362 David Sherwood <david.sherwood@arm.com>
22364 * omp-simd-clone.c (simd_clone_subparts): New function.
22365 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
22366 (ipa_simd_modify_function_body): Likewise.
22368 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22369 Alan Hayward <alan.hayward@arm.com>
22370 David Sherwood <david.sherwood@arm.com>
22372 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
22373 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
22374 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
22375 (expand_vector_condition, vector_element): Likewise.
22376 (subparts_gt): New function.
22377 (get_compute_type): Use subparts_gt.
22378 (count_type_subparts): Delete.
22379 (expand_vector_operations_1): Use subparts_gt instead of
22380 count_type_subparts.
22382 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22383 Alan Hayward <alan.hayward@arm.com>
22384 David Sherwood <david.sherwood@arm.com>
22386 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
22387 (vect_compile_time_alias): ...this new function. Do the calculation
22388 on poly_ints rather than trees.
22389 (vect_prune_runtime_alias_test_list): Update call accordingly.
22391 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22392 Alan Hayward <alan.hayward@arm.com>
22393 David Sherwood <david.sherwood@arm.com>
22395 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
22397 (vect_schedule_slp_instance): Likewise.
22399 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22400 Alan Hayward <alan.hayward@arm.com>
22401 David Sherwood <david.sherwood@arm.com>
22403 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
22404 constant and extern definitions for variable-length vectors.
22405 (vect_get_constant_vectors): Note that the number of units
22406 is known to be constant.
22408 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22409 Alan Hayward <alan.hayward@arm.com>
22410 David Sherwood <david.sherwood@arm.com>
22412 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
22413 of units as polynomial. Choose between WIDE and NARROW based
22416 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22417 Alan Hayward <alan.hayward@arm.com>
22418 David Sherwood <david.sherwood@arm.com>
22420 * tree-vect-stmts.c (simd_clone_subparts): New function.
22421 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
22423 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22424 Alan Hayward <alan.hayward@arm.com>
22425 David Sherwood <david.sherwood@arm.com>
22427 * tree-vect-stmts.c (vectorizable_call): Treat the number of
22428 vectors as polynomial. Use build_index_vector for
22429 IFN_GOMP_SIMD_LANE.
22431 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22432 Alan Hayward <alan.hayward@arm.com>
22433 David Sherwood <david.sherwood@arm.com>
22435 * tree-vect-stmts.c (get_load_store_type): Treat the number of
22436 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
22437 for variable-length vectors.
22438 (vectorizable_mask_load_store): Treat the number of units as
22439 polynomial, asserting that it is constant if the condition has
22440 already been enforced.
22441 (vectorizable_store, vectorizable_load): Likewise.
22443 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22444 Alan Hayward <alan.hayward@arm.com>
22445 David Sherwood <david.sherwood@arm.com>
22447 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
22448 of units as polynomial. Punt if we can't tell at compile time
22449 which vector contains the final result.
22451 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22452 Alan Hayward <alan.hayward@arm.com>
22453 David Sherwood <david.sherwood@arm.com>
22455 * tree-vect-loop.c (vectorizable_induction): Treat the number
22456 of units as polynomial. Punt on SLP inductions. Use an integer
22457 VEC_SERIES_EXPR for variable-length integer reductions. Use a
22458 cast of such a series for variable-length floating-point
22461 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22462 Alan Hayward <alan.hayward@arm.com>
22463 David Sherwood <david.sherwood@arm.com>
22465 * tree.h (build_index_vector): Declare.
22466 * tree.c (build_index_vector): New function.
22467 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
22468 of units as polynomial, forcibly converting it to a constant if
22469 vectorizable_reduction has already enforced the condition.
22470 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
22471 to create a {1,2,3,...} vector.
22472 (vectorizable_reduction): Treat the number of units as polynomial.
22473 Choose vectype_in based on the largest scalar element size rather
22474 than the smallest number of units. Enforce the restrictions
22477 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22478 Alan Hayward <alan.hayward@arm.com>
22479 David Sherwood <david.sherwood@arm.com>
22481 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
22482 number of units as polynomial.
22484 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22485 Alan Hayward <alan.hayward@arm.com>
22486 David Sherwood <david.sherwood@arm.com>
22488 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
22489 * target.def (autovectorize_vector_sizes): Return the vector sizes
22490 by pointer, using vector_sizes rather than a bitmask.
22491 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
22492 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
22493 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
22495 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
22496 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
22497 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
22498 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
22499 * omp-general.c (omp_max_vf): Likewise.
22500 * omp-low.c (omp_clause_aligned_alignment): Likewise.
22501 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
22502 * tree-vect-loop.c (vect_analyze_loop): Likewise.
22503 * tree-vect-slp.c (vect_slp_bb): Likewise.
22504 * doc/tm.texi: Regenerate.
22505 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
22507 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
22508 the vector size as a poly_uint64 rather than an unsigned int.
22509 (current_vector_size): Change from an unsigned int to a poly_uint64.
22510 (get_vectype_for_scalar_type): Update accordingly.
22511 * tree.h (build_truth_vector_type): Take the size and number of
22512 units as a poly_uint64 rather than an unsigned int.
22513 (build_vector_type): Add a temporary overload that takes
22514 the number of units as a poly_uint64 rather than an unsigned int.
22515 * tree.c (make_vector_type): Likewise.
22516 (build_truth_vector_type): Take the number of units as a poly_uint64
22517 rather than an unsigned int.
22519 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22520 Alan Hayward <alan.hayward@arm.com>
22521 David Sherwood <david.sherwood@arm.com>
22523 * target.def (get_mask_mode): Take the number of units and length
22524 as poly_uint64s rather than unsigned ints.
22525 * targhooks.h (default_get_mask_mode): Update accordingly.
22526 * targhooks.c (default_get_mask_mode): Likewise.
22527 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
22528 * doc/tm.texi: Regenerate.
22530 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22531 Alan Hayward <alan.hayward@arm.com>
22532 David Sherwood <david.sherwood@arm.com>
22534 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
22535 * omp-general.c (omp_max_vf): Likewise.
22536 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
22537 (expand_omp_simd): Handle polynomial safelen.
22538 * omp-low.c (omplow_simd_context): Add a default constructor.
22539 (omplow_simd_context::max_vf): Change from int to poly_uint64.
22540 (lower_rec_simd_input_clauses): Update accordingly.
22541 (lower_rec_input_clauses): Likewise.
22543 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22544 Alan Hayward <alan.hayward@arm.com>
22545 David Sherwood <david.sherwood@arm.com>
22547 * tree-vectorizer.h (vect_nunits_for_cost): New function.
22548 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
22549 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
22550 (vect_analyze_slp_cost): Likewise.
22551 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
22552 (vect_model_load_cost): Likewise.
22554 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22555 Alan Hayward <alan.hayward@arm.com>
22556 David Sherwood <david.sherwood@arm.com>
22558 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
22559 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
22560 from an unsigned int * to a poly_uint64_pod *.
22561 (calculate_unrolling_factor): New function.
22562 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
22564 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22565 Alan Hayward <alan.hayward@arm.com>
22566 David Sherwood <david.sherwood@arm.com>
22568 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
22569 from an unsigned int to a poly_uint64.
22570 (_loop_vec_info::slp_unrolling_factor): Likewise.
22571 (_loop_vec_info::vectorization_factor): Change from an int
22573 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
22574 (vect_get_num_vectors): New function.
22575 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
22576 (vect_get_num_copies): Use vect_get_num_vectors.
22577 (vect_analyze_data_ref_dependences): Change max_vf from an int *
22578 to an unsigned int *.
22579 (vect_analyze_data_refs): Change min_vf from an int * to a
22581 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
22582 than an unsigned HOST_WIDE_INT.
22583 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
22584 (vect_analyze_data_ref_dependence): Change max_vf from an int *
22585 to an unsigned int *.
22586 (vect_analyze_data_ref_dependences): Likewise.
22587 (vect_compute_data_ref_alignment): Handle polynomial vf.
22588 (vect_enhance_data_refs_alignment): Likewise.
22589 (vect_prune_runtime_alias_test_list): Likewise.
22590 (vect_shift_permute_load_chain): Likewise.
22591 (vect_supportable_dr_alignment): Likewise.
22592 (dependence_distance_ge_vf): Take the vectorization factor as a
22593 poly_uint64 rather than an unsigned HOST_WIDE_INT.
22594 (vect_analyze_data_refs): Change min_vf from an int * to a
22596 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
22597 vfm1 as a poly_uint64 rather than an int. Make the same change
22598 for the returned bound_scalar.
22599 (vect_gen_vector_loop_niters): Handle polynomial vf.
22600 (vect_do_peeling): Likewise. Update call to
22601 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
22602 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
22604 * tree-vect-loop.c (vect_determine_vectorization_factor)
22605 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
22606 (vect_get_known_peeling_cost): Likewise.
22607 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
22608 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
22609 (vect_transform_loop): Likewise. Use the lowest possible VF when
22610 updating the upper bounds of the loop.
22611 (vect_min_worthwhile_factor): Make static. Return an unsigned int
22612 rather than an int.
22613 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
22614 polynomial unroll factors.
22615 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
22616 (vect_make_slp_decision): Likewise.
22617 (vect_supported_load_permutation_p): Likewise, and polynomial
22619 (vect_analyze_slp_cost): Handle polynomial vf.
22620 (vect_slp_analyze_node_operations): Likewise.
22621 (vect_slp_analyze_bb_1): Likewise.
22622 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
22623 than an unsigned HOST_WIDE_INT.
22624 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
22625 (vectorizable_load): Handle polynomial vf.
22626 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
22628 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
22630 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22631 Alan Hayward <alan.hayward@arm.com>
22632 David Sherwood <david.sherwood@arm.com>
22634 * match.pd: Handle bit operations involving three constants
22635 and try to fold one pair.
22637 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22639 * tree-vect-loop-manip.c: Include gimple-fold.h.
22640 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
22641 niters_maybe_zero parameters. Handle other cases besides a step of 1.
22642 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
22643 Add a path that uses a step of VF instead of 1, but disable it
22645 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
22646 and niters_no_overflow parameters. Update calls to
22647 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
22648 Create a new SSA name if the latter choses to use a ste other
22649 than zero, and return it via niters_vector_mult_vf_var.
22650 * tree-vect-loop.c (vect_transform_loop): Update calls to
22651 vect_do_peeling, vect_gen_vector_loop_niters and
22652 slpeel_make_loop_iterate_ntimes.
22653 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
22654 (vect_gen_vector_loop_niters): Update declarations after above changes.
22656 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
22658 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
22659 128-bit round to integer instructions.
22660 (ceil<mode>2): Likewise.
22661 (btrunc<mode>2): Likewise.
22662 (round<mode>2): Likewise.
22664 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
22666 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
22667 unaligned VSX load/store on P8/P9.
22668 (expand_block_clear): Allow the use of unaligned VSX
22669 load/store on P8/P9.
22671 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
22673 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
22675 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
22676 swap associated with both a load and a store.
22678 2018-01-02 Andrew Waterman <andrew@sifive.com>
22680 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
22681 * config/riscv/riscv.md (clear_cache): Use it.
22683 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
22685 * web.c: Remove out-of-date comment.
22687 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
22689 * expr.c (fixup_args_size_notes): Check that any existing
22690 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
22691 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
22692 (emit_single_push_insn): ...here.
22694 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
22696 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
22697 (const_vector_encoded_nelts): New function.
22698 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
22699 (const_vector_int_elt, const_vector_elt): Declare.
22700 * emit-rtl.c (const_vector_int_elt_1): New function.
22701 (const_vector_elt): Likewise.
22702 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
22703 of CONST_VECTOR_ELT.
22705 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
22707 * expr.c: Include rtx-vector-builder.h.
22708 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
22709 directly on the tree encoding.
22710 (const_vector_from_tree): Likewise.
22711 * optabs.c: Include rtx-vector-builder.h.
22712 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
22713 sequence of "u" values.
22714 * vec-perm-indices.c: Include rtx-vector-builder.h.
22715 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
22716 directly on the vec_perm_indices encoding.
22718 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
22720 * doc/rtl.texi (const_vector): Describe new encoding scheme.
22721 * Makefile.in (OBJS): Add rtx-vector-builder.o.
22722 * rtx-vector-builder.h: New file.
22723 * rtx-vector-builder.c: Likewise.
22724 * rtl.h (rtx_def::u2): Add a const_vector field.
22725 (CONST_VECTOR_NPATTERNS): New macro.
22726 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
22727 (CONST_VECTOR_DUPLICATE_P): Likewise.
22728 (CONST_VECTOR_STEPPED_P): Likewise.
22729 (CONST_VECTOR_ENCODED_ELT): Likewise.
22730 (const_vec_duplicate_p): Check for a duplicated vector encoding.
22731 (unwrap_const_vec_duplicate): Likewise.
22732 (const_vec_series_p): Check for a non-duplicated vector encoding.
22733 Say that the function only returns true for integer vectors.
22734 * emit-rtl.c: Include rtx-vector-builder.h.
22735 (gen_const_vec_duplicate_1): Delete.
22736 (gen_const_vector): Call gen_const_vec_duplicate instead of
22737 gen_const_vec_duplicate_1.
22738 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
22739 (gen_const_vec_duplicate): Use rtx_vector_builder.
22740 (gen_const_vec_series): Likewise.
22741 (gen_rtx_CONST_VECTOR): Likewise.
22742 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
22743 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
22744 Build a new vector rather than modifying a CONST_VECTOR in-place.
22745 (handle_special_swappables): Update call accordingly.
22746 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
22747 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
22748 Build a new vector rather than modifying a CONST_VECTOR in-place.
22749 (handle_special_swappables): Update call accordingly.
22751 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
22753 * simplify-rtx.c (simplify_const_binary_operation): Use
22754 CONST_VECTOR_ELT instead of XVECEXP.
22756 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
22758 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
22759 the selector elements to be different from the data elements
22760 if the selector is a VECTOR_CST.
22761 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
22762 ssizetype for the selector.
22764 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
22766 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
22767 before testing each element individually.
22768 * tree-vect-generic.c (lower_vec_perm): Likewise.
22770 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
22772 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
22773 * selftest-run-tests.c (selftest::run_tests): Call it.
22774 * vector-builder.h (vector_builder::operator ==): New function.
22775 (vector_builder::operator !=): Likewise.
22776 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
22777 (vec_perm_indices::all_from_input_p): New function.
22778 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
22779 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
22780 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
22781 instead of reading the VECTOR_CST directly. Detect whether both
22782 vector inputs are the same before constructing the vec_perm_indices,
22783 and update the number of inputs argument accordingly. Use the
22784 utility functions added above. Only construct sel2 if we need to.
22786 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
22788 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
22789 the broadcast of the low byte.
22790 (expand_mult_highpart): Use an explicit encoding for the permutes.
22791 * optabs-query.c (can_mult_highpart_p): Likewise.
22792 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
22793 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
22794 (vectorizable_bswap): Likewise.
22795 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
22796 explicit encoding for the power-of-2 permutes.
22797 (vect_permute_store_chain): Likewise.
22798 (vect_grouped_load_supported): Likewise.
22799 (vect_permute_load_chain): Likewise.
22801 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
22803 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
22804 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
22805 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
22806 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
22807 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
22808 (vect_gen_perm_mask_any): Likewise.
22810 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
22812 * int-vector-builder.h: New file.
22813 * vec-perm-indices.h: Include int-vector-builder.h.
22814 (vec_perm_indices): Redefine as an int_vector_builder.
22815 (auto_vec_perm_indices): Delete.
22816 (vec_perm_builder): Redefine as a stand-alone class.
22817 (vec_perm_indices::vec_perm_indices): New function.
22818 (vec_perm_indices::clamp): Likewise.
22819 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
22820 (vec_perm_indices::new_vector): New function.
22821 (vec_perm_indices::new_expanded_vector): Update for new
22822 vec_perm_indices class.
22823 (vec_perm_indices::rotate_inputs): New function.
22824 (vec_perm_indices::all_in_range_p): Operate directly on the
22825 encoded form, without computing elided elements.
22826 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
22827 encoding. Update for new vec_perm_indices class.
22828 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
22829 the given vec_perm_builder.
22830 (expand_vec_perm_var): Update vec_perm_builder constructor.
22831 (expand_mult_highpart): Use vec_perm_builder instead of
22832 auto_vec_perm_indices.
22833 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
22834 vec_perm_indices instead of auto_vec_perm_indices. Use a single
22835 or double series encoding as appropriate.
22836 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
22837 vec_perm_indices instead of auto_vec_perm_indices.
22838 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
22839 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
22840 (vect_permute_store_chain): Likewise.
22841 (vect_grouped_load_supported): Likewise.
22842 (vect_permute_load_chain): Likewise.
22843 (vect_shift_permute_load_chain): Likewise.
22844 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
22845 (vect_transform_slp_perm_load): Likewise.
22846 (vect_schedule_slp_instance): Likewise.
22847 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
22848 (vectorizable_mask_load_store): Likewise.
22849 (vectorizable_bswap): Likewise.
22850 (vectorizable_store): Likewise.
22851 (vectorizable_load): Likewise.
22852 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
22853 vec_perm_indices instead of auto_vec_perm_indices. Use
22854 tree_to_vec_perm_builder to read the vector from a tree.
22855 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
22856 vec_perm_builder instead of a vec_perm_indices.
22857 (have_whole_vector_shift): Use vec_perm_builder and
22858 vec_perm_indices instead of auto_vec_perm_indices. Leave the
22859 truncation to calc_vec_perm_mask_for_shift.
22860 (vect_create_epilog_for_reduction): Likewise.
22861 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
22862 from auto_vec_perm_indices to vec_perm_indices.
22863 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
22864 instead of changing individual elements.
22865 (aarch64_vectorize_vec_perm_const): Use new_vector to install
22866 the vector in d.perm.
22867 * config/arm/arm.c (expand_vec_perm_d::perm): Change
22868 from auto_vec_perm_indices to vec_perm_indices.
22869 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
22870 instead of changing individual elements.
22871 (arm_vectorize_vec_perm_const): Use new_vector to install
22872 the vector in d.perm.
22873 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
22874 Update vec_perm_builder constructor.
22875 (rs6000_expand_interleave): Likewise.
22876 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
22877 (rs6000_expand_interleave): Likewise.
22879 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
22881 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
22882 to qimode could truncate the indices.
22883 * optabs.c (expand_vec_perm_var): Likewise.
22885 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
22887 * Makefile.in (OBJS): Add vec-perm-indices.o.
22888 * vec-perm-indices.h: New file.
22889 * vec-perm-indices.c: Likewise.
22890 * target.h (vec_perm_indices): Replace with a forward class
22892 (auto_vec_perm_indices): Move to vec-perm-indices.h.
22893 * optabs.h: Include vec-perm-indices.h.
22894 (expand_vec_perm): Delete.
22895 (selector_fits_mode_p, expand_vec_perm_var): Declare.
22896 (expand_vec_perm_const): Declare.
22897 * target.def (vec_perm_const_ok): Replace with...
22898 (vec_perm_const): ...this new hook.
22899 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
22900 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
22901 * doc/tm.texi: Regenerate.
22902 * optabs.def (vec_perm_const): Delete.
22903 * doc/md.texi (vec_perm_const): Likewise.
22904 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
22905 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
22906 expand_vec_perm for constant permutation vectors. Assert that
22907 the mode of variable permutation vectors is the integer equivalent
22908 of the mode that is being permuted.
22909 * optabs-query.h (selector_fits_mode_p): Declare.
22910 * optabs-query.c: Include vec-perm-indices.h.
22911 (selector_fits_mode_p): New function.
22912 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
22913 is defined, instead of checking whether the vec_perm_const_optab
22914 exists. Use targetm.vectorize.vec_perm_const instead of
22915 targetm.vectorize.vec_perm_const_ok. Check whether the indices
22916 fit in the vector mode before using a variable permute.
22917 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
22918 vec_perm_indices instead of an rtx.
22919 (expand_vec_perm): Replace with...
22920 (expand_vec_perm_const): ...this new function. Take the selector
22921 as a vec_perm_indices rather than an rtx. Also take the mode of
22922 the selector. Update call to shift_amt_for_vec_perm_mask.
22923 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
22924 Use vec_perm_indices::new_expanded_vector to expand the original
22925 selector into bytes. Check whether the indices fit in the vector
22926 mode before using a variable permute.
22927 (expand_vec_perm_var): Make global.
22928 (expand_mult_highpart): Use expand_vec_perm_const.
22929 * fold-const.c: Includes vec-perm-indices.h.
22930 * tree-ssa-forwprop.c: Likewise.
22931 * tree-vect-data-refs.c: Likewise.
22932 * tree-vect-generic.c: Likewise.
22933 * tree-vect-loop.c: Likewise.
22934 * tree-vect-slp.c: Likewise.
22935 * tree-vect-stmts.c: Likewise.
22936 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
22938 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
22939 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
22940 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
22941 (aarch64_vectorize_vec_perm_const): ...this new function.
22942 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
22943 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
22944 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
22945 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
22946 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
22947 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
22948 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
22950 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
22951 check for NEON modes.
22952 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
22953 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
22954 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
22955 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
22957 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
22958 the old VEC_PERM_CONST conditions.
22959 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
22960 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
22961 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
22962 (ia64_vectorize_vec_perm_const_ok): Merge into...
22963 (ia64_vectorize_vec_perm_const): ...this new function.
22964 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
22965 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
22966 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
22967 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
22968 * config/mips/mips.c (mips_expand_vec_perm_const)
22969 (mips_vectorize_vec_perm_const_ok): Merge into...
22970 (mips_vectorize_vec_perm_const): ...this new function.
22971 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
22972 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
22973 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
22974 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
22975 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
22976 (rs6000_expand_vec_perm_const): Delete.
22977 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
22979 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
22980 (altivec_expand_vec_perm_const_le): Take each operand individually.
22981 Operate on constant selectors rather than rtxes.
22982 (altivec_expand_vec_perm_const): Likewise. Update call to
22983 altivec_expand_vec_perm_const_le.
22984 (rs6000_expand_vec_perm_const): Delete.
22985 (rs6000_vectorize_vec_perm_const_ok): Delete.
22986 (rs6000_vectorize_vec_perm_const): New function.
22987 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
22988 an element count and rtx array.
22989 (rs6000_expand_extract_even): Update call accordingly.
22990 (rs6000_expand_interleave): Likewise.
22991 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
22992 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
22993 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
22994 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
22995 (rs6000_expand_vec_perm_const): Delete.
22996 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
22997 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
22998 (altivec_expand_vec_perm_const_le): Take each operand individually.
22999 Operate on constant selectors rather than rtxes.
23000 (altivec_expand_vec_perm_const): Likewise. Update call to
23001 altivec_expand_vec_perm_const_le.
23002 (rs6000_expand_vec_perm_const): Delete.
23003 (rs6000_vectorize_vec_perm_const_ok): Delete.
23004 (rs6000_vectorize_vec_perm_const): New function. Remove stray
23005 reference to the SPE evmerge intructions.
23006 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
23007 an element count and rtx array.
23008 (rs6000_expand_extract_even): Update call accordingly.
23009 (rs6000_expand_interleave): Likewise.
23010 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
23011 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
23013 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
23015 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23017 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
23018 vector mode and that that mode matches the mode of the data
23020 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
23021 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
23022 directly using expand_vec_perm_1 when forcing selectors into
23024 (expand_vec_perm_var): New function, split out from expand_vec_perm.
23026 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23028 * optabs-query.h (can_vec_perm_p): Delete.
23029 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
23030 * optabs-query.c (can_vec_perm_p): Split into...
23031 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
23032 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
23033 particular selector is valid.
23034 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
23035 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
23036 (vect_grouped_load_supported): Likewise.
23037 (vect_shift_permute_load_chain): Likewise.
23038 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
23039 (vect_transform_slp_perm_load): Likewise.
23040 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
23041 (vectorizable_bswap): Likewise.
23042 (vect_gen_perm_mask_checked): Likewise.
23043 * fold-const.c (fold_ternary_loc): Likewise. Don't take
23044 implementations of variable permutation vectors into account
23045 when deciding which selector to use.
23046 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
23047 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
23048 with a false third argument.
23049 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
23050 to test whether the constant selector is valid and can_vec_perm_var_p
23051 to test whether a variable selector is valid.
23053 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23055 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
23056 * optabs-query.c (can_vec_perm_p): Likewise.
23057 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
23058 instead of vec_perm_indices.
23059 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
23060 (vect_gen_perm_mask_checked): Likewise,
23061 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
23062 (vect_gen_perm_mask_checked): Likewise,
23064 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23066 * optabs-query.h (qimode_for_vec_perm): Declare.
23067 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
23068 (qimode_for_vec_perm): ...this new function.
23069 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
23071 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
23073 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
23074 does not have a conditional at the top.
23076 2018-01-02 Richard Biener <rguenther@suse.de>
23078 * ipa-inline.c (big_speedup_p): Fix expression.
23080 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
23083 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
23086 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
23090 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
23091 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
23092 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
23093 cond_taken_branch_cost 3->4.
23095 2018-01-01 Jakub Jelinek <jakub@redhat.com>
23097 PR tree-optimization/83581
23098 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
23099 TODO_cleanup_cfg if any changes have been made.
23101 PR middle-end/83608
23102 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
23103 convert_modes if target mode has the right side, but different mode
23106 PR middle-end/83609
23107 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
23108 last argument when extracting from CONCAT. If either from_real or
23109 from_imag is NULL, use expansion through memory. If result is not
23110 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
23111 the parts directly to inner mode, if even that fails, use expansion
23114 PR middle-end/83623
23115 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
23116 check for bswap in mode rather than HImode and use that in expand_unop
23119 Copyright (C) 2018 Free Software Foundation, Inc.
23121 Copying and distribution of this file, with or without modification,
23122 are permitted in any medium without royalty provided the copyright
23123 notice and this notice are preserved.